Spindle PID troubleshooting -- solved

Linuxcnc 2.8pre on mint.

Everything has been working fine for a while now. I started trying to fix my spindle speed (it's been running a little bit slower than programmed). Spindle is controlled from Mesa card via Analog 0-5V output.

I set the P=0.4, FF0=1.0, everything else 0.

See attached photo for halshow.

spindle-vel-fb-rps is 13.94723, and spindle-vel-cmd-rps is 16.66667
but, (see photo circled items), spindle actual rpm is 1293 when I commanded 1000.

What am I missing? Am I looking at the correct hal variables? Is there a hal variable to show the PID error?

Thanks.

Paul Tan.

Can your spindle do 4700 RPM at 5V input ?
Check the scaling so it corresponds to what it actually can do at 5V.

Yes, the spindle can do 5000 RPM. Scaling is not the issue. The spindle RPM is not linear to the output voltage (or, the output voltage might not be linear). I wanted to use the PID loop to adjust the output voltage accordingly. I.e., at low RPM, e.g. 500 RPM, I get actual RPM of about 420, at 1000 RPM, I get 950, at 2000 RPM I get 1950. At 4500 RPM, I get 4400 RPM. Not perfectly linear. I wanted to close that feedback loop.

What encoder is on the spindle and how is that scaled in hal ?

Encoder is built into the DMM DYN4 servo motor. I am scaling it based upon the pulley ratio. But, that's not the issue. My issue is why the PID error is wrong based upon what the system is seeing.

Hmmm... should I be asking this in the configuring linuxcnc section instead? or maybe in the HAL section? Asking the moderators here if it should be moved or not.

I can move it there, seems more appropriate .
Thank you.

First you need to make sure that the pid.s component is actually used. Check the hal file. If you attach it here people can check for you.

If that is all good, I would say to set P=0 and adjust FF0 so the actual spindle speed agrees well with the commanded speed. Judging from your experiments FF0 should be perhaps 1.05.

Then start to increase P until the error is as small as you want. You can go on with I and D. The error is in the pid.s.error pin.

Yes, you probably need to increase P, you have

P = 0.4

Which means if you are 100 RPM low the PID output only adds
the equivalent of 40 RPM to the commanded speed...

That's not the problem. The problem, if you look at the screenshot, is that it's overshooting the compensation, but the hal variables in the screenshot show that it's still below. Screenshot shows the spindle already at 1293 RPM, but the setting is set for 1000 RPM. This is with P=0.4. It should have attempted to decrease the spindle speed, but, if you look at the halshow output, it shows that spindle-vel-fb-rps is 13.94723, and spindle-vel-cmd-rps is 16.6667, so, it's behaving correctly if you look at the rps variables. BUT, the spindle-fb-rpm is 1315.024, and that is CLEARLY above the G code spindle speed of 1000 RPM.

I need someone to tell me what hal variables to look for to further debug this problem.

Thanks.

Paul Tan.