Throughput Testing Results
Contents
1. Throughput Testing Results 13 Application Note PMAC Setup Factory default V1 15a 3 AXIS Splined E C T and Phase Freq Streamlined Throughput bps Master Clock Freq Mhz Servo Freq Khz Internal External Binary External ASCII 20 2 26 1294 1211 546 20 4 52 877 817 347 20 9 04 0 0 0 40 2 26 2241 2241 1827 40 4 52 3644 3404 1530 40 9 04 2522 2342 1002 60 2 26 2247 2241 2241 60 4 52 4497 4497 2571 60 9 04 4913 4489 2055 PMAC Setup Factory default V1 15a 4 AXIS Splined E C T and Phase Freq Streamlined Throughput bps Master Clock Freq Mhz Servo Freq Khz Internal External Binary External ASCII 20 2 26 937 877 410 20 4 52 498 466 198 20 9 04 0 0 0 40 2 26 2250 2243 1429 40 4 52 2451 2433 1147 40 9 04 1485 1389 610 60 2 26 2246 2248 2241 60 4 52 4384 4198 2015 IPMAC Setup Factory default V1 15a 6 AXIS Splined E C T and Phase Freq Streamlined Throughput bps Master Clock Freq Mhz Servo Freq Khz Internal External Binary External ASCII 20 2 26 488 458 226 20 4 52 0 0 0 20 9 04 0 0 0 40 2 26 2001 1900 963 40 4 52 1435 1365 676 40 9 04 0 0 0 60 2 26 2242 2243 1569 60 4 52 2740 2569 1281 60 9 04 1635 1535 744 14 Throughput Testing Results Application Note 15 A Factory Default 2 Axis Program Master Clock Fr2. 8 Throughput Testing Results Application Note External ASCII Procedure This procedure checks the rate at which a PMAC program may be downloaded to the PMAC internal rotary buffer when moves are transferred thru the DPR in ASCII form Enhanced throughput may be attained using interrupt based handshaking Here the DPR fixed background buffer takes some time in background to update 1 Configure 116 and I17 116 10 117 115 2 Enable DPR Background stuff 10 11 149 1 Enable DPR background update 1159 1 1st motor cs only Define a rotary buffer internal to PMAC DEF ROT1500 Buffer size is dependant on number I17 and number of axis A conservative value buf_size num_axis lines_ahead 1 5 Make M Variable definition to point in unused DPR m1023 gt y D000 15 1 This serves as the end of program flag Make a far pointer to same location pmac_all_done unsigned int MK_FP 0xD400 0 Set pmac_all_done flag to 0 pmac_all_done 0 Open the rotary buffer OPEN ROT Change feedrate override feedrate delay 2000 Must put this in so feedrate override takes effect Begin execution of rotary program b1000r Begin download of program inc SPLINE1 TA01 m1023 0 For PC Timer Begin downloading NUM_LINES program lines timeout 0 run_time_error_flag FALSE timel hrt_read top while DProtBufFull id 1 AND lines_down lt NUM_LINES if number_of_axis 2 X1 else if numbe
3. E33 and the master clock frequency I10 holds the length of a servo interrupt cycle scaled so that 8 388 608 equals one millisecond Since I10 has a maximum value of 8 388 607 the servo interrupt cycle time should always be less than a millisecond unless making the basic unit of time on PMAC something other than a millisecond To have a servo sample time greater than one millisecond the sampling Imay be slowed in software with variable Ix60 Frequency can be checked on J4 pins 21 and 22 Note If E40 E43 are set up so that the card has a software address other than 0 the servo clock signal must be received over the serial port from card 0 so these jumpers have no effect 12 Throughput Testing Results Application Note E29 E33 Phase Clock Frequency Control Jumpers E29 through E33 control the speed of the phase clock and indirectly the servo clock which is divided down from the phase clock see E3 E6 No more than one of these five jumpers may be on at a time E29 E30 E31 E32 E33 Phase Clock Frequency Default and Physical Layout 19 6608 MHz 29 4912 MHz Master Clock Master Clock ON OFF OFF OFF OFF 2 26 KHz 3 39 KHz E29 OFF ON OFF OFF OFF 4 52 KHz 6 78 KHz E30 OFF OFF ON OFF OFF 9 04 KHz 13 55 KHz E31 OFF OFF OFF ON OFF 18 07 KHz 27 10 KHz E32 OFF OFF OFF OFF ON 36 14 KHz 54 21 KHz E33 Note If jumper E98 has bee
4. 3300 3085 1400 40 9 04 1980 1840 780 60 2 26 2247 2241 2241 60 4 52 4476 4476 2460 60 9 04 4476 4170 1860 16 Throughput Testing Results Application Note PMAC Setup Factory default PROM V1 15a 4 AXIS Splined Throughput bps Master Clock Freq Mhz Servo Freq Khz Internal External Binary External ASCII 20 2 26 767 718 340 20 4 52 319 0 0 20 9 04 0 0 0 40 2 26 2250 2243 1355 40 4 52 2300 2190 1052 40 9 04 1770 1062 470 60 2 26 2248 2247 2195 60 4 52 4260 4018 1920 60 9 04 2990 2820 1300 PMAC Setup Factory default PROM V1 15a 6 AXIS Splined Throughput bps Master Clock Freq Mhz Servo Freq Khz Internal External Binary External ASCII 20 2 26 400 370 177 20 4 52 0 0 0 20 9 04 0 0 0 40 2 26 1890 1795 920 40 4 52 1330 1256 625 40 9 04 0 0 0 60 2 26 2240 2243 1540 60 4 52 2240 2450 1246 60 9 04 1490 1390 680 Appendix E PMAC Throughput Results Circular Moves IPMAC Setup Factory default V1 15a 2 AXIS Circle Moves E C T and Phase Freq Streamlined SIF 2 2Khz Master Clock Freq Mhz Internal External Binary External ASCII 20 MHz Master Clock 438 437 318 40 MHz Master Clock 1426 1387 1002 60 MHz Master Clock 2213 2109 1550 Note Servo Interrupt Frequencies greater than 2 2 kHz were not attempted because previous results have show
5. a condition more frequently encountered with the lower speed 20Mhz PMAC than the 40 60 Mhz PMAC A Run Time Error occurs when the time at which the calculations for the next move are completed occurs after the move should have began At every Real Time Interrupt PMAC does motion planning if it is required then PLCO and finally PLCCO If this does not occur often enough the just described condition occurs It is for this reason that I8 Run Time Interrupt period for all test was set to 0 This mandates that Real Time Interrupts RTI occur at every servo cycle No PLCO PLCCO were used in the testing therefore the only thing done at RTI was to do if motion planning if needed This is the optimum setting for I8 when trying to optimize throughput since no significant processor time is taken unless it is required for the trajectory calculations When PMAC s throughput has been restricted by the feedrate override increasing the servo interrupt frequency will increase throughput The maximum feedrate override value can be determined by the following equation Max_Feedrate_Override 225 Servo_Interrupt_Frequency 2 26 KHz This condition for example usually occurred for the faster 40 60 MHz PMAC boards with the default servo interrupt frequency of 2 2 KHz In this scenario PMAC has computational power to spare yet since the motion program can only have a feedrate override of 225 and the minimum move time is lms throughput is restricted to 22
6. was described above The NC for Windows application uses only the external DPR binary rotary buffer for transferring motion programs to PMAC The two modes in which this is done each line converted to binary on the fly ASCI gt Binary or preprocessed Binary gt Binary were done for both 1 and 2 coordinate systems as well as the three PMAC master clock frequencies 20 40 amp 60 MHz Furthermore two host computer speeds 486 33 and 486 66 were evaluated Since the PC is the bottleneck for throughput in nearly all cases except for the 486 66 20 MHz PMAC case the faster the PC the higher the throughput Conditions for Testing PMAC The encoder conversion table and phase frequency were streamlined I8 was set to 0 and two PLCs were running PLC1 and PLC2 A file containing 10 000 SPLINE mode moves each of Ims was used to conduct the test The internal rotary buffer size was set to half that of the DPR rotary buffer size More on buffer size is discussed below 10 Throughput Testing Results Application Note Host Only two applications were running at the time of testing the NC for Windows Executive program and CNC EXE Obviously the more applications running the less time NC for Windows has to transfer the motion program The default rotary buffer sizes set by dprBinRot1 Size and dprBinRot2Size that were used are shown below Number of Coordinate DPR Rotary buffer sizes PMAC Internal Rotary Buffer Size
7. 20 9 04 0 0 0 40 2 26 2241 2243 1390 40 4 52 2539 2373 1112 40 9 04 1485 1389 610 60 2 26 2247 2248 2241 60 4 52 4410 4137 1946 60 9 04 3379 3137 1450 PMAC Setup Factory default V1 15A 6 AXIS Splined E C T Streamlined Throughput bps Master Clock Freq Mhz Servo Freq Khz Internal External Binary External ASCII 20 2 26 428 408 199 20 4 52 0 0 0 20 9 04 0 0 0 40 2 26 1930 1826 930 40 4 52 1380 1320 655 40 9 04 0 0 0 60 2 26 2247 2247 1540 60 4 52 2679 2521 1270 60 9 04 1635 1535 744 Appendix D PMAC Throughput Results No Optimizations The following results are from a PMAC prom version 1 15A which was not changed from factory default settings with the exception of enabling motors setting I8 0 PMAC Setup Factory default PROM V1 15a 2 AXIS Splined Max Throughput bps Master Clock Freq Mhz Servo Freq Khz Internal External Binary External ASCII 20 2 26 1623 1515 606 20 4 52 1086 997 367 20 9 04 0 0 0 40 2 26 2241 2241 2164 40 4 52 4476 4476 1837 40 9 04 3449 3167 1201 60 2 26 2240 2234 2243 60 4 52 4497 4478 3107 60 9 04 4495 4477 2452 PMAC Setup Factory default PROM V1 15a 3 AXIS Splined Throughput bps Master Clock Freq Mhz Servo Freq Khz Internal External Binary External ASCII 20 2 26 1096 1032 466 20 4 52 598 555 219 20 9 04 0 0 0 40 2 26 2241 2241 1711 40 4 52
8. 50 blocks per second How to Increase Throughput Before reading the conclusions deduced from the data it is suggested that the reader review PMAC s computational priorities see Computational Features in the PMAC User Manual User Guide Once a good understanding of how PMAC performs its multitasking is achieved the conclusions below will be better understood The following conclusion can be made from the data In order of most significant impact a higher throughput can always be attained by Optimization Throughput Increase Increasing the Master clock frequency 100 800 Decreasing the number the axis per block 200 700 Pre process the ASCII file to a binary one 80 100 Removing any unused entries from the Encoder Conversion Table 10 30 Using an internal buffer next fastest external binary slowest 10 15 external ASCII Setting the Phase frequency equal to the servo frequency 5 10 Change uncompiled PLCs to compiled PLCs Floating point operations run 2 3 time faster and fixed point operations run 20 30 times faster See Appendix A for detailed information on making selected optimizations Trends The following trends have been found Here we discuss the effects on throughput as a result of varying one parameter as all others remain fixed trajectory calculations Run time error occurs 2 Throughput Testing Results Application Note Increasing Servo Freque
9. Application Note PMAC Throughput Testing Results Research has been conducted to determine the rate at which PMAC can process a motion program This document is intended to give a PMAC user a general idea as to how fast a motion program can be processed based on the data attained and provide the steps that were used to determine this rate In addition this document explains the various boundaries that are reached when pushing PMAC to its computational limit and has suggestions to further increase the throughput A section devoted to PMAC NC for Windows is also included to explain throughput results attained for this application Throughput Fundamentals The rate at which PMAC processes a motion program is specified in units of blocks per second bps A block is considered to be one line in the motion program regardless of how many axis are in the line For example X10 X10Y20 X10Y20Z2 TAI etc are all considered one block A bit of categorizing is in order As far as PMAC is concerned motion programs are of two types external and internal Internal programs reside in a standard program buffer within PMAC s memory External programs programs generated and transferred by a host computer ultimately end up in one of PMAC s internal rotary buffers either in an ASCII format or binary binary only possible if using the Dual Ported Rams DPR Binary Rotary Buffer Furthermore external binary programs may be transferred from host to PMAC i
10. IN 2CS BIN gt BIN 1CS BIN gt BIN 2CS 20 MHz Master Clock 300 120 1100 350 40 MHz Master Clock 300 120 1100 350 60 MHz Master Clock 300 120 1100 350 Co re ee me oe ee ea e jarnar m GT Fhedeleag ap llimiae wo Soon add ned ies BB eo vate carar Clock iig L oo He marar Cock ide E so MH Moofer Clock 1700 1d lagga 1009 Tai ghpeut hoci Per beca ASCH gt BIN ICE ASM BI SS BOH BIR IC BIK SBIM JCE eh PMAC Setup Factory default V1 15F 2 AXIS Spline Moves E C T and Phase Freq Streamlined STF 2 Khz 426 66 PL Cs Master Clock Freq Mhz ASCH gt BIN ICS ASCII gt BIN 2CS BIN gt BIN 1CS BIN gt BIN 2CS 20 MHz Master Clk 1000 280 1400 550 40 MHz Master Clk 1000 310 2000 700 60 MHz Master Clk 1000 310 2000 700 18 Throughput Testing Results
11. MaxBlkPerDisp exists in the SYS INI initialization file it should be set to as large as the DPR rotary buffer size This parameter is used when converting RS274 to binary The CNC EXE application will convert and download until the DPR buffer has been filled or after MaxBlkPerDisp blocks have been processed MaxFeedOvrd MaxFeedOvrd restricts the maximum feedrate override value A default value of 200 restricts the user to a maximum of 2000 blocks per second Increase this value to 10 times the maximum number of blocks per second you wish to achieve Throughput Testing Results il Application Note Appendix A Making Optimizations Removing Unused Encoder Conversion Table Entries Removing unused entries in the encoder conversion table ECT and changing the phase frequency to the servo frequency are easily done The ECT has nine entries in it by default It assumes that there are eight motors with eight encoders as feedback and the last entry will use encoder number 4 as a TIMEBASE As an example to optimizing the ECT only using N motors not eight and not using the TIMEBASE entry To remove the unused entries in the ECT write a value of 0 to the ECT base address 1824 dec plus N in the Y address space If N 4 then this would be accomplished with the on line PMAC command WY 1828 0 The phase frequency can be set to the servo frequency by changing hardware jumpers E3 E6 and E 29 E33 For example to have a phase and servo fr
12. cation over the bus and DPR were established PMAC for all tests was configured by sending it the following command strings SHHee Card set to factory default state 158 1 Communication to ASCH DPR established 110 xxx Set to appropriate value a function of servo_frequency 1187 1 Set default Accel time 1188 0 Set the default S curve time 18 0 Set Real Time Interrupt frequency equal to the servo frequency amp 1 Setup motors and coordinate system 1 gt 1x Define all axis being used for the test 2 gt 1Ly And so on depending on the number of axis x y Z U V W i100 1 Activate all motors being used for a given test i200 1 And so on depending on the number of axis i111 0 So that motors don t really have to be hooked up i211 0 And so on depending on the number of axis i125 12c000 So that limits and amp faults do not have to be grounded in hardware i225 12c004_ And so on depending on the number of axis 1j Close the loop on all motors active for the given test 2j And so on depending on the number of axis Now the details of each testing method are described in pseudo code The nomenclature for the pseudo code is shown below Typeface Meaning Bold Strings sent to PMAC Bold Italic Pseudo Code Internal ASCII Procedure The following procedure determines how fast PMAC can run a program stored in a s
13. eq 40 Mhz ECT and Phase Frequency Minimized 9000 8000 7000 6000 5000 4000 3000 2000 Appendix C Throughput Data Encoder Conversion Table Minimized Internal Extemal Binary Method External Ascii IPMAC Setup Factory default V1 15a 2 AXIS E C T Streamlined Throughput bps Master Clock Freq Mhz _ Servo Freq Khz Internal External Binary External ASCII 20 2 26 1784 1674 695 20 4 52 1405 1310 516 20 9 04 0 0 0 40 2 26 2248 2241 2241 40 4 52 4480 4490 2066 40 9 04 4356 4039 1626 60 2 26 2248 2250 2247 60 4 52 4497 4480 3355 60 9 04 7696 6958 2988 IPMAC Setup Factory default V1 15a 3 AXIS Splined E C T Streamlined Throughput bps Master Clock Freq Mhz Servo Freq Khz Internal External Binary External ASCII 20 2 26 1196 1096 506 20 4 52 79 734 312 20 9 04 0 0 0 40 2 26 2241 2241 1787 40 4 52 3554 3311 1499 40 9 04 2522 2342 1002 60 2 26 2247 2247 2247 60 4 52 4497 4497 2522 60 9 04 4913 4489 2055 Throughput Testing Results 15 Application Note PMAC Setup Factory default V1 15A 4 AXIS Splined E C T Streamlined Throughput bps Master Clock Freq Mhz Servo Freq Khz Internal External Binary External ASCII 20 2 26 847 797 367 20 4 52 438 418 168
14. equency of 2 2 KHz place jumpers on E3 through E6 and put a jumper on E29 leaving E30 E33 without a jumper PMAC I Variable 110 specifies time between servo interrupts will not need to be changed from its default value in this case since 2 2KHz is the factory default servo frequency Additional info on these jumpers is given below E3 E6 Servo Clock Frequency Control The servo clock which determines how often the servo loop is closed is derived from the phase clock see E98 E29 E33 through a divide by N counter Jumpers E3 through E6 control this dividing function E3 E4 E5 E6 Servo Clock Phase Default and Physical Layout Clock Divided by N E3 E4 E5 E6 ON ON ON ON N divided by 1 OFF ON ON ON N divided by 2 ON OFF ON ON N divided by 3 OFF OFF ON ON N divided by 4 Only E5 and E6 ON 20MHz ON OFF ON ON N divided by 5 OFF ON OFF ON N divided by 6 Only E4 and E6 ON 30MHz ON OFF OFF ON N divided by 7 OFF OFF OFF ON N divided by 8 ON ON ON OFF N divided by 9 OFF ON ON OFF N divided by 10 ON OFF ON OFF N divided by 11 OFF OFF ON OFF N divided by 12 ON ON OFF OFF N divided by 13 OFF ON OFF OFF N divided by 14 ON OFF OFF OFF N divided by 15 OFF OFF OFF OFF N divided by 16 Note The setting of I Variable 110 should be adjusted to match the servo interrupt cycle time set by E98 E3 E6 E29
15. es a or Mt a Tesi Pearse a hieralian vale es terabon_waluec TE 24 sari of program Feedrale eedrabe Heralion valued Horeiion_valua inaran valum g hersien weaker a Throughput Testing Results 3 Application Note All necessary conditions for a successful test run are listed and then described below 1 Calculated run time was within 1 of the actual run time 2 Arun time error RTI did not occur 3 The watchdog timer did not trip 4 A servo interrupt error did not occur The first mode of failure extended actual run time occurs in both the external program types The time that the program should take to execute in milliseconds is calculated by the following formula run_time_calc NUM_LINES 2 100 0 feedrate 2 extra TA added in spline If PMAC takes 1 longer to run the program than the calculated value a failure is considered to have occurred Here the bottleneck is the rate at which PMAC can transfer data from the host not enough background time or if the host computer is slow the rate at which the host can transfer moves to PMAC In both cases the PMAC rotary buffer at times may not have any moves in it and extra time will be spent decelerating to stop possibly waiting for the next move and finally accelerating to speed A run time error occurs when the time at which the calculations for a move have completed is after the move should have started When this happens PMAC will ha
16. n changed to connect pins 2 3 default is 1 2 the phase clock frequency is exactly 1 2 that shown in the above table Note If E40 E43 are set so that the card has a software address other than 0 the phase clock signal must be received over the serial port from card 0 so these jumpers have no effect Appendix B Throughput Data Encoder Conversion Table and Phase Frequency Minimized Note The external binary tests for appendices B E were done in the conversion on the fly mode That is the ASCII program file was not preprocessed before the test was conducted rather each line of ASCII in the file was converted and downloaded as the motion program was running Preprocessing the file can improve throughput significantly as seen in the NC for Windows tests up to 100 when the host computer is the bottleneck The test results shown in appendices B E were done on a 486 66 PC with programs in DOS With this combination the host computer was never the slowest link in the process PMAC Setup Factory default V1 15a 2 AXIS Splined E C T and Phase Freq Streamlined Throughput bps Master Clock Freq Mhz Servo Freq Khz Internal External Binary External ASCII 20 2 26 1957 1826 768 20 4 52 1495 1395 555 20 9 04 0 0 0 40 2 26 2248 2241 2241 40 4 52 4496 4496 2100 40 9 04 4356 4039 1626 60 2 26 2248 2250 2247 60 4 52 4497 4496 3375 60 9 04 7796 6958 2988
17. n that this would not improve the throughput Increasing the servo interrupt frequency only enhances throughput when the bottleneck is the feedrate override In the 20 MHz case decreasing the servo interrupt frequency by one half will likely increase throughput although the servo performance will decline Throughput Testing Results 17 Application Note 1 15A 2 Axis Circular 113 TA TM Ims E C T and Phase Freq Minimized Fpp _ 20 MHE MORS Dlook oO 40 RHE Maila Clock oO 0 MH Miggie Clock maregi Extgrngl Binge Fecgengl Spell Melher Appendix F NC for Windows Throughput Results Spline Moves The NC for Windows application uses only the external DPR binary rotary buffer for transferring motion programs to PMAC The two modes in which this is done each line converted to binary on the fly ASCII gt Binary or preprocessed Binary gt Binary are shown for both 1 and 2 coordinate systems as well as the three PMAC master clock frequencies The difference in the two data sets below is only the speed of the host computer 486 33 and 486 66 Since the PC is the bottleneck for throughput in all cases except for the 486 66 20 MHz PMAC case the faster the PC the higher the throughput SIF 2 2Khz 486 33 2 PLCs PMAC Setup Factory default V1 15F 2 AXIS Spline Moves E C T and Phase Freq Streamlined Master Clock Freq Mhz ASCII gt BIN 1CS ASCII gt B
18. n two ways with or without preprocessing the ASCII program file That is when sending PMAC a program via the DPR binary rotary buffer the host may first convert the PMAC ASCII motion program file to a PMAC binary file and then send the binary file piece meal to the DPR Alternatively the host can convert and download each line of the ASCII file on the fly one line at a time All three forms internal and both external were tested under a variety of servo frequencies phase frequencies master clock frequencies number of axis in a single block and also with and without streamlining PMAC s encoder conversion table Appendices B D show data and bar charts going from the most optimal PMAC settings to the default factory settings respectively using DOS programs which used SPLINE mode Ims moves Appendix E has data and bar charts illustrating the maximum throughput for Circular moves with optimal PMAC settings Finally Appendix F has NC for Windows data and charts for splined 1ms moves Splined and circular moves were chosen since every move statement in the motion program is processed Whereas when PMAC is in Linear mode it may skip blocks depending on whether or not it has enough time to calculate the move As a result using Linear move mode would lead to erroneous values for the PMAC throughput PMAC Computational Limits What was found to limit the maximum throughput is discussed here Recognizing what is keeping PMAC from running faster i
19. ncy As one increases the servo interrupt frequency PMAC spends more time on 1 Sampling feedback devices converting data via the Encoder Conversion Table and computing the output based on the difference in the commanded Vs actual position 2 Foreground tasks a Motion planning b PLCs and or PLCCs none of which were used in this analysis and therefore less time becomes available for background tasks How the Data was attained The basic methodology in attaining the data is first discussed followed by exactly how PMAC was configured for the various tests that were conducted The data was attained by developing a DOS based application benchmrk c Available with ComLib For any given test the card was configured appropriately in both hardware servo_frequency and or Master clock frequency as well as software I variables Three functions were developed all dedicated testing routines for each of the three methods of running a PMAC program The main program would iteratively call one of these dedicated functions specifying the feedrate at which to run the PMAC program and the function would return the results of the test percent error when completed Depending on whether or not the last iteration ran successfully the feedrate would be varied according to the algorithm illustrated below Keep in mind the move times were set to 1 ms Bagin iealing Set laadeale 50 Peron est we S Yes Feedraie geceai
20. ng Results Application Note else timeout Buffer full or PMAC has a run time error while lines_down lt NUM_LINES if lines_down lt NUM_LINES if test ERR_DPR_BUFFERBUSY Report to screen that Buffer has had an error i e a bad command sent else Report that DPR rotary buffer has been filled if timeout lt THE_TIMEOUT Have not exceeded the timeout goto top else run_time_error_flag TRUE 12 Send the last line down test 1 timeout 0L while test lt 0 AND timeout lt THE_TIMEOUT test DPASCTStrToRot id m1023 1 For PC timer timeout J if timeout gt THE_TIMEOUT return 2 0 Program took to long for execution 13 Wait until program has completed running last statement assigns m1023 i 0 while pmac_all_done amp 0x8000 1 AND run_time_error_flag 0 AND i lt THE_TIMEOUT i time2 hrt_read 14 Close PMAC s rotary program buffer CLOSE 15 Timed out If yes a run time error occurred if run_time_error_flag i gt THE_TIMEOUT Display an error message return 1 0 16 Do necessary calculations run_time_pc time2 time1 1193 0 1000 0 bps lines_down 2 0 run_time_pc run_time_calc NUM_LINES 2 001 100 0 feedrate 2 for added TA times error run_time_pc run_time_calc run_time_calc 100 0 17 Display results Calculated run time run_time_calc Error error PMAC rotary program process rate bps 18 Return Error return error
21. ow i 0 A_TIMEOUT 100000 while pmac_all_done amp 0x8000 amp amp kbhit amp amp i lt A_LTIMEOUT i time2 hrt_read 9 Timed out if i gt A_TIMEOUT A run time error has occurred Exit procedure return failure condition return FALSE 10 Determine how long it program took to execute and whether or not a successful run run_time_pc time2 time1 1193 0 1000 0 PC CLOCK bps lines_down 3 0 run_time_pc run_time_calc NUM_LINES 2 0 001 100 0 feedrate in seconds percent_error run_time_pc run_time_calc run_time_calc 100 0 display run time calculated actual and error 11 Return error if error gt 1 return FALSE else return TRUE External Binary Procedure This procedure checks the rate at which a PMAC program may be downloaded to the DPR binary rotary buffer from ASCII code generated on the fly loading from a file has been found to be the same speed It has been found that a 66Mhz DX2 PC clone has been able to transfer program blocks from the PC to the DPR Rotary buffer at the following rates Axis Transfer Rate BPS 1 8681 2 6700 3 5900 4 4500 6 3500 6 Throughput Testing Results Application Note This is useful to know since if the BUS is not faster than PMAC s processing speed the host becomes the bottleneck in the transfer process PMAC program command processing rate Determined by dividing the number of block
22. r_of_axis 3 X1Y1 endif lines_down if lines_down lt NUM_LINES DPR rotary buffer has been filled Throughput Testing Results Application Note if DPsysRunTimeError id 1 A run time error has occurred return 1 0 i goto top while DProtBufFull id 1 Wait while the rotary buffer is full 12 Send the last line of the program M1023 1 For PC timer 13 Wait until program has completed running last statement assigns m1023 i 0 while pmac_all_done amp 0x8000 AND run_time_error_flag i time2 hrt_read 14 Close PMAC s rotary buffer CLOSE 15 How long did it take according to the PC run_time_pc time2 time1 1193 0 1000 0 bps lines_down 2 0 run_time_pc run_time_calc NUM_LINES 2 001 100 0 feedrate 2 for added TA times error run_time_pc run_time_calc run_time_calc 100 0 16 Display results Calculated run time run_time_calc Error error PMAC rotary program process rate bps 17 Return Error return error PMAC NC for Windows Throughput Testing Delta Tau s PMAC NC for Windows the first fully capable CNC controller providing an open architecture approach to machine tools was also tested for throughput See Appendix F This application as with any application that has to display data real time and respond to a multitasking system on a continual basis can be expected to have lower throughput than the dedicated non visual DOS based testing that
23. s Systems long integers 32 bit 48 bit words 1 2000 1000 2 1000 500 Timing of Program Execution The following PLC was downloaded to PMAC for timing how long the motion program took to execute 45 2 m1000 gt x 0 0 24 u m1001 gt x 0818 0 1 p1010 0 open plcl clear if m1001 if p1010 p1001 m1000 p1010 1 endif else 1 0 if p1010 1 p1002 m1000 p1010 0 p1003 p1002 p1001 000442 endif endif close i5 2 ENAPLC1 Optimizing SYS INI file Parameters Several parameters in the SYS INI file pertinent to throughput are now discussed dprBinRot1Size and dprBinRot2Size The rotary buffer sizes in these tests play a more important role in determining maximum throughput than in the DOS based throughput testing applications Why is this Once the PMAC rotary motion program execution has begun PMAC NC for Windows will update the display only when the DPR buffer has been filled or PMAC NC for Windows has downloaded MaxBlkPerDisp a sys ini parameter number of lines If the time it takes to update the display is longer than it takes PMAC to execute all the moves in the buffer then PMAC will bring all motors to a stop and wait for the next move which appears after the host is done updating the display If the condition just described occurs the run would be considered invalid since a good run is considered to have all moves completely blended with no dwells MaxBlkPerDisp If the parameter
24. s important in determining what actions if any may be taken to remove the bottleneck Keep in mind that none of the symptoms described below occurs the host computer may be the slowest link in the chain That is for external programs the rate at which the host can transfer the moves to PMAC is lower than the rate at which PMAC can process the moves For any given PMAC configuration the limit on the throughput may be due to one of following Symptom Throughput Bottleneck Remedies Watchdog trips card shuts down Not enough background time to keep the See Table 2 and the tail stops wagging watchdog timer happy A run time error occurs Program Not enough time to perform the motion Decrease I8 run status bit is TRUE Motors calculations in one real time interrupt Also see Table 2 move at the last commanded period Increase servo interrupt velocity run away The maximum feedrate override is frequency Program runs fine but can t increase restricted by the servo interrupt frequency feedrate See Appendix A for details on making optimizations Throughput Testing Results 1 Application Note A system that does not have enough background time will have the watch dog timer trip causing the card to shut down Background time is the time left over after higher priority tasks have been done i e Phasing Closing servo loops Motion planning PLCO PLCC0 etc Not having enough background time is
25. s sent by the difference in time from when the first block is sent and when the final block is sent The last program line is a m variable assignment This M Variable has been defined a m1023 gt y D000 15 1 a bit that is not being used in the DPR control panel area 1 10 11 Initialize and enable the PMAC DPR Binary rotary buffer if DProtBufInit id DPR_ROT_SIZE lt 0 Report an error and return 1 RETURN 1 DProtBuf id ON Define a rotary buffer internal to PMAC DEF ROT1500 Make M Variable definition to point in unused DPR this serves as the end of program flag m1023 gt y D000 15 1 Make a far pointer to same location pmac_all_done unsigned int MK_FP 0xD400 0 Set pmac_all_done flag to 0 pmac_all_done 0 pmac_all_done 0 Open PMAC s internal rotary buffer OPEN ROT Change the feedrate override value feedrate delay 2000 Must put this delay in so feedrate override takes effect before run Begin execution of rotary program b1000r Begin download of program clear inc SPLINE1 TAO1 m1023 0 For PC Timer Start timer timel hrt_read Begin downloading NUM_LINES program lines timeout 0 run_time_error_flag FALSE top do if number_of_axis 2 test DPASCTIStrToRot x1 else if number_of_axis 3 test DPASCIIStrToRot x1y1 J if test gt 0 Line sent to PMAC lines_down lines_down 1 timeout 0 Reset timeout Throughput Testi
26. tandard program buffer in ASCII form The routine first downloads the program begins execution and the timer and then waits for the last statement in the program to execute before stopping the timer The last statement in the routine sets a bit in DPR which is being polled 1 Make the M Variable definition to point in unused DPR m1023 gt y D000 15 1 This location is in DPR s control panel an unused location This serves as the end of program flag 2 Make a far pointer to same location pmac_all_done unsigned int MK_FP 0xD400 0 The code above is in C pmac_all_done points to PC segment D400 hex and offset 0 3 Set pmac_all_done flag to 0 pmac_all_done 0 4 Download program to PMAC Open prog1000 Clear INC SPLINE1 TA01 m1023 0 Set program_done flag to 0 do if number_of_axis 2 xlyl else if number_of_axis 3 xlylz1 Throughput Testing Results 5 Application Note endif lines_down lines_down 1 while test gt 0 AND lines_down lt NUM_LINES dwell0 So that Ms are synchronous m1023 1 Close 5 Change feedrate feedrate delay 2000 Must put this delay in so feedrate override takes effect before run 6 Begin execution of program and wait for acknowledge b1000r 7 Begin timer timel hrt_read 8 Wait until program has completed running or a timeout has occurred Recall the last statement is an m variable assignment which modifies the 15th bit of the word polled by the line bel
27. ually begins running the program and when the PC timer starts timing is considered negligible This is based on the fact that all tests used 3000 blocks of program code which resulted in a smallest run time of 670 ms In comparison the maximum response time of issuing a run command to PMAC is under 3 5 ms For both external programs the start time was considered to be just before the first program line was sent to a rotary buffer For all types a program was considered finished when the last line of code which was an m variable assignment statement occurred The M Variable was defined to point to DPR The host program would poll this memory location until it was assigned at which point the PC timer would stop Again the lag time between program termination and the stopping of the PC timer is considered negligible Timing could have also be done using PMAC s servo counter In this case program lines would have to be added to capture the servo counter at the start and end of the program The time would have been calculated as the difference between the ending and starting servo counter values times the servo interrupt period Actually this was done to compare timing techniques and it was found that the results consistently differed by less than 0 5 4 Throughput Testing Results Application Note Setting up PMAC The communication functions used in benchmrk c were from ComLib the Delta Tau PMAC C function communication library Communi
28. ve all motors continue to move at the last commanded velocity and the program never finishes In this case a time out counter in the testing algorithm determines that a run time error has occurred A watch dog timer will trip when insufficient background time was available for house keeping or the RTI is not occurring frequently enough won t be the cause of failure here or if the digital voltage goes below 4 5V This too is not a factor in the tests performed Servo errors will occur when not all the calculations necessary for a single servo cycle can be completed in a single servo period In this case the PMAC will report a servo interrupt error in the appropriate coordinate systems status word Nothing catastrophic occurs when this error is encountered PMAC will continue any unfinished servo calculations in the following servo interrupt The user will notice that the program will continue to run at exactly half the speed it should since every other servo cycle is now skipped Timing of Program Execution In all tests the means of timing was by use of a high resolution 50 microseconds PC based timer The timer was started at the beginning of program execution and stopped when the last line of code was executed For the internal ASCII program the beginning of program execution was considered to be the time immediately following the acknowledge character received by the host after sending PMAC a run command The lag time between when PMAC act
Download Pdf Manuals
Related Search