Ernst.Q Posted March 17, 2014 Posted March 17, 2014 I have a clipper drive unit and I am seeing strange peaks in the CPU resources graph. The phase frequency load (red bar) regularly jumps to up to four times the continous load. The peak values are depending on the frequency settings as described below. It peak seems to be related to the PWM frequency. The unit is running just by itself by 24PSU, no cables or busvoltage attached. If I do these tests with the same parameters on a Geobrick-LV with identical firmware version (but without Macro IC0) or with a stand-alone Clipper (just 5V PSU), I do not see this phenomenon, there are no clear peak loads observed. Unfortunatly, I cannot simply swap the Clipper boards. I100,4,100=1 ; I101,4,100=1 a) PWM=36kHz, phase=36kHz, servo=4.5kHz (I7000=817, I7001=1, I7002=7). Continuous phase load = 32%, peak load = 132% b) PWM=36kHz, phase=18kHz, servo=4.5kHz (I7000=817, I7001=3, I7002=3). Continuous phase load = 18%, peak load = 132% b) PWM=18kHz, phase=36kHz, servo=4.5kHz (I7000=1634, I7001=0, I7002=7). Continuous phase load = 32%, peak load = 66% Measurements are done without motors enabled.
Ernst.Q Posted March 18, 2014 Author Posted March 18, 2014 Using a modified version of the sample monitoring program in the SRM guide, I captured the maximum and minimum values of the timing registers. The variation of the actual phase time (X:$00037) was minimal, value between 119 and 122 tics (1 axis @ phase 36kHz) However, the hardware timer register at address X:$FFFF8C should always be about 1092 tics, but the minimum value observed was 270 tics (this is unexpected!). It actually was 270 in 0.16% of the samples. The value exactly explains the peak values on the CPU load calculation. I think the time measured indicates a relationship with the maxphase signal frequency.
curtwilson Posted March 18, 2014 Posted March 18, 2014 Do you have access to a scope? If so, can you see any glitches on the phase clock signal? (J5 pin 44)
Ernst.Q Posted March 19, 2014 Author Posted March 19, 2014 I checked the phase with the scope, I did not see any glitches. There was no variations in the frequency or pulse width. Looks pretty clean to me.
Ernst.Q Posted March 19, 2014 Author Posted March 19, 2014 I repeated the test with I7000=6527, I7001=7. Other parameters at default. sampling M70 (X:$FFFF8C,0,24) results in: - a correct value of about 34818 (phase period of 885 us) OR - a wrong value of about 2173 (0.15-0.3%). (period of 885/16) However, the value of M70 is never a value in between. If this was due to cross over, you would expect multiples of 2173 occurring, but they don't. Furthermore, the pulse trains of the phase on the scope are perfect, always the period of 885us.
Recommended Posts