Warner/Swasey QCM-6 lathe with Kflop/Kanalog

Thanks Tom, for your help. The spikes that I was referring to are the output section of the plot. They are also mirrored on the error plot as position error. As I increase the P gain, the frequency of the spikes increases, to the point that the axis motion is smoother and quieter (with the same move distance). With very low P gain (1.0 or less), the axis motion is noticeably jerky and noisy.

I will try the filter.

On the original machine drive to motor connection, there were rather large transformer looking devices labeled as inductors in line between drives and motors. Can you tell me what they are for? Are they or some form of them needed with the AMC drives? Could that have anything to do with the spikes I'm seeing in the output plot?

Sometimes with switching amplifiers inductors are added to increase the inductance to reduce current ripple. I think the old drives only switched at the line frequency of 120 Hz. The AMC drive switches at something like 20KHz so ripple would be much lower. I doubt if they would be needed.

I added the 500htz, Q 1.4 filter as you suggested. It changed the green output on the plot to a lesser intensity, but didn't change the frequency of the spikes.

Changing the move size changes the number of cycles. Appears to be between 4000 and 5000 counts per cycle, no matter the move size. That is, a 20,000 count move yields about 5 cycles and 40,000 yields about 10 cycles. One turn of the ballscrew is .200" or 16,000 counts. 80,000 counts = 1 inch.

Changing velocity does not seem to affect number of cycles, just spreads them out on the plot.

This AMC drive was used. According to the AMC spec sheet, it was shipped new set to current mode. When I got it, it was set to voltage mode. Am I to understand this means it is not using the tach? Should I try to switch it to current mode, or leave well enough alone? Thanks for your help!

It seems you might have some mechanical/physical issue that is occurring maybe 4 times per rev. Are the motors direct drive to the lead screw? Or is there gearing or a belt drive. If you rotate by hand can you feel anything? Possibly some issue with the motor or bushes. There might be 4 commutation cycles per rev.

Did you mean to say Velocity mode instead of Voltage mode? The output seems to correspond to Velocity. If the tach were not doing anything I would expect there to be more output at the beginning to accelerate and less or negative at the end of the move when decelerating.

I suppose the tach might have some commutation issue or other. You might disconnect it and use current/torque mode where the velocity is basically controlled by KFLOP instead of the drive and based on the encoder. Your encoder is quite high resolution so this should work quite well.

Motors are direct drive to ballscrew. I don't feel anything unusual when turning by hand. Motors have 6 brushes.

The AMC drive spec on sheet 1 lists Modes of Operation:
Current
Voltage
IR compensation
Velocity

Feedback supported: Tachometer (+/-60 VDC)

On sheet 5 it lists a Mode Selection Table:
Current
Voltage
IR Compensation
Tachometer

So, based on how the switches are set, it matches the table for voltage mode. I still get confused over current and velocity modes. Now they throw in voltage and IR compensation.

As a comparison, I connected this same drive to x axis. After initial tests, I set it up same as z axis. I get the same results as z axis. Same spikes, same position error. Position to Command plot follows great to me.

If you have the AMC manual please post it.

Voltage mode can act like a weak velocity mode. It you apply a voltage to an unloaded frictionless motor it will ramp up to a speed proportional to the applied voltage.

You didn't post a tuned plot. I'm guessing with the cyclic error still going on peak errors are probably +/-50 counts or +/-0.0006 inches moving at 120ipm. Not bad but I was hoping for better.

You might make a longer move so you can determine with more precision if it is exactly 4, 5, 6, or a non integer number of cycles per rev.

I just downloaded the 73 page manual from AMC here: https://www.a-m-c.com/support/reserved-discontinued/

Scroll down and choose the manual.

Select Hardware Downloads, then Classic Analog Drives HW Manual

I tried to download the AMC 30A8 datasheet, but it's too large. It's available on AMC website if you search.

Sorry. I meant upload.

I changed the drive over to Current mode per the Datasheet for the AMC 30A8. Disconnected the tach. Ran ground wires from the power supply chassis and from the AMC drive to the machine chassis ground block. Kflop, Kanalog, and Konnect boards are not grounded. I ran external 240VAC wiring to the drive power supply. Still have the 'spikes'.

The higher I raise the P gain, the lower the error, and the more frequent the spikes.

The following data and graphs refer to X axis, AMC 30A8 drive in Current Mode, no tach. Starting with a move size 0f 16,000, which is one ballscrew revolution, or .200" of axis movement. Low Pass 2nd Filter of 500Hz, Q1.4. Velocity = 40,000.

P=5, I=0.0001, D=100

More to come. Errands to do.

Changed P Gain to P=10

Changed P Gain to P=15

Changed Velocity from 40,000 to 20,000

Changed Move Size from 16,000 to 20,000

Changed Move Size from 20,000 to 30,000

Hi Charles,

I think that is classic servo instability. I wouldn't describe those as "spikes". Its like a poor driver trying to drive down the center of the road but making delayed corrections too severely and ends up weaving. As you increase the P Gain it gets closer to going unstable.

You will need to decrease the P Gain until the oscillation dampens out in several cycles.

Have you tried increasing D Gain until it goes unstable, then backing off D, then increasing P Gain?

At this point I have changed the drive from Z axis to X axis.
Changed to Current Mode.
Disconnected the tach.
Installed ground from Power Supply chassis to machine chassis.
Installed ground from AMC drive to machine chassis.
Powered drive power supply from external 240 VAC source and back to internal machine source.
Tried different parameter settings to demonstrate effect on 'spikes'.

Have not yet gotten rid of the 'spikes'.

As you suggested, dropped P Gain down to below oscillation. Still has Filter on. Move size 20,000.