Running servos in position mode instead of analog

JohnnyCNC wrote:

I'm running a DYN4 1.8K as a spindle motor geared 1:1.8 using velocity mode and +/-10V control. I rigid tap 3mm to 6mm with no problem. Never tried a DYN on an axis though.

Would you be able to use the halscope on the motor and check of there is a delay between velocity command or analog out and encoder velocity. Just curious.

Michael wrote:

Would you be able to use the halscope on the motor and check of there is a delay between velocity command or analog out and encoder velocity. Just curious.

Yes, it would sure be interesting to find out whether the 5 ms delay is in sending a motion command to the motor, or in getting the feedback from the encoder through the drive to the Mesa board. If the latter, maybe the encider could be hooked directly to the Mesa board and copied to the drive.

The delay sounds like the delay is “normal”. Oak boards adjust the delay during setup. Same with other drives with absolute encoders. Is there a way to do the same in LinuxCNC?
Still I would try to tune the drives after the Auto Tuning. I went way up with the Main Gain to 40, Speed Gain to 35, Torque Filter to 127, MaxAccel to 127 this really helped with the F-Errors at the start.

jmelson wrote:

Michael wrote:

Would you be able to use the halscope on the motor and check of there is a delay between velocity command or analog out and encoder velocity. Just curious.

Yes, it would sure be interesting to find out whether the 5 ms delay is in sending a motion command to the motor, or in getting the feedback from the encoder through the drive to the Mesa board. If the latter, maybe the encider could be hooked directly to the Mesa board and copied to the drive.

It thought about it too. Unfortunately they use a Absolute Encoder, aka serial. We may can use a FPGA to sniff the signal before it goes to the drive. Or generate a signal and read the output.

I tried auto tune. Didn't change.

Tried high gains, low gains and everything in-between. Spent weeks on these things. No change.

Tried analog signals and step signals to tye drives. No change.

The velocity command and analog output are dead on the with each other.

Not sure how we would adjust for the delay in the control. That defeats the purpose of synced motion and honestly not what I payed DMM for.

I have a heidenhain encoder I am going to mount directly to the motor shaft to hopefully show the DMM technician that the issue is with their product and get my money back. Just have not gotten around to it.

Please mount the encoder to the shaft. This will show if the delay is encoder output or the processing time from control input to execution.

Modern drives have a delay from control input to execute the command. I didn’t expect 5ms. but most non CNC applications that is more than acceptable. This doesn’t mean we should happy with it.
Centroid is compensating the delay from command sent to execution. This delay will always be the same. And the whole timeline will be delayed. PID control is possible after the first delay is absorbed and the control output is synced. The delay is measured during setup process and stored.
I don’t know how it is applied to the PID loop or the motion controller.

I am using these on a plasma and not connecting back to linuxcnc the encoders so I probably would not see the error and delays.

Good to note not to use DMM in the future but they are half the price of Yaskawa's, so if you don't need to close loop at linuxcnc I guess they would work.

Michael wrote:

Bevin's, did you close the position loop back to Linux? If so care to share any info or hal files?

No I didnt use it. There was a wicked delay I couldnt get rid of. I am not bringing the encoder back to linuxcnc. Plasma machine and it iws working fine.

I tried auto tune. Didn't change.

Tried high gains, low gains and everything in-between. Spent weeks on these things. No change.

Tried analog signals and step signals to tye drives. No change.

The velocity command and analog output are dead on the with each other.

Not sure how we would adjust for the delay in the control. That defeats the purpose of synced motion and honestly not what I payed DMM for.

I have a heidenhain encoder I am going to mount directly to the motor shaft to hopefully show the DMM technician that the issue is with their product and get my money back. Just have not gotten around to it.

Post here what you find out after mounting the encoder please.
I have an older gen of DMM servos on a lathe that have been working very reliably in step/dir mode. I also was not able to tune them in velocity mode most likely due to the delay and user errors (I was new to linuxcnc on that build). I was considering the DNY4s for a future build. I'm curious to see if a linear encoder or standard rotary encoder mounted to a ballscrew could be a workaround for velocity mode with Linuxcnc.

My schedule got a bit busy lately but I plan to be back on this project next week. I will let you know what I find out.