Need help with tuning Midi Drives with Metador Motors

Hi, I am busy revamping a Masterwood Project 310 with a Mesa 7i77 setup. This is my 3rd Mesa converted machine with my other two, a Biesse 336 and IMA 210 running production for almost a year now without problems.

The Masterwood ran fine with it's own outdate CN6 controller, but was a mission to program and could only interpolate 2 axis at a time so I opted to upgrade the control to LinuxCNC/Mesa.

But I can't seem to tune the Masterwood to get a f.error of below 0.5 mm at any speed.

The drives are Control Techniques Midi-Maestro 140's
The motors are Control Techniques Matador Brushed DC Servo with Tach feedback
Also on each axis is a Heidenhain Rod436 1000 line incremental encoder

The tachs is built into the motors and connects to the drives. The encoders are connected to the Mesa 7i77.

Whatever setting I tune in linuxcnc (P,I,D, FF1, FF2, Deadband, etc), i can't seem to get rid of oscillations and all the graphs from halscope seem to have a lot more noise then any of my other machines.

The closest I can get to a reasonable f.error of about 0.5mm is to tune the machine at a single feed rate of say 1m/min. When I increase or decrease the feed rate, then I have to re-tune the drives. The machine will run fine for a while at a single feed rate and then suddenly oscillation will start and then disappear again. It almost seem like the error gets compounded instead of cancelled.
This is happening on each axis(XYZ)

I have manually tuned the drives with a independent signal generator and oscilloscope as per the drives manual and i get decent PID curves as per the manual with the drive/tach/motor combo.
I have run the motors connected and disconnected from the balscrews (no visible difference in f.error performance)
I can run the motors at 40m/min or 0.1m/min with also no f.error performance increase.
I have opened and cleaned the Z and Y motors.

I have not measured the motors actual speed independently yet.
I have not replaced any motor bearings yet or serviced the encoders yet.

If anybody have any other ideas as to what I can try, short of replacing motors and drives. I mean they did work with the old Tecnos CN6 controller (DOS based). Also I am fairly familiar with linuxcnc and Mesa, but I could be missing something in the configuration. I have managed to tune the Biesse(yaskawa drives and motors) and IMA(Siemens Sinumerik drive and motors) to rapid at 70m/min and cut at up to 20m/min with both at about 0.003mm f-error)

Thanks
Roan

you may need more grounding to the motors
and use double shilded Signal cables sutch as for Plasmas

Can you post your hal/ini files and perhaps a plot of the following error?

Hi, here are my hal and ini files.

I am unfortunately not at the machine for the next two weeks, but I will see if i can get my assistant to post the halscope graphs.

My graphs do however look similar to the one below.


Also I am basically using the machines original wiring (twisted, shielded) and have tried extra grounding wires to see if I can eliminate noise, but no real difference.

thanks

I got my assistant to send me some videos and pics if it would help at all.

And here are some youtube vids:

#https://www.youtube.com/watch?v=g8ef7SHEHwI&feature=youtu.be
#https://www.youtube.com/watch?v=g8ef7SHEHwI&feature=youtu.be

I know the quality is not great, but it shows what I have been dealing with.

I could get the tuning better, but not much really and it never seems to be consistent.

Thanks

Try commenting out this line (and others like it for the other axis) and see if it helps,

If you look at the PID output in the plots, it does not seem that LinuxCNC is generating the oscillation (the PID output is almost flat relative to the velocity feedback), this suggests that the issue is with the drives velocity mode tuning, or some other motor/drive/tachometer problem.