Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
+A COMPUTER-CONTROLLED, 3-D, MAGNETIC FIELD MAPPING SYSTEM
Charles G. DolsLawrence Berkeley Laboratory
University of CaliforniaBerkeley, California
Abstract
From January 1963 until July 1970the Magnetic Measurements Group at LBLsurveyed the magnetic fields of 104 magnets with a measuring system controlledby a special-purpose digital computer. l ,2
In July of 1970 the special-purposecomputer was replaced by a Digital Equipment Corporation PDP-5. The modifiedsystem has demonstrated flexibility, convenience, and reliability during theexecution of 13 varied magnet fieldmeasuring programs.
Since March 1972 the system has beenmeasuring three orthogonal components ofmagnetic field at each grid intersectionin each pass.
This paper discusses some of theadvantages of the general-purpose computer and describes recent developments ofthe system.
Introduction
The immediate motivation for remodeling the magnet mapping apparatusthat is referred to in the abstract andthe associated references was the needfor an effective system for making performance studies of developmental magnetsfor the Berkeley 200-GeV Accelerator.When the responsibility for that accelerator was transferred to the NAL, themagnet group at Berkeley reduced thepriority of the remodeling project. Thegroup agreed on a revised plan and I proceeded with it as a fill-in job.
Because I was inexperienced in electronic logic circuitry and machinelanguage programming, I chose to build ahardware RIM* loader as a first-stepdidactic exercise.* "Hardware RIM loader:" The push of abutton transfers from a ROM (read-onlymemory) 15 l2-bit instructions that willcause a TT (teletype) to read paper tapeinto the computer core. In RIM (read-inmode) leach l2-bit equivalent from thetape IS preceded by its l2-bit coredestination address.
791
It turned out to be a very effectiveexercise and I heartily recommend it (orthe equivalent) for inexperienced otherswho are faced with similar projects.And, as a part of the development of acomputer-controlled mapper control system, a hardware RIM loader is very valuable when it replaces time-consumingswitch-register programming for computerdead starts.
General
Physically, the mapper system separates into two parts: (1) the triplerack and teletype illustrated in thephotograph, Fig. 1, and (2) the probepositioning apparatus illustrated in thephotographs, Figs. 4, 5, 6, and 7.
The block diagram (Fig. 2) suggestsa different binary division, i.e. (1) thecomputer with its satellites (teletype,hardware RIM loader, D/A converter,oscilloscope, and software) and (2) theinterface with its many satellites,including the probe-positioning apparatus. I will list the blocks and discussthem more or less individually.
The Computer and Its Satellites
The Computer
Our computer is a Digital EquipmentCorporation (DEC) PDP- 5. Its core memoryholds 4096 l2-bit words. The cycle timeis 6 microseconds. The "5" is the pioneer model of DEC's PDP-8 family andwhile its original cost and its size areboth an order of magnitude more than thatof a 1972 PDP-8, it has almost exactlythe same architecture and instructionset. The significant software for 5'sand 8's is completely compatible or iscompatible with trivial modificationwith both 5's and 8's.
The 5 communicates with the outsideworld through a teletype (TT), the D/Aconverter, and through Input/Outputcable connectors.
A weakness of the 5 is that itsprinted-circuit discrete-component structure includes a huge n';lJllber of ~ging,
heat-sensitive, germanIum transIstors.However, we feel we have successfullycompensated for this weakness by an exacting 'maintenance policy which requiresreplacements, repairs and/or adjustmentssuch that the machine not only operatesover the full range of supply voltagesspecified-roT margin checking,by DEC butdoes so i~ an ambient temperature above80 0 F (26 0 C) with its cabinet ventilating fan stopped.
Our 5 has been reliable in spite ofthe jarring it frequently gets in movesof (say) a half of a mile on a fork-lift(it is designed to be moved by forklifts) .
Telet~pe. The teletype (TT) is a standard odel ASR 33. It inputs to thecomputer through the keyboard and thepaper tape reader. It generates punchedpaper tape output and types under controlof the PDP-5.
Fig. 1. The comput~r mapper triple-rack and teletype.
792
Search cOil s
Flux standard
High speedpaper tape
reader
PDP-5computer 1-4-- ....
HardwareRIM
load er
D/Aconverter
Stepping motors,
limit switches,
and reference
switches
Fig. 2. Computer mapper block diagram.
Rim Loader. The hardware RIM loader ismostly,TTL integrated circuits and consists of a 16-word ROM, (starting addressand IS instructions), gates, inverters,a clock, and the level-converters forconverting TTL to PDP-s logic voltagelevels. The loader acts on the computerthrough the SR (switch register), the"Load Addres s" switch, and the "Depos it"switch in a sequence similar to that usedby a human operator when, if ever, heenters the 15 instructions manually.
D/A Converter and CRO. The cathode raytube oscilloscope is not at present usedin our mapping procedures. However, whenwe load the versatile program "Focal-69"to do engineering computing, we sometimesuse Focal's capacity to generate graphical displays on the CRO.
We have also displayed positions in"Space War" and "Tic Tac Toe" on the eRO.
Software. Standard software for mappingconsists of a Rim Loader for the HighSpeed Tape Reader, a BIN* loader, andthe main mapper program, which is nowcalled "MAP72."
* .A "BIN" loader reads a "blnary format"paper tape. In general BIN loaders aremuch more versatile than RIM loaders.They read RIM format and also load successively ascending addresses in corewhen a first address is given. Oursreads either from the TT or the highspeed reader, displays the address-beingloaded on the accumulator lights, andchecks the program's longitudinal parityagainst that determined at the time ofassembly.
793
Our dead-start procedure illustratessome of the functions of the software.With each of the 12 Switch-Register (SR)switches on the PDP-5 console at zero,the push button of the Hardware RIMloader is pushed. A teletype (TT) RIMloader goes into core. A RIM-formattape of the RIM-loader for the HighSpeed Tape Reader is positioned in theTT (switched to START); a tape with BINand MAP72 (in series) is positioned inthe H.S. Reader; the PDP-S console STARTswitch is depressed.
The TT RIM loader reads the H.S.RIM which is self-starting and which, inturn, reads the (RIM format) BIN programwhich is (again in turn) self-startingand which reads the binary-format MAP72program. At this point, in less thantwo minutes, the (almost invariable) indication of zero parity errors gives onea comfortable feeling -- in some contrastto the all-too-frequent result of preHSTR (High Speed Tape Reader) loadingattempts. It was frustrating to see theaccumulator lights displaying a parityerror, after 15 minutes of T T clacking.
MAP72 accepts TT Keyboard assignments of parameter values, automaticallycalibrates, positions the coil assembly,and records parameters and integratoroutput magnitudes on magnetic tape. Theprogram responds to 30 different on~ ortwo-character TT commands. Figure 3 isan example of TT output produced bythree of them. TP-CR- (-CR- = carriagereturn, the standard terminator) initiated typing of the "Title" and theParameter list. M-CR- accepted andechoed messages; -CR- before the end ofa message line puts that lineon the magnetic tape. CC-CR- resets anintegrator and (with I-second delays ateach step) reads the DVM, reverses theSLFS excitation, reads twice, restoresthe SLFS exci ta tion,*and reads again.The four readings are written on tape,averaged,and reported on the TT.
* See "Flux Standard" p. 7.
DATE 72/09/08AEC 6 DIGITSTITLE DEMONSTRATION LIST OF TI TLE AND PARAMETERS.MAP 33.t1.0 TPE 222.0 REC .7MOD 1.0 TME 501.8 MXD 1.0XRF .0 YRF 69.0 ZRF .0XMN -15.0 XMX 15.0 XDL J .0YMN 9.0 YMX 119.0 YDL 1.0ZMN -6.0 ZMX 6.0 ZDL 1.0CMN .0 CMX 2000.0 CDL 200.0CRF 1600.0 SHR 25.0 CMP .7XCR 10023.0 VCR 998.t1.0 ZCR 10000.0CTL .5 RAT 52.0 NPT 16.8BCI 16273.0XPN 10.0 YPN 9.0 ZPN 4.0CPN 1602.2 POT 3201 t. 0 rno .2CTRL/ S M
7 THI SIS AN EXAMPLE OF A "MESSAGE". ER 78 THI SIS ANOTHER. ER 89 THE FOLLOWING IS THE TELETYPE PRINTED RECORD OF A CALIBRATION. ER9
10CTRL/S CC -.2 -1012.4 -1012.5 -.i11010 ER 10 -1012.2 MAP 33.t1.0CTRL/S M
ANOTHER CALIBRATION WITH MORE RESOLUTION. ER11 HERE IS12
CTRL/S CC1010 ER 121010 ER 13CTRL/S TP
-.2-10161.4-1016.t1.0 -.tI.6-10160.3 -.2-10161.5-1016i1.0-10160.3 MAP 334.0
- 4.7
11
Fig. 3. An example of TT printing.
794
The major devices controlled by theprogram are very slow relative to thecomputer central processor: TT, 0.1 second per character; positioning motors,0.005 second per step; magnetic taperecorder 0.005 second per character;and DVM, 0.1 second per reading. Although the motors are not, in general,allowed to move while the DVM is counting, the use of program-interrupt allowsall of these devices to operate simultaneously. Thus,e.g., the teletype cantype positions and integrator outputsand record data on tape without delayingthe move, read; move, read; etc.sequence.
Map~er programs are written in theLBL verSIon of PAL, DEC's program assembly language, and assembled by LBL's CDC6600/7600 complex. This flexible assembly program was introduced into the computer center library by Robert Belshe,who is responsible for many of itsvaluable features.
The input to the assembly programconsists of: the basic mapper programwhich is stored at the computer centeron magnetic tape and/or LBL's PSS (Program Storage System); and program modifications punched into IBM cards. Theoutput consists of: the modified program on Mylar-reinforced paper tape, inbinary format and a "print-out. I? Theprint-out contains: a printed list ofcommand ,variable, and constant mnemonicswith programming comments; an errorlisting; a complete index of internalprogramming references; an error summary;and a core map (a display of the memoryallocations) .
A basic measuring sequence called a"Y-run" consists of a series of measurements of the three field components -Bx, By, and Bz, at even intervals ("YDL")along the Y-tube.
A Y-run begins and ends at Y ="YMN" (usually with the coils inside amu-metal shield to establish a zerofield reference). Measurements of thethree components are made as the coilassembly stops momentarily at each Yposition in steps of "YDL" to and including Y = "YMx. rr The direction of theY motion is then reversed and each position is remeasured as the coil assemblyreturns to Y = YMN.
A typical magnet mapping projectincludes:
1) Regulating magnet current toa specified value.
795
2) Ordering the computer to generate and record y-run data at eachX,Z grid intersection in the specifiedrange of X and Z.
3) Ordering the generation of X andZ "tie-in" runs.
4) Repeating 1,2, 3 at othervalues of magnet current.
5) Generating tie-in magnetizationruns by making Y-runs at fixed X and Zand varying magnet current.
Each tape record includes a listof parameter values. Calibrations areperformed frequently and automatically.
The Interface and Its Satellites
Interface. The interface interconnectsthe computer with the:
1. High Speed Photoelectric PaperTape Reader
2. Calendar and Clock3. Multiplex4. X, Y, and Z Integrators5. Flux Standard6. Digital Voltmeter7. Tape Recorder8. Probe-Positioning Apparatus.
The interface includes level converters to accommodate the PDP-5 logiclevels (-3, 0 volts) to the TTL circuitry (0, +3.6 volts) of the interface.The TTL integrated circuits of theinterface consist of a large number ofDigital Equipment Corporation M-seriesFlip Chip modules. Control connectionsto external stepping-motors and switchesare made through noise-insensitive Kseries modules. Switch positions (forexample) are interrogated through RCintegrator filters with a time constantof 16 ms.
M-series modules are used for gates,inverters, flip flops, "pseudo deviceselectors," bus drivers, buffer registers, etc. K-series boards includegates, flip flops, DC drivers, etc.
High Speed Tape Reader. Our high speedphotoelectrIc paper tape reader (HSTR)is a rather ancient Ferranti, model 196.I got two, this one and another, for spareparts when they were "excessed" fromtheir last job. This model was originally rated to read 200 characters persecond in start-stop mode. The ones Igot had been modified to read 400 characters per second. Not too unexpectedly,operation in start-stop mode at 400characters per second was marginal.
I replaced the 3600-rpm synchronous capstan drive motor with a l200-rpm motor,slowed the reel drives correspondingly,and we now have a reliable, 160 characters per second HSTR.
Calendar and Clock. The calendar andclock display date and time as month,day, hour (1-24), minutes, and 0.1minutes. The computer interrogates themfrequently, reports their positions,upon request, on the TT as e.g., DATEand TME in Figure 3, and (every record)on magnetic tape. Automatically recorded dates and times are valuable intracing operator errors and equipmentmalfunctions.
Multiplex. The Multiplex permits computer or manual control of connectionsof the SLFS into the input circuits ofthe integrators and connects the DVM input to a choice of shunt or integrator.Connections are made through shielded,guarded relays. Single point references are maintained by switching all input connections (e.g., Dymec High, Low,and Guard are switched).
Integrators. The three electronic integrators ln the mapper are LBL MOD-7l.The MOD-7l design reflects many years ofexperience with and development of integrators at Berkeley. This design incorporates careful guarding, isolation, andsimplification of the summing junction.We use Teflon dielectric capacitors toachieve low soakage and low temperaturecoefficient of capacitance.
An indication of the integratordrift is part of each measuring sequence.The first and last measurement of theintegrator output voltage in a Y-run (seeabove) is usually made with the searchcoil inside a mu-metal shield where themagnetic field is negligible. The closure (the difference between the firstand last readings) is used to adjust eachpoint in the measuring sequencetusing theassumption that the drift is linear withtime.
Table 1 contains integrator driftdata from a typical recent magnet measuring project with RC = 0.1 secondR ~ 51,000 ohms, C ~ 2 ~F, and NA,searchcoil area, ~ 4500 cm2 . The closureof 9 measuring sequences is listed alongwith the average closure and the deviation from the average of three consecutive measuring sequences of 2.5 minutesduration each. These samples were chosen completely at random. No integratoradjustments were made in thIS period.When the closure error is averaged overa measuring sequence, integrator driftcontributes less than 0.01% to measurement error. Note that these drift datawere taken when the apparatus was inservice in LBL's Building 6, the 184"synchrocyclotron bui lding. The electrical, thermal, and vibration "noise" levelwas high.
Fig. 4. The ena of the Y-tube 8howing the 3-D coil assembly,th~ precision rack, and the positioning mechanism.
796
Table I
Time: 72 June 21, 14081.45 730 101.65 650 720 -701.85 790 70
When the DVM is reading the outputvoltage of one of the electronic integrators, we have two integrators in seriesand the DVM output is conveniently insensitive to random spikes of electricalnoise.
Di~ital Voltmeter. We are using a DymecMo el 240lB integrating digital voltmeter (DVM). The nominal accuracy is0.01% and the O.l-volt range when usedwith a counting time of 1.0 second givesmagnet-current shunt voltage readingswith I-microvolt resolution.
The computer mapper has brought tous major gains in flexibility and reliability. Its flexibility permits us toseriously consider plans for expansionsof the system in many directions.
Stepping motors drive all three axes,but only the X and Z positions are closelyconstrained by the motor positions. One2-pitch screw is used for each X-driveand two 5-pitch screws for each Z-stand.The accuracy of X and Z positioning isdetermined by the precision of the screwsand the straightness of the guides.
The accuracy of Y-positioning isdetermined by the precision of the rackand by the epicyclic mechanism describedin Reference 2 and illustrated in thephotograph, Fig. 4. The cable take-upreel illustrated in the photograph,Fig. 7, is equipped with a "negator"spring to maintain constant tension onthe cable and a safety switch that interrupts the Y-motor operation if the cablefails to wind evenly.
Discussion
Flexibility. Ever since the first computer mapper job in July 1970 (with theold system standing by) every jobhas justified and/or required modifications, extensions, or improvements in themapper program. We have found that"patching" through the CTRL/s mode, ED(Examine, Deposit, or Dump) often couldbe done in a few minutes. And we couldgenerate an updated tape in one day.
are generated at each writing head whenit is pulsed. Thus, for example, magnetic defects in the tape could easilypass the echo check. However, we haveencountered very few defective tapes.
Probe-Positioning Apparatus. A rigidassembly of three cOlIs wlth orthogonalaxes (photograph, Fig. 4) is positioned ina cartesian coordinate system by two"Z-stands," two "X-drives," and the"Y - tube" (Photographs, Figs. 5 and 6) .
The stainless steel and aluminumZ-stands (Photograph, Fig. 5) move theX-drives simultaneously vertically inincrements ("ZDL") that are multiples of0.1 inch. The X-drives simultaneouslymove the Y-tube horizontally in increments of "XDL". The coil assembly ismoved reproducibly along the Y-tube inprecise increments.
5010
-70
9060
-150
FROMAVERAGE AVERAGECLOSURE CLOSURE(VVOLTS) (VVOLTS)
Drift Data forMod. 71 No. 7
DEVIATION
CLOSURE(llVOLTS)
72 June 19, 133925302500 24402290
72 June 17, 1142370330 320250
Time:7.607.607.60
Time:1. 451. 651. 85
A Sample ofIntegrator
Y-RUNMAXIMUMEXCURSION(VOLTS)
Flux Standard. The flux standard is aSLFS (square loop flux-standard -pronounced: slewfuss). A temperaturecontrolled core of ribbon-wound Deltamaxis normally saturated stably in somedirection. When the excitation is reversed, an accurately reproducible fluxchange links the output winding. SeeReference 3. The SLFS is in series witha search-coil and an integrator input.
Tape Recorder. The tape recorder is aKennedy Model 1400 with an Echo-checkoption. Writing is start-stop; speed is200 6-bit words per second; density is200 words per inch. It accommodates8-1/2 in. IBM-standard reels of magnetictape. It generates horizontal and vertical parity marks and reports the resultsof its own determination of horizontalparity. The echo check does not represent a full loop-closure; the "echoes"
797
Fig. S. The Z-stand supporting the X-drive and the V-tube.
798
Fig. 6. The probe-positioning apparatus in place for a job.
Fig. 7. The Y-tube motor-end assembly.
799
Reliability. The old mapper sufferedfrom some of the same weaknesses that thePDP-S has, e.g., aging, heat-sensitive,germanium transistors. However, theearlier mapper did not have the compensations provided by: -ri) thorough documentation, (b) improvements developedduring mass-production, (c) maintenancetechnicians working more-or-less continuously with the systems.
Plans. On a time scale of about a year,we plan to incorporate a faster DVM andMultiplex to permit detailed measurementof pulsed fields. We also plan to get acompatible l2-bit computer with 8K ofmemory. The larger memory will permitmore on-line data processing and errorchecking.
Acknowledgments
F. W. Macondray made the basic design decisions that determined the character of the computer mapper system, e.g.,the choice of K and M series modules.He sketched schematic wiring diagrams ofmost of the circuits of the interface andthe basic peripherals. I do not believethat I would have assembled this systemif I had not had Fred's major, basiccontribution.
M. L. Clinnick programmed valuablegeneral-purpose routines, e.g. dataformatting, input/output, and the examinedeposit-dump routines.
T. D. Robinson has made importantcontributions. He suggested, designed,built, and installed the Y-tube cabletake-up drum safety-switch describedabove. He has demonstrated craftsmanshipof the first order in machining precisionparts for the probe-positioning apparatus.
D. H. Nelson and P. G. Watson, whoeach had extensive experience with theearlier mapper systems, guided the development of the computer mapper with manysignificant observations and suggestions.
Paul Salz has been the key man inintegrator development.
I especially wish to acknowledgethe valuable assistance of KaarenDickman, J. G. Mathios, and D. H. Nelsonin preparing this report.
t This work was done under the auspices of theU.S. Atomic Energy Commission. It is beingissued as report LBL-1311.
800
1.
2.
3.
References
P. G. Watson, Brief Description ofOperation and Performance Specifications of the LRL Rapid MagneticField Mapping System, Universityof California Lawrence RadiationLaboratory Report UCID-195l(August 1963).
F. W. Macondray, The Rapid Mapper:A Data Acquisition System for LargeScale Magnetic Field Measurements,Proceedings of the Second Internatlonal Conference on MagnetTechnology, Oxford, 1967, p. 646.
F. w. Macondray, The "Square Loop"Flux Standard: A Precision PulseGenerator, Proceedings of theSecond Internatlonal Conference onMagnet Technology, Oxford, 1967~p. 639.