NASA Technical Paper 1838
A'Microprocessor-Based Table
Lookup Approach . for Magnetic Bearing Linearization
. -.
Nelson J. Groom and, James B. Miller .-
MAY 1981
NASA .. TECH LIBRARY KAFB, NM
NASA Technical Paper 1838
A Microprocessor-BasedTable Lookup Approach for. Magnetic BearingLinearization
Nelson J. Groom and James B. Miller LarrgleyResearch Ceuter Hatnpton,Virgilria
National Aeronautics and Space Administration
Scientific and Technical Information Branch
1981
SUMMARY
An approach forproduci ng a linear transfer chlar acteristic between force command and force output of a magnetic bearing actuator without flux biasing is pres.ented. The approach is mlcroprocessor based and uses a table 'lookup to generatedrive signals for the magnetiCbearing power driver. An experimental test setup used to demonstrate the feasibility of the approach is described, and testresults are presented. The testsetup contains bearing elements simi- lar to those used in a laboratory model annular momentum control device (AMCD) .
INTRODUCTION ,.
This paper describes an approach for producing a lineartransfer charac- teristic between the force command and force output of a magneticbearing actuator. The approach, which is microprocessor based and uses a table lookup togenerate drive signals for the magneticbearing actuator power driver, was investigated for application to a laboratory model annular momentum control device (AKD) . The laboratory model (described in ref. 1 ) was built to inves- tigate potential problem areas in implementing the AMCD concept and is being used aspart ofan AKD hardware technology development program. The basic AMCD concept is that of a rotatingannular rim,suspended by a minimum of three magneticbearing suspension stations and driven by a noncontactingelec- tromagneticspin motor. A detaileddiscussion of therationale for the AMCD configuration andof some of its potential applications is presented.in reference 2.
As described in reference 1 , themagnetic-bearing linearization technique used in the original laboratory model AMCD magnetic.suspension was to differen- tially control sets of magneticbearing elements about a permanent-magnet bias flux. Preliminarytests indicated that this approach (permanent-magnet flux biasing)presented a problem from a control system standpoint(ref. 3). For a given equivalent permanent-magnet stiffness, a minimum-bearing servo band- width is requiredfor stability. The existence of structural modes in thearea of thelaboratory model bearingservo crossover restricted the amountof bear- ingservo damping thatcould be achieved.
Because of the limitations of permanent-magnet fluxbiasing encountered with the laboratory modelAKD, a decision wasmade to explore an alternate approach tothe design of the magneticsuspension system (ref. 4). As a result, a new magneticsuspension system for the laboratory model has been designed, fabricated, and tested(ref. 5). Thenew system uses a zerobias flux approach forthe magneticbearing actuators. Analog multiplier and squareroot modules produce a direct solution to the ideal magnetic actuator force equation to provide a linearforce-current characteristic. (For furtherdiscussion of mag- neticbearing actuator control approaches, see ref. 6.) The accuracy of this approach is limited by theaccuracy with which theideal force equation approxi- mates the actual characteristics of theactuator and by theaccuracy of the analog components. This paper presents a zerobias flux linearization approach which is digital and which uses a lookup table constructed from measured actua- torcharacteristics. An experimental testsetup that was used to develop this approach is described, and test resultsare presented.
s YMBOLS
Dimensional quantitiesare presented in both SI Units and U.S. Customary Units. Measurementswere made in U.S. Customary Units.
force produced by bottanelectromagnet
force command for magneticbearing actuator
value of force command associated with mth line segment
force produced by topelectromagnet
microcanputersystem sample rate
= Fc(m+l ) - Fc(m) for all m, where m < N
= Fc - Fc (m) , where Fc (m) 6 Fc < Fc (m+l )
displacement of suspended element with respect to centered position in magnetic bearing actuator gaps
gap ofbottom electromagnet
magneticbearing actuator gap with suspendedelement centered
gapof top element
current in bottom electromagnet
current command for magneticbearing actuator
stored value of current command associated with mth line segment
current in topelectromagnet
electromagnetconstant
number of line segments
slope of mth line segment from (Fc (m) ,Ic (m) ) to (Fc (m+l ) ,Ic (m+l ) )
Abbreviations:
AD analog-to-digitalconverter
2 AMCDannular momentum controldevice
D/A digital-to-analogconverter dc directcurrent dc em f electromotiveemf force
APPROACH
Magnetic Bearing Control Approach
The magnetic bearing control approachis one which uses zero bias flux. Figure 1, a schematic representation of a magnetic bearing element ispair,
Suspended element
Figure 1.- Magnetic bearing actuator. presented in order to describe this approach. Included in the figureare top and bottom electromagnets, with currents IT and IB, respectively; a portion of the suspended element whichis centered between the electromagnets; and a position sensor which measures the displacementG of the suspended element with respect to the centered positionGo. In the zero bias flux control approach, one electromagnet at a time is controlled; top the electromagnet is controlled for an upward force and the bottom electromagnetis controlled for
3 a downward force (sinceeach electromagnet can physically produceonly a uni- directional force). Figure 1 indicates that if up is taken as the positive direction, the electromagnet gapsare
and
Gg=Go+G (2)
Assuming negligible fringing and ignoring nonlinearcore effects, the force pro- duced by each electromagnet is given by (ref.4)
and
The composite force-current characteristic of a zero bias flux actuator with the suspended element centeredis shown in figure2.
Y
Uppermagnet charactsristic * L'T Current
Figure 2.- Composite force-current characteristic of a zero bias flux magnetic actuator.
4 . , Linearization.. Approach .. .. The linearization approach is microprocessor based anduses: a lookuptable. The most straightforwardapproach for producing a linear transfer characteristic betweenforce command and force output wouldbe to solveequations (3) and (4) for theelectromagnet current required to producethe desired force. That is,
and
where Fc is the commanded force.Implementing the solutions of equations (5) and (6) digitallycould produce a more camputationallyaccurate result than the analogsolution but would not account for deviationof the actual bearingsfrom the ideal model. Thissolution would also require thatconsiderable computing power be dedicated to a relatively minorportion of the total controlsystem.
The tablelookup approach, which employs a one-to-onecorrespondence between force command Fc andoutput current command IC permits theoutput/ inputfunction to conform,within the limits ofquantization error, to the actual bearingcharacteristics. However, this formof table requires con- siderable memory space.The minimummemory requirement is obtained by select- ing the minimum set of straightline segments which approximate the output/ inputrelationship within the desired tolerance. The major disadvantageof thischoice of line segments is thatconsiderable time may be spentsearching forthe appropriate line segment.
Since in this application computational time is more critical than memory size,an approximation was selected whichrequires more memory buteliminates thesearch time. The minimum set of equallyspaced line segments N are selec- ted so that N = 2", andthe N + 1 pointsdefining the line segments are sepa- rated by equal force command steps AF~. If therange of Fc is scaled to a B bit binary word (B > n) , the n most significant bits of, Fc uniquely identifythe N linesegments. In actual computation these bits identify'the
5 lookup table data associated with themth force commanddata point Fc(m) such that F, (m) 2 F, Fc (m+l ) . The B - n least significant bits of F, repre- sent the difference 6Fc between F, and F, (m) . (See fig. 3. )
line seqment m I
Figure 3.- Table lookup approach.
The data contained in the lookup table consistof the slope SLOPE(m) of each line segment m between (Fc(m) ,Ic (m)) and (Fc (m+l) ,Ic (m+l) ) and of the current command Ic(m) corresponding to F,(m). The location of the appropri- ate data is obtained directly from the force command word by masking and shift- ing. Calculation of output current command IC at G = 0 requires a single multiplication of 6Fc by SLOPE(m) and addition of this result to Ic(m). Since actual variation of electromagnet current gapwith was observed to be linear, as indicated by equations (5) and (6), calculation of IC for other values of G is accomplished by multiplying IC by the appropriate bearing gap. For a more detailed descriptionof the table lookup algorithm, see appendix A.
6 HARDWARe TESTS
Description of Test System
The test system consisted of a magnetic bearing test fixture connected to a microcomputersystem as shown infigure 4. Thissystem was used to obtain the data required to develop the lookup table and to obtain data on the perfor- mance ofthe proposed approach. A descriptionof the current driver shown in figure 4 is givenin appendix B.
Figure 4.- Magneticbearing test system.
Magneticbearing test fixture.- The magneticbearing test fixture shown in figure 5 consistsof a magneticbearing element pair; an equivalent "suspended"element, which is connectedthrough a pairof load cells to thebase of thefixture; and a pair ofposition sensors. Thesuspended element can be set to any desired vertical position in the magnetic bearing gap using the adjusting screws mountedon the load cells. The positionsensors are used to measure the bearinggaps. The magneticbearing elements have the same dimensions as the originalmagnetic bearing elements delivered with the laboratory model AMCD which is describedin reference 1. Two maindifferences exist between the test fixturebearing elements and the original laboratory model elements: (1) the core material ofthe test fixturebearing elements is SAE 1010 soft steel (as opposed to a lower loss silicon core iron used inthe original elements), and (2) the test fixturebearing elements contain no permanent-magnet material.
7 Figure 5.- Magnetic bearing test fixture.
The load cells, shown in figure 5, arestrain-gage bridge instrumented bending beams.The output of thebridge is a voltage which is directly pro- portionalto the load applied to the beam.The cells were connectedas shown in figure 6. They have a loadrange of 244.48 N (+lo lb) and a nominal scale factor of k0.045 mV/N-V (k0.2 mV/lb-V). Scalefactor and offsetdiffer from cell to cell and vary with changes in test fixture configuration, power supply voltage, and temperature;therefore, software was designed to provide for periodic system calibration.
Calibration was accmplished by applying a sequence of known loads to each cell and performing a first-order least-squares fit to the resulting data. Typical raw calibrationdata for one cellare given in figure 7.
8 A/ D converter Mi crocornputer
Figure 6.- Load cell connection diagram.
TEST LORD( L6 1 -1 -6 -S -q -3 -2 -I
Figure 7.- Typical calibration datafor a given load cell.
9 Microcomputer system.- Thesame microcomputer system was used for both characterization and performance tests. The major system components include (7) microcomputer and memory, (2) analoginterface unit, (3) cartridgedata recorder, and (4) portabledata terminal (fig. 4).
The microcomputer is an Intel’ 8080 microprocessor-based computer which includesserial and parallel ports, programmable hardware timers, and 2000 bytes ofrandom access memory (RAM). This processor was selectedfor the initial phase of developnentbecause it is wellsupported in hardware and software. Availablesoftware includes FORTRAN aswell as assemblylanguage and the manu- facturer‘shigh-level language.
Mostof the test programs were ‘written in FORTRAN with assemblylanguage hardware drivers. However, thetable lookup routine was written in assembly languagebecause of its time-criticalnature. A listing of thetable lookup routine is presented in appendix A. Memory expansionboards containing 24 000 bytes of RAM memory were added to permit run time storage of large FORTRAN programs and substantial test data.
The analog interface unit provides up to 32 multiplexed A/D inputchannels andtwo D/A outputchannels. The inputchannels provide the processor with position and force data from the test fixture position sensors and load cells and provideexternal command signals during those tests in which an analogforce command is used. Thetwo outputchannels are used to supplythe current command tothe bearing current driver. The analog interface unit wasprogrammed to per- form conversions in twomodes. During characterizationtests, conversions were initiated underprogram control, and a flag on the interface was set at end of conversion. During thoseportions of the performance test when thetable lookup algorithm was running in a real-timeenvironment, the interface wasprogrammed to start a conversionsequence on the rising edge of a hardware timer and to interruptthe processor at the endof conversion. This techniqueprovides a stable sampling rate and does notrequire software timing loops.
The datarecorder was used for temporary storage and fortransport of test programs and testresults. This recorderpermitted flexible use of the microcomputer system at a location which was remote from themicroprocessor development system used forsoftware generation.
The dataterminal provided run time parameter selection,control, and monitoring of the system test programs.
Description of Tests
Characterization tests.- Characterization tests were required to establish thevalues of lookup tabledata. These data were collected by the microproces- sor system and transferred to a programmable desk-top calculator for analysis
lUse ofnames of manufacturers in this report doesnot constitute an official endorsement of such manufacturers,either expressed or implied, by the National Aeronautics and Space Administration.
10 and reduction. A FORTRAN test program applied a 2048-pint sequence of equally spaced current drive signals to the magneticbearing test fixture and measured the actual force registered by the test fixture load cells at amanually selec- ted rim position. This process was repeatedfor a series ofrim positions. These positions were varied from -0.127 cm (-0.050 in.) to 0.127 cm (0.050 in.) in increments of 0.0127 cm (0.005 in.). The resultingforce vs currentdata were converted by linear interpolation to a2048-point current vs forcetable with equalforce steps. These data were systematically reduced to forman N-point minimum size lookup table (N = 2") capable of approximatingthe measured datato within k1 percent. This tolerance is within the limits of the existing analogsolution to the ideal force equation.
Performance tests.- Two tests were performed to evaluate the assembly language interpolation routine and the table lookuptechnique:
1. Linearity tests (static) - A sequence of force commands were applied to the test fixture throughthe linearization algorithm by a FORTRAN test pro- gram.The sequence of force commandswas applied in the same direction and over the same range as in thecharacterization tests. The force produced on the rim was measured by the load cells andcompared with theinput command.
2. Frequency response tests (dynamic) - A sinusoidal force-command signal was appliedto the test fixture through thelinearization algorithm, and the outputcurrent response was observed. The response measurements were performed by a frequencyresponse analyzer. Theforce-command signal was given a dc offset so that measurements could be performed independently on upperand lower bearingelements. For this portion of thetest, the characterization data were replaced by a linear table to permit direct comparison of inputforce command with currentoutput. This replacement was necessitated by a mechanicalresonance of the test fixture which preventeduse of load cells for frequencyresponse measurements. The frequency was measured overa bandwidth slightly largerthan halfthe sampling frequency of thealgorithm.
TEST RElSULTS AND DISCUSSION
Characterization Tests
The force,current, and gap relationshipsobtained frm thedata collected duringcharacterization tests are similar to those obtained from theideal electrmagnetforce equations (eqs. (3) and (4) 1. However, measured data deviated from idealrelationships near thezero force points, and differences between constantsfor upper and lower bearingelements were observed.Figure 8 is a plot of the actual force data resulting from the application of asequence of equally spaced currentdrive signals. Each curverepresents 2048 datapoints
I BERR I NG FORCE( N 1
Figure'8.-Force-current data taken from magneticbearing test fixture.
taken at a given gap setting. The force vs currentdata with equalcurrent steps were converted by linear interpolation to current vs force data with equal forcesteps. Analysis of theresulting data indicated that the relationship between current and gap for a givenforce was sufficiently linear to permit gap compensation to be accomplished by multiplying the output current by the gap (as in eqs. (5) and (6) ) . Systematicreduction of thedata for G = 0 resulted in a 128-pint table that was capable of reproducing the original data to within 0.5 percent,as illustrated in figure 9. System memory requirementsfor this size table are quite reasonable.
Performance Tests
Linearity.- The main results of the table lookup algorithm linearity tests are shown in figures 10 and 11 . Figure 10 shows the relation between force command input and the measured forceoutput of the system. This particularplot is for G = 0, but the same result was obtainedfor other values of G. The actualpercentage force error can be seen in figure 11. These results, although still within acceptable limits (less than 21 percenterror), do notagree com- pletely with theerrors predicted during data reduction (shown in fig. 9).
12 Figure 9.- Error associated with reductionof original data (for G = 0) to a 128-point table.
FORCE CIUtili 0 -3r-- -2 -I
k
-121 1, -12 -10 -E -6 -L1 -2 E 2 6 E IE I2 FORCE COMHRND( N)
Figure 10.- Static output/input characteristics of table lookup algorithm.
13 -I ' ' I I I I I I -12' -IE -E -6 -6' -2 B 2 'i 6 E IE 12 FORCECOMMRND(N 1
Figure 11 .- Percentage force error of table lookupalgorithm.
In particular, relatively large errors occur at the maximum positive and negative excursions of theforce command. Since no attempt wasmade to minimize the effects of hysteresis in thetest fixture, considerable residual magnetic flux can exist in theelectromagnet cores and in theequivalent "suspended"element. Variations in the magnetic history of the test fixture are the probable cause of theseerrors. One possible way to reduce theeffects of hysteresis on accuracy wouldbe to use flux feedback in the power driverloop instead of cur- rentfeedback. Another more obviousapproach wouldbe to uselow-hysteresis material in both theelectromagnet core and the suspendedelement magnetic cir- cuit material.
Frequency response.- The table lookup algorithmrequired approximately a 10-percentgreater execution time than the 1.85-ms worst-casevalue predicted by evaluatingthe microprocessor instruction execution times. This loss of timeoccurred during the test because the table lookupalgorithm instructions and data were stored in RAM memory on a memory expansionboard. The increased executiontime limited the system sample rate f, to a maximumof approxi- mately 490 Hz ratherthan the 540 Hz which was predicted.
Results of thefrequency response test are summarized in figure 12. The bandwidth exceeds 700 Hz, which was consideredadequate for this portion of the bearing system. The magnitude is flat from dc to approximately fs/4 and then rolls offgradually to approximately fs/2, which is thetheoretical frequency limit for a sampled data system.Since this algorithmretains no history of thesignals and has an almost fixedexecution time, the phase response varies linearly with frequency to approximately fJ2.
14 e
I on lL
Frequency. Hz
(a) Magnitude.
I80
150
IZO
10
60
1
8
I IO IO0 II
Frequency, Hz (b) Phase.
Figure 12.- Frequency response (magnitude and phase) of table lookup algorithm at 490-Hz repetition rate.
15 CONCLUDING REMARKS
A microprocessor based table lookupapproach for magneticbearing lineari- zation without flux biasing has been presented, and an experimental test setup used to demonstratethe feasibility of theconcept has been described.Results obtained with the experimental test setup generally showed very close agreement with theoreticalpredictions. Using a 128-point table,the table lookup algo- rithm produced a linear transfer characteristic between force command and force output of the test fixture magneticbearing actuator to within ?l-percenterror. The frequencyresponse of thealgorithm was greaterthan 100 Hz, which should be adequatefor this portion of thebearing system.
This approach when used as an innerloop for the actuators could form the basis€or an all-digital magneticsuspension control system. One application would be thelaboratory model annular momentum controldevice (AMCD). An all- digital system would allowcontrol-system parameter changes to bemade in soft- ware without requiring the circuit component changes and circuit rewiring which arenecessary with existing analogsystems. Also, advanced controllerdesign approachescould be more easily implemented.
Langley Research Center NationalAeronautics and Space Administration Hampton, VA 23665 March 1 9, 1 981
16 APPENDIX A
TABm LOOKUP ALGORITHM
Algorithm Description
Thisappendix presents a listing,in assembly language, of the table look- upalgorithm used to drivethe magnetic bearing test fixture. The algorithm performs a nonlineartransformation of an input force commandFc and a mea- a sured rim displacement G intocurrent commands IC to eitherthe top or bottom bearingelement. The transformation approximates the solutions of equa- tions (5) and(6) in the main text. Thelookup table contains128 four-byte 1 datagroups. Each of thegroups consists of a 16-bitvalue of SLOPE(m) and a 16-bitvalue of current Ic(m) obtained from measurementof the actual bear- ingforce-current characteristic with the rim centered (i.e., GT = $ = Go).
The routine is interruptdriven and is called upon completion of the force command A/D conversionwhich is initiated by a hardware real-time clock. After readingthe force command, the processor sets upthe analog interface and initi- ates conversion of the rim displacement. Because o€ theconfiguration of the analoginterface, the force commandFc and rim displacement G are 12-bit two's complementnumbers which are signextended to 16 bits.
The least significant 5 bits of Fc are removedand stored, sincethey containthe value 6Fc.The remaining 7 bitsidentify the 128 data groups. Thesebits are appropriatelyshifted, converted to offsetbinary, and added to thelookup-table base address to produce amemory pointer. The stored slope value is loadedusing this memory pointer,and the pointer is incremented.The slope value is thenmultiplied by6Fc (an8-bit by 16-bitmultiplication rou- tine is used to conserve time). Thisresult is added to thestored value of Ic(m), which is now addressed by the memory pointer. The result,Ic(Fc,Go), is thecurrent command required to produce a desiredforce by eitherbearing element when the rim is centered (GT = $ = Go). Sincethe required current is directlyproportional to thebearing gap, the current command forany rim displacement is givenby the followingrelationships:
where IC = f(Fc,Go).After substitution of the table lookupalgorithm for IC andthe expressions for GT and GB from equations(1) and (2) in the maintext, these equations became
17 APPENDIX A
IC(m) + SLOPE(m1
Ic(m) + SLOPE(m) 6Fc - ) rG1 ")
Since Go is a constant,these equations can be rewrittenas
The lookup table slope and currententries are predivided by Go toeliminate a real-timemultiplication by l/Go.
Because of thescaling of inputs and outputs(see table Al) and the need to perform all calculations on 16-bit or less integer numbers, thestored look- up tablevalues are scaled as shown in table A2.
TABLE A1 .- SCALINGOF INPUTS AND OUTPUTS
Variable Range Analog scale factor
FC k11 .1 2 N 0.9 v/N k2.5 lb 4 V/lb
IC k5 A 1 V/A
G k0.127 cm d78.7 v/cm t0.05 in. 200 V/in.
aThesevalues include the sum ofboth position sensors. %?hisvalue includes a softwaremultiplication by 2 (line 95).
18 APPENDIX A
TABLE A2.- SCALING OF TABLE VALUE
Variable Units
~
SLOPE (m) A/N
A/1 b
aThe (0.5) factor accounts for the difference between the total scale factors for Fc and IC. bThe (28) and (2' 6, factors are employedto maxi- mize the number of significant bits while maintaining intermediate and final results within a 16-bit integer format. These factors are removed by implied shifts during the two multiplications. CFactors (Go)(32.2) (1 03) and (Go)(5) (2l 4, are for Go in SI (cm) and U.S. Customary Units, respectively.
19 APPENDIX A Program Listing f6lW :Fl:LW. ARI pRINT(:Fl:LMIKIR. LST)OGJECT(:Fl:LOOMJP. OBJ)PAGELIIDTH(80) EJECT
Lcc OBJ LIE SOURCE STATEMENT
1;
3; 4; 5 us 6 ; SAVEMRIN PROGRRM REGISTERS 8 INTRU. ;SAVE A & PSY 9 ;SFWE BC 10 ;SFWE DE 11 ;MK 12 i LOADFORCE COMMANDAND i SETUP FOR GRP MEASUREMENT E 15 ;LW FORCE(N) 16 iLMwp c"EL E 17 ;SELECT W CHAHNEL E 18 ;STWT GftP CONvEEIffl 19 j GENERATEDELTA FORCE COMMAND j ANDTABLE ADDRESS OFFSET 22 ;CLEFR B 23 ;HIGHFORCE DATA > R 24 ;MOUT HIM 4 BITS 25 ;STORE DBll TWDB6 > H 26 ;LOU FORCE MTA) A 27 ;lWSK OUT HIGH 3 BITS 28 ;DELTA FORCE > C 29 ;HIVE LOUDATA > A (LOW 5 BITS tl ASKED) 38 ;CWIE DB11 TWCB8 UITH DB7 TMU DES
98lB 07 31 RLC I 981C 07 32 RLC ;pEFRRANGE BYTE BBlD 07 33 RLC ; TOFGlUi WlE 07 34 RLC j DBl0, DB9,DB8, D87, PB6, D85,8, 0 98lF 17 35 RAL i 88288 36 mw L,fl ;CWERT TO e821 3F 37 m ; OFFSET #22 17 38 RAC ; BIEFIRY 0023 E681 39 RH1 0lH ; Fm 8825 67 40 mW H,A , 41 42 j COMPUTEADDRESS OF SLOPECN) 8826 llw E 43 LXI DJTABLE ;LOAD Ti%LESTART ADDRESS 8629 23 44 IW H ;ADJUST 'TaE' ADDRESS YFilUE m23 45 IW H I 6828 19 46 IXlDD ;ADD TflBLE STmT TO OFFSET 47 48 49 i LORDSLOPE
20
J APPENDIX A
ISIS-IImm m RSSEWLER, v3. e MEPRGE 2
LOC CBJ LIR SWRCE STATErmT .. m5E 58 Iww Elfl ;LIYW)Locl BYTE SLWE(N) BBX, 23 51 IW H ;IHCRRlEHT POINTER FOR HIM BYTE BB2E 56 52 rn D,fl ;LOADHIGHBYTE SLOPE(N) BBZF 23 53 IW H ; INCREMENT WINTER FOR CllRRMT A DDRESS 54 55 i MULTIPLY SLOPECN) BY DELTA FORCE 56 i (SPECIAL 8BIT X 16BIT MULT. > 0038 E5 57 WH ;MADDRESS .. 0031 EB 58 XCHG ;W MILTIPLICWSLOPE) TO K .. 0032 noeee 59 SHLD TEW ;SAVE MLTIPLIrn IN TW 0035 2iEW 6e LXI H,E" ;LOAD CWXE CQMER MSS 0038 3689 61 WI~,e9~ ;Lm CYCLEMKlHTER 00m 11m 62 LXI D,W ;CLEAR TEMPORARY RESUT BBm 79 63 CW: mw FI,c 003 1F 64 RfR ;ROTATE WTIPLIER(DELTAF) 003 4F 65 rn C,A 8848 35 66 mfl COUHTER;DECREIQNTCYCLE 8841 cm 67 JZ FINI ;TEST FW WTIPLYCWLETE W4D24FM 68 JK SKIP ;.IMPIF HUTIPLIER BIT 0 8847 2AoBBB 69 LHCD TEW ;GET IILLTIPLICRN BB4R 19 76 DADD i AHDADD WBEB 71 XCHG ;SAVE PARTIFlC PRODOCT Be4c 2lCE60 n LXI H,B" iRELof#) 8N-M MSS BB4F 7i7 73 SKIP: mW hD 8856 1F 74 RH? ;ROTATE 06-51 57 75 mW D,A 8852 78 76 rx3v A,E i TEtPORARY 0853 1F 77 RAR 8854 5F 78 IWd E,A ; REWT 8855 c33m 79 .RIP LOOP i LOOP 8855 E1 88 FINI: POP H ; RESTORERDDRESS 81 82 j COMPUTE CURRENT COMMAND FROM 83 ; BASEICN) AND DELTA I 6859 4€ 84 rn c,n ,LORDLOCl BYTE BRSEI(N) 885A 23 85 INX H 8858 46 86 HOV BI# iLWD HIGH BYTE BAKI(N) mEB a7 XCHG ;WDELTA1 TO M me9 88 DfDB ;ADD BAKI(N) TO DELTA1 00x44 89 mW B,H ;WE CURRENT COW?@ W4D 98 m C,L ; TO EC 91 92 i CORRECT FOR BEARING GAP. VARIATION 8868- E 93 LDASTATUS ;CLEAR RTII288 EOC FLFwi 8863 2ABBBB E94 LKD WIN ;LOAD GFP w629 95 DfDH ,Dam GFR 8867 #e5 96 MI R5H ;LW FORCE CUMND CWNIEL 8869 328888 E 97 STR WXKR ;KLECT FORCE CWfW CHAHIEL m6C 111621 98 LXI D,GW;LOAD "IWL GAQ YFlCuE WBB ORA ;TEST FOR OF CURREHT .COlWi 99 B SIGH ,. ,_ D 8678 F5 168 PuYlpRI ;SAVE SIGH RAG 0071 F27F00 161 JP UPPER ;JW TO UPqER BEARIffi COIL 182 APPENDIX A
ISIS-I1 8688/8685 WlCRo RS-ER, V3.0 WEPAGE 3
LIHE YMlCE STATEWIT
103 j COMPUTE LOWER BEARING COIL GAP 194 LOKR: DAD D ;ADD W(T) TO KMIW GAP 185 DUB i 186 m A,C i TME 107 m i muTE 108 m C,A i 189 WT' hB i WllE ue rn icF ill nOV B,A ; CUWKNT 112 JW FIML i JW TO FIHAL CURRDlT CO"D C WTATIOH 113 114 115 j COMPUTE UPPER BEARING COIL GAP 116 IJPPER: nOV A, E ;SUBTRFICT 117 SllBL ; W!T) 118 w L,fl ;FROM 119 HOV AID i WIW 128 SBBH ; Gfp AT 121 w3v H,A i CENTER 122 123 j MULTIPLY GHP BY CURRENT COMMRND 124 FIWC: RllD TEW ;STM WTIPLICAN IN TEW 125 LXI H,BHUI ;STDRE 126 HVI 14,llH; i BIT COUHT 127 LXI D,W ; INITACIZE RESlLT 128 LOOPI: m A, B ;ROTATE 129 RAR i 130 #w B,R i MTIPLIER 131 mv A,C 132 RfR. ;RIGHT 133 mv C,A 134 DCR! ;DECREMMT BIT CWT 135 JZ FIN11 ;ME? TEN OUTPUT 136 JNC SKIB ;W IF M) CARRY FKM ROTATE 137 LHCD TEW ; OTHRHIK 138 DAOD ; ADD MTIPLICRH 139 XCHG RERLT 140 LXI H, BHun ;RESTORE BIT CURT POIHTER 141SKIPI: WV A,D ;ROTATE 142 m ; TW 143 MW D,A i RESUT 144 mW A,E ; RIGHT 145 rn i 146 KJV EJA 147 JW LoBi LOW;REPEAT 148 149 j CHECK FOR OVERCURRENT VALUE 150 j AND OUTPUT RESULT TO CORRECT PORT 151 FINIl: XCHG ;5" CURRMT TO HL 152 mW RH ;LCW CU?REHT 153 CPI 08H ;TEST CllRRRiT VALUE 154 JC OK ;JuR IF OK 155 LXI tt 07FFH ;CURREHT=WIM CURREHT 156 OK: wp pscl ;RESTORE CURRRlT SIN
22 APPENDIX A
1
ISIS-11 meme5 m RSSRBLER, v3. e MEPAGE 4
LOC OBJ LIE SURE STATEENT
157 JP WTPOS ;JUP IF POSITIVE E 158 WD CURNEG ;OUTPUT REWT TO LOMR BEfiRIffi D/R 159 JW REM ;RETURH E 168 WTWS: SHLD CURPOS ;WTPUT RESUT TO UPPER BEARING D/A 161 RETURN: PW H ;RESTORE HL 162 POP D ;RESTOPE DE 163 WPB ;RESTORE Bc 164 iI41 A,M ;LOAD END OF INTERLRT 165 WT eoAH ; WTPIJT EOIC 166 PC? PSM iw A & PW 167 RET .; RETlRN 168 169 178 171 17-2~~1:DU e 173 TRIP: DU e 174 GAP EQIJ 8478 175 Et&
EXTERNAC SmS WIN Em COHV EM MJXiXH? E W0 STATUS E W
USER SyneoLS WIN EMBHvl AWE C#NV Em CURNEGE0BW CURWS E FIHAL A 8685 FIN1 A €@S8 FIN11 A eeAE GAP A 2116 INTRU A 8888 Loop Rem Loop1 A8898 LMR A e874 IlJXRM E 88W OK A em WTrnS A m2 RETW A 8Bc5 RIP A WF SKIP1 A 88A5 STATUSE W4 TABLE E 848e TEW A BeDe UPPER A 487F
23 c .. APPENDIX B
MAGNETIC BEARING CURRENT DRIVER
A schematic diagram of the current driver used in themagnetic bearing test system is shown in figure B1. The driver is capable of supplying up to
5.8 kR I.;"
?+15 V
Figure B1 .- Magnetic bearing current driver.
5 A to each bearing coil and has a gain of 1 AD. Drive is provided by the LH 0002 currentamplifier at low current levels ( 24 APPENDIX B 201- ~~- IO Frequency. HZ (a) Magnitude. ! IO3 Frequency. Hz (b) Phase. Figure B2.- Frequency response (magnitudeand phase) of magnetic bearing current driver. 25