[LinuxCNC/linuxcnc Issue#444] PWM is not the same for M4 and M3

#1 – SebKuzminsky 于 2018-06-19

Yep, this looks like another bug.

StepConf configures the spindle pwmgen with a small positive offset, and the pwmgen component (incorrectly, i think) always adds this offset to the duty cycle. So negative values on the .value pin are closer to zero than you’d think, just like your video so clearly shows.

The pwmgen manpage claims to “add/subtract” to offset from the computed duty cycle, presumably following the sign of the .value pin. (That manpage also incorrectly lists many pwmgen pins as params, but we converted them to pins back in 2008…)

git log shows that pwmgen has always behaved in this buggy way, and it took a careful observer like @Masznotwor to spot the problem. This may mean that the pwmgen component isn’t widely used. Maybe most folks use Mesa hardware if they need pwm outputs?

#2 – SebKuzminsky 于 2018-06-19

Nope, I misread “underlined +” as the plus/minus symbol “±”. The pwmgen manpage does say that the offset is always added to the duty-cycle, whether the .value pin is positive or negative.

I think the pwmgen offset is intended to avoid the dead zone around 0% duty cycle, and for that reason i think the offset should be subtracted from the duty cycle when the input value is negative.

Am I missing some other use of the offset?

For now I think @Masznotwor (and anyone with a pwmgen-controlled spindle) would be best served by setting the offset to 0 and commanding a slightly higher S-word to spin the spindle up.

#3 – andypugh 于 2018-06-19

Offset could be used to balance an unbalanced device. I think that “offset” probably should be added regardless of the sign of value. The problem may be that stepgen is attempting to use it as a “deadband” instead.

The offset in PID is useful for balancing the weight of the head / table on a mill. I don’t know if there are applications where it would, instead, be appropriate to apply it to the PWMgen.

#4 – Masznotwor 于 2018-06-19

Do not be afraid that I know English poorly, I also understand what is going on in HAL and INI files, so I can be wrong:
I think that it should be done simply:
the component generates PWM and that’s it
the direction of rotation is not to influence PWM
And if it is something you need, it should rather be run as an option if it causes problems

I think that many people use one-way spindles (water spindle) and only use PWM CW

I had not noticed the error but the previous lathes had an encoder and it may have affected. I do not know, I just did not go down that low speed

here the problem seems to be solved, unfortunately I do not know how to apply it and whether it is a good solution
https://forum.linuxcnc.org/49-basic-configuration/28700-lpt-spindel-pwm-different-speed-with-m3-and-m4

I have such ingredients

loadrt pwmgen output_type=1

addf parport.0.read base-thread
addf stepgen.make-pulses base-thread
addf pwmgen.make-pulses base-thread
addf parport.0.write base-thread
addf parport.0.reset base-thread

addf stepgen.capture-position servo-thread
addf motion-command-handler servo-thread
addf motion-controller servo-thread
addf stepgen.update-freq servo-thread
addf pwmgen.update servo-thread

net spindle-cmd-rpm => pwmgen.0.value
net spindle-on <= motion.spindle-on => pwmgen.0.enable
net spindle-pwm <= pwmgen.0.pwm setp pwmgen.0.pwm-freq 100.0 setp pwmgen.0.scale 2984.21052632 setp pwmgen.0.offset 0.05 setp pwmgen.0.dither-pwm true net spindle-cmd-rpm <= motion.spindle-speed-out net spindle-cmd-rpm-abs <= motion.spindle-speed-out-abs net spindle-cmd-rps <= motion.spindle-speed-out-rps net spindle-cmd-rps-abs <= motion.spindle-speed-out-rps-abs net spindle-at-speed => motion.spindle-at-speed
net spindle-ccw <= motion.spindle-reverse I actually checked, you can not notice at higher revolutions because then the m3 / m4 differential is about 1V on the transducer