Upload
evette
View
50
Download
0
Embed Size (px)
DESCRIPTION
Monday Semin ar. IT Survey. LHCb B Field Map. Géraldine Conti. EPFL, the 21 th of May 2007. Parameterized LHCb B Field Map. Reminder of the Goal. MC Parameterization. Iterative Polynomial Fitting Method. One-go 3D Fitting Method. Choice of Regions. - PowerPoint PPT Presentation
Citation preview
1
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
LHCb B Field MapLHCb B Field Map
Géraldine Conti
IT SurveyIT Survey
EPFL, the 21th of May 2007
Monday Seminar
2
Parameterized Parameterized LHCb B Field LHCb B Field
MapMap
3
Outline …
Analytical Field Service
Reminder of the Goal
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
MC Parameterization
Iterative Polynomial Fitting Method
One-go 3D Fitting Method
Choice of Regions
Results in the Acceptance Region
Residual Parameterization
Measurement Matching MethodPreliminary Results
Analytic Parameterization
4
B measurement Campaign (Dec. 2005)
B Measurements done…
by 3D Hall probes arranged in a fixed configuration on a movable support
obtained for the two polarities
cover most of LHCb acceptance: Upstream : x = [-1.0, 1.0] m y = [-0.4, 0.4] m
Magnet : x = [-2.7, 2.4] m y = [-1.0, 1.0] m
Downstream : x = [-2.5, 2.5] m y = [-1.7, 1.7] m
(with demagnetization cycle)
(fine grid of 8x8x10 cm)
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
5
Reminder of the Goal
1) TOSCA simulated B field map (grid of 10x10x10 cm)
2) Measurements (They don’t cover all the acceptance)
Why is a parameterized B field map needed ?
1) The real measurements can be compared with MC data to model the residuals : Residual parameterization.
Provide an accurate determination of the B field map as close as possible to the measurements
What is available to perform the parameterizations?
Goal
2) The extrapolation of the Residual parameterization can be realized for the regions where no measurement is available.
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
6
B-Field Map based B-Field Map based on MC dataon MC data
(MC Parameterization)
7
Principle of the Method (eg. By component) :
Iterative Multipolynomial Fitting Method (1)
Fit By as a function of y with x,z fixed. By(y) for x=10cm and z=830cm
Fit Ai as a function of z with x fixed.
Fit Bj as a function of x.
At the end: N·M·P coefficients to cover the map.
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
N coefficients Ai for each slice.
M coefficients Bj
for each slice.P coefficients Ck.
Pol(4)A0
A1
A2
A3
A4
B0
B1
B2
B3
B4z
z
z
z
z x
x
x
x
x
Ai(z) for x=210cm : Bj(x) for A0 :
Pol(4)
8
Iterative Multipolynomial Fitting Method (2)
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
The method works well, but not easy to achieve the final precision because of the difficulty of optimizing iteratively.
Downstream region is… Magnet region… Residuals ∆(BMC - Banalytic) at z=830cm Bx(y) for x=90cm and z=[350,680] cm
…reasonably described.
…needs Fourier parameterization due to oscillatory pattern.
9
One-go 3D Fitting Method (1)
Least square procedure : Determine the cn coefficients and the Fn functions, such that :
Modified Gram-Schmidt orthogonalisation algorithm to keep the functions Fn that significantly reduce S :
Basis change to have Wn functions which are orthogonal between them.
Wn
Fn
WnFn
To decide if the Nth function is to be kept, the projection of Fn
on Wn is measured and should be greater than a given value to contribute significantly to the reduction of S (the angle should be greater than a given value) . See H. Wind, Yellow report, vol.72-21,CERN,1972 ; H. Wind, Yellow report EP/81-12,CERN,1981
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
is minimal, with
Fn are not orthogonal between them.
10
Implementation of the method (2)
An implementation of the method is available in ROOT (TMultiDimFit Class). Some technical problems have been encountered, but solved thanks to René Brun.
Configuration of the optimization :
A) Type of Polynomials : Monomials, Legendre, Chebyshev
C) Main limits to the number of terms in the parameterization :
1) Max. of terms in the final parameterization
2) Max. of powers for each variable x,y,z to be considered
3) Min. angle
B) Relative error accepted :
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
11
Compromise between relative precision, small number of regions, small number of terms in the parameterization.
Regions Definition for the B field Maps (1)
Regions definitions as simple as possible.
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
QuickTime™ and aTIFF (LZW) decompressor
are needed to see this picture.
acceptance angles z coordinate x and y coordinates
Lots of different cuttings tested :
Cuttings with respect to B field gradient have been tested too, but the geometry of the cuttings was too complicate.
12
Regions Definition for the B field Maps (2)
Cutting depends mainly on only one variable (z)
Downstream Magnet
8m
Upstream
3m 10m- 0.5m
4m
1 1.5 2 2.4 4.1 4.7 5.1 6.1 6.7 7.3 8.5 9 z(m)
y(m)
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
The same cutting is chosen for Bx,By and Bz components.
13
Regions Definition for the B field Maps (3)
However, in the magnet region, [ymax-30cm,ymax] values have been fitted separately, but with the same z cuts to improve the fits.
Bx for x=1.3m and z=4.8m By for x=1.3m and z=4.8m
MC Parameterizations involve 50 to 150 terms, depending on the B field fluctuations.
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
14
By map Result (x=0m, y=0m) (1)
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
MC parameterized By
TOSCA simulated By
Very good matching !
15
By map Result : Relative Precision (2)
Upstream Magnet Downstream
Relative precision on By < 0.001 inside the three regions
Relative precision :
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
16
Bx map Result (x=0.4m, y=0.4m)
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
MC parameterized By
TOSCA simulated By
17
Bz map Result (x=0.4m, y=0.4m)
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
MC parameterized By
TOSCA simulated By
18
Continuity at the region boundaries
The relative discrepancy between MC parameterizations at the boundary of two regions is in the same order of the fit precision (~10-3).
Discrepancy between By parameterizations at the 14 boundaries Discrepancy between the two By parameterizations at the z=610cm boundary
Relative Discrepancy :
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
By boundary at z=610cm
19
Out of acceptance regions
The Downstream region has been already parameterized.
Problems (peaks) are encountered for the Upstream and Magnet regions to find a good parameterization, because we are inside material.
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
Upstream Magnet Downstream
Parameterization needed, but with a less acurate precision.
20
Matching with the Matching with the measurementsmeasurements
(Residual Parameterization)
21
Clean-up of the measurements
Clean-up of the measurements done in the three regions (started by Adlene Hicheur).
Clean-up
Some non-physical behaviours observed in the measurements, which can perturb the parameterizations.
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
22
Matching Method
Parameterize the residuals (Bmeasurements - Banalytic values) : Analytic Parameterization = MC Parameterization + Residual Parameterization
Extrapolate the Residual Parameterization to regions where no measurement is available.
Calculate the B values with the MC Parameterizations at the same measurement coordinates (x,y,z).
Measurements Analytic values Residuals
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
23
By Residual Results (1)
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
x=4cm, y=4cm, Magnetx=4cm, y=4cm, Upstream
InterpolationMC Parameterization
Analytic ParameterizationMeasurements
(MC + Residual Parameterization)
Non negligible corrections for the most important B component (By) near the centre (x~0cm and y~cm) !
24
By Residual Results (2)
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
x=100cm, y=4cm, Magnet x=204cm, y=164cm
x=140cm, y=140cm
InterpolationMC Parameterization
Analytic ParameterizationMeasurements
(MC + Residual Parameterization)
The most important discrepancies between MC data and measurements are found for big x and/or y values
25
Reverse Polarisation By Residuals (3)
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
x=100cm, y=4cm, Magnet
InterpolationMC Parameterization
Analytic Parameterization-(Reverse Polarisation Measurements)
(MC + Residual Parameterization)
The values obtained with the Analytic Parameterization (found for positive polarisation measurements) seems to match well with the reverse polarisation measurements. More comparisons still have to be done…
26
Analytic Field Analytic Field ServiceService
27
Analytic Field Service
Speed tests forseen with tracks Several scenarios :
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
It has been implemented in Det/Magnet, but is not available in CVS yet.
However, according to preliminary tests, the speed of the B value access seems to be an issue…
1) Analytic Parameterization is faster or of the order of time of the interpolation method :
Best scenario : faster access and more accurate B values .
2) Analytic Parameterization is slower than the interpolation :
Only the Residual Parameterization could be used with the interpolation method to give more accurate B values .
A new file with B values used by the interpolation could begenerated with the help of the Analytic Parameterization .
28
Conclusions and Perspectives
Residual parameterizations for Bx and Bz.
Extrapolate the Residual parameterizations to regions where no measurements are available.
Parameterization based on MC simulation data has been succesfully performed in the acceptance region by the One-go 3D Fitting Method and reaches a rel. Prec. < 10-3 for the By component in all the 3 regions.
MC parameterization of the « Out of acceptance » Magnet and Upstream regions.
On-going / to do :
May 21, 2007 Monday Seminar, EPFL
Géraldine Conti
Speed tests of the Analytic Service with tracks.
Parameterization of the By residuals gives the expected more accurate B values with respect to the measurements.