AMC drive tuning

I just swapped out my old SD drives with 4 AMC BE25A20H drives running in brush mode with tach feedback.
I first tried all cards individually with 1 axis to confirm dip sw and pot settings. Adjusted so there was no drift and squeal. After final wiring all motor settings appeared incorrect. Motors are having a slight drift and squeal. Is it normal to see a change in those settings as more drives are added and more power is demanded from the supply? Interestingly, if I try to lock by hand one of the slowly drifting motor shafts, all motors seem to increase their rotation. Is this normal?
I assume this goes away after the balance is readjusted. How will this affect tuning if I add a 4th axis down the road?
Each drive has its dc fed directly from the large cap.

Control input is hooked up to the 7i77 +-10v, Axis is on, drives belts removed from the motors and system responds/corrects properly to table movement and it's linear scales.

Is there an easy way to output 0v on the 7i77 to re-tune the balance on the drives in case the encoders are trying to actively change that output during testing? Or do I need to redo the balance with the 7i77 +/-10 disconnected? The drive manual suggest grounding both reference inputs during the initial setup. Can I simply pull the terminals on the 7i77 and ground on that end or does it have to be on the drive end?

Last, I'm having a senior moment with the axis PID calibration assistant under machine menu. I can't initiate movement to test. Arrow jog keys used?

OT-CNC wrote:

I just swapped out my old SD drives with 4 AMC BE25A20H drives running in brush mode with tach feedback.
I first tried all cards individually with 1 axis to confirm dip sw and pot settings. Adjusted so there was no drift and squeal. After final wiring all motor settings appeared incorrect. Motors are having a slight drift and squeal. Is it normal to see a change in those settings as more drives are added and more power is demanded from the supply?

Yes, analog drives with no posiioning loop will drift, and the drift will change due to temperature, load, currents flowing in the entire system, etc. Don't worry over a slow drift until you have the encoders going into LinuxCNC and the positioning loop running.

Also, you want the drives to be disabled anytime LinuxCNC is not in the "machine ON" state, so that they are never left in that drifting condition.

Jon

Jon,

Thanks for the explanation.
The belts are back on and I'm trying to tune with linuxcnc now. I'm running into some issues. First movement is rough. I went from P50 down to P30 and I still get oscillations after traveling a longer distance at 20 ipm. Short moves seem fine. I'm following tommylight's servo 'tuning how to. What's really odd is that ccw direction for both x and y is relatively smooth but cw is rougher. Manually moving the table both directions are equally smooth when powered off. What is causing this? Previous setup worked fine. When powered up, trying to rotate the handwheel, I can feel the it fight back more in one direction than the other. Do I need to re-adjust the balance pot on the drives? Joint.0.f-error on the scope was almost a third greater in the rough cw direction.
Is it possible that the drive's internal gain it too high? I backed off one turn from squealing at rest, no load on motor.

OT-CNC wrote:

Jon,

Thanks for the explanation.
The belts are back on and I'm trying to tune with linuxcnc now. I'm running into some issues. First movement is rough. I went from P50 down to P30 and I still get oscillations after traveling a longer distance at 20 ipm.

The velocity servo drive with a tachometer provides much of the gain, so very low P gain should work fine.
First, set P quite low (2-5) and adjust FF1 for minimal error during moves. Once you get the FF1 giving good velocity
tracking, then you can turn P back up, but stay low enough so things are stable. Apply a little bit of D to improve stabiltity.

Jon

Something is way off. I took a quick video of the cw/ccw mismatch. When I power the drives to test without any +-10v control input I have a locked wheel that feels symmetrical in response to me pushing/pulling on the handwheel as expected. The movement is smooth also when applying a +/- 1.5v battery voltage to drive control input (not shown).
When I connect to linuxcnc, it's rougher in one direction than the other. When I push on the wheel, it fights back violently, when pulling on the wheel it's soft. You can see that mid video. Would a poorly tuned system not be symmetrical?
I'm puzzled as I must have had beginners luck on the 1st try with the old drives. Every thing just worked. Is there some incompatibility with these drives and the 7i77?

1. Try running with just FF1 (that is,P,I,D =0. This will eliminate feedback device issues)
beware, that this will drift slowly when stopped
2. Can you post your hal/ini files?

I tried running with PID with all 0 and varying the FF1. I would get joint following errors. Bumping the P up past 9 would go into instability. This is after turning off net x-vel-cmd => pid.x.command-deriv
With it on P could go as high as 35.
I need to get back to basics here, 1st, PCW, can you tell me if I missed something about the 7i77 compatibility with the drive input specs? Here is from your manual and attached is the drive AMC manual I have the positive and ground correct. The shielding drain however, is going from the drive end to a central grounding bar. Not attached at the 7i77. This is how I ran it before with no issues on the old drives. Can you tell from the drive schematics if it reads differential inputs? I have no additional grounding on the drive other than chassis ground. EDIT (I just re-read the connector info and it states differential input. So that should be correct)

The 7I77 provides six channels of +-10V analog servo interface on connector TB5.
Minimum load resistance is 2K Ohms. Normally analog servo drives use differential inputs
to avoid ground loops. Suggested wiring is shielded twisted pair with7I77 AOUTN to drive
AIN+, 7I77 GNDN (adjacent to AOUTN) to drive AIN- as the twisted pair and 7I77 GNDN
connected to the shield at the 7I77 end only. The drives common or GND signal should be
connected to the 7I77 power GND with a separate wire

Second, I need a refresher on when to use the output_scale numbers in the ini. I'm trying to work with a 200ipm max and around 125v dc drive. Or should I be leaving the output scale set at 1?
I'll upload ini and hal after I get back to the shop.

1. Yes, the drive has differential inputs so 7I77 AOUT --> REF+, 7I77 GND --> REF-
2. You probably need to loosen the following error limits to tune the drives properly
FF1 is the primary tuning term for velocity mode drives. I suggest that you don't connect anything to the PIDs command derivative pin (comment out any line that includes this pin)

The ideal way to set output scale is in machine units/second at 10V. This means FF1
= 1.0 and all PID parameters are in actual engineering units.

Thank you PCW.
I will turn off all net x-vel-cmd => pid.x.command-deriv in the hal, set output_scale to 1, relax following errors and try again with FF1.

Why would the control loop not be symmetrical even in a poorly tuned system? Is there an explanation to that? Jogging left vrs right wouldn't you expect the errors to be the same?

My guess would be an encoder issue, thats why its important to test with FF1 only

(and if you do know the machine velocity (in IPS) at 10V is better to use this as the output scale values so FF1==1.00)