Upload
independent
View
0
Download
0
Embed Size (px)
Citation preview
Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101
The accuracy of the ATLAS muon X-ray tomograph
R Avramidou J Berbiers C Boudineau C Dechelette D DrakoulakosC Fabjan S Grau E Gschwendtner J-M Maugain H Rieder
S Rangod F Rohrbach E Sbrissa E Sedykh I Sedykh Y SmirnovL Vertogradov I Vichou
CERN 1211 Geneva 23 Switzerland
Received 15 April 2002 received in revised form 20 August 2002 accepted 20 August 2002
Abstract
A gigantic detector the ATLAS project is under construction at CERN for particle physics research at the Large
Hadron Collider which is to be ready by 2006 An X-ray tomograph has been developed designed and constructed at
CERN in order to control the mechanical quality of the ATLAS muon chambers We reached a measurement accuracy
of 2 mm systematic and 2 mm statistical uncertainties in the horizontal and vertical directions in the working area 220 cm
(horizontal) 60 cm (vertical) Here we describe in detail the fundamental approach of the basic principle chosen to
achieve such good accuracy In order to crosscheck our precision key results of measurements are presented
r 2003 Elsevier Science BV All rights reserved
PACS 0785F 8759F
Keywords X-ray instrumentation Computerized tomography LHC accelerator ATLAS experiment Muon chamber detector
Quality control
1 Introduction
11 Basic physics motivation
A new collider the Large Hadron Collider(LHC) is under construction at CERN and willbe able to deliver proton collisions at 14TeV atluminosities up to 1034 cm2 s1 This is a newenergy frontier it opens a great opportunity forparticle physics towards new discoveries in our
understanding of matter It concerns the mostfundamental questions of the day in particlephysics like the origin of differences in the particlemasses which may be explained with the existenceof a new neutral boson the Higgs Each particlemass quarks and leptons and the Higgs itselfwould then be regarded as the measure of theparticle coupling strength to the Higgs field carriedby the Higgs boson Another domain of researchat the LHC collider is the supersymmetry theorywhich is currently considered as the strongestcompetitor for extensions of the Standard ModelSUSY predicts the existence of partners for allknown particles quarksndashsquarks leptonsndashsleptons
Corresponding author 1224 route Bellevue F-01280 Pre-
vessin-Moens France Fax 41-22-767-4500
E-mail address francoisrohrbachcernch (F Rohrbach)
0168-900203$ - see front matter r 2003 Elsevier Science BV All rights reserved
PII S 0 1 6 8 - 9 0 0 2 ( 0 2 ) 0 1 6 1 1 - X
gluonsndashgluinos etc with masses well suited to theLHC energy range The Minimal SupersymmetricStandard Model calls for more Higgs states(neutral and charged Higgs) Among many otherexperimental physics goals the hunt for SUSYpartners will also be a hot challenge for thedetectors
Entering a new energy domain may also lead tounexpected discoveries as has already occurredmany times in particle physics experiments There-fore the detectors should be ready for all newphysics and open to the greatest discovery to bemade at LHC a surprise
However at the same time the new machinecreates a new challenge for the experiments veryhigh luminosity high radiation environment andvery small cross-sections for discoveries lead tounprecedented experimental conditions in terms ofthe data acquisition rate and signal over back-ground
After several years of Research and Develop-ment large LHC collaborations started the con-struction phase of the huge detectors needed inorder to exploit the full potential of the LHCcollider for physics searches
12 ATLAS at CERN
One of the two very large detectors underpreparation for LHC is the ATLAS detector [1ndash3]
ATLAS has been designed as a general-purposedetector for LHC with efficient and high-resolu-tion tracking excellent calorimetry and precisemuon detection The inner detector is embedded ina 2T superconducting solenoid coil placed insidethe calorimetry The calorimeter is installedbetween the inner tracker and the muon spectro-meter which forms the outer shell of the ATLASdetector
The muon spectrometer surrounding the calori-meter is optimized not only for Higgs search butalso to leave open the possibility to tackle manyother kinds of searches in order to exploit the fullpotential of the LHC It has been designed as ahigh-resolution high-efficiency muon spectro-meter with stand-alone triggering and precisemomentum-measurement capability over a widerange of transverse momenta covering with high
hermeticity a large range of pseudorapidity Thiswas achieved by using an original well-adaptedmagnetic field configuration large superconduct-ing air-core toroids arranged in an eight-foldsymmetry with a bending power ranging from25 Tm at Z frac14 0 to 8Tm at the maximum pseudo-rapidity covered by the ATLAS muon detector
The detector will fill a huge cavern(26m 47m 30m) surrounding the collisionarea about 90m under the ground surface
ATLAS is a very large world-wide collabora-tion at present it involves 150 institutions withabout 2000 participants In terms of size cost andtimescale this detector will be the biggest everbuilt in the world for high-energy physics experi-ments
13 The ATLAS muon spectrometer
The instrumentation of the ATLAS muonspectrometer (see the Muon Spectrometer Techni-cal Design Report [4]) as a sub-detector ofATLAS has been designed by a collaborationfrom 44 institutes in 13 countries
High-momentum final-state muons are amongstthe most promising and robust signatures forsearching for the Higgs The clearest signal forHiggs production would be through one of theleptonic decay modes of the produced Z bosons(H-ZZ-m+mm+m) provided the Higgs masswere above E184GeV [2m(Z)] The branchingratio of this decay mode is small (011) but thesignature of this kind of event will be clear andunambiguous
The performance required for physics in mo-mentum and mass resolutions is at the level of 1The transverse-momentum resolution should beconstant over the full rapidity range
14 Layout
The spectrometer is shown in Figs 1ndash3The design philosophy was to adopt a safe and
industrial approach matching all the physicsrequirements and experimental constraints forgetting a reliable safe and robust muon spectro-meter able to run for at least 10 years withoutmajor trouble
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10184
The conceptual layout of the spectrometer isbased on a system of three large superconductingair-core toroid magnets instrumented withseparate-function trigger and high-precisiontracking chambers This magnet configurationprovides a field that is mostly orthogonalto the muon trajectories producing a maximumbending power which is needed for momen-
tum resolution and charge-sign determinationat the highest energies while minimizing thedegradation of resolution due to multiple scatter-ing
Last but not least another important considera-tion is to produce an economical design thatallows a total surface of about 5000m2 to becovered at an affordable cost
chamberschambers
chambers
chambers
Cathode stripResistive plate
Thin gap
Monitored drift tube
l = ~ 46m
Oslash ~ 24 m
P
beam
P
beam
Fig 1 Three-dimensional sketch of the ATLAS muon spectrometer layout
Fig 2 Side view of one quarter of the muon spectrometer
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 85
15 Momentum measurement
Charged particles with high momentum (above300GeVc) crossing the toroid magnets of themuon spectrometer will have very large radius ofcurvature producing small saggita signals(sE1525mm) over the long muon trajectories(5ndash15m) between three measuring stations Inorder to achieve the required precision in themomentum measurement the saggita must bedetermined with a precision of 15ndash50 mm overmost of the pseudorapidity range a precisionmeasurement of the track coordinates in theprincipal bending direction of the magnetic fieldis provided by Monitored Drift Tubes (MDTs)and Cathode Strip Chambers (CSCs) The MDTchambers are arranged in three layers all aroundthe calorimeter in order to determine the momen-tum with the best possible resolution The CSCsare used for the very forward area where theparticle flux is too high for the drift chambers
With this layout a three-point measurement isavailable (see Fig 4) by installing three stations ofmultilayer MDT chambers covering the fullrapidity gap with high hermeticity The driftMDT chambers are called lsquomonitoredrsquo because acontinuous monitoring of the internal chamberdeformations is done In addition to this on-linesurvey the position in space of each chamber isalso monitored with the required precision(r20 mm)
With this layout the momentum resolution istypically 2ndash3 at 100GeVc over most of thekinematic range
The muon spectrometer is designed for amomentum resolution DpT=pTo1104 pT wherepT is in GeV for pT gt 300GeV To achieve thisresolution by a three-point measurement with thesize and bending power of the ATLAS toroidseach point must be measured with an accuracy ofabout 50 mm This sets the scale for the require-ments on the intrinsic resolution the mechanical
End-captoroid
Barrel toroidcoils
Calorimeters
MDT chambersResistive plate chambers
Inner detector
oslash ~
24 m
Fig 3 Cross view of the muon spectrometer
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10186
precision and the survey accuracy of the muonchambers
The momentum resolution of the spectrometeris limited by the intrinsic detector resolutionMDT calibration errors chamber positioninguncertainties multiple scattering and statisticalfluctuations of energy loss At smaller momenta(below about 300GeVc) the resolution is limitedby multiple scattering to a few per cent at highermomenta (above 300GeVc) it is dominated bychamber precision and alignment
16 The MDT chambers
The task in numbers is impressive the totalnumber of MDTs is E1200 making a total area ofabout 5000m2 The total number of readoutchannels (number of drift tubes) is E400 000 andthe total length of the tubes is E1100 km
The basic detection elements of the MDTchambers are drift chambers made of aluminiumtubes of 30mm diameter and 400 mm wall thick-ness with a 50 mm diameter central WndashRe wireThe tubes are operated with a non-flammableArndashCO2 mixture at 3 bar absolute pressure
The tubes are produced by extrusion from ahard aluminium alloy and are available commer-cially They are closed by endplugs which provideaccurate positioning of the anode wires wiretension gas tightness and electrical and gasconnections The drift tubes can be manufacturedto tight mechanical tolerances which are well-matched to their intrinsic resolution propertiesmostly using automated assembly procedures Thetube lengths vary from 70 to 630 cm
For a good linear spacendashtime relation with amaximum drift time of B500ns the envisagedworking point provides a small Lorentz angle andgood ageing properties due to low gas amplification
X [c
m]
Z [cm]Collision vertex
Pseudo-rapidity Beam lineaxis
Fig 4 Longitudinal cross-section in the bending plane of the spectrometer showing the barrel and end-cap magnet air-coil toroid
configuration It shows the pseudo-rapidity coverage of the muon spectrometer from 0 to 28 and a sketch of the layout principle of the
three detecting muon stations showing the trajectories of a few GeVc positively and negatively charged particles
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 87
Using the timendashdrift distance relationship the single-wire resolution is typically 80mm when operated athigh gas pressure (3bar)
The muon chamber layout was optimized toreduce the number of different chamber sizes
The barrel chambers are of rectangular shapewith sizes from 2 to 10m2
The end-cap chambers are of trapezoidal shapeTheir sizes range from 1 to 10m2 for individualchamber modules and up to 30m2 when severalof them are preassembled for installation Thechamber design must guarantee reliability andstability of construction and operation for theanticipated lifetime of the experiment in anirradiated environment The basic design isshown in Fig 5
To improve the resolution of a chamber beyondthe single-wire limit and to achieve adequateredundancy for pattern recognition the MDTchambers are constructed from 2 4 monolayersof drift tubes for the inner and 2 3 monolayersfor the middle and outer stations The tubes are
arranged in multilayers of three or four mono-layers respectively on either side of a rigidsupport structure The support structures thelsquospacer framesrsquo provide for accurate positioningof the two multilayers with respect to each otherand for mechanical integrity under effects oftemperature and gravity As an example ofdifficulties encountered when seeking to reach therequired resolution the drift tubes which are notmounted vertically are bent slightly in order tomatch their shape to the gravitational sag of thewires
The structural components of the spacerframes are three lsquocrossplatesrsquo to which thedrift tube multilayers are attached and twolsquolong beamsrsquo connecting the crossplates Theframes need to be constructed to a moderatemechanical accuracy of 705mm only accuratepositioning of the drift tubes is provided bythe assembly procedure The chambers willbe attached to the rail structures of the spectro-meter by three-point kinematic supports Oncea chamber is installed in its final location in
Cross plate
Multilayer
In-plane alignment
Longitudinal beam
3-layerMultilayer
micro track
Tube cross section
BE
+
Fig 5 Schematic drawing of a rectangular MDT chamber constructed from multilayers of three monolayers each for installation in
the barrel spectrometer The chambers for the end-cap are of trapezoidal shape but are of similar design otherwise In the toroidal
magnetic field the tubes are oriented essentially parallel to the field lines
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10188
the spectrometer mechanical deformations aremonitored by an in-plane optical system hencethe name lsquoMDTrsquo chambers The spacer framessupport most of the components of the alignmentsystem
MDT chambers are made for robust andreliable operation This is achieved thanks to themechanical concept which provides a solid separa-tion of detecting cells and a good electricalshielding of neighbouring wires
Chambers must withstand very long storagetime followed by very long running periods Inorder to reach the required momentum resolutionall MDT chambers must be constructed with20 mm mechanical accuracy
17 Quality assurance for MDTs
The MDT project in ATLAS is on the scale of avery large industrial project aiming at an excep-tional quality of construction in terms of accuracy(r20 mm) material reliability delivery and assem-bly schedule The X-ray tomograph quality-con-trol platform may be considered as one of the keyinstruments needed for the MDT
The production of the MDT chambers is basedon the concept of very high mechanical precisionand the use of X-ray tomography as a quality-control platform for the mechanically achievedprecision
2 X-ray tomography
The development design construction andtuning of an X-ray tomograph suited forthe MDT quality control was the aim of theX-ray Quality-Control lsquoX-QCrsquo project (see Refs[5ndash11])
21 Basic principle
The choice of X-ray tomography as a quality-control tool for the MDT project was based ontwo essential features
The ability of X-rays to detect 50 mm tungstenwires hidden inside the aluminium drift tubes of
the muon detector Sharp absorption spikes areobserved either as a shadow seen by an X-raydetector placed under the chamber (passivemode) or as a sharp peak signal detected by thechamber itself (active mode) see Fig 6
The possibility to produce narrow collimatedX-ray beams allowing one to reach detectionprofiles with the required peak position detec-tion accuracy (1ndash2 mm)
The value of the absorption coefficient for X-rayphotons is very sensitive to attenuating materialsand to the X-ray energy If d is the density Z theatomic number of the material and W the energyof X-rays the mass absorption coefficient m=dvaries as EethZ=W THORN3 at 40 keV The optimumenergy range for this application is between 40and 60 keV
22 Basic procedure
A narrow X-ray beam is precisely moved acrossa section of the MDT chamber A scintillatorinstalled under the chamber records the absorp-tion profile during the scanning along the multi-layer cross-section of the chamber The data areanalysed and provide projective measurementStereo-measurements are possible when usingdifferent inclination angles of the X-ray beamFrom the data registered a two-dimensional mapof wire positions can be reconstructed Everychamber is scanned in a few positions along thetube length A full calibration of the muonchambers produces a list of all the wire coordi-nates in two dimensions Using the scanningresults for a few sections along the chambers athree-dimensional map can finally be made
After the analysis of raw data the results arestored in a database providing a full list of the wireposition and accuracy together with the para-meters of the chamber
We can measure the MDT chamberrsquos wireposition using two different methods
The lsquopassiversquo method This is based on therelatively good transparency of a MDT cham-ber to X-rays and on the strong absorptionof the 50 mm tungsten anode wires The out-put intensity of the X-ray beam is measured
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 89
by a scintillator Fig 6 shows a typicalcounting rate of the scintillator versus the X
coordinate1
The lsquoactiversquo method This second procedure isbased on the position-dependence efficiencyof drift tubes for X-rays When the X-raybeam hits the wire photoelectrons Augerelectrons and fluorescence are produced ata much higher rate than in the gas or in the
walls These radiations have a high prob-ability of escaping from the surface of the wireand of being detected by direct ionization in thegas
Fig 6 shows the wire signal from a tubebelonging to the bottom layer of the chamberThe three wires in the upper layers above theactive tube are clearly seen as absorption peaks bythe active tube they produce a shadow effectwhich is observed as inverted peaks comparedto the tube wire signal itself The relative intensityof the signals is not only proportional to theabsorption power of the material but also tothe probability of photoelectrons to escape fromthe material and to ionize the detecting gas ofthe drift tube
Passive mode
Layer 4 Active mode
X coordinate (mm)
Co
un
tin
g r
ate
(kH
z)
4 absorption lines seen
3 absorption lines seen
1 peak wire signal
Fig 6 Passive and active scanning methods
1 In this paper we exclusively use the natural X-ray
tomograph coordinate system (see Ref [12]) X (scanning
direction in the horizontal plane) Y (in the horizontal plane
perpendicular to the scanning direction) and Z (the vertical
coordinate) The MDT chambers in ATLAS use another system
of coordinates x (along the tubes) y (normal to chamber plane)
and z (across the tubes) Consequently the correspondence is as
follows X-zY-xZ-y
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10190
3 The X-ray tomograph
31 The challenge
The maximum width of the standardized MDTchambers has been fixed at 2160mm Thereforethe maximum scanning span of the X-ray tomo-graph has been chosen to be 2200mm The mainchallenge was to provide in short time to theconstructors reliable results on the wire position-ing of different chamber prototypes built with20 mm accuracy specification The X-ray tomo-graph had to consequently achieve a betteraccuracy aiming at 10 mm and be built withinstringent time and money constraints
32 Design considerations
The measurement of the position of the wires intwo dimensions along a cross-section of achamber was performed by using a cart movingon well-polished rails carrying two X-ray sourcesand their collimating systems producing beamswith opposite and fixed inclination anglesThe selection of the two inclination angles Y1Y2 the so-called lsquoscanning anglersquo relative to theperpendicular to the chamber plane was animportant issue under consideration for tworeasons
They define the stereo-angle (jY1j thorn jY2j) andconsequently the two-dimensional accuracy ofthe position of the wires
The absorption peak pattern of the wireswithin a corridor defined by the Y angle mustshow well-separated absorption signals Thispattern is very sensitive to the Y angle anddepends upon the chamber structure character-istics ie the height of the spacer the symmetrybetween the two multilayers and the horizontaland vertical wire pitch
The most convenient choice is to align thebeams in the natural 7301 corridors of themultilayers The angles must be chosen a little bithigher or lower in order to separate the wire peaksignals In the described set-up they are 73141
The geometric reconstruction requires theknowledge of four mechanical parameters the
linear scanning position the vertical straightnessthe angular displacement around the Y -axis (pitchangle) and the angular variation around the X -axis(roll angle) The excellent control of the cartposition (better than 1 mm) was crucial for achiev-ing high reconstruction accuracy We chose for themain structure of the X-ray tomograph standardhigh-precision mechanics combined with ultimateprecision survey tools in order to monitor the mainparameters controlling the X-ray beam positionduring scanning In this respect industrial laserinterferometers2 were used for monitoring Z (thevertical straightness) X (the horizontal displace-ment) DY (the pitch angle variation) and the rollangle
Reliable industrial choices were made wheneverpossible for the construction of the other parts andfor the software of the X-ray tomograph
33 Construction
Fig 7 shows the X-ray tomograph designed andbuilt at CERN Owing to the necessity of a verystable environment the X-ray tomograph wasinstalled in a clean room in building 188 groundfloor (one of the most stable ground floors atCERN) with temperature control and humiditycontrol
The X-ray tomograph consists of
A fixed 3500mm iron portico equipped with aprecise motorized cart rolling on rails along theX -axis
A rolling cart equipped with two collimated X-ray beams at 7314 with respect to the Z-axisThe 6281 stereo-angle has been chosen on thebasis of our previous experience and fits thetube corridors perfectly provided the spacersbetween the multilayers are correctly chosenThe X-ray beams have a cross-section of about30 mm 8mm and a divergence of about
2The basic principle of industrial interferometry used here is
to count the number of fringes of the interference between a
fixed reference laser beam and a second laser beam which is
reflected by a mirror installed on the moving chart Industry
suggests a few optical schemes of the main and reference beams
to allow measurements of longitudinal transversal (with respect
to the laser beam direction) and angular variations
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 91
60 mrad the longer axis being aligned along thewires to be measured
Four interferometers which monitor the mainparameters of the scanner movement the linearX value the pitch angle (rotation around the Y -axis) and two vertical straightnesses (theirbeams are parallel with distance about 160mmbetween them) The last two measurementsallow the calculation of the yaw (rotationaround the Z-axis) The horizontal straightnessis not monitored as it would have only asecond-order effect in the corrections Theinterferometers allow monitoring (2000 mea-surements per second) with an accuracy about03 mm for linear 03 mrad for angular para-meters over the full range of the scanner
Two calibration rulers each split into two parts(Fig 8) and placed under and above thechamber to be measured Every calibrationruler consists of two layers of nine ceramicplates (Fig 9) with six gold strips on eachcomposing a regular grid of 54 equivalent wireswith 30mm pitch The distance between thoselayers is 45mm The ceramic plates are installedon a carbon fibre support The calibrationrulers are used for the online (ie it is made for
every scan) reconstruction of the geometricalparameters of the X-ray tomograph and tocheck the stability and reliability of the X-raytomograph
A set of computer-controlled motors allowingthe alignment of the chamber on the X-raytomograph and choosing of the cross-sectionplane of the chamber One small moving cart isenslaved to the X-ray cart supporting twoscintillators optimized for the X-ray beamenergy spectrum following the beams inorder to record two shadowgrams simulta-neously
The control of the X-ray tomograph cart andchamber movements the data-acquisition andparameter monitoring are supported by adistributed online system based on VME PCSun computers connected via PCI MXI sharedmemory interface using softwarewritten on CC++ and LabVIEW
Two X-ray high-voltage power supplies (up to60 kV 50mA) The two high-voltage cablesfeeding the X-ray tubes move in synchronismwith the scanner on a separate cart in orderto ensure a perfect stability of the X-ray rollingcart
and collimating systems
Support of the upper calibration ruler
Chamber
Porticowithrails
Cart with shielded X-ray tubes
Z
X
Z
X
Fig 7 The X-ray tomograph during the mechanical control of a layers muon
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10192
4 Online calibration
Measurements with an accuracy of a fewmicrons require a precise control of all thegeometrical parameters of the X-ray tomographduring the measurements This requires the con-tinuous monitoring of the absolute position of theinterferometer laser beams in an external system ofcoordinates and of the origin of all laser axes Thislast point is not an easy requirement becauseduring the installation of the chambers on the
X-ray tomograph or during maintenance workthe trajectory of the laser beams may be frequentlyinterrupted for a short time by some object whichautomatically leads to the loss of the zeroreference In order to avoid this problem thecalibration and zero reset must be done for everyscan
During the scanning the wires of the referencerulers must be seen by the X-ray beams of theX-ray tomograph without overlap from thechamber wires The reference wires are previouslymeasured independently on a very-high-precisionoptical bench (E1 mm) During the scanning theshadow patterns of both the wires of the referencerulers and of the chamber are registered simulta-neously The basic geometrical parameters of theX-ray tomograph are then calculated for each scanusing the recorded shadowgram together with theoptical results of the position of the reference rulerwires [13] This approach allows a reliablepermanent control of the accuracy of the measure-ments Such a calibration scheme provides thebasis for the calculation of the position of allchamber wires
Let us stress that all the measurements are donein the same scan all calculations for the chamberand reference ruler wires are made with the same
Fig 8 One half of the calibration ruler (consisting of nine elements fixed on the carbon fibre tube) during assembling and
measurements
Fig 9 One element of the calibration ruler each ceramic plate
has six narrow gold strips faking wires
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 93
geometrical parameters The two reference rulersare measured with a known accuracy As aconsequence the chamber which is lying betweenthe two reference rulers is measured with the sameaccuracy Moreover in the offline analysis [13] thesame functions are used for the geometricalcalculation of the reference ruler and the chamberwires The X-ray tomograph accuracy is estimatedusing the residuals of the difference between thepredictions of the position of each wire of thereference rulers as given by the X-ray beams andtheir actual known position This method allowsone to control the accuracy of the X-ray tomo-graph scan-by-scan and consequently to detect anyloss of performance it allows for example toreveal small displacements or failures of theinterferometer system
Let us stress that the calibration is performedfor every scan and therefore for every scan thegeometrical parameters are calculated and theX-ray tomograph accuracy is checked
While studying the X-ray tomograph perfor-mance soon after its construction it became clearthat the precision of the optical measurements ofthe calibration rulers was not high enough toachieve the wanted accuracy It was later con-firmed that the optical measurements performedby the CERN Metrology Group using conven-tional industrial equipment were not preciseenough Therefore we had to develop a dedicatedmethod in order to improve the accuracy
Geometrical characteristics of the referencerulers may be divided into two groups the firstgroup consists of the parameters which can becorrected using direct X-ray tomograph measure-ments and the second group consists of para-meters for which other external reference devicesare needed In the first group we can note thatsome systematic errors in the position of individualcalibration ruler wires are well detected by theX-ray tomograph and those components of theoptical measurement errors are corrected by usingthe data coming from several scans As an examplefor the second group of parameters the straight-ness of the monitoring X-ray beams does not allowthe detection of possible errors in the opticalmeasurements of the vertical and horizontal shiftsbetween the layers of the calibration rulers If such
an error exists it leads to an error in the verticalscale and in the orthogonality of the system ofcoordinates of the X-ray tomograph
We call these two groups of optical measure-ment errors lsquonon-linearrsquo and lsquolinearrsquo errors3
respectively An autocalibration approach wasdeveloped to improve the accuracy of the opticalmeasurements It consists of two steps correctionof non-linear components in optical measurementsand determination (plus correction) of the linearparameters
The first step does not require special commentsAfter the geometrical reconstruction of the cali-bration ruler wires determined from several scans(in order to improve statistics) the residualsbetween the predictions by every X-ray beamand the optical measurements are calculatedUsing corresponding values from both non-paral-lel beams the two-dimensional correction ofposition of every calibration ruler wire may befound Because the pattern recognition maychange after the corrections an additional itera-tion may be done
5 Autocalibration correction of the linear errors in
the optical measurements of the calibration ruler
wire positions
51 Theoretical aspects
Let us consider the question of the linearsystematics effect of the absolute scales the scalesfor the X - and Z-axis and their relative angle (non-orthogonality of the system of coordinates) The X
scale is determined by the linear interferometer(measuring the position of the cart along the scan)the corresponding laser beam is parallel to theOX -axis (with second-order errors due to the non-parallelism of the laser beam and scan direction)The OZ scale and the non-orthogonality errors ofthe coordinate system however are completelydefined by the optical measurement errors They
3We have chosen the names in that way because the lsquolinearrsquo
(non-linear) errors in the optical measurements are propagated
to the chamber wire position errors which are linear (non-
linear) functions of the wire coordinates
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10194
cannot be determined from the X-ray tomographmeasurements themselves During the geometryfit the calculated effects of any change in the OZ
scale and in the non-orthogonality of the axes maybe completely compensated with correspondingchanges of the geometrical parameters of the X-ray beams Therefore some external measurementsare necessary
Let us consider three hypothetical parallel wiresA B C of the same rigid chamber (see Fig 10 thewires are perpendicular to the plane of picture andonly their projections A B C are seen) The pointsA B C should not lie on the same straight lineLet us imagine that they are measured by the X-ray tomograph in the original position (Fig 10a)and after rotation of the chamber by 901 aroundthe wiresrsquo direction (Fig 10b) If the X-raytomograph has a Z scale error instead of thepositions of the points A B C the positions ofpoints A0 B0 C0 and points A00 B00 C00 in therotated orientation will be found Putting bothmeasurements in the same system of coordinates
(we need three points because the angle aboutwhich these three points are rotated is not knownprecisely and in general does not equal 901) wecan estimate the Z scale error (Fig 10c)
A similar analysis applies for the non-orthogon-ality error Three wires A B C are measured aspoints A0 B0 C0 (Fig 11a) and as points A00 B00C00 in the 901 orientation (Fig 11b) Putting bothmeasurements in the same system of coordinates(Fig 11c) we can evaluate the non-orthogonalityerror introduced by the X-ray tomograph
Having determined the values of the Z scale andthe non-orthogonality errors we can calculate andthen correct the horizontal and vertical shiftsbetween the layers of the reference ruler becausethese shifts are the source of the errors in the Z
scale and non-orthogonality
52 Practical realization
In order to realize the described approach asmall calibration chamber (gauge chamber) was
A B
C
A B
C
AA
C
B
C
B
B
BA
A
C C
O X
Z
(a) (b) (c)
Fig 11 Non-orthogonality error calibration
A B
C
A
B
B
C
C
A
A
C
BCA
B
AB
C
O X
Z
(a) (b) (c)
Fig 10 Z scale error calibration
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 95
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
gluonsndashgluinos etc with masses well suited to theLHC energy range The Minimal SupersymmetricStandard Model calls for more Higgs states(neutral and charged Higgs) Among many otherexperimental physics goals the hunt for SUSYpartners will also be a hot challenge for thedetectors
Entering a new energy domain may also lead tounexpected discoveries as has already occurredmany times in particle physics experiments There-fore the detectors should be ready for all newphysics and open to the greatest discovery to bemade at LHC a surprise
However at the same time the new machinecreates a new challenge for the experiments veryhigh luminosity high radiation environment andvery small cross-sections for discoveries lead tounprecedented experimental conditions in terms ofthe data acquisition rate and signal over back-ground
After several years of Research and Develop-ment large LHC collaborations started the con-struction phase of the huge detectors needed inorder to exploit the full potential of the LHCcollider for physics searches
12 ATLAS at CERN
One of the two very large detectors underpreparation for LHC is the ATLAS detector [1ndash3]
ATLAS has been designed as a general-purposedetector for LHC with efficient and high-resolu-tion tracking excellent calorimetry and precisemuon detection The inner detector is embedded ina 2T superconducting solenoid coil placed insidethe calorimetry The calorimeter is installedbetween the inner tracker and the muon spectro-meter which forms the outer shell of the ATLASdetector
The muon spectrometer surrounding the calori-meter is optimized not only for Higgs search butalso to leave open the possibility to tackle manyother kinds of searches in order to exploit the fullpotential of the LHC It has been designed as ahigh-resolution high-efficiency muon spectro-meter with stand-alone triggering and precisemomentum-measurement capability over a widerange of transverse momenta covering with high
hermeticity a large range of pseudorapidity Thiswas achieved by using an original well-adaptedmagnetic field configuration large superconduct-ing air-core toroids arranged in an eight-foldsymmetry with a bending power ranging from25 Tm at Z frac14 0 to 8Tm at the maximum pseudo-rapidity covered by the ATLAS muon detector
The detector will fill a huge cavern(26m 47m 30m) surrounding the collisionarea about 90m under the ground surface
ATLAS is a very large world-wide collabora-tion at present it involves 150 institutions withabout 2000 participants In terms of size cost andtimescale this detector will be the biggest everbuilt in the world for high-energy physics experi-ments
13 The ATLAS muon spectrometer
The instrumentation of the ATLAS muonspectrometer (see the Muon Spectrometer Techni-cal Design Report [4]) as a sub-detector ofATLAS has been designed by a collaborationfrom 44 institutes in 13 countries
High-momentum final-state muons are amongstthe most promising and robust signatures forsearching for the Higgs The clearest signal forHiggs production would be through one of theleptonic decay modes of the produced Z bosons(H-ZZ-m+mm+m) provided the Higgs masswere above E184GeV [2m(Z)] The branchingratio of this decay mode is small (011) but thesignature of this kind of event will be clear andunambiguous
The performance required for physics in mo-mentum and mass resolutions is at the level of 1The transverse-momentum resolution should beconstant over the full rapidity range
14 Layout
The spectrometer is shown in Figs 1ndash3The design philosophy was to adopt a safe and
industrial approach matching all the physicsrequirements and experimental constraints forgetting a reliable safe and robust muon spectro-meter able to run for at least 10 years withoutmajor trouble
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10184
The conceptual layout of the spectrometer isbased on a system of three large superconductingair-core toroid magnets instrumented withseparate-function trigger and high-precisiontracking chambers This magnet configurationprovides a field that is mostly orthogonalto the muon trajectories producing a maximumbending power which is needed for momen-
tum resolution and charge-sign determinationat the highest energies while minimizing thedegradation of resolution due to multiple scatter-ing
Last but not least another important considera-tion is to produce an economical design thatallows a total surface of about 5000m2 to becovered at an affordable cost
chamberschambers
chambers
chambers
Cathode stripResistive plate
Thin gap
Monitored drift tube
l = ~ 46m
Oslash ~ 24 m
P
beam
P
beam
Fig 1 Three-dimensional sketch of the ATLAS muon spectrometer layout
Fig 2 Side view of one quarter of the muon spectrometer
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 85
15 Momentum measurement
Charged particles with high momentum (above300GeVc) crossing the toroid magnets of themuon spectrometer will have very large radius ofcurvature producing small saggita signals(sE1525mm) over the long muon trajectories(5ndash15m) between three measuring stations Inorder to achieve the required precision in themomentum measurement the saggita must bedetermined with a precision of 15ndash50 mm overmost of the pseudorapidity range a precisionmeasurement of the track coordinates in theprincipal bending direction of the magnetic fieldis provided by Monitored Drift Tubes (MDTs)and Cathode Strip Chambers (CSCs) The MDTchambers are arranged in three layers all aroundthe calorimeter in order to determine the momen-tum with the best possible resolution The CSCsare used for the very forward area where theparticle flux is too high for the drift chambers
With this layout a three-point measurement isavailable (see Fig 4) by installing three stations ofmultilayer MDT chambers covering the fullrapidity gap with high hermeticity The driftMDT chambers are called lsquomonitoredrsquo because acontinuous monitoring of the internal chamberdeformations is done In addition to this on-linesurvey the position in space of each chamber isalso monitored with the required precision(r20 mm)
With this layout the momentum resolution istypically 2ndash3 at 100GeVc over most of thekinematic range
The muon spectrometer is designed for amomentum resolution DpT=pTo1104 pT wherepT is in GeV for pT gt 300GeV To achieve thisresolution by a three-point measurement with thesize and bending power of the ATLAS toroidseach point must be measured with an accuracy ofabout 50 mm This sets the scale for the require-ments on the intrinsic resolution the mechanical
End-captoroid
Barrel toroidcoils
Calorimeters
MDT chambersResistive plate chambers
Inner detector
oslash ~
24 m
Fig 3 Cross view of the muon spectrometer
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10186
precision and the survey accuracy of the muonchambers
The momentum resolution of the spectrometeris limited by the intrinsic detector resolutionMDT calibration errors chamber positioninguncertainties multiple scattering and statisticalfluctuations of energy loss At smaller momenta(below about 300GeVc) the resolution is limitedby multiple scattering to a few per cent at highermomenta (above 300GeVc) it is dominated bychamber precision and alignment
16 The MDT chambers
The task in numbers is impressive the totalnumber of MDTs is E1200 making a total area ofabout 5000m2 The total number of readoutchannels (number of drift tubes) is E400 000 andthe total length of the tubes is E1100 km
The basic detection elements of the MDTchambers are drift chambers made of aluminiumtubes of 30mm diameter and 400 mm wall thick-ness with a 50 mm diameter central WndashRe wireThe tubes are operated with a non-flammableArndashCO2 mixture at 3 bar absolute pressure
The tubes are produced by extrusion from ahard aluminium alloy and are available commer-cially They are closed by endplugs which provideaccurate positioning of the anode wires wiretension gas tightness and electrical and gasconnections The drift tubes can be manufacturedto tight mechanical tolerances which are well-matched to their intrinsic resolution propertiesmostly using automated assembly procedures Thetube lengths vary from 70 to 630 cm
For a good linear spacendashtime relation with amaximum drift time of B500ns the envisagedworking point provides a small Lorentz angle andgood ageing properties due to low gas amplification
X [c
m]
Z [cm]Collision vertex
Pseudo-rapidity Beam lineaxis
Fig 4 Longitudinal cross-section in the bending plane of the spectrometer showing the barrel and end-cap magnet air-coil toroid
configuration It shows the pseudo-rapidity coverage of the muon spectrometer from 0 to 28 and a sketch of the layout principle of the
three detecting muon stations showing the trajectories of a few GeVc positively and negatively charged particles
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 87
Using the timendashdrift distance relationship the single-wire resolution is typically 80mm when operated athigh gas pressure (3bar)
The muon chamber layout was optimized toreduce the number of different chamber sizes
The barrel chambers are of rectangular shapewith sizes from 2 to 10m2
The end-cap chambers are of trapezoidal shapeTheir sizes range from 1 to 10m2 for individualchamber modules and up to 30m2 when severalof them are preassembled for installation Thechamber design must guarantee reliability andstability of construction and operation for theanticipated lifetime of the experiment in anirradiated environment The basic design isshown in Fig 5
To improve the resolution of a chamber beyondthe single-wire limit and to achieve adequateredundancy for pattern recognition the MDTchambers are constructed from 2 4 monolayersof drift tubes for the inner and 2 3 monolayersfor the middle and outer stations The tubes are
arranged in multilayers of three or four mono-layers respectively on either side of a rigidsupport structure The support structures thelsquospacer framesrsquo provide for accurate positioningof the two multilayers with respect to each otherand for mechanical integrity under effects oftemperature and gravity As an example ofdifficulties encountered when seeking to reach therequired resolution the drift tubes which are notmounted vertically are bent slightly in order tomatch their shape to the gravitational sag of thewires
The structural components of the spacerframes are three lsquocrossplatesrsquo to which thedrift tube multilayers are attached and twolsquolong beamsrsquo connecting the crossplates Theframes need to be constructed to a moderatemechanical accuracy of 705mm only accuratepositioning of the drift tubes is provided bythe assembly procedure The chambers willbe attached to the rail structures of the spectro-meter by three-point kinematic supports Oncea chamber is installed in its final location in
Cross plate
Multilayer
In-plane alignment
Longitudinal beam
3-layerMultilayer
micro track
Tube cross section
BE
+
Fig 5 Schematic drawing of a rectangular MDT chamber constructed from multilayers of three monolayers each for installation in
the barrel spectrometer The chambers for the end-cap are of trapezoidal shape but are of similar design otherwise In the toroidal
magnetic field the tubes are oriented essentially parallel to the field lines
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10188
the spectrometer mechanical deformations aremonitored by an in-plane optical system hencethe name lsquoMDTrsquo chambers The spacer framessupport most of the components of the alignmentsystem
MDT chambers are made for robust andreliable operation This is achieved thanks to themechanical concept which provides a solid separa-tion of detecting cells and a good electricalshielding of neighbouring wires
Chambers must withstand very long storagetime followed by very long running periods Inorder to reach the required momentum resolutionall MDT chambers must be constructed with20 mm mechanical accuracy
17 Quality assurance for MDTs
The MDT project in ATLAS is on the scale of avery large industrial project aiming at an excep-tional quality of construction in terms of accuracy(r20 mm) material reliability delivery and assem-bly schedule The X-ray tomograph quality-con-trol platform may be considered as one of the keyinstruments needed for the MDT
The production of the MDT chambers is basedon the concept of very high mechanical precisionand the use of X-ray tomography as a quality-control platform for the mechanically achievedprecision
2 X-ray tomography
The development design construction andtuning of an X-ray tomograph suited forthe MDT quality control was the aim of theX-ray Quality-Control lsquoX-QCrsquo project (see Refs[5ndash11])
21 Basic principle
The choice of X-ray tomography as a quality-control tool for the MDT project was based ontwo essential features
The ability of X-rays to detect 50 mm tungstenwires hidden inside the aluminium drift tubes of
the muon detector Sharp absorption spikes areobserved either as a shadow seen by an X-raydetector placed under the chamber (passivemode) or as a sharp peak signal detected by thechamber itself (active mode) see Fig 6
The possibility to produce narrow collimatedX-ray beams allowing one to reach detectionprofiles with the required peak position detec-tion accuracy (1ndash2 mm)
The value of the absorption coefficient for X-rayphotons is very sensitive to attenuating materialsand to the X-ray energy If d is the density Z theatomic number of the material and W the energyof X-rays the mass absorption coefficient m=dvaries as EethZ=W THORN3 at 40 keV The optimumenergy range for this application is between 40and 60 keV
22 Basic procedure
A narrow X-ray beam is precisely moved acrossa section of the MDT chamber A scintillatorinstalled under the chamber records the absorp-tion profile during the scanning along the multi-layer cross-section of the chamber The data areanalysed and provide projective measurementStereo-measurements are possible when usingdifferent inclination angles of the X-ray beamFrom the data registered a two-dimensional mapof wire positions can be reconstructed Everychamber is scanned in a few positions along thetube length A full calibration of the muonchambers produces a list of all the wire coordi-nates in two dimensions Using the scanningresults for a few sections along the chambers athree-dimensional map can finally be made
After the analysis of raw data the results arestored in a database providing a full list of the wireposition and accuracy together with the para-meters of the chamber
We can measure the MDT chamberrsquos wireposition using two different methods
The lsquopassiversquo method This is based on therelatively good transparency of a MDT cham-ber to X-rays and on the strong absorptionof the 50 mm tungsten anode wires The out-put intensity of the X-ray beam is measured
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 89
by a scintillator Fig 6 shows a typicalcounting rate of the scintillator versus the X
coordinate1
The lsquoactiversquo method This second procedure isbased on the position-dependence efficiencyof drift tubes for X-rays When the X-raybeam hits the wire photoelectrons Augerelectrons and fluorescence are produced ata much higher rate than in the gas or in the
walls These radiations have a high prob-ability of escaping from the surface of the wireand of being detected by direct ionization in thegas
Fig 6 shows the wire signal from a tubebelonging to the bottom layer of the chamberThe three wires in the upper layers above theactive tube are clearly seen as absorption peaks bythe active tube they produce a shadow effectwhich is observed as inverted peaks comparedto the tube wire signal itself The relative intensityof the signals is not only proportional to theabsorption power of the material but also tothe probability of photoelectrons to escape fromthe material and to ionize the detecting gas ofthe drift tube
Passive mode
Layer 4 Active mode
X coordinate (mm)
Co
un
tin
g r
ate
(kH
z)
4 absorption lines seen
3 absorption lines seen
1 peak wire signal
Fig 6 Passive and active scanning methods
1 In this paper we exclusively use the natural X-ray
tomograph coordinate system (see Ref [12]) X (scanning
direction in the horizontal plane) Y (in the horizontal plane
perpendicular to the scanning direction) and Z (the vertical
coordinate) The MDT chambers in ATLAS use another system
of coordinates x (along the tubes) y (normal to chamber plane)
and z (across the tubes) Consequently the correspondence is as
follows X-zY-xZ-y
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10190
3 The X-ray tomograph
31 The challenge
The maximum width of the standardized MDTchambers has been fixed at 2160mm Thereforethe maximum scanning span of the X-ray tomo-graph has been chosen to be 2200mm The mainchallenge was to provide in short time to theconstructors reliable results on the wire position-ing of different chamber prototypes built with20 mm accuracy specification The X-ray tomo-graph had to consequently achieve a betteraccuracy aiming at 10 mm and be built withinstringent time and money constraints
32 Design considerations
The measurement of the position of the wires intwo dimensions along a cross-section of achamber was performed by using a cart movingon well-polished rails carrying two X-ray sourcesand their collimating systems producing beamswith opposite and fixed inclination anglesThe selection of the two inclination angles Y1Y2 the so-called lsquoscanning anglersquo relative to theperpendicular to the chamber plane was animportant issue under consideration for tworeasons
They define the stereo-angle (jY1j thorn jY2j) andconsequently the two-dimensional accuracy ofthe position of the wires
The absorption peak pattern of the wireswithin a corridor defined by the Y angle mustshow well-separated absorption signals Thispattern is very sensitive to the Y angle anddepends upon the chamber structure character-istics ie the height of the spacer the symmetrybetween the two multilayers and the horizontaland vertical wire pitch
The most convenient choice is to align thebeams in the natural 7301 corridors of themultilayers The angles must be chosen a little bithigher or lower in order to separate the wire peaksignals In the described set-up they are 73141
The geometric reconstruction requires theknowledge of four mechanical parameters the
linear scanning position the vertical straightnessthe angular displacement around the Y -axis (pitchangle) and the angular variation around the X -axis(roll angle) The excellent control of the cartposition (better than 1 mm) was crucial for achiev-ing high reconstruction accuracy We chose for themain structure of the X-ray tomograph standardhigh-precision mechanics combined with ultimateprecision survey tools in order to monitor the mainparameters controlling the X-ray beam positionduring scanning In this respect industrial laserinterferometers2 were used for monitoring Z (thevertical straightness) X (the horizontal displace-ment) DY (the pitch angle variation) and the rollangle
Reliable industrial choices were made wheneverpossible for the construction of the other parts andfor the software of the X-ray tomograph
33 Construction
Fig 7 shows the X-ray tomograph designed andbuilt at CERN Owing to the necessity of a verystable environment the X-ray tomograph wasinstalled in a clean room in building 188 groundfloor (one of the most stable ground floors atCERN) with temperature control and humiditycontrol
The X-ray tomograph consists of
A fixed 3500mm iron portico equipped with aprecise motorized cart rolling on rails along theX -axis
A rolling cart equipped with two collimated X-ray beams at 7314 with respect to the Z-axisThe 6281 stereo-angle has been chosen on thebasis of our previous experience and fits thetube corridors perfectly provided the spacersbetween the multilayers are correctly chosenThe X-ray beams have a cross-section of about30 mm 8mm and a divergence of about
2The basic principle of industrial interferometry used here is
to count the number of fringes of the interference between a
fixed reference laser beam and a second laser beam which is
reflected by a mirror installed on the moving chart Industry
suggests a few optical schemes of the main and reference beams
to allow measurements of longitudinal transversal (with respect
to the laser beam direction) and angular variations
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 91
60 mrad the longer axis being aligned along thewires to be measured
Four interferometers which monitor the mainparameters of the scanner movement the linearX value the pitch angle (rotation around the Y -axis) and two vertical straightnesses (theirbeams are parallel with distance about 160mmbetween them) The last two measurementsallow the calculation of the yaw (rotationaround the Z-axis) The horizontal straightnessis not monitored as it would have only asecond-order effect in the corrections Theinterferometers allow monitoring (2000 mea-surements per second) with an accuracy about03 mm for linear 03 mrad for angular para-meters over the full range of the scanner
Two calibration rulers each split into two parts(Fig 8) and placed under and above thechamber to be measured Every calibrationruler consists of two layers of nine ceramicplates (Fig 9) with six gold strips on eachcomposing a regular grid of 54 equivalent wireswith 30mm pitch The distance between thoselayers is 45mm The ceramic plates are installedon a carbon fibre support The calibrationrulers are used for the online (ie it is made for
every scan) reconstruction of the geometricalparameters of the X-ray tomograph and tocheck the stability and reliability of the X-raytomograph
A set of computer-controlled motors allowingthe alignment of the chamber on the X-raytomograph and choosing of the cross-sectionplane of the chamber One small moving cart isenslaved to the X-ray cart supporting twoscintillators optimized for the X-ray beamenergy spectrum following the beams inorder to record two shadowgrams simulta-neously
The control of the X-ray tomograph cart andchamber movements the data-acquisition andparameter monitoring are supported by adistributed online system based on VME PCSun computers connected via PCI MXI sharedmemory interface using softwarewritten on CC++ and LabVIEW
Two X-ray high-voltage power supplies (up to60 kV 50mA) The two high-voltage cablesfeeding the X-ray tubes move in synchronismwith the scanner on a separate cart in orderto ensure a perfect stability of the X-ray rollingcart
and collimating systems
Support of the upper calibration ruler
Chamber
Porticowithrails
Cart with shielded X-ray tubes
Z
X
Z
X
Fig 7 The X-ray tomograph during the mechanical control of a layers muon
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10192
4 Online calibration
Measurements with an accuracy of a fewmicrons require a precise control of all thegeometrical parameters of the X-ray tomographduring the measurements This requires the con-tinuous monitoring of the absolute position of theinterferometer laser beams in an external system ofcoordinates and of the origin of all laser axes Thislast point is not an easy requirement becauseduring the installation of the chambers on the
X-ray tomograph or during maintenance workthe trajectory of the laser beams may be frequentlyinterrupted for a short time by some object whichautomatically leads to the loss of the zeroreference In order to avoid this problem thecalibration and zero reset must be done for everyscan
During the scanning the wires of the referencerulers must be seen by the X-ray beams of theX-ray tomograph without overlap from thechamber wires The reference wires are previouslymeasured independently on a very-high-precisionoptical bench (E1 mm) During the scanning theshadow patterns of both the wires of the referencerulers and of the chamber are registered simulta-neously The basic geometrical parameters of theX-ray tomograph are then calculated for each scanusing the recorded shadowgram together with theoptical results of the position of the reference rulerwires [13] This approach allows a reliablepermanent control of the accuracy of the measure-ments Such a calibration scheme provides thebasis for the calculation of the position of allchamber wires
Let us stress that all the measurements are donein the same scan all calculations for the chamberand reference ruler wires are made with the same
Fig 8 One half of the calibration ruler (consisting of nine elements fixed on the carbon fibre tube) during assembling and
measurements
Fig 9 One element of the calibration ruler each ceramic plate
has six narrow gold strips faking wires
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 93
geometrical parameters The two reference rulersare measured with a known accuracy As aconsequence the chamber which is lying betweenthe two reference rulers is measured with the sameaccuracy Moreover in the offline analysis [13] thesame functions are used for the geometricalcalculation of the reference ruler and the chamberwires The X-ray tomograph accuracy is estimatedusing the residuals of the difference between thepredictions of the position of each wire of thereference rulers as given by the X-ray beams andtheir actual known position This method allowsone to control the accuracy of the X-ray tomo-graph scan-by-scan and consequently to detect anyloss of performance it allows for example toreveal small displacements or failures of theinterferometer system
Let us stress that the calibration is performedfor every scan and therefore for every scan thegeometrical parameters are calculated and theX-ray tomograph accuracy is checked
While studying the X-ray tomograph perfor-mance soon after its construction it became clearthat the precision of the optical measurements ofthe calibration rulers was not high enough toachieve the wanted accuracy It was later con-firmed that the optical measurements performedby the CERN Metrology Group using conven-tional industrial equipment were not preciseenough Therefore we had to develop a dedicatedmethod in order to improve the accuracy
Geometrical characteristics of the referencerulers may be divided into two groups the firstgroup consists of the parameters which can becorrected using direct X-ray tomograph measure-ments and the second group consists of para-meters for which other external reference devicesare needed In the first group we can note thatsome systematic errors in the position of individualcalibration ruler wires are well detected by theX-ray tomograph and those components of theoptical measurement errors are corrected by usingthe data coming from several scans As an examplefor the second group of parameters the straight-ness of the monitoring X-ray beams does not allowthe detection of possible errors in the opticalmeasurements of the vertical and horizontal shiftsbetween the layers of the calibration rulers If such
an error exists it leads to an error in the verticalscale and in the orthogonality of the system ofcoordinates of the X-ray tomograph
We call these two groups of optical measure-ment errors lsquonon-linearrsquo and lsquolinearrsquo errors3
respectively An autocalibration approach wasdeveloped to improve the accuracy of the opticalmeasurements It consists of two steps correctionof non-linear components in optical measurementsand determination (plus correction) of the linearparameters
The first step does not require special commentsAfter the geometrical reconstruction of the cali-bration ruler wires determined from several scans(in order to improve statistics) the residualsbetween the predictions by every X-ray beamand the optical measurements are calculatedUsing corresponding values from both non-paral-lel beams the two-dimensional correction ofposition of every calibration ruler wire may befound Because the pattern recognition maychange after the corrections an additional itera-tion may be done
5 Autocalibration correction of the linear errors in
the optical measurements of the calibration ruler
wire positions
51 Theoretical aspects
Let us consider the question of the linearsystematics effect of the absolute scales the scalesfor the X - and Z-axis and their relative angle (non-orthogonality of the system of coordinates) The X
scale is determined by the linear interferometer(measuring the position of the cart along the scan)the corresponding laser beam is parallel to theOX -axis (with second-order errors due to the non-parallelism of the laser beam and scan direction)The OZ scale and the non-orthogonality errors ofthe coordinate system however are completelydefined by the optical measurement errors They
3We have chosen the names in that way because the lsquolinearrsquo
(non-linear) errors in the optical measurements are propagated
to the chamber wire position errors which are linear (non-
linear) functions of the wire coordinates
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10194
cannot be determined from the X-ray tomographmeasurements themselves During the geometryfit the calculated effects of any change in the OZ
scale and in the non-orthogonality of the axes maybe completely compensated with correspondingchanges of the geometrical parameters of the X-ray beams Therefore some external measurementsare necessary
Let us consider three hypothetical parallel wiresA B C of the same rigid chamber (see Fig 10 thewires are perpendicular to the plane of picture andonly their projections A B C are seen) The pointsA B C should not lie on the same straight lineLet us imagine that they are measured by the X-ray tomograph in the original position (Fig 10a)and after rotation of the chamber by 901 aroundthe wiresrsquo direction (Fig 10b) If the X-raytomograph has a Z scale error instead of thepositions of the points A B C the positions ofpoints A0 B0 C0 and points A00 B00 C00 in therotated orientation will be found Putting bothmeasurements in the same system of coordinates
(we need three points because the angle aboutwhich these three points are rotated is not knownprecisely and in general does not equal 901) wecan estimate the Z scale error (Fig 10c)
A similar analysis applies for the non-orthogon-ality error Three wires A B C are measured aspoints A0 B0 C0 (Fig 11a) and as points A00 B00C00 in the 901 orientation (Fig 11b) Putting bothmeasurements in the same system of coordinates(Fig 11c) we can evaluate the non-orthogonalityerror introduced by the X-ray tomograph
Having determined the values of the Z scale andthe non-orthogonality errors we can calculate andthen correct the horizontal and vertical shiftsbetween the layers of the reference ruler becausethese shifts are the source of the errors in the Z
scale and non-orthogonality
52 Practical realization
In order to realize the described approach asmall calibration chamber (gauge chamber) was
A B
C
A B
C
AA
C
B
C
B
B
BA
A
C C
O X
Z
(a) (b) (c)
Fig 11 Non-orthogonality error calibration
A B
C
A
B
B
C
C
A
A
C
BCA
B
AB
C
O X
Z
(a) (b) (c)
Fig 10 Z scale error calibration
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 95
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
The conceptual layout of the spectrometer isbased on a system of three large superconductingair-core toroid magnets instrumented withseparate-function trigger and high-precisiontracking chambers This magnet configurationprovides a field that is mostly orthogonalto the muon trajectories producing a maximumbending power which is needed for momen-
tum resolution and charge-sign determinationat the highest energies while minimizing thedegradation of resolution due to multiple scatter-ing
Last but not least another important considera-tion is to produce an economical design thatallows a total surface of about 5000m2 to becovered at an affordable cost
chamberschambers
chambers
chambers
Cathode stripResistive plate
Thin gap
Monitored drift tube
l = ~ 46m
Oslash ~ 24 m
P
beam
P
beam
Fig 1 Three-dimensional sketch of the ATLAS muon spectrometer layout
Fig 2 Side view of one quarter of the muon spectrometer
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 85
15 Momentum measurement
Charged particles with high momentum (above300GeVc) crossing the toroid magnets of themuon spectrometer will have very large radius ofcurvature producing small saggita signals(sE1525mm) over the long muon trajectories(5ndash15m) between three measuring stations Inorder to achieve the required precision in themomentum measurement the saggita must bedetermined with a precision of 15ndash50 mm overmost of the pseudorapidity range a precisionmeasurement of the track coordinates in theprincipal bending direction of the magnetic fieldis provided by Monitored Drift Tubes (MDTs)and Cathode Strip Chambers (CSCs) The MDTchambers are arranged in three layers all aroundthe calorimeter in order to determine the momen-tum with the best possible resolution The CSCsare used for the very forward area where theparticle flux is too high for the drift chambers
With this layout a three-point measurement isavailable (see Fig 4) by installing three stations ofmultilayer MDT chambers covering the fullrapidity gap with high hermeticity The driftMDT chambers are called lsquomonitoredrsquo because acontinuous monitoring of the internal chamberdeformations is done In addition to this on-linesurvey the position in space of each chamber isalso monitored with the required precision(r20 mm)
With this layout the momentum resolution istypically 2ndash3 at 100GeVc over most of thekinematic range
The muon spectrometer is designed for amomentum resolution DpT=pTo1104 pT wherepT is in GeV for pT gt 300GeV To achieve thisresolution by a three-point measurement with thesize and bending power of the ATLAS toroidseach point must be measured with an accuracy ofabout 50 mm This sets the scale for the require-ments on the intrinsic resolution the mechanical
End-captoroid
Barrel toroidcoils
Calorimeters
MDT chambersResistive plate chambers
Inner detector
oslash ~
24 m
Fig 3 Cross view of the muon spectrometer
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10186
precision and the survey accuracy of the muonchambers
The momentum resolution of the spectrometeris limited by the intrinsic detector resolutionMDT calibration errors chamber positioninguncertainties multiple scattering and statisticalfluctuations of energy loss At smaller momenta(below about 300GeVc) the resolution is limitedby multiple scattering to a few per cent at highermomenta (above 300GeVc) it is dominated bychamber precision and alignment
16 The MDT chambers
The task in numbers is impressive the totalnumber of MDTs is E1200 making a total area ofabout 5000m2 The total number of readoutchannels (number of drift tubes) is E400 000 andthe total length of the tubes is E1100 km
The basic detection elements of the MDTchambers are drift chambers made of aluminiumtubes of 30mm diameter and 400 mm wall thick-ness with a 50 mm diameter central WndashRe wireThe tubes are operated with a non-flammableArndashCO2 mixture at 3 bar absolute pressure
The tubes are produced by extrusion from ahard aluminium alloy and are available commer-cially They are closed by endplugs which provideaccurate positioning of the anode wires wiretension gas tightness and electrical and gasconnections The drift tubes can be manufacturedto tight mechanical tolerances which are well-matched to their intrinsic resolution propertiesmostly using automated assembly procedures Thetube lengths vary from 70 to 630 cm
For a good linear spacendashtime relation with amaximum drift time of B500ns the envisagedworking point provides a small Lorentz angle andgood ageing properties due to low gas amplification
X [c
m]
Z [cm]Collision vertex
Pseudo-rapidity Beam lineaxis
Fig 4 Longitudinal cross-section in the bending plane of the spectrometer showing the barrel and end-cap magnet air-coil toroid
configuration It shows the pseudo-rapidity coverage of the muon spectrometer from 0 to 28 and a sketch of the layout principle of the
three detecting muon stations showing the trajectories of a few GeVc positively and negatively charged particles
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 87
Using the timendashdrift distance relationship the single-wire resolution is typically 80mm when operated athigh gas pressure (3bar)
The muon chamber layout was optimized toreduce the number of different chamber sizes
The barrel chambers are of rectangular shapewith sizes from 2 to 10m2
The end-cap chambers are of trapezoidal shapeTheir sizes range from 1 to 10m2 for individualchamber modules and up to 30m2 when severalof them are preassembled for installation Thechamber design must guarantee reliability andstability of construction and operation for theanticipated lifetime of the experiment in anirradiated environment The basic design isshown in Fig 5
To improve the resolution of a chamber beyondthe single-wire limit and to achieve adequateredundancy for pattern recognition the MDTchambers are constructed from 2 4 monolayersof drift tubes for the inner and 2 3 monolayersfor the middle and outer stations The tubes are
arranged in multilayers of three or four mono-layers respectively on either side of a rigidsupport structure The support structures thelsquospacer framesrsquo provide for accurate positioningof the two multilayers with respect to each otherand for mechanical integrity under effects oftemperature and gravity As an example ofdifficulties encountered when seeking to reach therequired resolution the drift tubes which are notmounted vertically are bent slightly in order tomatch their shape to the gravitational sag of thewires
The structural components of the spacerframes are three lsquocrossplatesrsquo to which thedrift tube multilayers are attached and twolsquolong beamsrsquo connecting the crossplates Theframes need to be constructed to a moderatemechanical accuracy of 705mm only accuratepositioning of the drift tubes is provided bythe assembly procedure The chambers willbe attached to the rail structures of the spectro-meter by three-point kinematic supports Oncea chamber is installed in its final location in
Cross plate
Multilayer
In-plane alignment
Longitudinal beam
3-layerMultilayer
micro track
Tube cross section
BE
+
Fig 5 Schematic drawing of a rectangular MDT chamber constructed from multilayers of three monolayers each for installation in
the barrel spectrometer The chambers for the end-cap are of trapezoidal shape but are of similar design otherwise In the toroidal
magnetic field the tubes are oriented essentially parallel to the field lines
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10188
the spectrometer mechanical deformations aremonitored by an in-plane optical system hencethe name lsquoMDTrsquo chambers The spacer framessupport most of the components of the alignmentsystem
MDT chambers are made for robust andreliable operation This is achieved thanks to themechanical concept which provides a solid separa-tion of detecting cells and a good electricalshielding of neighbouring wires
Chambers must withstand very long storagetime followed by very long running periods Inorder to reach the required momentum resolutionall MDT chambers must be constructed with20 mm mechanical accuracy
17 Quality assurance for MDTs
The MDT project in ATLAS is on the scale of avery large industrial project aiming at an excep-tional quality of construction in terms of accuracy(r20 mm) material reliability delivery and assem-bly schedule The X-ray tomograph quality-con-trol platform may be considered as one of the keyinstruments needed for the MDT
The production of the MDT chambers is basedon the concept of very high mechanical precisionand the use of X-ray tomography as a quality-control platform for the mechanically achievedprecision
2 X-ray tomography
The development design construction andtuning of an X-ray tomograph suited forthe MDT quality control was the aim of theX-ray Quality-Control lsquoX-QCrsquo project (see Refs[5ndash11])
21 Basic principle
The choice of X-ray tomography as a quality-control tool for the MDT project was based ontwo essential features
The ability of X-rays to detect 50 mm tungstenwires hidden inside the aluminium drift tubes of
the muon detector Sharp absorption spikes areobserved either as a shadow seen by an X-raydetector placed under the chamber (passivemode) or as a sharp peak signal detected by thechamber itself (active mode) see Fig 6
The possibility to produce narrow collimatedX-ray beams allowing one to reach detectionprofiles with the required peak position detec-tion accuracy (1ndash2 mm)
The value of the absorption coefficient for X-rayphotons is very sensitive to attenuating materialsand to the X-ray energy If d is the density Z theatomic number of the material and W the energyof X-rays the mass absorption coefficient m=dvaries as EethZ=W THORN3 at 40 keV The optimumenergy range for this application is between 40and 60 keV
22 Basic procedure
A narrow X-ray beam is precisely moved acrossa section of the MDT chamber A scintillatorinstalled under the chamber records the absorp-tion profile during the scanning along the multi-layer cross-section of the chamber The data areanalysed and provide projective measurementStereo-measurements are possible when usingdifferent inclination angles of the X-ray beamFrom the data registered a two-dimensional mapof wire positions can be reconstructed Everychamber is scanned in a few positions along thetube length A full calibration of the muonchambers produces a list of all the wire coordi-nates in two dimensions Using the scanningresults for a few sections along the chambers athree-dimensional map can finally be made
After the analysis of raw data the results arestored in a database providing a full list of the wireposition and accuracy together with the para-meters of the chamber
We can measure the MDT chamberrsquos wireposition using two different methods
The lsquopassiversquo method This is based on therelatively good transparency of a MDT cham-ber to X-rays and on the strong absorptionof the 50 mm tungsten anode wires The out-put intensity of the X-ray beam is measured
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 89
by a scintillator Fig 6 shows a typicalcounting rate of the scintillator versus the X
coordinate1
The lsquoactiversquo method This second procedure isbased on the position-dependence efficiencyof drift tubes for X-rays When the X-raybeam hits the wire photoelectrons Augerelectrons and fluorescence are produced ata much higher rate than in the gas or in the
walls These radiations have a high prob-ability of escaping from the surface of the wireand of being detected by direct ionization in thegas
Fig 6 shows the wire signal from a tubebelonging to the bottom layer of the chamberThe three wires in the upper layers above theactive tube are clearly seen as absorption peaks bythe active tube they produce a shadow effectwhich is observed as inverted peaks comparedto the tube wire signal itself The relative intensityof the signals is not only proportional to theabsorption power of the material but also tothe probability of photoelectrons to escape fromthe material and to ionize the detecting gas ofthe drift tube
Passive mode
Layer 4 Active mode
X coordinate (mm)
Co
un
tin
g r
ate
(kH
z)
4 absorption lines seen
3 absorption lines seen
1 peak wire signal
Fig 6 Passive and active scanning methods
1 In this paper we exclusively use the natural X-ray
tomograph coordinate system (see Ref [12]) X (scanning
direction in the horizontal plane) Y (in the horizontal plane
perpendicular to the scanning direction) and Z (the vertical
coordinate) The MDT chambers in ATLAS use another system
of coordinates x (along the tubes) y (normal to chamber plane)
and z (across the tubes) Consequently the correspondence is as
follows X-zY-xZ-y
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10190
3 The X-ray tomograph
31 The challenge
The maximum width of the standardized MDTchambers has been fixed at 2160mm Thereforethe maximum scanning span of the X-ray tomo-graph has been chosen to be 2200mm The mainchallenge was to provide in short time to theconstructors reliable results on the wire position-ing of different chamber prototypes built with20 mm accuracy specification The X-ray tomo-graph had to consequently achieve a betteraccuracy aiming at 10 mm and be built withinstringent time and money constraints
32 Design considerations
The measurement of the position of the wires intwo dimensions along a cross-section of achamber was performed by using a cart movingon well-polished rails carrying two X-ray sourcesand their collimating systems producing beamswith opposite and fixed inclination anglesThe selection of the two inclination angles Y1Y2 the so-called lsquoscanning anglersquo relative to theperpendicular to the chamber plane was animportant issue under consideration for tworeasons
They define the stereo-angle (jY1j thorn jY2j) andconsequently the two-dimensional accuracy ofthe position of the wires
The absorption peak pattern of the wireswithin a corridor defined by the Y angle mustshow well-separated absorption signals Thispattern is very sensitive to the Y angle anddepends upon the chamber structure character-istics ie the height of the spacer the symmetrybetween the two multilayers and the horizontaland vertical wire pitch
The most convenient choice is to align thebeams in the natural 7301 corridors of themultilayers The angles must be chosen a little bithigher or lower in order to separate the wire peaksignals In the described set-up they are 73141
The geometric reconstruction requires theknowledge of four mechanical parameters the
linear scanning position the vertical straightnessthe angular displacement around the Y -axis (pitchangle) and the angular variation around the X -axis(roll angle) The excellent control of the cartposition (better than 1 mm) was crucial for achiev-ing high reconstruction accuracy We chose for themain structure of the X-ray tomograph standardhigh-precision mechanics combined with ultimateprecision survey tools in order to monitor the mainparameters controlling the X-ray beam positionduring scanning In this respect industrial laserinterferometers2 were used for monitoring Z (thevertical straightness) X (the horizontal displace-ment) DY (the pitch angle variation) and the rollangle
Reliable industrial choices were made wheneverpossible for the construction of the other parts andfor the software of the X-ray tomograph
33 Construction
Fig 7 shows the X-ray tomograph designed andbuilt at CERN Owing to the necessity of a verystable environment the X-ray tomograph wasinstalled in a clean room in building 188 groundfloor (one of the most stable ground floors atCERN) with temperature control and humiditycontrol
The X-ray tomograph consists of
A fixed 3500mm iron portico equipped with aprecise motorized cart rolling on rails along theX -axis
A rolling cart equipped with two collimated X-ray beams at 7314 with respect to the Z-axisThe 6281 stereo-angle has been chosen on thebasis of our previous experience and fits thetube corridors perfectly provided the spacersbetween the multilayers are correctly chosenThe X-ray beams have a cross-section of about30 mm 8mm and a divergence of about
2The basic principle of industrial interferometry used here is
to count the number of fringes of the interference between a
fixed reference laser beam and a second laser beam which is
reflected by a mirror installed on the moving chart Industry
suggests a few optical schemes of the main and reference beams
to allow measurements of longitudinal transversal (with respect
to the laser beam direction) and angular variations
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 91
60 mrad the longer axis being aligned along thewires to be measured
Four interferometers which monitor the mainparameters of the scanner movement the linearX value the pitch angle (rotation around the Y -axis) and two vertical straightnesses (theirbeams are parallel with distance about 160mmbetween them) The last two measurementsallow the calculation of the yaw (rotationaround the Z-axis) The horizontal straightnessis not monitored as it would have only asecond-order effect in the corrections Theinterferometers allow monitoring (2000 mea-surements per second) with an accuracy about03 mm for linear 03 mrad for angular para-meters over the full range of the scanner
Two calibration rulers each split into two parts(Fig 8) and placed under and above thechamber to be measured Every calibrationruler consists of two layers of nine ceramicplates (Fig 9) with six gold strips on eachcomposing a regular grid of 54 equivalent wireswith 30mm pitch The distance between thoselayers is 45mm The ceramic plates are installedon a carbon fibre support The calibrationrulers are used for the online (ie it is made for
every scan) reconstruction of the geometricalparameters of the X-ray tomograph and tocheck the stability and reliability of the X-raytomograph
A set of computer-controlled motors allowingthe alignment of the chamber on the X-raytomograph and choosing of the cross-sectionplane of the chamber One small moving cart isenslaved to the X-ray cart supporting twoscintillators optimized for the X-ray beamenergy spectrum following the beams inorder to record two shadowgrams simulta-neously
The control of the X-ray tomograph cart andchamber movements the data-acquisition andparameter monitoring are supported by adistributed online system based on VME PCSun computers connected via PCI MXI sharedmemory interface using softwarewritten on CC++ and LabVIEW
Two X-ray high-voltage power supplies (up to60 kV 50mA) The two high-voltage cablesfeeding the X-ray tubes move in synchronismwith the scanner on a separate cart in orderto ensure a perfect stability of the X-ray rollingcart
and collimating systems
Support of the upper calibration ruler
Chamber
Porticowithrails
Cart with shielded X-ray tubes
Z
X
Z
X
Fig 7 The X-ray tomograph during the mechanical control of a layers muon
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10192
4 Online calibration
Measurements with an accuracy of a fewmicrons require a precise control of all thegeometrical parameters of the X-ray tomographduring the measurements This requires the con-tinuous monitoring of the absolute position of theinterferometer laser beams in an external system ofcoordinates and of the origin of all laser axes Thislast point is not an easy requirement becauseduring the installation of the chambers on the
X-ray tomograph or during maintenance workthe trajectory of the laser beams may be frequentlyinterrupted for a short time by some object whichautomatically leads to the loss of the zeroreference In order to avoid this problem thecalibration and zero reset must be done for everyscan
During the scanning the wires of the referencerulers must be seen by the X-ray beams of theX-ray tomograph without overlap from thechamber wires The reference wires are previouslymeasured independently on a very-high-precisionoptical bench (E1 mm) During the scanning theshadow patterns of both the wires of the referencerulers and of the chamber are registered simulta-neously The basic geometrical parameters of theX-ray tomograph are then calculated for each scanusing the recorded shadowgram together with theoptical results of the position of the reference rulerwires [13] This approach allows a reliablepermanent control of the accuracy of the measure-ments Such a calibration scheme provides thebasis for the calculation of the position of allchamber wires
Let us stress that all the measurements are donein the same scan all calculations for the chamberand reference ruler wires are made with the same
Fig 8 One half of the calibration ruler (consisting of nine elements fixed on the carbon fibre tube) during assembling and
measurements
Fig 9 One element of the calibration ruler each ceramic plate
has six narrow gold strips faking wires
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 93
geometrical parameters The two reference rulersare measured with a known accuracy As aconsequence the chamber which is lying betweenthe two reference rulers is measured with the sameaccuracy Moreover in the offline analysis [13] thesame functions are used for the geometricalcalculation of the reference ruler and the chamberwires The X-ray tomograph accuracy is estimatedusing the residuals of the difference between thepredictions of the position of each wire of thereference rulers as given by the X-ray beams andtheir actual known position This method allowsone to control the accuracy of the X-ray tomo-graph scan-by-scan and consequently to detect anyloss of performance it allows for example toreveal small displacements or failures of theinterferometer system
Let us stress that the calibration is performedfor every scan and therefore for every scan thegeometrical parameters are calculated and theX-ray tomograph accuracy is checked
While studying the X-ray tomograph perfor-mance soon after its construction it became clearthat the precision of the optical measurements ofthe calibration rulers was not high enough toachieve the wanted accuracy It was later con-firmed that the optical measurements performedby the CERN Metrology Group using conven-tional industrial equipment were not preciseenough Therefore we had to develop a dedicatedmethod in order to improve the accuracy
Geometrical characteristics of the referencerulers may be divided into two groups the firstgroup consists of the parameters which can becorrected using direct X-ray tomograph measure-ments and the second group consists of para-meters for which other external reference devicesare needed In the first group we can note thatsome systematic errors in the position of individualcalibration ruler wires are well detected by theX-ray tomograph and those components of theoptical measurement errors are corrected by usingthe data coming from several scans As an examplefor the second group of parameters the straight-ness of the monitoring X-ray beams does not allowthe detection of possible errors in the opticalmeasurements of the vertical and horizontal shiftsbetween the layers of the calibration rulers If such
an error exists it leads to an error in the verticalscale and in the orthogonality of the system ofcoordinates of the X-ray tomograph
We call these two groups of optical measure-ment errors lsquonon-linearrsquo and lsquolinearrsquo errors3
respectively An autocalibration approach wasdeveloped to improve the accuracy of the opticalmeasurements It consists of two steps correctionof non-linear components in optical measurementsand determination (plus correction) of the linearparameters
The first step does not require special commentsAfter the geometrical reconstruction of the cali-bration ruler wires determined from several scans(in order to improve statistics) the residualsbetween the predictions by every X-ray beamand the optical measurements are calculatedUsing corresponding values from both non-paral-lel beams the two-dimensional correction ofposition of every calibration ruler wire may befound Because the pattern recognition maychange after the corrections an additional itera-tion may be done
5 Autocalibration correction of the linear errors in
the optical measurements of the calibration ruler
wire positions
51 Theoretical aspects
Let us consider the question of the linearsystematics effect of the absolute scales the scalesfor the X - and Z-axis and their relative angle (non-orthogonality of the system of coordinates) The X
scale is determined by the linear interferometer(measuring the position of the cart along the scan)the corresponding laser beam is parallel to theOX -axis (with second-order errors due to the non-parallelism of the laser beam and scan direction)The OZ scale and the non-orthogonality errors ofthe coordinate system however are completelydefined by the optical measurement errors They
3We have chosen the names in that way because the lsquolinearrsquo
(non-linear) errors in the optical measurements are propagated
to the chamber wire position errors which are linear (non-
linear) functions of the wire coordinates
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10194
cannot be determined from the X-ray tomographmeasurements themselves During the geometryfit the calculated effects of any change in the OZ
scale and in the non-orthogonality of the axes maybe completely compensated with correspondingchanges of the geometrical parameters of the X-ray beams Therefore some external measurementsare necessary
Let us consider three hypothetical parallel wiresA B C of the same rigid chamber (see Fig 10 thewires are perpendicular to the plane of picture andonly their projections A B C are seen) The pointsA B C should not lie on the same straight lineLet us imagine that they are measured by the X-ray tomograph in the original position (Fig 10a)and after rotation of the chamber by 901 aroundthe wiresrsquo direction (Fig 10b) If the X-raytomograph has a Z scale error instead of thepositions of the points A B C the positions ofpoints A0 B0 C0 and points A00 B00 C00 in therotated orientation will be found Putting bothmeasurements in the same system of coordinates
(we need three points because the angle aboutwhich these three points are rotated is not knownprecisely and in general does not equal 901) wecan estimate the Z scale error (Fig 10c)
A similar analysis applies for the non-orthogon-ality error Three wires A B C are measured aspoints A0 B0 C0 (Fig 11a) and as points A00 B00C00 in the 901 orientation (Fig 11b) Putting bothmeasurements in the same system of coordinates(Fig 11c) we can evaluate the non-orthogonalityerror introduced by the X-ray tomograph
Having determined the values of the Z scale andthe non-orthogonality errors we can calculate andthen correct the horizontal and vertical shiftsbetween the layers of the reference ruler becausethese shifts are the source of the errors in the Z
scale and non-orthogonality
52 Practical realization
In order to realize the described approach asmall calibration chamber (gauge chamber) was
A B
C
A B
C
AA
C
B
C
B
B
BA
A
C C
O X
Z
(a) (b) (c)
Fig 11 Non-orthogonality error calibration
A B
C
A
B
B
C
C
A
A
C
BCA
B
AB
C
O X
Z
(a) (b) (c)
Fig 10 Z scale error calibration
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 95
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
15 Momentum measurement
Charged particles with high momentum (above300GeVc) crossing the toroid magnets of themuon spectrometer will have very large radius ofcurvature producing small saggita signals(sE1525mm) over the long muon trajectories(5ndash15m) between three measuring stations Inorder to achieve the required precision in themomentum measurement the saggita must bedetermined with a precision of 15ndash50 mm overmost of the pseudorapidity range a precisionmeasurement of the track coordinates in theprincipal bending direction of the magnetic fieldis provided by Monitored Drift Tubes (MDTs)and Cathode Strip Chambers (CSCs) The MDTchambers are arranged in three layers all aroundthe calorimeter in order to determine the momen-tum with the best possible resolution The CSCsare used for the very forward area where theparticle flux is too high for the drift chambers
With this layout a three-point measurement isavailable (see Fig 4) by installing three stations ofmultilayer MDT chambers covering the fullrapidity gap with high hermeticity The driftMDT chambers are called lsquomonitoredrsquo because acontinuous monitoring of the internal chamberdeformations is done In addition to this on-linesurvey the position in space of each chamber isalso monitored with the required precision(r20 mm)
With this layout the momentum resolution istypically 2ndash3 at 100GeVc over most of thekinematic range
The muon spectrometer is designed for amomentum resolution DpT=pTo1104 pT wherepT is in GeV for pT gt 300GeV To achieve thisresolution by a three-point measurement with thesize and bending power of the ATLAS toroidseach point must be measured with an accuracy ofabout 50 mm This sets the scale for the require-ments on the intrinsic resolution the mechanical
End-captoroid
Barrel toroidcoils
Calorimeters
MDT chambersResistive plate chambers
Inner detector
oslash ~
24 m
Fig 3 Cross view of the muon spectrometer
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10186
precision and the survey accuracy of the muonchambers
The momentum resolution of the spectrometeris limited by the intrinsic detector resolutionMDT calibration errors chamber positioninguncertainties multiple scattering and statisticalfluctuations of energy loss At smaller momenta(below about 300GeVc) the resolution is limitedby multiple scattering to a few per cent at highermomenta (above 300GeVc) it is dominated bychamber precision and alignment
16 The MDT chambers
The task in numbers is impressive the totalnumber of MDTs is E1200 making a total area ofabout 5000m2 The total number of readoutchannels (number of drift tubes) is E400 000 andthe total length of the tubes is E1100 km
The basic detection elements of the MDTchambers are drift chambers made of aluminiumtubes of 30mm diameter and 400 mm wall thick-ness with a 50 mm diameter central WndashRe wireThe tubes are operated with a non-flammableArndashCO2 mixture at 3 bar absolute pressure
The tubes are produced by extrusion from ahard aluminium alloy and are available commer-cially They are closed by endplugs which provideaccurate positioning of the anode wires wiretension gas tightness and electrical and gasconnections The drift tubes can be manufacturedto tight mechanical tolerances which are well-matched to their intrinsic resolution propertiesmostly using automated assembly procedures Thetube lengths vary from 70 to 630 cm
For a good linear spacendashtime relation with amaximum drift time of B500ns the envisagedworking point provides a small Lorentz angle andgood ageing properties due to low gas amplification
X [c
m]
Z [cm]Collision vertex
Pseudo-rapidity Beam lineaxis
Fig 4 Longitudinal cross-section in the bending plane of the spectrometer showing the barrel and end-cap magnet air-coil toroid
configuration It shows the pseudo-rapidity coverage of the muon spectrometer from 0 to 28 and a sketch of the layout principle of the
three detecting muon stations showing the trajectories of a few GeVc positively and negatively charged particles
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 87
Using the timendashdrift distance relationship the single-wire resolution is typically 80mm when operated athigh gas pressure (3bar)
The muon chamber layout was optimized toreduce the number of different chamber sizes
The barrel chambers are of rectangular shapewith sizes from 2 to 10m2
The end-cap chambers are of trapezoidal shapeTheir sizes range from 1 to 10m2 for individualchamber modules and up to 30m2 when severalof them are preassembled for installation Thechamber design must guarantee reliability andstability of construction and operation for theanticipated lifetime of the experiment in anirradiated environment The basic design isshown in Fig 5
To improve the resolution of a chamber beyondthe single-wire limit and to achieve adequateredundancy for pattern recognition the MDTchambers are constructed from 2 4 monolayersof drift tubes for the inner and 2 3 monolayersfor the middle and outer stations The tubes are
arranged in multilayers of three or four mono-layers respectively on either side of a rigidsupport structure The support structures thelsquospacer framesrsquo provide for accurate positioningof the two multilayers with respect to each otherand for mechanical integrity under effects oftemperature and gravity As an example ofdifficulties encountered when seeking to reach therequired resolution the drift tubes which are notmounted vertically are bent slightly in order tomatch their shape to the gravitational sag of thewires
The structural components of the spacerframes are three lsquocrossplatesrsquo to which thedrift tube multilayers are attached and twolsquolong beamsrsquo connecting the crossplates Theframes need to be constructed to a moderatemechanical accuracy of 705mm only accuratepositioning of the drift tubes is provided bythe assembly procedure The chambers willbe attached to the rail structures of the spectro-meter by three-point kinematic supports Oncea chamber is installed in its final location in
Cross plate
Multilayer
In-plane alignment
Longitudinal beam
3-layerMultilayer
micro track
Tube cross section
BE
+
Fig 5 Schematic drawing of a rectangular MDT chamber constructed from multilayers of three monolayers each for installation in
the barrel spectrometer The chambers for the end-cap are of trapezoidal shape but are of similar design otherwise In the toroidal
magnetic field the tubes are oriented essentially parallel to the field lines
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10188
the spectrometer mechanical deformations aremonitored by an in-plane optical system hencethe name lsquoMDTrsquo chambers The spacer framessupport most of the components of the alignmentsystem
MDT chambers are made for robust andreliable operation This is achieved thanks to themechanical concept which provides a solid separa-tion of detecting cells and a good electricalshielding of neighbouring wires
Chambers must withstand very long storagetime followed by very long running periods Inorder to reach the required momentum resolutionall MDT chambers must be constructed with20 mm mechanical accuracy
17 Quality assurance for MDTs
The MDT project in ATLAS is on the scale of avery large industrial project aiming at an excep-tional quality of construction in terms of accuracy(r20 mm) material reliability delivery and assem-bly schedule The X-ray tomograph quality-con-trol platform may be considered as one of the keyinstruments needed for the MDT
The production of the MDT chambers is basedon the concept of very high mechanical precisionand the use of X-ray tomography as a quality-control platform for the mechanically achievedprecision
2 X-ray tomography
The development design construction andtuning of an X-ray tomograph suited forthe MDT quality control was the aim of theX-ray Quality-Control lsquoX-QCrsquo project (see Refs[5ndash11])
21 Basic principle
The choice of X-ray tomography as a quality-control tool for the MDT project was based ontwo essential features
The ability of X-rays to detect 50 mm tungstenwires hidden inside the aluminium drift tubes of
the muon detector Sharp absorption spikes areobserved either as a shadow seen by an X-raydetector placed under the chamber (passivemode) or as a sharp peak signal detected by thechamber itself (active mode) see Fig 6
The possibility to produce narrow collimatedX-ray beams allowing one to reach detectionprofiles with the required peak position detec-tion accuracy (1ndash2 mm)
The value of the absorption coefficient for X-rayphotons is very sensitive to attenuating materialsand to the X-ray energy If d is the density Z theatomic number of the material and W the energyof X-rays the mass absorption coefficient m=dvaries as EethZ=W THORN3 at 40 keV The optimumenergy range for this application is between 40and 60 keV
22 Basic procedure
A narrow X-ray beam is precisely moved acrossa section of the MDT chamber A scintillatorinstalled under the chamber records the absorp-tion profile during the scanning along the multi-layer cross-section of the chamber The data areanalysed and provide projective measurementStereo-measurements are possible when usingdifferent inclination angles of the X-ray beamFrom the data registered a two-dimensional mapof wire positions can be reconstructed Everychamber is scanned in a few positions along thetube length A full calibration of the muonchambers produces a list of all the wire coordi-nates in two dimensions Using the scanningresults for a few sections along the chambers athree-dimensional map can finally be made
After the analysis of raw data the results arestored in a database providing a full list of the wireposition and accuracy together with the para-meters of the chamber
We can measure the MDT chamberrsquos wireposition using two different methods
The lsquopassiversquo method This is based on therelatively good transparency of a MDT cham-ber to X-rays and on the strong absorptionof the 50 mm tungsten anode wires The out-put intensity of the X-ray beam is measured
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 89
by a scintillator Fig 6 shows a typicalcounting rate of the scintillator versus the X
coordinate1
The lsquoactiversquo method This second procedure isbased on the position-dependence efficiencyof drift tubes for X-rays When the X-raybeam hits the wire photoelectrons Augerelectrons and fluorescence are produced ata much higher rate than in the gas or in the
walls These radiations have a high prob-ability of escaping from the surface of the wireand of being detected by direct ionization in thegas
Fig 6 shows the wire signal from a tubebelonging to the bottom layer of the chamberThe three wires in the upper layers above theactive tube are clearly seen as absorption peaks bythe active tube they produce a shadow effectwhich is observed as inverted peaks comparedto the tube wire signal itself The relative intensityof the signals is not only proportional to theabsorption power of the material but also tothe probability of photoelectrons to escape fromthe material and to ionize the detecting gas ofthe drift tube
Passive mode
Layer 4 Active mode
X coordinate (mm)
Co
un
tin
g r
ate
(kH
z)
4 absorption lines seen
3 absorption lines seen
1 peak wire signal
Fig 6 Passive and active scanning methods
1 In this paper we exclusively use the natural X-ray
tomograph coordinate system (see Ref [12]) X (scanning
direction in the horizontal plane) Y (in the horizontal plane
perpendicular to the scanning direction) and Z (the vertical
coordinate) The MDT chambers in ATLAS use another system
of coordinates x (along the tubes) y (normal to chamber plane)
and z (across the tubes) Consequently the correspondence is as
follows X-zY-xZ-y
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10190
3 The X-ray tomograph
31 The challenge
The maximum width of the standardized MDTchambers has been fixed at 2160mm Thereforethe maximum scanning span of the X-ray tomo-graph has been chosen to be 2200mm The mainchallenge was to provide in short time to theconstructors reliable results on the wire position-ing of different chamber prototypes built with20 mm accuracy specification The X-ray tomo-graph had to consequently achieve a betteraccuracy aiming at 10 mm and be built withinstringent time and money constraints
32 Design considerations
The measurement of the position of the wires intwo dimensions along a cross-section of achamber was performed by using a cart movingon well-polished rails carrying two X-ray sourcesand their collimating systems producing beamswith opposite and fixed inclination anglesThe selection of the two inclination angles Y1Y2 the so-called lsquoscanning anglersquo relative to theperpendicular to the chamber plane was animportant issue under consideration for tworeasons
They define the stereo-angle (jY1j thorn jY2j) andconsequently the two-dimensional accuracy ofthe position of the wires
The absorption peak pattern of the wireswithin a corridor defined by the Y angle mustshow well-separated absorption signals Thispattern is very sensitive to the Y angle anddepends upon the chamber structure character-istics ie the height of the spacer the symmetrybetween the two multilayers and the horizontaland vertical wire pitch
The most convenient choice is to align thebeams in the natural 7301 corridors of themultilayers The angles must be chosen a little bithigher or lower in order to separate the wire peaksignals In the described set-up they are 73141
The geometric reconstruction requires theknowledge of four mechanical parameters the
linear scanning position the vertical straightnessthe angular displacement around the Y -axis (pitchangle) and the angular variation around the X -axis(roll angle) The excellent control of the cartposition (better than 1 mm) was crucial for achiev-ing high reconstruction accuracy We chose for themain structure of the X-ray tomograph standardhigh-precision mechanics combined with ultimateprecision survey tools in order to monitor the mainparameters controlling the X-ray beam positionduring scanning In this respect industrial laserinterferometers2 were used for monitoring Z (thevertical straightness) X (the horizontal displace-ment) DY (the pitch angle variation) and the rollangle
Reliable industrial choices were made wheneverpossible for the construction of the other parts andfor the software of the X-ray tomograph
33 Construction
Fig 7 shows the X-ray tomograph designed andbuilt at CERN Owing to the necessity of a verystable environment the X-ray tomograph wasinstalled in a clean room in building 188 groundfloor (one of the most stable ground floors atCERN) with temperature control and humiditycontrol
The X-ray tomograph consists of
A fixed 3500mm iron portico equipped with aprecise motorized cart rolling on rails along theX -axis
A rolling cart equipped with two collimated X-ray beams at 7314 with respect to the Z-axisThe 6281 stereo-angle has been chosen on thebasis of our previous experience and fits thetube corridors perfectly provided the spacersbetween the multilayers are correctly chosenThe X-ray beams have a cross-section of about30 mm 8mm and a divergence of about
2The basic principle of industrial interferometry used here is
to count the number of fringes of the interference between a
fixed reference laser beam and a second laser beam which is
reflected by a mirror installed on the moving chart Industry
suggests a few optical schemes of the main and reference beams
to allow measurements of longitudinal transversal (with respect
to the laser beam direction) and angular variations
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 91
60 mrad the longer axis being aligned along thewires to be measured
Four interferometers which monitor the mainparameters of the scanner movement the linearX value the pitch angle (rotation around the Y -axis) and two vertical straightnesses (theirbeams are parallel with distance about 160mmbetween them) The last two measurementsallow the calculation of the yaw (rotationaround the Z-axis) The horizontal straightnessis not monitored as it would have only asecond-order effect in the corrections Theinterferometers allow monitoring (2000 mea-surements per second) with an accuracy about03 mm for linear 03 mrad for angular para-meters over the full range of the scanner
Two calibration rulers each split into two parts(Fig 8) and placed under and above thechamber to be measured Every calibrationruler consists of two layers of nine ceramicplates (Fig 9) with six gold strips on eachcomposing a regular grid of 54 equivalent wireswith 30mm pitch The distance between thoselayers is 45mm The ceramic plates are installedon a carbon fibre support The calibrationrulers are used for the online (ie it is made for
every scan) reconstruction of the geometricalparameters of the X-ray tomograph and tocheck the stability and reliability of the X-raytomograph
A set of computer-controlled motors allowingthe alignment of the chamber on the X-raytomograph and choosing of the cross-sectionplane of the chamber One small moving cart isenslaved to the X-ray cart supporting twoscintillators optimized for the X-ray beamenergy spectrum following the beams inorder to record two shadowgrams simulta-neously
The control of the X-ray tomograph cart andchamber movements the data-acquisition andparameter monitoring are supported by adistributed online system based on VME PCSun computers connected via PCI MXI sharedmemory interface using softwarewritten on CC++ and LabVIEW
Two X-ray high-voltage power supplies (up to60 kV 50mA) The two high-voltage cablesfeeding the X-ray tubes move in synchronismwith the scanner on a separate cart in orderto ensure a perfect stability of the X-ray rollingcart
and collimating systems
Support of the upper calibration ruler
Chamber
Porticowithrails
Cart with shielded X-ray tubes
Z
X
Z
X
Fig 7 The X-ray tomograph during the mechanical control of a layers muon
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10192
4 Online calibration
Measurements with an accuracy of a fewmicrons require a precise control of all thegeometrical parameters of the X-ray tomographduring the measurements This requires the con-tinuous monitoring of the absolute position of theinterferometer laser beams in an external system ofcoordinates and of the origin of all laser axes Thislast point is not an easy requirement becauseduring the installation of the chambers on the
X-ray tomograph or during maintenance workthe trajectory of the laser beams may be frequentlyinterrupted for a short time by some object whichautomatically leads to the loss of the zeroreference In order to avoid this problem thecalibration and zero reset must be done for everyscan
During the scanning the wires of the referencerulers must be seen by the X-ray beams of theX-ray tomograph without overlap from thechamber wires The reference wires are previouslymeasured independently on a very-high-precisionoptical bench (E1 mm) During the scanning theshadow patterns of both the wires of the referencerulers and of the chamber are registered simulta-neously The basic geometrical parameters of theX-ray tomograph are then calculated for each scanusing the recorded shadowgram together with theoptical results of the position of the reference rulerwires [13] This approach allows a reliablepermanent control of the accuracy of the measure-ments Such a calibration scheme provides thebasis for the calculation of the position of allchamber wires
Let us stress that all the measurements are donein the same scan all calculations for the chamberand reference ruler wires are made with the same
Fig 8 One half of the calibration ruler (consisting of nine elements fixed on the carbon fibre tube) during assembling and
measurements
Fig 9 One element of the calibration ruler each ceramic plate
has six narrow gold strips faking wires
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 93
geometrical parameters The two reference rulersare measured with a known accuracy As aconsequence the chamber which is lying betweenthe two reference rulers is measured with the sameaccuracy Moreover in the offline analysis [13] thesame functions are used for the geometricalcalculation of the reference ruler and the chamberwires The X-ray tomograph accuracy is estimatedusing the residuals of the difference between thepredictions of the position of each wire of thereference rulers as given by the X-ray beams andtheir actual known position This method allowsone to control the accuracy of the X-ray tomo-graph scan-by-scan and consequently to detect anyloss of performance it allows for example toreveal small displacements or failures of theinterferometer system
Let us stress that the calibration is performedfor every scan and therefore for every scan thegeometrical parameters are calculated and theX-ray tomograph accuracy is checked
While studying the X-ray tomograph perfor-mance soon after its construction it became clearthat the precision of the optical measurements ofthe calibration rulers was not high enough toachieve the wanted accuracy It was later con-firmed that the optical measurements performedby the CERN Metrology Group using conven-tional industrial equipment were not preciseenough Therefore we had to develop a dedicatedmethod in order to improve the accuracy
Geometrical characteristics of the referencerulers may be divided into two groups the firstgroup consists of the parameters which can becorrected using direct X-ray tomograph measure-ments and the second group consists of para-meters for which other external reference devicesare needed In the first group we can note thatsome systematic errors in the position of individualcalibration ruler wires are well detected by theX-ray tomograph and those components of theoptical measurement errors are corrected by usingthe data coming from several scans As an examplefor the second group of parameters the straight-ness of the monitoring X-ray beams does not allowthe detection of possible errors in the opticalmeasurements of the vertical and horizontal shiftsbetween the layers of the calibration rulers If such
an error exists it leads to an error in the verticalscale and in the orthogonality of the system ofcoordinates of the X-ray tomograph
We call these two groups of optical measure-ment errors lsquonon-linearrsquo and lsquolinearrsquo errors3
respectively An autocalibration approach wasdeveloped to improve the accuracy of the opticalmeasurements It consists of two steps correctionof non-linear components in optical measurementsand determination (plus correction) of the linearparameters
The first step does not require special commentsAfter the geometrical reconstruction of the cali-bration ruler wires determined from several scans(in order to improve statistics) the residualsbetween the predictions by every X-ray beamand the optical measurements are calculatedUsing corresponding values from both non-paral-lel beams the two-dimensional correction ofposition of every calibration ruler wire may befound Because the pattern recognition maychange after the corrections an additional itera-tion may be done
5 Autocalibration correction of the linear errors in
the optical measurements of the calibration ruler
wire positions
51 Theoretical aspects
Let us consider the question of the linearsystematics effect of the absolute scales the scalesfor the X - and Z-axis and their relative angle (non-orthogonality of the system of coordinates) The X
scale is determined by the linear interferometer(measuring the position of the cart along the scan)the corresponding laser beam is parallel to theOX -axis (with second-order errors due to the non-parallelism of the laser beam and scan direction)The OZ scale and the non-orthogonality errors ofthe coordinate system however are completelydefined by the optical measurement errors They
3We have chosen the names in that way because the lsquolinearrsquo
(non-linear) errors in the optical measurements are propagated
to the chamber wire position errors which are linear (non-
linear) functions of the wire coordinates
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10194
cannot be determined from the X-ray tomographmeasurements themselves During the geometryfit the calculated effects of any change in the OZ
scale and in the non-orthogonality of the axes maybe completely compensated with correspondingchanges of the geometrical parameters of the X-ray beams Therefore some external measurementsare necessary
Let us consider three hypothetical parallel wiresA B C of the same rigid chamber (see Fig 10 thewires are perpendicular to the plane of picture andonly their projections A B C are seen) The pointsA B C should not lie on the same straight lineLet us imagine that they are measured by the X-ray tomograph in the original position (Fig 10a)and after rotation of the chamber by 901 aroundthe wiresrsquo direction (Fig 10b) If the X-raytomograph has a Z scale error instead of thepositions of the points A B C the positions ofpoints A0 B0 C0 and points A00 B00 C00 in therotated orientation will be found Putting bothmeasurements in the same system of coordinates
(we need three points because the angle aboutwhich these three points are rotated is not knownprecisely and in general does not equal 901) wecan estimate the Z scale error (Fig 10c)
A similar analysis applies for the non-orthogon-ality error Three wires A B C are measured aspoints A0 B0 C0 (Fig 11a) and as points A00 B00C00 in the 901 orientation (Fig 11b) Putting bothmeasurements in the same system of coordinates(Fig 11c) we can evaluate the non-orthogonalityerror introduced by the X-ray tomograph
Having determined the values of the Z scale andthe non-orthogonality errors we can calculate andthen correct the horizontal and vertical shiftsbetween the layers of the reference ruler becausethese shifts are the source of the errors in the Z
scale and non-orthogonality
52 Practical realization
In order to realize the described approach asmall calibration chamber (gauge chamber) was
A B
C
A B
C
AA
C
B
C
B
B
BA
A
C C
O X
Z
(a) (b) (c)
Fig 11 Non-orthogonality error calibration
A B
C
A
B
B
C
C
A
A
C
BCA
B
AB
C
O X
Z
(a) (b) (c)
Fig 10 Z scale error calibration
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 95
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
precision and the survey accuracy of the muonchambers
The momentum resolution of the spectrometeris limited by the intrinsic detector resolutionMDT calibration errors chamber positioninguncertainties multiple scattering and statisticalfluctuations of energy loss At smaller momenta(below about 300GeVc) the resolution is limitedby multiple scattering to a few per cent at highermomenta (above 300GeVc) it is dominated bychamber precision and alignment
16 The MDT chambers
The task in numbers is impressive the totalnumber of MDTs is E1200 making a total area ofabout 5000m2 The total number of readoutchannels (number of drift tubes) is E400 000 andthe total length of the tubes is E1100 km
The basic detection elements of the MDTchambers are drift chambers made of aluminiumtubes of 30mm diameter and 400 mm wall thick-ness with a 50 mm diameter central WndashRe wireThe tubes are operated with a non-flammableArndashCO2 mixture at 3 bar absolute pressure
The tubes are produced by extrusion from ahard aluminium alloy and are available commer-cially They are closed by endplugs which provideaccurate positioning of the anode wires wiretension gas tightness and electrical and gasconnections The drift tubes can be manufacturedto tight mechanical tolerances which are well-matched to their intrinsic resolution propertiesmostly using automated assembly procedures Thetube lengths vary from 70 to 630 cm
For a good linear spacendashtime relation with amaximum drift time of B500ns the envisagedworking point provides a small Lorentz angle andgood ageing properties due to low gas amplification
X [c
m]
Z [cm]Collision vertex
Pseudo-rapidity Beam lineaxis
Fig 4 Longitudinal cross-section in the bending plane of the spectrometer showing the barrel and end-cap magnet air-coil toroid
configuration It shows the pseudo-rapidity coverage of the muon spectrometer from 0 to 28 and a sketch of the layout principle of the
three detecting muon stations showing the trajectories of a few GeVc positively and negatively charged particles
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 87
Using the timendashdrift distance relationship the single-wire resolution is typically 80mm when operated athigh gas pressure (3bar)
The muon chamber layout was optimized toreduce the number of different chamber sizes
The barrel chambers are of rectangular shapewith sizes from 2 to 10m2
The end-cap chambers are of trapezoidal shapeTheir sizes range from 1 to 10m2 for individualchamber modules and up to 30m2 when severalof them are preassembled for installation Thechamber design must guarantee reliability andstability of construction and operation for theanticipated lifetime of the experiment in anirradiated environment The basic design isshown in Fig 5
To improve the resolution of a chamber beyondthe single-wire limit and to achieve adequateredundancy for pattern recognition the MDTchambers are constructed from 2 4 monolayersof drift tubes for the inner and 2 3 monolayersfor the middle and outer stations The tubes are
arranged in multilayers of three or four mono-layers respectively on either side of a rigidsupport structure The support structures thelsquospacer framesrsquo provide for accurate positioningof the two multilayers with respect to each otherand for mechanical integrity under effects oftemperature and gravity As an example ofdifficulties encountered when seeking to reach therequired resolution the drift tubes which are notmounted vertically are bent slightly in order tomatch their shape to the gravitational sag of thewires
The structural components of the spacerframes are three lsquocrossplatesrsquo to which thedrift tube multilayers are attached and twolsquolong beamsrsquo connecting the crossplates Theframes need to be constructed to a moderatemechanical accuracy of 705mm only accuratepositioning of the drift tubes is provided bythe assembly procedure The chambers willbe attached to the rail structures of the spectro-meter by three-point kinematic supports Oncea chamber is installed in its final location in
Cross plate
Multilayer
In-plane alignment
Longitudinal beam
3-layerMultilayer
micro track
Tube cross section
BE
+
Fig 5 Schematic drawing of a rectangular MDT chamber constructed from multilayers of three monolayers each for installation in
the barrel spectrometer The chambers for the end-cap are of trapezoidal shape but are of similar design otherwise In the toroidal
magnetic field the tubes are oriented essentially parallel to the field lines
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10188
the spectrometer mechanical deformations aremonitored by an in-plane optical system hencethe name lsquoMDTrsquo chambers The spacer framessupport most of the components of the alignmentsystem
MDT chambers are made for robust andreliable operation This is achieved thanks to themechanical concept which provides a solid separa-tion of detecting cells and a good electricalshielding of neighbouring wires
Chambers must withstand very long storagetime followed by very long running periods Inorder to reach the required momentum resolutionall MDT chambers must be constructed with20 mm mechanical accuracy
17 Quality assurance for MDTs
The MDT project in ATLAS is on the scale of avery large industrial project aiming at an excep-tional quality of construction in terms of accuracy(r20 mm) material reliability delivery and assem-bly schedule The X-ray tomograph quality-con-trol platform may be considered as one of the keyinstruments needed for the MDT
The production of the MDT chambers is basedon the concept of very high mechanical precisionand the use of X-ray tomography as a quality-control platform for the mechanically achievedprecision
2 X-ray tomography
The development design construction andtuning of an X-ray tomograph suited forthe MDT quality control was the aim of theX-ray Quality-Control lsquoX-QCrsquo project (see Refs[5ndash11])
21 Basic principle
The choice of X-ray tomography as a quality-control tool for the MDT project was based ontwo essential features
The ability of X-rays to detect 50 mm tungstenwires hidden inside the aluminium drift tubes of
the muon detector Sharp absorption spikes areobserved either as a shadow seen by an X-raydetector placed under the chamber (passivemode) or as a sharp peak signal detected by thechamber itself (active mode) see Fig 6
The possibility to produce narrow collimatedX-ray beams allowing one to reach detectionprofiles with the required peak position detec-tion accuracy (1ndash2 mm)
The value of the absorption coefficient for X-rayphotons is very sensitive to attenuating materialsand to the X-ray energy If d is the density Z theatomic number of the material and W the energyof X-rays the mass absorption coefficient m=dvaries as EethZ=W THORN3 at 40 keV The optimumenergy range for this application is between 40and 60 keV
22 Basic procedure
A narrow X-ray beam is precisely moved acrossa section of the MDT chamber A scintillatorinstalled under the chamber records the absorp-tion profile during the scanning along the multi-layer cross-section of the chamber The data areanalysed and provide projective measurementStereo-measurements are possible when usingdifferent inclination angles of the X-ray beamFrom the data registered a two-dimensional mapof wire positions can be reconstructed Everychamber is scanned in a few positions along thetube length A full calibration of the muonchambers produces a list of all the wire coordi-nates in two dimensions Using the scanningresults for a few sections along the chambers athree-dimensional map can finally be made
After the analysis of raw data the results arestored in a database providing a full list of the wireposition and accuracy together with the para-meters of the chamber
We can measure the MDT chamberrsquos wireposition using two different methods
The lsquopassiversquo method This is based on therelatively good transparency of a MDT cham-ber to X-rays and on the strong absorptionof the 50 mm tungsten anode wires The out-put intensity of the X-ray beam is measured
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 89
by a scintillator Fig 6 shows a typicalcounting rate of the scintillator versus the X
coordinate1
The lsquoactiversquo method This second procedure isbased on the position-dependence efficiencyof drift tubes for X-rays When the X-raybeam hits the wire photoelectrons Augerelectrons and fluorescence are produced ata much higher rate than in the gas or in the
walls These radiations have a high prob-ability of escaping from the surface of the wireand of being detected by direct ionization in thegas
Fig 6 shows the wire signal from a tubebelonging to the bottom layer of the chamberThe three wires in the upper layers above theactive tube are clearly seen as absorption peaks bythe active tube they produce a shadow effectwhich is observed as inverted peaks comparedto the tube wire signal itself The relative intensityof the signals is not only proportional to theabsorption power of the material but also tothe probability of photoelectrons to escape fromthe material and to ionize the detecting gas ofthe drift tube
Passive mode
Layer 4 Active mode
X coordinate (mm)
Co
un
tin
g r
ate
(kH
z)
4 absorption lines seen
3 absorption lines seen
1 peak wire signal
Fig 6 Passive and active scanning methods
1 In this paper we exclusively use the natural X-ray
tomograph coordinate system (see Ref [12]) X (scanning
direction in the horizontal plane) Y (in the horizontal plane
perpendicular to the scanning direction) and Z (the vertical
coordinate) The MDT chambers in ATLAS use another system
of coordinates x (along the tubes) y (normal to chamber plane)
and z (across the tubes) Consequently the correspondence is as
follows X-zY-xZ-y
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10190
3 The X-ray tomograph
31 The challenge
The maximum width of the standardized MDTchambers has been fixed at 2160mm Thereforethe maximum scanning span of the X-ray tomo-graph has been chosen to be 2200mm The mainchallenge was to provide in short time to theconstructors reliable results on the wire position-ing of different chamber prototypes built with20 mm accuracy specification The X-ray tomo-graph had to consequently achieve a betteraccuracy aiming at 10 mm and be built withinstringent time and money constraints
32 Design considerations
The measurement of the position of the wires intwo dimensions along a cross-section of achamber was performed by using a cart movingon well-polished rails carrying two X-ray sourcesand their collimating systems producing beamswith opposite and fixed inclination anglesThe selection of the two inclination angles Y1Y2 the so-called lsquoscanning anglersquo relative to theperpendicular to the chamber plane was animportant issue under consideration for tworeasons
They define the stereo-angle (jY1j thorn jY2j) andconsequently the two-dimensional accuracy ofthe position of the wires
The absorption peak pattern of the wireswithin a corridor defined by the Y angle mustshow well-separated absorption signals Thispattern is very sensitive to the Y angle anddepends upon the chamber structure character-istics ie the height of the spacer the symmetrybetween the two multilayers and the horizontaland vertical wire pitch
The most convenient choice is to align thebeams in the natural 7301 corridors of themultilayers The angles must be chosen a little bithigher or lower in order to separate the wire peaksignals In the described set-up they are 73141
The geometric reconstruction requires theknowledge of four mechanical parameters the
linear scanning position the vertical straightnessthe angular displacement around the Y -axis (pitchangle) and the angular variation around the X -axis(roll angle) The excellent control of the cartposition (better than 1 mm) was crucial for achiev-ing high reconstruction accuracy We chose for themain structure of the X-ray tomograph standardhigh-precision mechanics combined with ultimateprecision survey tools in order to monitor the mainparameters controlling the X-ray beam positionduring scanning In this respect industrial laserinterferometers2 were used for monitoring Z (thevertical straightness) X (the horizontal displace-ment) DY (the pitch angle variation) and the rollangle
Reliable industrial choices were made wheneverpossible for the construction of the other parts andfor the software of the X-ray tomograph
33 Construction
Fig 7 shows the X-ray tomograph designed andbuilt at CERN Owing to the necessity of a verystable environment the X-ray tomograph wasinstalled in a clean room in building 188 groundfloor (one of the most stable ground floors atCERN) with temperature control and humiditycontrol
The X-ray tomograph consists of
A fixed 3500mm iron portico equipped with aprecise motorized cart rolling on rails along theX -axis
A rolling cart equipped with two collimated X-ray beams at 7314 with respect to the Z-axisThe 6281 stereo-angle has been chosen on thebasis of our previous experience and fits thetube corridors perfectly provided the spacersbetween the multilayers are correctly chosenThe X-ray beams have a cross-section of about30 mm 8mm and a divergence of about
2The basic principle of industrial interferometry used here is
to count the number of fringes of the interference between a
fixed reference laser beam and a second laser beam which is
reflected by a mirror installed on the moving chart Industry
suggests a few optical schemes of the main and reference beams
to allow measurements of longitudinal transversal (with respect
to the laser beam direction) and angular variations
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 91
60 mrad the longer axis being aligned along thewires to be measured
Four interferometers which monitor the mainparameters of the scanner movement the linearX value the pitch angle (rotation around the Y -axis) and two vertical straightnesses (theirbeams are parallel with distance about 160mmbetween them) The last two measurementsallow the calculation of the yaw (rotationaround the Z-axis) The horizontal straightnessis not monitored as it would have only asecond-order effect in the corrections Theinterferometers allow monitoring (2000 mea-surements per second) with an accuracy about03 mm for linear 03 mrad for angular para-meters over the full range of the scanner
Two calibration rulers each split into two parts(Fig 8) and placed under and above thechamber to be measured Every calibrationruler consists of two layers of nine ceramicplates (Fig 9) with six gold strips on eachcomposing a regular grid of 54 equivalent wireswith 30mm pitch The distance between thoselayers is 45mm The ceramic plates are installedon a carbon fibre support The calibrationrulers are used for the online (ie it is made for
every scan) reconstruction of the geometricalparameters of the X-ray tomograph and tocheck the stability and reliability of the X-raytomograph
A set of computer-controlled motors allowingthe alignment of the chamber on the X-raytomograph and choosing of the cross-sectionplane of the chamber One small moving cart isenslaved to the X-ray cart supporting twoscintillators optimized for the X-ray beamenergy spectrum following the beams inorder to record two shadowgrams simulta-neously
The control of the X-ray tomograph cart andchamber movements the data-acquisition andparameter monitoring are supported by adistributed online system based on VME PCSun computers connected via PCI MXI sharedmemory interface using softwarewritten on CC++ and LabVIEW
Two X-ray high-voltage power supplies (up to60 kV 50mA) The two high-voltage cablesfeeding the X-ray tubes move in synchronismwith the scanner on a separate cart in orderto ensure a perfect stability of the X-ray rollingcart
and collimating systems
Support of the upper calibration ruler
Chamber
Porticowithrails
Cart with shielded X-ray tubes
Z
X
Z
X
Fig 7 The X-ray tomograph during the mechanical control of a layers muon
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10192
4 Online calibration
Measurements with an accuracy of a fewmicrons require a precise control of all thegeometrical parameters of the X-ray tomographduring the measurements This requires the con-tinuous monitoring of the absolute position of theinterferometer laser beams in an external system ofcoordinates and of the origin of all laser axes Thislast point is not an easy requirement becauseduring the installation of the chambers on the
X-ray tomograph or during maintenance workthe trajectory of the laser beams may be frequentlyinterrupted for a short time by some object whichautomatically leads to the loss of the zeroreference In order to avoid this problem thecalibration and zero reset must be done for everyscan
During the scanning the wires of the referencerulers must be seen by the X-ray beams of theX-ray tomograph without overlap from thechamber wires The reference wires are previouslymeasured independently on a very-high-precisionoptical bench (E1 mm) During the scanning theshadow patterns of both the wires of the referencerulers and of the chamber are registered simulta-neously The basic geometrical parameters of theX-ray tomograph are then calculated for each scanusing the recorded shadowgram together with theoptical results of the position of the reference rulerwires [13] This approach allows a reliablepermanent control of the accuracy of the measure-ments Such a calibration scheme provides thebasis for the calculation of the position of allchamber wires
Let us stress that all the measurements are donein the same scan all calculations for the chamberand reference ruler wires are made with the same
Fig 8 One half of the calibration ruler (consisting of nine elements fixed on the carbon fibre tube) during assembling and
measurements
Fig 9 One element of the calibration ruler each ceramic plate
has six narrow gold strips faking wires
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 93
geometrical parameters The two reference rulersare measured with a known accuracy As aconsequence the chamber which is lying betweenthe two reference rulers is measured with the sameaccuracy Moreover in the offline analysis [13] thesame functions are used for the geometricalcalculation of the reference ruler and the chamberwires The X-ray tomograph accuracy is estimatedusing the residuals of the difference between thepredictions of the position of each wire of thereference rulers as given by the X-ray beams andtheir actual known position This method allowsone to control the accuracy of the X-ray tomo-graph scan-by-scan and consequently to detect anyloss of performance it allows for example toreveal small displacements or failures of theinterferometer system
Let us stress that the calibration is performedfor every scan and therefore for every scan thegeometrical parameters are calculated and theX-ray tomograph accuracy is checked
While studying the X-ray tomograph perfor-mance soon after its construction it became clearthat the precision of the optical measurements ofthe calibration rulers was not high enough toachieve the wanted accuracy It was later con-firmed that the optical measurements performedby the CERN Metrology Group using conven-tional industrial equipment were not preciseenough Therefore we had to develop a dedicatedmethod in order to improve the accuracy
Geometrical characteristics of the referencerulers may be divided into two groups the firstgroup consists of the parameters which can becorrected using direct X-ray tomograph measure-ments and the second group consists of para-meters for which other external reference devicesare needed In the first group we can note thatsome systematic errors in the position of individualcalibration ruler wires are well detected by theX-ray tomograph and those components of theoptical measurement errors are corrected by usingthe data coming from several scans As an examplefor the second group of parameters the straight-ness of the monitoring X-ray beams does not allowthe detection of possible errors in the opticalmeasurements of the vertical and horizontal shiftsbetween the layers of the calibration rulers If such
an error exists it leads to an error in the verticalscale and in the orthogonality of the system ofcoordinates of the X-ray tomograph
We call these two groups of optical measure-ment errors lsquonon-linearrsquo and lsquolinearrsquo errors3
respectively An autocalibration approach wasdeveloped to improve the accuracy of the opticalmeasurements It consists of two steps correctionof non-linear components in optical measurementsand determination (plus correction) of the linearparameters
The first step does not require special commentsAfter the geometrical reconstruction of the cali-bration ruler wires determined from several scans(in order to improve statistics) the residualsbetween the predictions by every X-ray beamand the optical measurements are calculatedUsing corresponding values from both non-paral-lel beams the two-dimensional correction ofposition of every calibration ruler wire may befound Because the pattern recognition maychange after the corrections an additional itera-tion may be done
5 Autocalibration correction of the linear errors in
the optical measurements of the calibration ruler
wire positions
51 Theoretical aspects
Let us consider the question of the linearsystematics effect of the absolute scales the scalesfor the X - and Z-axis and their relative angle (non-orthogonality of the system of coordinates) The X
scale is determined by the linear interferometer(measuring the position of the cart along the scan)the corresponding laser beam is parallel to theOX -axis (with second-order errors due to the non-parallelism of the laser beam and scan direction)The OZ scale and the non-orthogonality errors ofthe coordinate system however are completelydefined by the optical measurement errors They
3We have chosen the names in that way because the lsquolinearrsquo
(non-linear) errors in the optical measurements are propagated
to the chamber wire position errors which are linear (non-
linear) functions of the wire coordinates
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10194
cannot be determined from the X-ray tomographmeasurements themselves During the geometryfit the calculated effects of any change in the OZ
scale and in the non-orthogonality of the axes maybe completely compensated with correspondingchanges of the geometrical parameters of the X-ray beams Therefore some external measurementsare necessary
Let us consider three hypothetical parallel wiresA B C of the same rigid chamber (see Fig 10 thewires are perpendicular to the plane of picture andonly their projections A B C are seen) The pointsA B C should not lie on the same straight lineLet us imagine that they are measured by the X-ray tomograph in the original position (Fig 10a)and after rotation of the chamber by 901 aroundthe wiresrsquo direction (Fig 10b) If the X-raytomograph has a Z scale error instead of thepositions of the points A B C the positions ofpoints A0 B0 C0 and points A00 B00 C00 in therotated orientation will be found Putting bothmeasurements in the same system of coordinates
(we need three points because the angle aboutwhich these three points are rotated is not knownprecisely and in general does not equal 901) wecan estimate the Z scale error (Fig 10c)
A similar analysis applies for the non-orthogon-ality error Three wires A B C are measured aspoints A0 B0 C0 (Fig 11a) and as points A00 B00C00 in the 901 orientation (Fig 11b) Putting bothmeasurements in the same system of coordinates(Fig 11c) we can evaluate the non-orthogonalityerror introduced by the X-ray tomograph
Having determined the values of the Z scale andthe non-orthogonality errors we can calculate andthen correct the horizontal and vertical shiftsbetween the layers of the reference ruler becausethese shifts are the source of the errors in the Z
scale and non-orthogonality
52 Practical realization
In order to realize the described approach asmall calibration chamber (gauge chamber) was
A B
C
A B
C
AA
C
B
C
B
B
BA
A
C C
O X
Z
(a) (b) (c)
Fig 11 Non-orthogonality error calibration
A B
C
A
B
B
C
C
A
A
C
BCA
B
AB
C
O X
Z
(a) (b) (c)
Fig 10 Z scale error calibration
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 95
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
Using the timendashdrift distance relationship the single-wire resolution is typically 80mm when operated athigh gas pressure (3bar)
The muon chamber layout was optimized toreduce the number of different chamber sizes
The barrel chambers are of rectangular shapewith sizes from 2 to 10m2
The end-cap chambers are of trapezoidal shapeTheir sizes range from 1 to 10m2 for individualchamber modules and up to 30m2 when severalof them are preassembled for installation Thechamber design must guarantee reliability andstability of construction and operation for theanticipated lifetime of the experiment in anirradiated environment The basic design isshown in Fig 5
To improve the resolution of a chamber beyondthe single-wire limit and to achieve adequateredundancy for pattern recognition the MDTchambers are constructed from 2 4 monolayersof drift tubes for the inner and 2 3 monolayersfor the middle and outer stations The tubes are
arranged in multilayers of three or four mono-layers respectively on either side of a rigidsupport structure The support structures thelsquospacer framesrsquo provide for accurate positioningof the two multilayers with respect to each otherand for mechanical integrity under effects oftemperature and gravity As an example ofdifficulties encountered when seeking to reach therequired resolution the drift tubes which are notmounted vertically are bent slightly in order tomatch their shape to the gravitational sag of thewires
The structural components of the spacerframes are three lsquocrossplatesrsquo to which thedrift tube multilayers are attached and twolsquolong beamsrsquo connecting the crossplates Theframes need to be constructed to a moderatemechanical accuracy of 705mm only accuratepositioning of the drift tubes is provided bythe assembly procedure The chambers willbe attached to the rail structures of the spectro-meter by three-point kinematic supports Oncea chamber is installed in its final location in
Cross plate
Multilayer
In-plane alignment
Longitudinal beam
3-layerMultilayer
micro track
Tube cross section
BE
+
Fig 5 Schematic drawing of a rectangular MDT chamber constructed from multilayers of three monolayers each for installation in
the barrel spectrometer The chambers for the end-cap are of trapezoidal shape but are of similar design otherwise In the toroidal
magnetic field the tubes are oriented essentially parallel to the field lines
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10188
the spectrometer mechanical deformations aremonitored by an in-plane optical system hencethe name lsquoMDTrsquo chambers The spacer framessupport most of the components of the alignmentsystem
MDT chambers are made for robust andreliable operation This is achieved thanks to themechanical concept which provides a solid separa-tion of detecting cells and a good electricalshielding of neighbouring wires
Chambers must withstand very long storagetime followed by very long running periods Inorder to reach the required momentum resolutionall MDT chambers must be constructed with20 mm mechanical accuracy
17 Quality assurance for MDTs
The MDT project in ATLAS is on the scale of avery large industrial project aiming at an excep-tional quality of construction in terms of accuracy(r20 mm) material reliability delivery and assem-bly schedule The X-ray tomograph quality-con-trol platform may be considered as one of the keyinstruments needed for the MDT
The production of the MDT chambers is basedon the concept of very high mechanical precisionand the use of X-ray tomography as a quality-control platform for the mechanically achievedprecision
2 X-ray tomography
The development design construction andtuning of an X-ray tomograph suited forthe MDT quality control was the aim of theX-ray Quality-Control lsquoX-QCrsquo project (see Refs[5ndash11])
21 Basic principle
The choice of X-ray tomography as a quality-control tool for the MDT project was based ontwo essential features
The ability of X-rays to detect 50 mm tungstenwires hidden inside the aluminium drift tubes of
the muon detector Sharp absorption spikes areobserved either as a shadow seen by an X-raydetector placed under the chamber (passivemode) or as a sharp peak signal detected by thechamber itself (active mode) see Fig 6
The possibility to produce narrow collimatedX-ray beams allowing one to reach detectionprofiles with the required peak position detec-tion accuracy (1ndash2 mm)
The value of the absorption coefficient for X-rayphotons is very sensitive to attenuating materialsand to the X-ray energy If d is the density Z theatomic number of the material and W the energyof X-rays the mass absorption coefficient m=dvaries as EethZ=W THORN3 at 40 keV The optimumenergy range for this application is between 40and 60 keV
22 Basic procedure
A narrow X-ray beam is precisely moved acrossa section of the MDT chamber A scintillatorinstalled under the chamber records the absorp-tion profile during the scanning along the multi-layer cross-section of the chamber The data areanalysed and provide projective measurementStereo-measurements are possible when usingdifferent inclination angles of the X-ray beamFrom the data registered a two-dimensional mapof wire positions can be reconstructed Everychamber is scanned in a few positions along thetube length A full calibration of the muonchambers produces a list of all the wire coordi-nates in two dimensions Using the scanningresults for a few sections along the chambers athree-dimensional map can finally be made
After the analysis of raw data the results arestored in a database providing a full list of the wireposition and accuracy together with the para-meters of the chamber
We can measure the MDT chamberrsquos wireposition using two different methods
The lsquopassiversquo method This is based on therelatively good transparency of a MDT cham-ber to X-rays and on the strong absorptionof the 50 mm tungsten anode wires The out-put intensity of the X-ray beam is measured
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 89
by a scintillator Fig 6 shows a typicalcounting rate of the scintillator versus the X
coordinate1
The lsquoactiversquo method This second procedure isbased on the position-dependence efficiencyof drift tubes for X-rays When the X-raybeam hits the wire photoelectrons Augerelectrons and fluorescence are produced ata much higher rate than in the gas or in the
walls These radiations have a high prob-ability of escaping from the surface of the wireand of being detected by direct ionization in thegas
Fig 6 shows the wire signal from a tubebelonging to the bottom layer of the chamberThe three wires in the upper layers above theactive tube are clearly seen as absorption peaks bythe active tube they produce a shadow effectwhich is observed as inverted peaks comparedto the tube wire signal itself The relative intensityof the signals is not only proportional to theabsorption power of the material but also tothe probability of photoelectrons to escape fromthe material and to ionize the detecting gas ofthe drift tube
Passive mode
Layer 4 Active mode
X coordinate (mm)
Co
un
tin
g r
ate
(kH
z)
4 absorption lines seen
3 absorption lines seen
1 peak wire signal
Fig 6 Passive and active scanning methods
1 In this paper we exclusively use the natural X-ray
tomograph coordinate system (see Ref [12]) X (scanning
direction in the horizontal plane) Y (in the horizontal plane
perpendicular to the scanning direction) and Z (the vertical
coordinate) The MDT chambers in ATLAS use another system
of coordinates x (along the tubes) y (normal to chamber plane)
and z (across the tubes) Consequently the correspondence is as
follows X-zY-xZ-y
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10190
3 The X-ray tomograph
31 The challenge
The maximum width of the standardized MDTchambers has been fixed at 2160mm Thereforethe maximum scanning span of the X-ray tomo-graph has been chosen to be 2200mm The mainchallenge was to provide in short time to theconstructors reliable results on the wire position-ing of different chamber prototypes built with20 mm accuracy specification The X-ray tomo-graph had to consequently achieve a betteraccuracy aiming at 10 mm and be built withinstringent time and money constraints
32 Design considerations
The measurement of the position of the wires intwo dimensions along a cross-section of achamber was performed by using a cart movingon well-polished rails carrying two X-ray sourcesand their collimating systems producing beamswith opposite and fixed inclination anglesThe selection of the two inclination angles Y1Y2 the so-called lsquoscanning anglersquo relative to theperpendicular to the chamber plane was animportant issue under consideration for tworeasons
They define the stereo-angle (jY1j thorn jY2j) andconsequently the two-dimensional accuracy ofthe position of the wires
The absorption peak pattern of the wireswithin a corridor defined by the Y angle mustshow well-separated absorption signals Thispattern is very sensitive to the Y angle anddepends upon the chamber structure character-istics ie the height of the spacer the symmetrybetween the two multilayers and the horizontaland vertical wire pitch
The most convenient choice is to align thebeams in the natural 7301 corridors of themultilayers The angles must be chosen a little bithigher or lower in order to separate the wire peaksignals In the described set-up they are 73141
The geometric reconstruction requires theknowledge of four mechanical parameters the
linear scanning position the vertical straightnessthe angular displacement around the Y -axis (pitchangle) and the angular variation around the X -axis(roll angle) The excellent control of the cartposition (better than 1 mm) was crucial for achiev-ing high reconstruction accuracy We chose for themain structure of the X-ray tomograph standardhigh-precision mechanics combined with ultimateprecision survey tools in order to monitor the mainparameters controlling the X-ray beam positionduring scanning In this respect industrial laserinterferometers2 were used for monitoring Z (thevertical straightness) X (the horizontal displace-ment) DY (the pitch angle variation) and the rollangle
Reliable industrial choices were made wheneverpossible for the construction of the other parts andfor the software of the X-ray tomograph
33 Construction
Fig 7 shows the X-ray tomograph designed andbuilt at CERN Owing to the necessity of a verystable environment the X-ray tomograph wasinstalled in a clean room in building 188 groundfloor (one of the most stable ground floors atCERN) with temperature control and humiditycontrol
The X-ray tomograph consists of
A fixed 3500mm iron portico equipped with aprecise motorized cart rolling on rails along theX -axis
A rolling cart equipped with two collimated X-ray beams at 7314 with respect to the Z-axisThe 6281 stereo-angle has been chosen on thebasis of our previous experience and fits thetube corridors perfectly provided the spacersbetween the multilayers are correctly chosenThe X-ray beams have a cross-section of about30 mm 8mm and a divergence of about
2The basic principle of industrial interferometry used here is
to count the number of fringes of the interference between a
fixed reference laser beam and a second laser beam which is
reflected by a mirror installed on the moving chart Industry
suggests a few optical schemes of the main and reference beams
to allow measurements of longitudinal transversal (with respect
to the laser beam direction) and angular variations
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 91
60 mrad the longer axis being aligned along thewires to be measured
Four interferometers which monitor the mainparameters of the scanner movement the linearX value the pitch angle (rotation around the Y -axis) and two vertical straightnesses (theirbeams are parallel with distance about 160mmbetween them) The last two measurementsallow the calculation of the yaw (rotationaround the Z-axis) The horizontal straightnessis not monitored as it would have only asecond-order effect in the corrections Theinterferometers allow monitoring (2000 mea-surements per second) with an accuracy about03 mm for linear 03 mrad for angular para-meters over the full range of the scanner
Two calibration rulers each split into two parts(Fig 8) and placed under and above thechamber to be measured Every calibrationruler consists of two layers of nine ceramicplates (Fig 9) with six gold strips on eachcomposing a regular grid of 54 equivalent wireswith 30mm pitch The distance between thoselayers is 45mm The ceramic plates are installedon a carbon fibre support The calibrationrulers are used for the online (ie it is made for
every scan) reconstruction of the geometricalparameters of the X-ray tomograph and tocheck the stability and reliability of the X-raytomograph
A set of computer-controlled motors allowingthe alignment of the chamber on the X-raytomograph and choosing of the cross-sectionplane of the chamber One small moving cart isenslaved to the X-ray cart supporting twoscintillators optimized for the X-ray beamenergy spectrum following the beams inorder to record two shadowgrams simulta-neously
The control of the X-ray tomograph cart andchamber movements the data-acquisition andparameter monitoring are supported by adistributed online system based on VME PCSun computers connected via PCI MXI sharedmemory interface using softwarewritten on CC++ and LabVIEW
Two X-ray high-voltage power supplies (up to60 kV 50mA) The two high-voltage cablesfeeding the X-ray tubes move in synchronismwith the scanner on a separate cart in orderto ensure a perfect stability of the X-ray rollingcart
and collimating systems
Support of the upper calibration ruler
Chamber
Porticowithrails
Cart with shielded X-ray tubes
Z
X
Z
X
Fig 7 The X-ray tomograph during the mechanical control of a layers muon
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10192
4 Online calibration
Measurements with an accuracy of a fewmicrons require a precise control of all thegeometrical parameters of the X-ray tomographduring the measurements This requires the con-tinuous monitoring of the absolute position of theinterferometer laser beams in an external system ofcoordinates and of the origin of all laser axes Thislast point is not an easy requirement becauseduring the installation of the chambers on the
X-ray tomograph or during maintenance workthe trajectory of the laser beams may be frequentlyinterrupted for a short time by some object whichautomatically leads to the loss of the zeroreference In order to avoid this problem thecalibration and zero reset must be done for everyscan
During the scanning the wires of the referencerulers must be seen by the X-ray beams of theX-ray tomograph without overlap from thechamber wires The reference wires are previouslymeasured independently on a very-high-precisionoptical bench (E1 mm) During the scanning theshadow patterns of both the wires of the referencerulers and of the chamber are registered simulta-neously The basic geometrical parameters of theX-ray tomograph are then calculated for each scanusing the recorded shadowgram together with theoptical results of the position of the reference rulerwires [13] This approach allows a reliablepermanent control of the accuracy of the measure-ments Such a calibration scheme provides thebasis for the calculation of the position of allchamber wires
Let us stress that all the measurements are donein the same scan all calculations for the chamberand reference ruler wires are made with the same
Fig 8 One half of the calibration ruler (consisting of nine elements fixed on the carbon fibre tube) during assembling and
measurements
Fig 9 One element of the calibration ruler each ceramic plate
has six narrow gold strips faking wires
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 93
geometrical parameters The two reference rulersare measured with a known accuracy As aconsequence the chamber which is lying betweenthe two reference rulers is measured with the sameaccuracy Moreover in the offline analysis [13] thesame functions are used for the geometricalcalculation of the reference ruler and the chamberwires The X-ray tomograph accuracy is estimatedusing the residuals of the difference between thepredictions of the position of each wire of thereference rulers as given by the X-ray beams andtheir actual known position This method allowsone to control the accuracy of the X-ray tomo-graph scan-by-scan and consequently to detect anyloss of performance it allows for example toreveal small displacements or failures of theinterferometer system
Let us stress that the calibration is performedfor every scan and therefore for every scan thegeometrical parameters are calculated and theX-ray tomograph accuracy is checked
While studying the X-ray tomograph perfor-mance soon after its construction it became clearthat the precision of the optical measurements ofthe calibration rulers was not high enough toachieve the wanted accuracy It was later con-firmed that the optical measurements performedby the CERN Metrology Group using conven-tional industrial equipment were not preciseenough Therefore we had to develop a dedicatedmethod in order to improve the accuracy
Geometrical characteristics of the referencerulers may be divided into two groups the firstgroup consists of the parameters which can becorrected using direct X-ray tomograph measure-ments and the second group consists of para-meters for which other external reference devicesare needed In the first group we can note thatsome systematic errors in the position of individualcalibration ruler wires are well detected by theX-ray tomograph and those components of theoptical measurement errors are corrected by usingthe data coming from several scans As an examplefor the second group of parameters the straight-ness of the monitoring X-ray beams does not allowthe detection of possible errors in the opticalmeasurements of the vertical and horizontal shiftsbetween the layers of the calibration rulers If such
an error exists it leads to an error in the verticalscale and in the orthogonality of the system ofcoordinates of the X-ray tomograph
We call these two groups of optical measure-ment errors lsquonon-linearrsquo and lsquolinearrsquo errors3
respectively An autocalibration approach wasdeveloped to improve the accuracy of the opticalmeasurements It consists of two steps correctionof non-linear components in optical measurementsand determination (plus correction) of the linearparameters
The first step does not require special commentsAfter the geometrical reconstruction of the cali-bration ruler wires determined from several scans(in order to improve statistics) the residualsbetween the predictions by every X-ray beamand the optical measurements are calculatedUsing corresponding values from both non-paral-lel beams the two-dimensional correction ofposition of every calibration ruler wire may befound Because the pattern recognition maychange after the corrections an additional itera-tion may be done
5 Autocalibration correction of the linear errors in
the optical measurements of the calibration ruler
wire positions
51 Theoretical aspects
Let us consider the question of the linearsystematics effect of the absolute scales the scalesfor the X - and Z-axis and their relative angle (non-orthogonality of the system of coordinates) The X
scale is determined by the linear interferometer(measuring the position of the cart along the scan)the corresponding laser beam is parallel to theOX -axis (with second-order errors due to the non-parallelism of the laser beam and scan direction)The OZ scale and the non-orthogonality errors ofthe coordinate system however are completelydefined by the optical measurement errors They
3We have chosen the names in that way because the lsquolinearrsquo
(non-linear) errors in the optical measurements are propagated
to the chamber wire position errors which are linear (non-
linear) functions of the wire coordinates
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10194
cannot be determined from the X-ray tomographmeasurements themselves During the geometryfit the calculated effects of any change in the OZ
scale and in the non-orthogonality of the axes maybe completely compensated with correspondingchanges of the geometrical parameters of the X-ray beams Therefore some external measurementsare necessary
Let us consider three hypothetical parallel wiresA B C of the same rigid chamber (see Fig 10 thewires are perpendicular to the plane of picture andonly their projections A B C are seen) The pointsA B C should not lie on the same straight lineLet us imagine that they are measured by the X-ray tomograph in the original position (Fig 10a)and after rotation of the chamber by 901 aroundthe wiresrsquo direction (Fig 10b) If the X-raytomograph has a Z scale error instead of thepositions of the points A B C the positions ofpoints A0 B0 C0 and points A00 B00 C00 in therotated orientation will be found Putting bothmeasurements in the same system of coordinates
(we need three points because the angle aboutwhich these three points are rotated is not knownprecisely and in general does not equal 901) wecan estimate the Z scale error (Fig 10c)
A similar analysis applies for the non-orthogon-ality error Three wires A B C are measured aspoints A0 B0 C0 (Fig 11a) and as points A00 B00C00 in the 901 orientation (Fig 11b) Putting bothmeasurements in the same system of coordinates(Fig 11c) we can evaluate the non-orthogonalityerror introduced by the X-ray tomograph
Having determined the values of the Z scale andthe non-orthogonality errors we can calculate andthen correct the horizontal and vertical shiftsbetween the layers of the reference ruler becausethese shifts are the source of the errors in the Z
scale and non-orthogonality
52 Practical realization
In order to realize the described approach asmall calibration chamber (gauge chamber) was
A B
C
A B
C
AA
C
B
C
B
B
BA
A
C C
O X
Z
(a) (b) (c)
Fig 11 Non-orthogonality error calibration
A B
C
A
B
B
C
C
A
A
C
BCA
B
AB
C
O X
Z
(a) (b) (c)
Fig 10 Z scale error calibration
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 95
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
the spectrometer mechanical deformations aremonitored by an in-plane optical system hencethe name lsquoMDTrsquo chambers The spacer framessupport most of the components of the alignmentsystem
MDT chambers are made for robust andreliable operation This is achieved thanks to themechanical concept which provides a solid separa-tion of detecting cells and a good electricalshielding of neighbouring wires
Chambers must withstand very long storagetime followed by very long running periods Inorder to reach the required momentum resolutionall MDT chambers must be constructed with20 mm mechanical accuracy
17 Quality assurance for MDTs
The MDT project in ATLAS is on the scale of avery large industrial project aiming at an excep-tional quality of construction in terms of accuracy(r20 mm) material reliability delivery and assem-bly schedule The X-ray tomograph quality-con-trol platform may be considered as one of the keyinstruments needed for the MDT
The production of the MDT chambers is basedon the concept of very high mechanical precisionand the use of X-ray tomography as a quality-control platform for the mechanically achievedprecision
2 X-ray tomography
The development design construction andtuning of an X-ray tomograph suited forthe MDT quality control was the aim of theX-ray Quality-Control lsquoX-QCrsquo project (see Refs[5ndash11])
21 Basic principle
The choice of X-ray tomography as a quality-control tool for the MDT project was based ontwo essential features
The ability of X-rays to detect 50 mm tungstenwires hidden inside the aluminium drift tubes of
the muon detector Sharp absorption spikes areobserved either as a shadow seen by an X-raydetector placed under the chamber (passivemode) or as a sharp peak signal detected by thechamber itself (active mode) see Fig 6
The possibility to produce narrow collimatedX-ray beams allowing one to reach detectionprofiles with the required peak position detec-tion accuracy (1ndash2 mm)
The value of the absorption coefficient for X-rayphotons is very sensitive to attenuating materialsand to the X-ray energy If d is the density Z theatomic number of the material and W the energyof X-rays the mass absorption coefficient m=dvaries as EethZ=W THORN3 at 40 keV The optimumenergy range for this application is between 40and 60 keV
22 Basic procedure
A narrow X-ray beam is precisely moved acrossa section of the MDT chamber A scintillatorinstalled under the chamber records the absorp-tion profile during the scanning along the multi-layer cross-section of the chamber The data areanalysed and provide projective measurementStereo-measurements are possible when usingdifferent inclination angles of the X-ray beamFrom the data registered a two-dimensional mapof wire positions can be reconstructed Everychamber is scanned in a few positions along thetube length A full calibration of the muonchambers produces a list of all the wire coordi-nates in two dimensions Using the scanningresults for a few sections along the chambers athree-dimensional map can finally be made
After the analysis of raw data the results arestored in a database providing a full list of the wireposition and accuracy together with the para-meters of the chamber
We can measure the MDT chamberrsquos wireposition using two different methods
The lsquopassiversquo method This is based on therelatively good transparency of a MDT cham-ber to X-rays and on the strong absorptionof the 50 mm tungsten anode wires The out-put intensity of the X-ray beam is measured
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 89
by a scintillator Fig 6 shows a typicalcounting rate of the scintillator versus the X
coordinate1
The lsquoactiversquo method This second procedure isbased on the position-dependence efficiencyof drift tubes for X-rays When the X-raybeam hits the wire photoelectrons Augerelectrons and fluorescence are produced ata much higher rate than in the gas or in the
walls These radiations have a high prob-ability of escaping from the surface of the wireand of being detected by direct ionization in thegas
Fig 6 shows the wire signal from a tubebelonging to the bottom layer of the chamberThe three wires in the upper layers above theactive tube are clearly seen as absorption peaks bythe active tube they produce a shadow effectwhich is observed as inverted peaks comparedto the tube wire signal itself The relative intensityof the signals is not only proportional to theabsorption power of the material but also tothe probability of photoelectrons to escape fromthe material and to ionize the detecting gas ofthe drift tube
Passive mode
Layer 4 Active mode
X coordinate (mm)
Co
un
tin
g r
ate
(kH
z)
4 absorption lines seen
3 absorption lines seen
1 peak wire signal
Fig 6 Passive and active scanning methods
1 In this paper we exclusively use the natural X-ray
tomograph coordinate system (see Ref [12]) X (scanning
direction in the horizontal plane) Y (in the horizontal plane
perpendicular to the scanning direction) and Z (the vertical
coordinate) The MDT chambers in ATLAS use another system
of coordinates x (along the tubes) y (normal to chamber plane)
and z (across the tubes) Consequently the correspondence is as
follows X-zY-xZ-y
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10190
3 The X-ray tomograph
31 The challenge
The maximum width of the standardized MDTchambers has been fixed at 2160mm Thereforethe maximum scanning span of the X-ray tomo-graph has been chosen to be 2200mm The mainchallenge was to provide in short time to theconstructors reliable results on the wire position-ing of different chamber prototypes built with20 mm accuracy specification The X-ray tomo-graph had to consequently achieve a betteraccuracy aiming at 10 mm and be built withinstringent time and money constraints
32 Design considerations
The measurement of the position of the wires intwo dimensions along a cross-section of achamber was performed by using a cart movingon well-polished rails carrying two X-ray sourcesand their collimating systems producing beamswith opposite and fixed inclination anglesThe selection of the two inclination angles Y1Y2 the so-called lsquoscanning anglersquo relative to theperpendicular to the chamber plane was animportant issue under consideration for tworeasons
They define the stereo-angle (jY1j thorn jY2j) andconsequently the two-dimensional accuracy ofthe position of the wires
The absorption peak pattern of the wireswithin a corridor defined by the Y angle mustshow well-separated absorption signals Thispattern is very sensitive to the Y angle anddepends upon the chamber structure character-istics ie the height of the spacer the symmetrybetween the two multilayers and the horizontaland vertical wire pitch
The most convenient choice is to align thebeams in the natural 7301 corridors of themultilayers The angles must be chosen a little bithigher or lower in order to separate the wire peaksignals In the described set-up they are 73141
The geometric reconstruction requires theknowledge of four mechanical parameters the
linear scanning position the vertical straightnessthe angular displacement around the Y -axis (pitchangle) and the angular variation around the X -axis(roll angle) The excellent control of the cartposition (better than 1 mm) was crucial for achiev-ing high reconstruction accuracy We chose for themain structure of the X-ray tomograph standardhigh-precision mechanics combined with ultimateprecision survey tools in order to monitor the mainparameters controlling the X-ray beam positionduring scanning In this respect industrial laserinterferometers2 were used for monitoring Z (thevertical straightness) X (the horizontal displace-ment) DY (the pitch angle variation) and the rollangle
Reliable industrial choices were made wheneverpossible for the construction of the other parts andfor the software of the X-ray tomograph
33 Construction
Fig 7 shows the X-ray tomograph designed andbuilt at CERN Owing to the necessity of a verystable environment the X-ray tomograph wasinstalled in a clean room in building 188 groundfloor (one of the most stable ground floors atCERN) with temperature control and humiditycontrol
The X-ray tomograph consists of
A fixed 3500mm iron portico equipped with aprecise motorized cart rolling on rails along theX -axis
A rolling cart equipped with two collimated X-ray beams at 7314 with respect to the Z-axisThe 6281 stereo-angle has been chosen on thebasis of our previous experience and fits thetube corridors perfectly provided the spacersbetween the multilayers are correctly chosenThe X-ray beams have a cross-section of about30 mm 8mm and a divergence of about
2The basic principle of industrial interferometry used here is
to count the number of fringes of the interference between a
fixed reference laser beam and a second laser beam which is
reflected by a mirror installed on the moving chart Industry
suggests a few optical schemes of the main and reference beams
to allow measurements of longitudinal transversal (with respect
to the laser beam direction) and angular variations
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 91
60 mrad the longer axis being aligned along thewires to be measured
Four interferometers which monitor the mainparameters of the scanner movement the linearX value the pitch angle (rotation around the Y -axis) and two vertical straightnesses (theirbeams are parallel with distance about 160mmbetween them) The last two measurementsallow the calculation of the yaw (rotationaround the Z-axis) The horizontal straightnessis not monitored as it would have only asecond-order effect in the corrections Theinterferometers allow monitoring (2000 mea-surements per second) with an accuracy about03 mm for linear 03 mrad for angular para-meters over the full range of the scanner
Two calibration rulers each split into two parts(Fig 8) and placed under and above thechamber to be measured Every calibrationruler consists of two layers of nine ceramicplates (Fig 9) with six gold strips on eachcomposing a regular grid of 54 equivalent wireswith 30mm pitch The distance between thoselayers is 45mm The ceramic plates are installedon a carbon fibre support The calibrationrulers are used for the online (ie it is made for
every scan) reconstruction of the geometricalparameters of the X-ray tomograph and tocheck the stability and reliability of the X-raytomograph
A set of computer-controlled motors allowingthe alignment of the chamber on the X-raytomograph and choosing of the cross-sectionplane of the chamber One small moving cart isenslaved to the X-ray cart supporting twoscintillators optimized for the X-ray beamenergy spectrum following the beams inorder to record two shadowgrams simulta-neously
The control of the X-ray tomograph cart andchamber movements the data-acquisition andparameter monitoring are supported by adistributed online system based on VME PCSun computers connected via PCI MXI sharedmemory interface using softwarewritten on CC++ and LabVIEW
Two X-ray high-voltage power supplies (up to60 kV 50mA) The two high-voltage cablesfeeding the X-ray tubes move in synchronismwith the scanner on a separate cart in orderto ensure a perfect stability of the X-ray rollingcart
and collimating systems
Support of the upper calibration ruler
Chamber
Porticowithrails
Cart with shielded X-ray tubes
Z
X
Z
X
Fig 7 The X-ray tomograph during the mechanical control of a layers muon
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10192
4 Online calibration
Measurements with an accuracy of a fewmicrons require a precise control of all thegeometrical parameters of the X-ray tomographduring the measurements This requires the con-tinuous monitoring of the absolute position of theinterferometer laser beams in an external system ofcoordinates and of the origin of all laser axes Thislast point is not an easy requirement becauseduring the installation of the chambers on the
X-ray tomograph or during maintenance workthe trajectory of the laser beams may be frequentlyinterrupted for a short time by some object whichautomatically leads to the loss of the zeroreference In order to avoid this problem thecalibration and zero reset must be done for everyscan
During the scanning the wires of the referencerulers must be seen by the X-ray beams of theX-ray tomograph without overlap from thechamber wires The reference wires are previouslymeasured independently on a very-high-precisionoptical bench (E1 mm) During the scanning theshadow patterns of both the wires of the referencerulers and of the chamber are registered simulta-neously The basic geometrical parameters of theX-ray tomograph are then calculated for each scanusing the recorded shadowgram together with theoptical results of the position of the reference rulerwires [13] This approach allows a reliablepermanent control of the accuracy of the measure-ments Such a calibration scheme provides thebasis for the calculation of the position of allchamber wires
Let us stress that all the measurements are donein the same scan all calculations for the chamberand reference ruler wires are made with the same
Fig 8 One half of the calibration ruler (consisting of nine elements fixed on the carbon fibre tube) during assembling and
measurements
Fig 9 One element of the calibration ruler each ceramic plate
has six narrow gold strips faking wires
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 93
geometrical parameters The two reference rulersare measured with a known accuracy As aconsequence the chamber which is lying betweenthe two reference rulers is measured with the sameaccuracy Moreover in the offline analysis [13] thesame functions are used for the geometricalcalculation of the reference ruler and the chamberwires The X-ray tomograph accuracy is estimatedusing the residuals of the difference between thepredictions of the position of each wire of thereference rulers as given by the X-ray beams andtheir actual known position This method allowsone to control the accuracy of the X-ray tomo-graph scan-by-scan and consequently to detect anyloss of performance it allows for example toreveal small displacements or failures of theinterferometer system
Let us stress that the calibration is performedfor every scan and therefore for every scan thegeometrical parameters are calculated and theX-ray tomograph accuracy is checked
While studying the X-ray tomograph perfor-mance soon after its construction it became clearthat the precision of the optical measurements ofthe calibration rulers was not high enough toachieve the wanted accuracy It was later con-firmed that the optical measurements performedby the CERN Metrology Group using conven-tional industrial equipment were not preciseenough Therefore we had to develop a dedicatedmethod in order to improve the accuracy
Geometrical characteristics of the referencerulers may be divided into two groups the firstgroup consists of the parameters which can becorrected using direct X-ray tomograph measure-ments and the second group consists of para-meters for which other external reference devicesare needed In the first group we can note thatsome systematic errors in the position of individualcalibration ruler wires are well detected by theX-ray tomograph and those components of theoptical measurement errors are corrected by usingthe data coming from several scans As an examplefor the second group of parameters the straight-ness of the monitoring X-ray beams does not allowthe detection of possible errors in the opticalmeasurements of the vertical and horizontal shiftsbetween the layers of the calibration rulers If such
an error exists it leads to an error in the verticalscale and in the orthogonality of the system ofcoordinates of the X-ray tomograph
We call these two groups of optical measure-ment errors lsquonon-linearrsquo and lsquolinearrsquo errors3
respectively An autocalibration approach wasdeveloped to improve the accuracy of the opticalmeasurements It consists of two steps correctionof non-linear components in optical measurementsand determination (plus correction) of the linearparameters
The first step does not require special commentsAfter the geometrical reconstruction of the cali-bration ruler wires determined from several scans(in order to improve statistics) the residualsbetween the predictions by every X-ray beamand the optical measurements are calculatedUsing corresponding values from both non-paral-lel beams the two-dimensional correction ofposition of every calibration ruler wire may befound Because the pattern recognition maychange after the corrections an additional itera-tion may be done
5 Autocalibration correction of the linear errors in
the optical measurements of the calibration ruler
wire positions
51 Theoretical aspects
Let us consider the question of the linearsystematics effect of the absolute scales the scalesfor the X - and Z-axis and their relative angle (non-orthogonality of the system of coordinates) The X
scale is determined by the linear interferometer(measuring the position of the cart along the scan)the corresponding laser beam is parallel to theOX -axis (with second-order errors due to the non-parallelism of the laser beam and scan direction)The OZ scale and the non-orthogonality errors ofthe coordinate system however are completelydefined by the optical measurement errors They
3We have chosen the names in that way because the lsquolinearrsquo
(non-linear) errors in the optical measurements are propagated
to the chamber wire position errors which are linear (non-
linear) functions of the wire coordinates
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10194
cannot be determined from the X-ray tomographmeasurements themselves During the geometryfit the calculated effects of any change in the OZ
scale and in the non-orthogonality of the axes maybe completely compensated with correspondingchanges of the geometrical parameters of the X-ray beams Therefore some external measurementsare necessary
Let us consider three hypothetical parallel wiresA B C of the same rigid chamber (see Fig 10 thewires are perpendicular to the plane of picture andonly their projections A B C are seen) The pointsA B C should not lie on the same straight lineLet us imagine that they are measured by the X-ray tomograph in the original position (Fig 10a)and after rotation of the chamber by 901 aroundthe wiresrsquo direction (Fig 10b) If the X-raytomograph has a Z scale error instead of thepositions of the points A B C the positions ofpoints A0 B0 C0 and points A00 B00 C00 in therotated orientation will be found Putting bothmeasurements in the same system of coordinates
(we need three points because the angle aboutwhich these three points are rotated is not knownprecisely and in general does not equal 901) wecan estimate the Z scale error (Fig 10c)
A similar analysis applies for the non-orthogon-ality error Three wires A B C are measured aspoints A0 B0 C0 (Fig 11a) and as points A00 B00C00 in the 901 orientation (Fig 11b) Putting bothmeasurements in the same system of coordinates(Fig 11c) we can evaluate the non-orthogonalityerror introduced by the X-ray tomograph
Having determined the values of the Z scale andthe non-orthogonality errors we can calculate andthen correct the horizontal and vertical shiftsbetween the layers of the reference ruler becausethese shifts are the source of the errors in the Z
scale and non-orthogonality
52 Practical realization
In order to realize the described approach asmall calibration chamber (gauge chamber) was
A B
C
A B
C
AA
C
B
C
B
B
BA
A
C C
O X
Z
(a) (b) (c)
Fig 11 Non-orthogonality error calibration
A B
C
A
B
B
C
C
A
A
C
BCA
B
AB
C
O X
Z
(a) (b) (c)
Fig 10 Z scale error calibration
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 95
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
by a scintillator Fig 6 shows a typicalcounting rate of the scintillator versus the X
coordinate1
The lsquoactiversquo method This second procedure isbased on the position-dependence efficiencyof drift tubes for X-rays When the X-raybeam hits the wire photoelectrons Augerelectrons and fluorescence are produced ata much higher rate than in the gas or in the
walls These radiations have a high prob-ability of escaping from the surface of the wireand of being detected by direct ionization in thegas
Fig 6 shows the wire signal from a tubebelonging to the bottom layer of the chamberThe three wires in the upper layers above theactive tube are clearly seen as absorption peaks bythe active tube they produce a shadow effectwhich is observed as inverted peaks comparedto the tube wire signal itself The relative intensityof the signals is not only proportional to theabsorption power of the material but also tothe probability of photoelectrons to escape fromthe material and to ionize the detecting gas ofthe drift tube
Passive mode
Layer 4 Active mode
X coordinate (mm)
Co
un
tin
g r
ate
(kH
z)
4 absorption lines seen
3 absorption lines seen
1 peak wire signal
Fig 6 Passive and active scanning methods
1 In this paper we exclusively use the natural X-ray
tomograph coordinate system (see Ref [12]) X (scanning
direction in the horizontal plane) Y (in the horizontal plane
perpendicular to the scanning direction) and Z (the vertical
coordinate) The MDT chambers in ATLAS use another system
of coordinates x (along the tubes) y (normal to chamber plane)
and z (across the tubes) Consequently the correspondence is as
follows X-zY-xZ-y
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10190
3 The X-ray tomograph
31 The challenge
The maximum width of the standardized MDTchambers has been fixed at 2160mm Thereforethe maximum scanning span of the X-ray tomo-graph has been chosen to be 2200mm The mainchallenge was to provide in short time to theconstructors reliable results on the wire position-ing of different chamber prototypes built with20 mm accuracy specification The X-ray tomo-graph had to consequently achieve a betteraccuracy aiming at 10 mm and be built withinstringent time and money constraints
32 Design considerations
The measurement of the position of the wires intwo dimensions along a cross-section of achamber was performed by using a cart movingon well-polished rails carrying two X-ray sourcesand their collimating systems producing beamswith opposite and fixed inclination anglesThe selection of the two inclination angles Y1Y2 the so-called lsquoscanning anglersquo relative to theperpendicular to the chamber plane was animportant issue under consideration for tworeasons
They define the stereo-angle (jY1j thorn jY2j) andconsequently the two-dimensional accuracy ofthe position of the wires
The absorption peak pattern of the wireswithin a corridor defined by the Y angle mustshow well-separated absorption signals Thispattern is very sensitive to the Y angle anddepends upon the chamber structure character-istics ie the height of the spacer the symmetrybetween the two multilayers and the horizontaland vertical wire pitch
The most convenient choice is to align thebeams in the natural 7301 corridors of themultilayers The angles must be chosen a little bithigher or lower in order to separate the wire peaksignals In the described set-up they are 73141
The geometric reconstruction requires theknowledge of four mechanical parameters the
linear scanning position the vertical straightnessthe angular displacement around the Y -axis (pitchangle) and the angular variation around the X -axis(roll angle) The excellent control of the cartposition (better than 1 mm) was crucial for achiev-ing high reconstruction accuracy We chose for themain structure of the X-ray tomograph standardhigh-precision mechanics combined with ultimateprecision survey tools in order to monitor the mainparameters controlling the X-ray beam positionduring scanning In this respect industrial laserinterferometers2 were used for monitoring Z (thevertical straightness) X (the horizontal displace-ment) DY (the pitch angle variation) and the rollangle
Reliable industrial choices were made wheneverpossible for the construction of the other parts andfor the software of the X-ray tomograph
33 Construction
Fig 7 shows the X-ray tomograph designed andbuilt at CERN Owing to the necessity of a verystable environment the X-ray tomograph wasinstalled in a clean room in building 188 groundfloor (one of the most stable ground floors atCERN) with temperature control and humiditycontrol
The X-ray tomograph consists of
A fixed 3500mm iron portico equipped with aprecise motorized cart rolling on rails along theX -axis
A rolling cart equipped with two collimated X-ray beams at 7314 with respect to the Z-axisThe 6281 stereo-angle has been chosen on thebasis of our previous experience and fits thetube corridors perfectly provided the spacersbetween the multilayers are correctly chosenThe X-ray beams have a cross-section of about30 mm 8mm and a divergence of about
2The basic principle of industrial interferometry used here is
to count the number of fringes of the interference between a
fixed reference laser beam and a second laser beam which is
reflected by a mirror installed on the moving chart Industry
suggests a few optical schemes of the main and reference beams
to allow measurements of longitudinal transversal (with respect
to the laser beam direction) and angular variations
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 91
60 mrad the longer axis being aligned along thewires to be measured
Four interferometers which monitor the mainparameters of the scanner movement the linearX value the pitch angle (rotation around the Y -axis) and two vertical straightnesses (theirbeams are parallel with distance about 160mmbetween them) The last two measurementsallow the calculation of the yaw (rotationaround the Z-axis) The horizontal straightnessis not monitored as it would have only asecond-order effect in the corrections Theinterferometers allow monitoring (2000 mea-surements per second) with an accuracy about03 mm for linear 03 mrad for angular para-meters over the full range of the scanner
Two calibration rulers each split into two parts(Fig 8) and placed under and above thechamber to be measured Every calibrationruler consists of two layers of nine ceramicplates (Fig 9) with six gold strips on eachcomposing a regular grid of 54 equivalent wireswith 30mm pitch The distance between thoselayers is 45mm The ceramic plates are installedon a carbon fibre support The calibrationrulers are used for the online (ie it is made for
every scan) reconstruction of the geometricalparameters of the X-ray tomograph and tocheck the stability and reliability of the X-raytomograph
A set of computer-controlled motors allowingthe alignment of the chamber on the X-raytomograph and choosing of the cross-sectionplane of the chamber One small moving cart isenslaved to the X-ray cart supporting twoscintillators optimized for the X-ray beamenergy spectrum following the beams inorder to record two shadowgrams simulta-neously
The control of the X-ray tomograph cart andchamber movements the data-acquisition andparameter monitoring are supported by adistributed online system based on VME PCSun computers connected via PCI MXI sharedmemory interface using softwarewritten on CC++ and LabVIEW
Two X-ray high-voltage power supplies (up to60 kV 50mA) The two high-voltage cablesfeeding the X-ray tubes move in synchronismwith the scanner on a separate cart in orderto ensure a perfect stability of the X-ray rollingcart
and collimating systems
Support of the upper calibration ruler
Chamber
Porticowithrails
Cart with shielded X-ray tubes
Z
X
Z
X
Fig 7 The X-ray tomograph during the mechanical control of a layers muon
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10192
4 Online calibration
Measurements with an accuracy of a fewmicrons require a precise control of all thegeometrical parameters of the X-ray tomographduring the measurements This requires the con-tinuous monitoring of the absolute position of theinterferometer laser beams in an external system ofcoordinates and of the origin of all laser axes Thislast point is not an easy requirement becauseduring the installation of the chambers on the
X-ray tomograph or during maintenance workthe trajectory of the laser beams may be frequentlyinterrupted for a short time by some object whichautomatically leads to the loss of the zeroreference In order to avoid this problem thecalibration and zero reset must be done for everyscan
During the scanning the wires of the referencerulers must be seen by the X-ray beams of theX-ray tomograph without overlap from thechamber wires The reference wires are previouslymeasured independently on a very-high-precisionoptical bench (E1 mm) During the scanning theshadow patterns of both the wires of the referencerulers and of the chamber are registered simulta-neously The basic geometrical parameters of theX-ray tomograph are then calculated for each scanusing the recorded shadowgram together with theoptical results of the position of the reference rulerwires [13] This approach allows a reliablepermanent control of the accuracy of the measure-ments Such a calibration scheme provides thebasis for the calculation of the position of allchamber wires
Let us stress that all the measurements are donein the same scan all calculations for the chamberand reference ruler wires are made with the same
Fig 8 One half of the calibration ruler (consisting of nine elements fixed on the carbon fibre tube) during assembling and
measurements
Fig 9 One element of the calibration ruler each ceramic plate
has six narrow gold strips faking wires
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 93
geometrical parameters The two reference rulersare measured with a known accuracy As aconsequence the chamber which is lying betweenthe two reference rulers is measured with the sameaccuracy Moreover in the offline analysis [13] thesame functions are used for the geometricalcalculation of the reference ruler and the chamberwires The X-ray tomograph accuracy is estimatedusing the residuals of the difference between thepredictions of the position of each wire of thereference rulers as given by the X-ray beams andtheir actual known position This method allowsone to control the accuracy of the X-ray tomo-graph scan-by-scan and consequently to detect anyloss of performance it allows for example toreveal small displacements or failures of theinterferometer system
Let us stress that the calibration is performedfor every scan and therefore for every scan thegeometrical parameters are calculated and theX-ray tomograph accuracy is checked
While studying the X-ray tomograph perfor-mance soon after its construction it became clearthat the precision of the optical measurements ofthe calibration rulers was not high enough toachieve the wanted accuracy It was later con-firmed that the optical measurements performedby the CERN Metrology Group using conven-tional industrial equipment were not preciseenough Therefore we had to develop a dedicatedmethod in order to improve the accuracy
Geometrical characteristics of the referencerulers may be divided into two groups the firstgroup consists of the parameters which can becorrected using direct X-ray tomograph measure-ments and the second group consists of para-meters for which other external reference devicesare needed In the first group we can note thatsome systematic errors in the position of individualcalibration ruler wires are well detected by theX-ray tomograph and those components of theoptical measurement errors are corrected by usingthe data coming from several scans As an examplefor the second group of parameters the straight-ness of the monitoring X-ray beams does not allowthe detection of possible errors in the opticalmeasurements of the vertical and horizontal shiftsbetween the layers of the calibration rulers If such
an error exists it leads to an error in the verticalscale and in the orthogonality of the system ofcoordinates of the X-ray tomograph
We call these two groups of optical measure-ment errors lsquonon-linearrsquo and lsquolinearrsquo errors3
respectively An autocalibration approach wasdeveloped to improve the accuracy of the opticalmeasurements It consists of two steps correctionof non-linear components in optical measurementsand determination (plus correction) of the linearparameters
The first step does not require special commentsAfter the geometrical reconstruction of the cali-bration ruler wires determined from several scans(in order to improve statistics) the residualsbetween the predictions by every X-ray beamand the optical measurements are calculatedUsing corresponding values from both non-paral-lel beams the two-dimensional correction ofposition of every calibration ruler wire may befound Because the pattern recognition maychange after the corrections an additional itera-tion may be done
5 Autocalibration correction of the linear errors in
the optical measurements of the calibration ruler
wire positions
51 Theoretical aspects
Let us consider the question of the linearsystematics effect of the absolute scales the scalesfor the X - and Z-axis and their relative angle (non-orthogonality of the system of coordinates) The X
scale is determined by the linear interferometer(measuring the position of the cart along the scan)the corresponding laser beam is parallel to theOX -axis (with second-order errors due to the non-parallelism of the laser beam and scan direction)The OZ scale and the non-orthogonality errors ofthe coordinate system however are completelydefined by the optical measurement errors They
3We have chosen the names in that way because the lsquolinearrsquo
(non-linear) errors in the optical measurements are propagated
to the chamber wire position errors which are linear (non-
linear) functions of the wire coordinates
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10194
cannot be determined from the X-ray tomographmeasurements themselves During the geometryfit the calculated effects of any change in the OZ
scale and in the non-orthogonality of the axes maybe completely compensated with correspondingchanges of the geometrical parameters of the X-ray beams Therefore some external measurementsare necessary
Let us consider three hypothetical parallel wiresA B C of the same rigid chamber (see Fig 10 thewires are perpendicular to the plane of picture andonly their projections A B C are seen) The pointsA B C should not lie on the same straight lineLet us imagine that they are measured by the X-ray tomograph in the original position (Fig 10a)and after rotation of the chamber by 901 aroundthe wiresrsquo direction (Fig 10b) If the X-raytomograph has a Z scale error instead of thepositions of the points A B C the positions ofpoints A0 B0 C0 and points A00 B00 C00 in therotated orientation will be found Putting bothmeasurements in the same system of coordinates
(we need three points because the angle aboutwhich these three points are rotated is not knownprecisely and in general does not equal 901) wecan estimate the Z scale error (Fig 10c)
A similar analysis applies for the non-orthogon-ality error Three wires A B C are measured aspoints A0 B0 C0 (Fig 11a) and as points A00 B00C00 in the 901 orientation (Fig 11b) Putting bothmeasurements in the same system of coordinates(Fig 11c) we can evaluate the non-orthogonalityerror introduced by the X-ray tomograph
Having determined the values of the Z scale andthe non-orthogonality errors we can calculate andthen correct the horizontal and vertical shiftsbetween the layers of the reference ruler becausethese shifts are the source of the errors in the Z
scale and non-orthogonality
52 Practical realization
In order to realize the described approach asmall calibration chamber (gauge chamber) was
A B
C
A B
C
AA
C
B
C
B
B
BA
A
C C
O X
Z
(a) (b) (c)
Fig 11 Non-orthogonality error calibration
A B
C
A
B
B
C
C
A
A
C
BCA
B
AB
C
O X
Z
(a) (b) (c)
Fig 10 Z scale error calibration
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 95
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
3 The X-ray tomograph
31 The challenge
The maximum width of the standardized MDTchambers has been fixed at 2160mm Thereforethe maximum scanning span of the X-ray tomo-graph has been chosen to be 2200mm The mainchallenge was to provide in short time to theconstructors reliable results on the wire position-ing of different chamber prototypes built with20 mm accuracy specification The X-ray tomo-graph had to consequently achieve a betteraccuracy aiming at 10 mm and be built withinstringent time and money constraints
32 Design considerations
The measurement of the position of the wires intwo dimensions along a cross-section of achamber was performed by using a cart movingon well-polished rails carrying two X-ray sourcesand their collimating systems producing beamswith opposite and fixed inclination anglesThe selection of the two inclination angles Y1Y2 the so-called lsquoscanning anglersquo relative to theperpendicular to the chamber plane was animportant issue under consideration for tworeasons
They define the stereo-angle (jY1j thorn jY2j) andconsequently the two-dimensional accuracy ofthe position of the wires
The absorption peak pattern of the wireswithin a corridor defined by the Y angle mustshow well-separated absorption signals Thispattern is very sensitive to the Y angle anddepends upon the chamber structure character-istics ie the height of the spacer the symmetrybetween the two multilayers and the horizontaland vertical wire pitch
The most convenient choice is to align thebeams in the natural 7301 corridors of themultilayers The angles must be chosen a little bithigher or lower in order to separate the wire peaksignals In the described set-up they are 73141
The geometric reconstruction requires theknowledge of four mechanical parameters the
linear scanning position the vertical straightnessthe angular displacement around the Y -axis (pitchangle) and the angular variation around the X -axis(roll angle) The excellent control of the cartposition (better than 1 mm) was crucial for achiev-ing high reconstruction accuracy We chose for themain structure of the X-ray tomograph standardhigh-precision mechanics combined with ultimateprecision survey tools in order to monitor the mainparameters controlling the X-ray beam positionduring scanning In this respect industrial laserinterferometers2 were used for monitoring Z (thevertical straightness) X (the horizontal displace-ment) DY (the pitch angle variation) and the rollangle
Reliable industrial choices were made wheneverpossible for the construction of the other parts andfor the software of the X-ray tomograph
33 Construction
Fig 7 shows the X-ray tomograph designed andbuilt at CERN Owing to the necessity of a verystable environment the X-ray tomograph wasinstalled in a clean room in building 188 groundfloor (one of the most stable ground floors atCERN) with temperature control and humiditycontrol
The X-ray tomograph consists of
A fixed 3500mm iron portico equipped with aprecise motorized cart rolling on rails along theX -axis
A rolling cart equipped with two collimated X-ray beams at 7314 with respect to the Z-axisThe 6281 stereo-angle has been chosen on thebasis of our previous experience and fits thetube corridors perfectly provided the spacersbetween the multilayers are correctly chosenThe X-ray beams have a cross-section of about30 mm 8mm and a divergence of about
2The basic principle of industrial interferometry used here is
to count the number of fringes of the interference between a
fixed reference laser beam and a second laser beam which is
reflected by a mirror installed on the moving chart Industry
suggests a few optical schemes of the main and reference beams
to allow measurements of longitudinal transversal (with respect
to the laser beam direction) and angular variations
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 91
60 mrad the longer axis being aligned along thewires to be measured
Four interferometers which monitor the mainparameters of the scanner movement the linearX value the pitch angle (rotation around the Y -axis) and two vertical straightnesses (theirbeams are parallel with distance about 160mmbetween them) The last two measurementsallow the calculation of the yaw (rotationaround the Z-axis) The horizontal straightnessis not monitored as it would have only asecond-order effect in the corrections Theinterferometers allow monitoring (2000 mea-surements per second) with an accuracy about03 mm for linear 03 mrad for angular para-meters over the full range of the scanner
Two calibration rulers each split into two parts(Fig 8) and placed under and above thechamber to be measured Every calibrationruler consists of two layers of nine ceramicplates (Fig 9) with six gold strips on eachcomposing a regular grid of 54 equivalent wireswith 30mm pitch The distance between thoselayers is 45mm The ceramic plates are installedon a carbon fibre support The calibrationrulers are used for the online (ie it is made for
every scan) reconstruction of the geometricalparameters of the X-ray tomograph and tocheck the stability and reliability of the X-raytomograph
A set of computer-controlled motors allowingthe alignment of the chamber on the X-raytomograph and choosing of the cross-sectionplane of the chamber One small moving cart isenslaved to the X-ray cart supporting twoscintillators optimized for the X-ray beamenergy spectrum following the beams inorder to record two shadowgrams simulta-neously
The control of the X-ray tomograph cart andchamber movements the data-acquisition andparameter monitoring are supported by adistributed online system based on VME PCSun computers connected via PCI MXI sharedmemory interface using softwarewritten on CC++ and LabVIEW
Two X-ray high-voltage power supplies (up to60 kV 50mA) The two high-voltage cablesfeeding the X-ray tubes move in synchronismwith the scanner on a separate cart in orderto ensure a perfect stability of the X-ray rollingcart
and collimating systems
Support of the upper calibration ruler
Chamber
Porticowithrails
Cart with shielded X-ray tubes
Z
X
Z
X
Fig 7 The X-ray tomograph during the mechanical control of a layers muon
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10192
4 Online calibration
Measurements with an accuracy of a fewmicrons require a precise control of all thegeometrical parameters of the X-ray tomographduring the measurements This requires the con-tinuous monitoring of the absolute position of theinterferometer laser beams in an external system ofcoordinates and of the origin of all laser axes Thislast point is not an easy requirement becauseduring the installation of the chambers on the
X-ray tomograph or during maintenance workthe trajectory of the laser beams may be frequentlyinterrupted for a short time by some object whichautomatically leads to the loss of the zeroreference In order to avoid this problem thecalibration and zero reset must be done for everyscan
During the scanning the wires of the referencerulers must be seen by the X-ray beams of theX-ray tomograph without overlap from thechamber wires The reference wires are previouslymeasured independently on a very-high-precisionoptical bench (E1 mm) During the scanning theshadow patterns of both the wires of the referencerulers and of the chamber are registered simulta-neously The basic geometrical parameters of theX-ray tomograph are then calculated for each scanusing the recorded shadowgram together with theoptical results of the position of the reference rulerwires [13] This approach allows a reliablepermanent control of the accuracy of the measure-ments Such a calibration scheme provides thebasis for the calculation of the position of allchamber wires
Let us stress that all the measurements are donein the same scan all calculations for the chamberand reference ruler wires are made with the same
Fig 8 One half of the calibration ruler (consisting of nine elements fixed on the carbon fibre tube) during assembling and
measurements
Fig 9 One element of the calibration ruler each ceramic plate
has six narrow gold strips faking wires
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 93
geometrical parameters The two reference rulersare measured with a known accuracy As aconsequence the chamber which is lying betweenthe two reference rulers is measured with the sameaccuracy Moreover in the offline analysis [13] thesame functions are used for the geometricalcalculation of the reference ruler and the chamberwires The X-ray tomograph accuracy is estimatedusing the residuals of the difference between thepredictions of the position of each wire of thereference rulers as given by the X-ray beams andtheir actual known position This method allowsone to control the accuracy of the X-ray tomo-graph scan-by-scan and consequently to detect anyloss of performance it allows for example toreveal small displacements or failures of theinterferometer system
Let us stress that the calibration is performedfor every scan and therefore for every scan thegeometrical parameters are calculated and theX-ray tomograph accuracy is checked
While studying the X-ray tomograph perfor-mance soon after its construction it became clearthat the precision of the optical measurements ofthe calibration rulers was not high enough toachieve the wanted accuracy It was later con-firmed that the optical measurements performedby the CERN Metrology Group using conven-tional industrial equipment were not preciseenough Therefore we had to develop a dedicatedmethod in order to improve the accuracy
Geometrical characteristics of the referencerulers may be divided into two groups the firstgroup consists of the parameters which can becorrected using direct X-ray tomograph measure-ments and the second group consists of para-meters for which other external reference devicesare needed In the first group we can note thatsome systematic errors in the position of individualcalibration ruler wires are well detected by theX-ray tomograph and those components of theoptical measurement errors are corrected by usingthe data coming from several scans As an examplefor the second group of parameters the straight-ness of the monitoring X-ray beams does not allowthe detection of possible errors in the opticalmeasurements of the vertical and horizontal shiftsbetween the layers of the calibration rulers If such
an error exists it leads to an error in the verticalscale and in the orthogonality of the system ofcoordinates of the X-ray tomograph
We call these two groups of optical measure-ment errors lsquonon-linearrsquo and lsquolinearrsquo errors3
respectively An autocalibration approach wasdeveloped to improve the accuracy of the opticalmeasurements It consists of two steps correctionof non-linear components in optical measurementsand determination (plus correction) of the linearparameters
The first step does not require special commentsAfter the geometrical reconstruction of the cali-bration ruler wires determined from several scans(in order to improve statistics) the residualsbetween the predictions by every X-ray beamand the optical measurements are calculatedUsing corresponding values from both non-paral-lel beams the two-dimensional correction ofposition of every calibration ruler wire may befound Because the pattern recognition maychange after the corrections an additional itera-tion may be done
5 Autocalibration correction of the linear errors in
the optical measurements of the calibration ruler
wire positions
51 Theoretical aspects
Let us consider the question of the linearsystematics effect of the absolute scales the scalesfor the X - and Z-axis and their relative angle (non-orthogonality of the system of coordinates) The X
scale is determined by the linear interferometer(measuring the position of the cart along the scan)the corresponding laser beam is parallel to theOX -axis (with second-order errors due to the non-parallelism of the laser beam and scan direction)The OZ scale and the non-orthogonality errors ofthe coordinate system however are completelydefined by the optical measurement errors They
3We have chosen the names in that way because the lsquolinearrsquo
(non-linear) errors in the optical measurements are propagated
to the chamber wire position errors which are linear (non-
linear) functions of the wire coordinates
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10194
cannot be determined from the X-ray tomographmeasurements themselves During the geometryfit the calculated effects of any change in the OZ
scale and in the non-orthogonality of the axes maybe completely compensated with correspondingchanges of the geometrical parameters of the X-ray beams Therefore some external measurementsare necessary
Let us consider three hypothetical parallel wiresA B C of the same rigid chamber (see Fig 10 thewires are perpendicular to the plane of picture andonly their projections A B C are seen) The pointsA B C should not lie on the same straight lineLet us imagine that they are measured by the X-ray tomograph in the original position (Fig 10a)and after rotation of the chamber by 901 aroundthe wiresrsquo direction (Fig 10b) If the X-raytomograph has a Z scale error instead of thepositions of the points A B C the positions ofpoints A0 B0 C0 and points A00 B00 C00 in therotated orientation will be found Putting bothmeasurements in the same system of coordinates
(we need three points because the angle aboutwhich these three points are rotated is not knownprecisely and in general does not equal 901) wecan estimate the Z scale error (Fig 10c)
A similar analysis applies for the non-orthogon-ality error Three wires A B C are measured aspoints A0 B0 C0 (Fig 11a) and as points A00 B00C00 in the 901 orientation (Fig 11b) Putting bothmeasurements in the same system of coordinates(Fig 11c) we can evaluate the non-orthogonalityerror introduced by the X-ray tomograph
Having determined the values of the Z scale andthe non-orthogonality errors we can calculate andthen correct the horizontal and vertical shiftsbetween the layers of the reference ruler becausethese shifts are the source of the errors in the Z
scale and non-orthogonality
52 Practical realization
In order to realize the described approach asmall calibration chamber (gauge chamber) was
A B
C
A B
C
AA
C
B
C
B
B
BA
A
C C
O X
Z
(a) (b) (c)
Fig 11 Non-orthogonality error calibration
A B
C
A
B
B
C
C
A
A
C
BCA
B
AB
C
O X
Z
(a) (b) (c)
Fig 10 Z scale error calibration
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 95
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
60 mrad the longer axis being aligned along thewires to be measured
Four interferometers which monitor the mainparameters of the scanner movement the linearX value the pitch angle (rotation around the Y -axis) and two vertical straightnesses (theirbeams are parallel with distance about 160mmbetween them) The last two measurementsallow the calculation of the yaw (rotationaround the Z-axis) The horizontal straightnessis not monitored as it would have only asecond-order effect in the corrections Theinterferometers allow monitoring (2000 mea-surements per second) with an accuracy about03 mm for linear 03 mrad for angular para-meters over the full range of the scanner
Two calibration rulers each split into two parts(Fig 8) and placed under and above thechamber to be measured Every calibrationruler consists of two layers of nine ceramicplates (Fig 9) with six gold strips on eachcomposing a regular grid of 54 equivalent wireswith 30mm pitch The distance between thoselayers is 45mm The ceramic plates are installedon a carbon fibre support The calibrationrulers are used for the online (ie it is made for
every scan) reconstruction of the geometricalparameters of the X-ray tomograph and tocheck the stability and reliability of the X-raytomograph
A set of computer-controlled motors allowingthe alignment of the chamber on the X-raytomograph and choosing of the cross-sectionplane of the chamber One small moving cart isenslaved to the X-ray cart supporting twoscintillators optimized for the X-ray beamenergy spectrum following the beams inorder to record two shadowgrams simulta-neously
The control of the X-ray tomograph cart andchamber movements the data-acquisition andparameter monitoring are supported by adistributed online system based on VME PCSun computers connected via PCI MXI sharedmemory interface using softwarewritten on CC++ and LabVIEW
Two X-ray high-voltage power supplies (up to60 kV 50mA) The two high-voltage cablesfeeding the X-ray tubes move in synchronismwith the scanner on a separate cart in orderto ensure a perfect stability of the X-ray rollingcart
and collimating systems
Support of the upper calibration ruler
Chamber
Porticowithrails
Cart with shielded X-ray tubes
Z
X
Z
X
Fig 7 The X-ray tomograph during the mechanical control of a layers muon
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10192
4 Online calibration
Measurements with an accuracy of a fewmicrons require a precise control of all thegeometrical parameters of the X-ray tomographduring the measurements This requires the con-tinuous monitoring of the absolute position of theinterferometer laser beams in an external system ofcoordinates and of the origin of all laser axes Thislast point is not an easy requirement becauseduring the installation of the chambers on the
X-ray tomograph or during maintenance workthe trajectory of the laser beams may be frequentlyinterrupted for a short time by some object whichautomatically leads to the loss of the zeroreference In order to avoid this problem thecalibration and zero reset must be done for everyscan
During the scanning the wires of the referencerulers must be seen by the X-ray beams of theX-ray tomograph without overlap from thechamber wires The reference wires are previouslymeasured independently on a very-high-precisionoptical bench (E1 mm) During the scanning theshadow patterns of both the wires of the referencerulers and of the chamber are registered simulta-neously The basic geometrical parameters of theX-ray tomograph are then calculated for each scanusing the recorded shadowgram together with theoptical results of the position of the reference rulerwires [13] This approach allows a reliablepermanent control of the accuracy of the measure-ments Such a calibration scheme provides thebasis for the calculation of the position of allchamber wires
Let us stress that all the measurements are donein the same scan all calculations for the chamberand reference ruler wires are made with the same
Fig 8 One half of the calibration ruler (consisting of nine elements fixed on the carbon fibre tube) during assembling and
measurements
Fig 9 One element of the calibration ruler each ceramic plate
has six narrow gold strips faking wires
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 93
geometrical parameters The two reference rulersare measured with a known accuracy As aconsequence the chamber which is lying betweenthe two reference rulers is measured with the sameaccuracy Moreover in the offline analysis [13] thesame functions are used for the geometricalcalculation of the reference ruler and the chamberwires The X-ray tomograph accuracy is estimatedusing the residuals of the difference between thepredictions of the position of each wire of thereference rulers as given by the X-ray beams andtheir actual known position This method allowsone to control the accuracy of the X-ray tomo-graph scan-by-scan and consequently to detect anyloss of performance it allows for example toreveal small displacements or failures of theinterferometer system
Let us stress that the calibration is performedfor every scan and therefore for every scan thegeometrical parameters are calculated and theX-ray tomograph accuracy is checked
While studying the X-ray tomograph perfor-mance soon after its construction it became clearthat the precision of the optical measurements ofthe calibration rulers was not high enough toachieve the wanted accuracy It was later con-firmed that the optical measurements performedby the CERN Metrology Group using conven-tional industrial equipment were not preciseenough Therefore we had to develop a dedicatedmethod in order to improve the accuracy
Geometrical characteristics of the referencerulers may be divided into two groups the firstgroup consists of the parameters which can becorrected using direct X-ray tomograph measure-ments and the second group consists of para-meters for which other external reference devicesare needed In the first group we can note thatsome systematic errors in the position of individualcalibration ruler wires are well detected by theX-ray tomograph and those components of theoptical measurement errors are corrected by usingthe data coming from several scans As an examplefor the second group of parameters the straight-ness of the monitoring X-ray beams does not allowthe detection of possible errors in the opticalmeasurements of the vertical and horizontal shiftsbetween the layers of the calibration rulers If such
an error exists it leads to an error in the verticalscale and in the orthogonality of the system ofcoordinates of the X-ray tomograph
We call these two groups of optical measure-ment errors lsquonon-linearrsquo and lsquolinearrsquo errors3
respectively An autocalibration approach wasdeveloped to improve the accuracy of the opticalmeasurements It consists of two steps correctionof non-linear components in optical measurementsand determination (plus correction) of the linearparameters
The first step does not require special commentsAfter the geometrical reconstruction of the cali-bration ruler wires determined from several scans(in order to improve statistics) the residualsbetween the predictions by every X-ray beamand the optical measurements are calculatedUsing corresponding values from both non-paral-lel beams the two-dimensional correction ofposition of every calibration ruler wire may befound Because the pattern recognition maychange after the corrections an additional itera-tion may be done
5 Autocalibration correction of the linear errors in
the optical measurements of the calibration ruler
wire positions
51 Theoretical aspects
Let us consider the question of the linearsystematics effect of the absolute scales the scalesfor the X - and Z-axis and their relative angle (non-orthogonality of the system of coordinates) The X
scale is determined by the linear interferometer(measuring the position of the cart along the scan)the corresponding laser beam is parallel to theOX -axis (with second-order errors due to the non-parallelism of the laser beam and scan direction)The OZ scale and the non-orthogonality errors ofthe coordinate system however are completelydefined by the optical measurement errors They
3We have chosen the names in that way because the lsquolinearrsquo
(non-linear) errors in the optical measurements are propagated
to the chamber wire position errors which are linear (non-
linear) functions of the wire coordinates
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10194
cannot be determined from the X-ray tomographmeasurements themselves During the geometryfit the calculated effects of any change in the OZ
scale and in the non-orthogonality of the axes maybe completely compensated with correspondingchanges of the geometrical parameters of the X-ray beams Therefore some external measurementsare necessary
Let us consider three hypothetical parallel wiresA B C of the same rigid chamber (see Fig 10 thewires are perpendicular to the plane of picture andonly their projections A B C are seen) The pointsA B C should not lie on the same straight lineLet us imagine that they are measured by the X-ray tomograph in the original position (Fig 10a)and after rotation of the chamber by 901 aroundthe wiresrsquo direction (Fig 10b) If the X-raytomograph has a Z scale error instead of thepositions of the points A B C the positions ofpoints A0 B0 C0 and points A00 B00 C00 in therotated orientation will be found Putting bothmeasurements in the same system of coordinates
(we need three points because the angle aboutwhich these three points are rotated is not knownprecisely and in general does not equal 901) wecan estimate the Z scale error (Fig 10c)
A similar analysis applies for the non-orthogon-ality error Three wires A B C are measured aspoints A0 B0 C0 (Fig 11a) and as points A00 B00C00 in the 901 orientation (Fig 11b) Putting bothmeasurements in the same system of coordinates(Fig 11c) we can evaluate the non-orthogonalityerror introduced by the X-ray tomograph
Having determined the values of the Z scale andthe non-orthogonality errors we can calculate andthen correct the horizontal and vertical shiftsbetween the layers of the reference ruler becausethese shifts are the source of the errors in the Z
scale and non-orthogonality
52 Practical realization
In order to realize the described approach asmall calibration chamber (gauge chamber) was
A B
C
A B
C
AA
C
B
C
B
B
BA
A
C C
O X
Z
(a) (b) (c)
Fig 11 Non-orthogonality error calibration
A B
C
A
B
B
C
C
A
A
C
BCA
B
AB
C
O X
Z
(a) (b) (c)
Fig 10 Z scale error calibration
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 95
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
4 Online calibration
Measurements with an accuracy of a fewmicrons require a precise control of all thegeometrical parameters of the X-ray tomographduring the measurements This requires the con-tinuous monitoring of the absolute position of theinterferometer laser beams in an external system ofcoordinates and of the origin of all laser axes Thislast point is not an easy requirement becauseduring the installation of the chambers on the
X-ray tomograph or during maintenance workthe trajectory of the laser beams may be frequentlyinterrupted for a short time by some object whichautomatically leads to the loss of the zeroreference In order to avoid this problem thecalibration and zero reset must be done for everyscan
During the scanning the wires of the referencerulers must be seen by the X-ray beams of theX-ray tomograph without overlap from thechamber wires The reference wires are previouslymeasured independently on a very-high-precisionoptical bench (E1 mm) During the scanning theshadow patterns of both the wires of the referencerulers and of the chamber are registered simulta-neously The basic geometrical parameters of theX-ray tomograph are then calculated for each scanusing the recorded shadowgram together with theoptical results of the position of the reference rulerwires [13] This approach allows a reliablepermanent control of the accuracy of the measure-ments Such a calibration scheme provides thebasis for the calculation of the position of allchamber wires
Let us stress that all the measurements are donein the same scan all calculations for the chamberand reference ruler wires are made with the same
Fig 8 One half of the calibration ruler (consisting of nine elements fixed on the carbon fibre tube) during assembling and
measurements
Fig 9 One element of the calibration ruler each ceramic plate
has six narrow gold strips faking wires
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 93
geometrical parameters The two reference rulersare measured with a known accuracy As aconsequence the chamber which is lying betweenthe two reference rulers is measured with the sameaccuracy Moreover in the offline analysis [13] thesame functions are used for the geometricalcalculation of the reference ruler and the chamberwires The X-ray tomograph accuracy is estimatedusing the residuals of the difference between thepredictions of the position of each wire of thereference rulers as given by the X-ray beams andtheir actual known position This method allowsone to control the accuracy of the X-ray tomo-graph scan-by-scan and consequently to detect anyloss of performance it allows for example toreveal small displacements or failures of theinterferometer system
Let us stress that the calibration is performedfor every scan and therefore for every scan thegeometrical parameters are calculated and theX-ray tomograph accuracy is checked
While studying the X-ray tomograph perfor-mance soon after its construction it became clearthat the precision of the optical measurements ofthe calibration rulers was not high enough toachieve the wanted accuracy It was later con-firmed that the optical measurements performedby the CERN Metrology Group using conven-tional industrial equipment were not preciseenough Therefore we had to develop a dedicatedmethod in order to improve the accuracy
Geometrical characteristics of the referencerulers may be divided into two groups the firstgroup consists of the parameters which can becorrected using direct X-ray tomograph measure-ments and the second group consists of para-meters for which other external reference devicesare needed In the first group we can note thatsome systematic errors in the position of individualcalibration ruler wires are well detected by theX-ray tomograph and those components of theoptical measurement errors are corrected by usingthe data coming from several scans As an examplefor the second group of parameters the straight-ness of the monitoring X-ray beams does not allowthe detection of possible errors in the opticalmeasurements of the vertical and horizontal shiftsbetween the layers of the calibration rulers If such
an error exists it leads to an error in the verticalscale and in the orthogonality of the system ofcoordinates of the X-ray tomograph
We call these two groups of optical measure-ment errors lsquonon-linearrsquo and lsquolinearrsquo errors3
respectively An autocalibration approach wasdeveloped to improve the accuracy of the opticalmeasurements It consists of two steps correctionof non-linear components in optical measurementsand determination (plus correction) of the linearparameters
The first step does not require special commentsAfter the geometrical reconstruction of the cali-bration ruler wires determined from several scans(in order to improve statistics) the residualsbetween the predictions by every X-ray beamand the optical measurements are calculatedUsing corresponding values from both non-paral-lel beams the two-dimensional correction ofposition of every calibration ruler wire may befound Because the pattern recognition maychange after the corrections an additional itera-tion may be done
5 Autocalibration correction of the linear errors in
the optical measurements of the calibration ruler
wire positions
51 Theoretical aspects
Let us consider the question of the linearsystematics effect of the absolute scales the scalesfor the X - and Z-axis and their relative angle (non-orthogonality of the system of coordinates) The X
scale is determined by the linear interferometer(measuring the position of the cart along the scan)the corresponding laser beam is parallel to theOX -axis (with second-order errors due to the non-parallelism of the laser beam and scan direction)The OZ scale and the non-orthogonality errors ofthe coordinate system however are completelydefined by the optical measurement errors They
3We have chosen the names in that way because the lsquolinearrsquo
(non-linear) errors in the optical measurements are propagated
to the chamber wire position errors which are linear (non-
linear) functions of the wire coordinates
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10194
cannot be determined from the X-ray tomographmeasurements themselves During the geometryfit the calculated effects of any change in the OZ
scale and in the non-orthogonality of the axes maybe completely compensated with correspondingchanges of the geometrical parameters of the X-ray beams Therefore some external measurementsare necessary
Let us consider three hypothetical parallel wiresA B C of the same rigid chamber (see Fig 10 thewires are perpendicular to the plane of picture andonly their projections A B C are seen) The pointsA B C should not lie on the same straight lineLet us imagine that they are measured by the X-ray tomograph in the original position (Fig 10a)and after rotation of the chamber by 901 aroundthe wiresrsquo direction (Fig 10b) If the X-raytomograph has a Z scale error instead of thepositions of the points A B C the positions ofpoints A0 B0 C0 and points A00 B00 C00 in therotated orientation will be found Putting bothmeasurements in the same system of coordinates
(we need three points because the angle aboutwhich these three points are rotated is not knownprecisely and in general does not equal 901) wecan estimate the Z scale error (Fig 10c)
A similar analysis applies for the non-orthogon-ality error Three wires A B C are measured aspoints A0 B0 C0 (Fig 11a) and as points A00 B00C00 in the 901 orientation (Fig 11b) Putting bothmeasurements in the same system of coordinates(Fig 11c) we can evaluate the non-orthogonalityerror introduced by the X-ray tomograph
Having determined the values of the Z scale andthe non-orthogonality errors we can calculate andthen correct the horizontal and vertical shiftsbetween the layers of the reference ruler becausethese shifts are the source of the errors in the Z
scale and non-orthogonality
52 Practical realization
In order to realize the described approach asmall calibration chamber (gauge chamber) was
A B
C
A B
C
AA
C
B
C
B
B
BA
A
C C
O X
Z
(a) (b) (c)
Fig 11 Non-orthogonality error calibration
A B
C
A
B
B
C
C
A
A
C
BCA
B
AB
C
O X
Z
(a) (b) (c)
Fig 10 Z scale error calibration
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 95
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
geometrical parameters The two reference rulersare measured with a known accuracy As aconsequence the chamber which is lying betweenthe two reference rulers is measured with the sameaccuracy Moreover in the offline analysis [13] thesame functions are used for the geometricalcalculation of the reference ruler and the chamberwires The X-ray tomograph accuracy is estimatedusing the residuals of the difference between thepredictions of the position of each wire of thereference rulers as given by the X-ray beams andtheir actual known position This method allowsone to control the accuracy of the X-ray tomo-graph scan-by-scan and consequently to detect anyloss of performance it allows for example toreveal small displacements or failures of theinterferometer system
Let us stress that the calibration is performedfor every scan and therefore for every scan thegeometrical parameters are calculated and theX-ray tomograph accuracy is checked
While studying the X-ray tomograph perfor-mance soon after its construction it became clearthat the precision of the optical measurements ofthe calibration rulers was not high enough toachieve the wanted accuracy It was later con-firmed that the optical measurements performedby the CERN Metrology Group using conven-tional industrial equipment were not preciseenough Therefore we had to develop a dedicatedmethod in order to improve the accuracy
Geometrical characteristics of the referencerulers may be divided into two groups the firstgroup consists of the parameters which can becorrected using direct X-ray tomograph measure-ments and the second group consists of para-meters for which other external reference devicesare needed In the first group we can note thatsome systematic errors in the position of individualcalibration ruler wires are well detected by theX-ray tomograph and those components of theoptical measurement errors are corrected by usingthe data coming from several scans As an examplefor the second group of parameters the straight-ness of the monitoring X-ray beams does not allowthe detection of possible errors in the opticalmeasurements of the vertical and horizontal shiftsbetween the layers of the calibration rulers If such
an error exists it leads to an error in the verticalscale and in the orthogonality of the system ofcoordinates of the X-ray tomograph
We call these two groups of optical measure-ment errors lsquonon-linearrsquo and lsquolinearrsquo errors3
respectively An autocalibration approach wasdeveloped to improve the accuracy of the opticalmeasurements It consists of two steps correctionof non-linear components in optical measurementsand determination (plus correction) of the linearparameters
The first step does not require special commentsAfter the geometrical reconstruction of the cali-bration ruler wires determined from several scans(in order to improve statistics) the residualsbetween the predictions by every X-ray beamand the optical measurements are calculatedUsing corresponding values from both non-paral-lel beams the two-dimensional correction ofposition of every calibration ruler wire may befound Because the pattern recognition maychange after the corrections an additional itera-tion may be done
5 Autocalibration correction of the linear errors in
the optical measurements of the calibration ruler
wire positions
51 Theoretical aspects
Let us consider the question of the linearsystematics effect of the absolute scales the scalesfor the X - and Z-axis and their relative angle (non-orthogonality of the system of coordinates) The X
scale is determined by the linear interferometer(measuring the position of the cart along the scan)the corresponding laser beam is parallel to theOX -axis (with second-order errors due to the non-parallelism of the laser beam and scan direction)The OZ scale and the non-orthogonality errors ofthe coordinate system however are completelydefined by the optical measurement errors They
3We have chosen the names in that way because the lsquolinearrsquo
(non-linear) errors in the optical measurements are propagated
to the chamber wire position errors which are linear (non-
linear) functions of the wire coordinates
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10194
cannot be determined from the X-ray tomographmeasurements themselves During the geometryfit the calculated effects of any change in the OZ
scale and in the non-orthogonality of the axes maybe completely compensated with correspondingchanges of the geometrical parameters of the X-ray beams Therefore some external measurementsare necessary
Let us consider three hypothetical parallel wiresA B C of the same rigid chamber (see Fig 10 thewires are perpendicular to the plane of picture andonly their projections A B C are seen) The pointsA B C should not lie on the same straight lineLet us imagine that they are measured by the X-ray tomograph in the original position (Fig 10a)and after rotation of the chamber by 901 aroundthe wiresrsquo direction (Fig 10b) If the X-raytomograph has a Z scale error instead of thepositions of the points A B C the positions ofpoints A0 B0 C0 and points A00 B00 C00 in therotated orientation will be found Putting bothmeasurements in the same system of coordinates
(we need three points because the angle aboutwhich these three points are rotated is not knownprecisely and in general does not equal 901) wecan estimate the Z scale error (Fig 10c)
A similar analysis applies for the non-orthogon-ality error Three wires A B C are measured aspoints A0 B0 C0 (Fig 11a) and as points A00 B00C00 in the 901 orientation (Fig 11b) Putting bothmeasurements in the same system of coordinates(Fig 11c) we can evaluate the non-orthogonalityerror introduced by the X-ray tomograph
Having determined the values of the Z scale andthe non-orthogonality errors we can calculate andthen correct the horizontal and vertical shiftsbetween the layers of the reference ruler becausethese shifts are the source of the errors in the Z
scale and non-orthogonality
52 Practical realization
In order to realize the described approach asmall calibration chamber (gauge chamber) was
A B
C
A B
C
AA
C
B
C
B
B
BA
A
C C
O X
Z
(a) (b) (c)
Fig 11 Non-orthogonality error calibration
A B
C
A
B
B
C
C
A
A
C
BCA
B
AB
C
O X
Z
(a) (b) (c)
Fig 10 Z scale error calibration
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 95
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
cannot be determined from the X-ray tomographmeasurements themselves During the geometryfit the calculated effects of any change in the OZ
scale and in the non-orthogonality of the axes maybe completely compensated with correspondingchanges of the geometrical parameters of the X-ray beams Therefore some external measurementsare necessary
Let us consider three hypothetical parallel wiresA B C of the same rigid chamber (see Fig 10 thewires are perpendicular to the plane of picture andonly their projections A B C are seen) The pointsA B C should not lie on the same straight lineLet us imagine that they are measured by the X-ray tomograph in the original position (Fig 10a)and after rotation of the chamber by 901 aroundthe wiresrsquo direction (Fig 10b) If the X-raytomograph has a Z scale error instead of thepositions of the points A B C the positions ofpoints A0 B0 C0 and points A00 B00 C00 in therotated orientation will be found Putting bothmeasurements in the same system of coordinates
(we need three points because the angle aboutwhich these three points are rotated is not knownprecisely and in general does not equal 901) wecan estimate the Z scale error (Fig 10c)
A similar analysis applies for the non-orthogon-ality error Three wires A B C are measured aspoints A0 B0 C0 (Fig 11a) and as points A00 B00C00 in the 901 orientation (Fig 11b) Putting bothmeasurements in the same system of coordinates(Fig 11c) we can evaluate the non-orthogonalityerror introduced by the X-ray tomograph
Having determined the values of the Z scale andthe non-orthogonality errors we can calculate andthen correct the horizontal and vertical shiftsbetween the layers of the reference ruler becausethese shifts are the source of the errors in the Z
scale and non-orthogonality
52 Practical realization
In order to realize the described approach asmall calibration chamber (gauge chamber) was
A B
C
A B
C
AA
C
B
C
B
B
BA
A
C C
O X
Z
(a) (b) (c)
Fig 11 Non-orthogonality error calibration
A B
C
A
B
B
C
C
A
A
C
BCA
B
AB
C
O X
Z
(a) (b) (c)
Fig 10 Z scale error calibration
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 95
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
made of a rigid carbon and aluminium frame(Fig 12) It consists of two horizontal layers ofwires about 200mm long with 60mm betweenwires and 300mm between layers It covers aworking area of B300 300mm2 One layercontains five wires and the other one contains sixwires placed with a spacing convenient forperforming the optical measurements
In order to reduce the statistical errors thegauge chamber is measured in eight differentorientations four rotated by 901 around the wiredirection and four after a 1801 rotation aroundthe vertical axis of the chamber To describe theposition of a system of coordinates attached to the
gauge chamber for each orientation we introducethree parameters xCj zCj j frac14 1yN frac14 8 in ourcase (see Fig 12) The position of every wire of thegauge chamber may be described in the gaugechamber system of coordinates as (M frac14 11 in ourcase) To determine the X-ray tomograph con-tribution to the measurement errors due to non-orthogonality and Z scale error we have to entertwo additional parameters bZ the vertical scalingdistortion factor (along Z) and gNO the angleerror due to the non-orthogonality A global fit isdone for finding the parameters (bZ gNO xCj zCj aCj and xWj zWj) by a minimization of differencesof coordinates of the same wires in the common
5 wires up shifted by 60 mm in X
6 wires down shifted by 60 mm in X
X
Z
X
Z
Xj
Zj
zCj
xCj
Cj
Fig 12 Gauge chamber
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10196
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
system of coordinates for all chamber orientationsIn our case we use for the minimization the least-squares method because the measurement errorsare Gaussian
Having determined the bZ gNO parameters wecorrect the shifts between the layers of thecalibration rulers We repeat this analysis withthe corrected values until we have small enougherror values (less than the errors of theseparameters estimated in the fit procedure) for thebZ gNO parameters
6 X-ray tomograph accuracy
The estimation of the accuracy of the X-raytomograph can be obtained in different waysautocalibration online calibration and actualmeasurements
61 Autocalibration results
The autocalibration procedure using the gaugechamber allowed us not only to calibrate the X-raytomograph but also to estimate parameters char-acterizing the accuracy of the X-ray tomographWe consider the results of the autocalibration asthe best procedure for the estimation of theaccuracy of the X-ray tomograph because wemeasure the same gauge chamber in many posi-tions and orientations Indeed if the X-raytomograph introduces the errors to the measure-ments of the gauge chamber wire coordinates wewill see such errors when comparing the results ofthe autocalibration scans performed for the gaugechamber in many orientations because the errorswill apply to different wires and different direc-tions (in the gauge chamber system of coordi-nates) To maximally highlight the effect we use asmany orientations as possible But the problem ofthe wire reconstruction during the analysis restrictstheir number to eight
Measurements with rotations during autocali-bration (in the same place usually at the centre ofthe working space) also allow the positions of thegauge chamber wires in its system of coordinatesto be obtained Comparison of measurements of
the gauge chamber in different places of the X-raytomograph working space allows the estimation ofthe accuracy over a whole working space Such acomparison is one of the key points of theautocalibration approach
The results of the autocalibration sessions of theX-ray tomograph performed over several years arequite similar (Fig 13) Statistical fluctuation of thewire positions of the gauge chamber in allorientations and working space places is foundto reach 2 mm standard deviation The estimationsof the errors of the vertical scale is 19 mmm and ofthe non-orthogonality 21 mmm (21 mrad)
We have to stress that all results are based onthe interferometry measurements ie we assumethat they are ideal otherwise we have to correct allvalues above But because the precision of theinterferometer measurements in our set-up isbetter than 03 mm (specification provided by themanufacturers) therefore the interferometer errorsare negligible with respect to the 2 mm of the X-raytomograph accuracy
62 Online calibration
The result of the on-line calibration is also usedto estimate and control the accuracy of the X-raytomograph For every scan we measure theposition of the wires of the calibration rulersusually there are 600ndash650 measurements by bothbeams Using those measurements 38 parameters(the general geometry of the X-ray tomograph ieX-ray beams with respect to each other and to theinterferometer laser beams and sensors and three-dimensional positions of four calibration rulerstwo on both lower and upper supports) have to befound This is enough to determine the X-raytomograph geometrical parameters with highprecision The root mean square (rms) of theresiduals between the real position of the wirepeaks (Fig 6) along the scan direction andprediction of the same position using the para-meters found and the interferometry measure-ments for the carrier with the X-ray beams are 15ndash2 mm for the calibration scans Owing to the X-rayabsorption for the scans with the real chambersthe residual rms is a bit worse 2ndash3 mm Thesenumbers are exactly what is expected as a
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 97
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
contribution from the statistical errors of theestimation of peak positions the other effects(for example systematics the interferometer er-rors instability of hardware parameters) are verysmall
Let us point out the key feature of ourapproach the calibration ruler structure whichconsists of four layers of wires (two above and twounder the chamber) covering all the working spaceof the X-ray tomograph We can predict theposition of all those wires surrounding the
measured chamber with a precision better than2ndash3 mm Therefore we can measure the chamberwires inside the calibration structure with the sameaccuracy because they are seen by the X-raytomograph as peaks very similar to the calibrationruler ones
63 Measurements of real chambers
The measurements of real chambers (see Refs[14ndash16]) give a lot of possibilities to crosscheck the
Calibration with rotations after optical corrections
-150
-100
-50
0
50
100
150
-150 -100 -50 0 50 100 150
Residuals map for all scansX (mm)
Z (
mm
)
Fig 13 Typical result of the autocalibration procedure The centres of circles correspond to the wires of the gauge chamber The
measured shifts of the wires from nominal positions for all calibration scans are shown as short lines starting from the centres of wires
Directions of the lines exactly correspond to the real shifts and lengths are proportional to the shift values The scale is indicated by the
circles which correspond to 5 mm of the shift value The standard deviation of the X-ray tomograph measurement errors is estimated as
16mm Uncertainties of the non-orthogonality and Z scale errors are estimated as 21mmm and 19mmm standard deviations
correspondingly
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash10198
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
X-ray tomograph accuracy Let us show some ofthem
The barrel chambers allow the comparison ofthe X-ray tomograph measurements in two posi-tions (original and flipped ie rotated by aroundwire direction) because they are rigid enough tokeep their shape In fact those comparisons arevery similar to the autocalibration calculationsexcept that the area of checking is larger
Some chambers had different tensions of thewires producing sagging variations This effect isdetected very well for scans in the middle cross-section of the chamber For example for one ofthe biggest chambers an expected effect of the wiresag difference in the middle of the chamber was85 mm The value obtained using the X-raytomograph measurement was 8673 mm
The alignment platforms glued onto the tubes ofthe muon chamber consist of frames with opticallymeasured wires aligned to the chamber wires Allthose wires are measured simultaneously in thesame scan The sags of the tubes on which thealignment platforms are glued and the sag ofthe wires inside of those tubes are different Hencethe additional shifts between the alignment plat-form wires and the chamber wires can be seen bythe X-ray tomograph Although the platformposition is not known (it must be measured bythe X-ray tomograph) we can measure the changeof the shift for the chamber with and without sag
compensation4 For example the biggest change ofthe shift which we had observed was estimatedtheoretically as 121 mm On the X-ray tomographat one chamber end this effect was measured forthe two platforms as 120706 and 121719 mmand at the other chamber end as 128730 and131745 mm
For scans in intermediate cross-sections (be-tween the scan sections near the chamber edges)some parameters describing a quality of thechamber may be calculated as a linear combina-tion of the same parameters found for scans nearthe edges Those calculations are very interestingin the case of big difference for some parameter forthe scans near the edges Fig 14 illustrates such aneffect It shows a deviation from the straight line ofthe horizontal shift between the chamber multi-layers as a function of the scan position
In all such effects the measurements of the X-raytomograph are very consistent with expectedvalues Until now no difference has been detectedbetween theoretical estimations of the effects andcalculations from the X-ray tomograph results (of
Horizontal shift vs scan section
-1500
-1000
-500
0
500
1000
1500
2000
-2000 -1000 0 1000 2000
Scan position (mm)
Sh
ift
(microm
)Deviation from straight line
-6
-4
-2
0
2
4
6
-2000 -1000 0 1000 2000
Scan position (mm)
Dev
iati
on
(microm
)
Fig 14 Illustration of the crosscheck of the X-ray tomograph accuracy using its measurements of the horizontal shifts between
multilayers for a chamber which has a big error in that parameter On the left-hand picture the linear dependence of the horizontal
shift from the scan position may be seen and on the right-hand one the deviations from the straight line are shown The difference
values are distributed with the 24 mm standard deviation For comparison a typical estimation in the X-ray tomograph analysis of the
horizontal shift measurement error is around 21 mm standard deviation
4The sag compensation of the chamber serves to decrease the
difference in the gravitational deflection of the chamber tubes
and wires The barrel chambers have a special mechanism to
additionally support the middle of the chamber with respect to
the edges of the chamber Such a design is necessary to
compensate the sag differences of the chambers in different
orientation in ATLAS
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 99
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
course up to statistical errors which are alwaysestimated)
7 Conclusion
Let us summarize the main methodological andpractical results obtained after 6 years of develop-ment of the X-ray tomograph
The accuracy of the X-ray tomograph is betterthan 2 mm (statistical error) and 2 mm (systematicerror) for both vertical and horizontal wirecoordinates in the working area 220 cm (horizon-tal) 60 cm (vertical)
The adopted measurement procedure allows adecrease in the required accuracy for the opticalmeasurements of the calibration rulers Indeed westarted the autocalibration procedure using opticalmeasurements which were corrected afterwardsAlternatively we could start with for examplenominal (foreseen by design) wire positions of thecalibration rulers and correct them5 An absoluteprecision of the horizontal coordinate of the wiresis guaranteed by the interferometers The verticalcoordinate and non-orthogonality of the X-raytomograph is provided by the autocalibrationprocedure As a consequence we were able toestimate the precision of the optical measurementsto about 6 mm when considering the errors for bothvertical and horizontal wire coordinates togetherThe final information from the CERN metrologygroup about the precision of the metrology tableused for the optical measurements matched ourestimation horizontal precision about 3 mm andvertical precision about 11 mm
The measurement procedure adopted allows theautomatic adaptation of the geometry reconstruc-tion procedure to the geometrical parameters
change zero values of the interferometers posi-tions of the calibration rulers X-ray beams etc Itmeans that the analysis procedure is not changedin the case of a small modification done duringmaintenance to the X-ray tomograph layout
As a consequence the proposed scheme for themeasurements of the chambers of the ATLASMuon system allows the automated serial scanningof the chambers with maximal scanning rate Thescanning rate in such a case is restricted mostly bysome physical limitation of the set-up (countingrate of the scintillators necessary statistics toobtain enough precision etc)
The chosen procedure allows the automaticcontrol of the precision of the X-ray tomographfor every scan with checking each time of theresiduals of the calibration ruler wires
If the 1200 muon chambers could be measuredbefore ATLAS installation all chambers would bequalified below 5 mm accuracy even if they wouldnot have been built according to specificationsUnfortunately for this purpose at least threemore X-ray tomographs would be necessarybecause one single X-ray tomograph cannotmeasure all the chambers within schedule
Acknowledgements
We would like to thank all the members of theX-ray tomograph group the ECP-EOS EST-MFEP-ATM CERN groups for excellent technicalassistance during design construction and exploi-tation of the X-ray tomograph the ATLASCollaboration and particularly the MDT Muongroup for continuous support and fruitful discus-sions and advice
References
[1] ATLAS experiment on WWW httpatlasinfocernch80
ATLASWelcomehtml
[2] ATLAS technical proposal CERNLHCC94-43 LHCC
P2 15 December 1994
[3] The ATLAS collaboration ATLAS technical proposal for
a general-purpose pp experiment at the Large Hadron
Collider at CERN CERNLHCC94-43 LHCCP2 15
December 1994
5 In reality the wires of the calibration rulers were not
parallel with the required accuracy This led to the necessity to
know projections on the planes YOZ and XOY of the angles
between wires The optical measurements were used for that
Hence we do not need to optically measure the average
positions of the calibration ruler wires but we have to measure
the non-parallelism of the wires Accuracy of those measure-
ments may be much lower because first the corresponding
values are used in the second-order corrections and second
they are done in local area (wire length is 30mm)
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101100
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101
[4] ATLAS muon spectrometer technical design report
CERNLHCC97-22 1997
[5] L Vertogradov High precision wire positions measuring
in the drift tube package by the X-ray scanner (XTomo-
graph) ATLAS note ATL-MUON-94-41 1994
[6] G Alexeev et al XTOMO A prototype of the X-ray
tomograph for high precision measurements of the MDT
muon chambers ATLAS note ATL-MUON-97-142 1997
[7] D Drakoulakos et al XTOMO One-dimensional X-ray
scanning of a 8-layer MDT prototype using the lsquopassive
modersquo ATLAS note ATL-MUON-97-151 1997
[8] D Drakoulakos et al XTOMO2 stereo-measurements of
the MDT muon chambers using a high precision X-ray
tomograph ATLAS note ATL-MUON-97-155 1997
[9] J Berbier et al X-ray tomograph prototype for MDT
quality control ATLAS note ATL-MUON-97-174 1997
[10] D Drakoulakos et al The high-precision X-ray tomo-
graph for quality control of the ATLAS MDT muon
spectrometer Proceedings of the Seventh Asia Pacific
Physics Conference 19ndash23 August 1997 Beijing pp 183ndash
188
[11] J Berbier et al Nucl Instr and Meth A 419 (1998) 342
[12] D Drakoulakos et al The high precision X-ray
tomograph for quality control of the ATLAS MDT muon
spectrometer CERN-OPEN-97-023 30 July 1997
[13] E Gschwendtner F Rohrbach Y Sedykh Analysis and
results from measurements on an X-ray tomograph of
large full-scale MDT prototypes ATLAS note ATL-
MUON-98-175 1997
[14] R Avramidou et al Calibration of the X-ray tomograph
ATLAS note ATL-MUON-2001-008 2001
[15] C Bini G Ciapetti Analysis of the X-ray tomograph
measurements of Calypso prototype ATLAS note ATL-
MUON-97-187 1997
[16] X Tomo WWW reference address httpxtomohome-
cernchxtomo
R Avramidou et al Nuclear Instruments and Methods in Physics Research A 496 (2003) 83ndash101 101