Set up using AMC AB15A100 drives, brushed DC motor Prototrak Plus with Encoders

What is the current setting of the SW1-4 switch? It should be in the off position.

Todd, thanks! Attached is a pic with ch 5 "hm2_5i25.0.7i77.0.1.analogout0" at same units 200m/div as all others, except ch 2 "pid.x.error" units are 1m/div.

Many thanks!

Todd, Yes, all SW1 settings are OFF.

Thanks!

Todd, thanks! This attached trace looks the best today. The channel 2 (teal color) units are 1m/div; channels 1,3 & 4 are 200m/div & channel 5 is 500 m/div. Channels 3 & 5 are very close trace to each other, if the units were set the same. Channel 2 "pid.x.error" will go above (as shown) the zero/datum line, but can be made to go below zero/datum line. What is channel 2 showing & I need this to flatten out or follow the other traces?? The speed of the feed appears to look slow for 60"/min, what do I look at or how to calculate the feed rate from Hal Oscope?
Naturally, helpful suggestions are welcomed.
Many thanks!

Edit: This attached trace has high P=135, with the "pid.x.error" below the "zero" line (the trace has a different "zero" point than the others.
 

You want to minimize the error (channel 2). Perfect would be for the error to be a flat line at zero, but that is impossible. There will always be peaks, valleys and caused by disturbances from inertia, cutting forces, friction... Your goal is to minimize them and reduce oscillations from servo dithering. You want the average of the error to be zero, so some below and some above.

In a system with not enough P, the oscillations will become shorter both in amplitude and frequency as P is increased. But after a certain point the system will become unstable and adding more P will begin to increase the amplitude of the oscillations, while the frequency will continue to become shorter.

Adding D can improve system dampening and settling time, but at some point adding more D will also cause instability.

Todd, Many, many thanks!! This trace looks good, ch2 units 500u/div; heavy oscillation on the "pid.x.error" (consider the units, maybe not so bad), 1/2 below zero!! How close is this??

Many thanks!

In the 2nd trace you added to your previous post you will notice that there is a significant negative error during the cruse. That is because you increased the FF1 from near 2 to 3. (Remember how I said to leave FF1 alone once you find the right setting.)

In this last trace you certainly have the best error in a trace you've posted yet, with peaks of about 0.001 at the jerk points.

The jerk points will always be the location of the greatest error, because Linuxcnc's motion planner has no jerk limiting and makes acceleration changes with infinite jerk. Jerk is the 3rd order derivative of the position command and is the rate of change of acceleration. (It is one of Linuxcnc's biggest warts.) It is a difficult problem to fix involving some very high level math, but I digress.

See if you can use a little D to improve the settling time. Reducing and or eliminating the wavy lines durring the cruse phase. If you can use D to smooth that out, great. It probably won't do much for the large spike of error at the jerk points though, but more P might make them smaller.

PS. Change the gain of the orange and green traces in Halscope so they don't go off of the screen (the most interesting part of those two traces are not visible.)

Todd, many thanks!! The FF1 was used to lower the ch 2 (pid.x.error) closer to the datum. I used "P" until it would go no lower, then used FF1 to lower it more. Should I re-do this part? I need to leave FF1 at what value?
Attached are some traces in .zip folder (1.5 MB (1,504,709 bytes)) that have "P" and "D" adjusted, if folder loads you will see some change. I am looking at the channel 2 trace?
The units in traces are ch1 & 4=200m/div; ch2 + 500u/div; ch 3 & 5 = 500m/div.
Many thanks!

Tuning D is not easy. there is often only a small window of what works. and outside of that you may not see much change, and too much and too little often may not look all that different. But when you find that sweet spot it is almost magic.

I'd say that your P=370 tests looked the best with D=.4 and .04 a bit of a toss, up.

What happens with a little more D with P=370? Try .6, .8, 1.0. If that looks like it is going the wrong direction, try .3, .2, .1.

Todd, Thank you!! When you talk "D" & the sweet spot, how do you recognize it, as you are going up in the values?
I have many traces, the attached pics are in the range you asked about. If you continue higher "D" the very start of ch2 "trace" gets a larger & larger spike as you go up in "D". I could use just a clue to be shore I got the correct idea. The ch2, needs to be split on the "zero" datum; ch4 (encoder velocity) needs to follow (lay on top of) the red trace, exactly, all traces, w/o oscillation. Is there other traces I need to watch?
Many thanks.

Units are the same as last group.