Bridgeport Servo tuning

Unfortunately for me it appears that the impedance of the scope is a problem for the PID output - it doesn't function correctly when I attach a scope across the PID output. Oh well, it certainly appears that there's some delay in response of the amplifier. I may do some more testing on that.

There should be no problem reading the analog out with a scope.
A normal scope probe is 10M which should have basically no effect on the 7I77s analog output (which as a OP amp has about a 1 Ohm output impedance)

You do have to make sure the scope probe ground lead is connected the 7I77s ground and the scope probe is connected to AOUT. Even if the drive has differential inputs, the scope and 7I77 have unbalanced signals so must share a common ground.

Ah - that must be my problem. Will try again over the weekend.

OK, tried the scope again and I still can't do that, but for another better-understood reason. My oscilloscope input channels have a common shield and the servo drive signal has an isolated output.

I was able to make a very interesting plot by putting in a step function, as was suggested a while ago. I set all the gains to 0 except for FF1, which I left at the value I had determined during tuning earlier. I turned up MAX_ACCEL to 99999. The delay between the PWM signal and motor motion is much more manageable than we thought it might be - about 6 ms. I haven't quite figured out the ramifications on tuning yet but it does seem that I might need to set up the inverse deadband function. I will see if I can find where the threshold voltage is.

Scale 10ms/div

To be clear, I was assuming that you had trouble measuring the 7I77
outputs with the oscilloscope (this should work fine). Its generally not safe to measure the motor drive PWM output with a standard scope probe as the motor drive PWM is basically differential, high voltage, and not neccesarily ground referenced so not only is there possibly no ground signal that makes sense, you may exceed the ratings of your scope probe.

IMHO a 6 MS delay needs to be fixed. as it will make tuning and decent performance nearly impossible. Depending on how much time you want to spend on that drive you could replace it or try and trouble shoot /adjust it.

For a basic assesment of what that drive is doing I would disconnect its motor from the mill, and see if it is actually working in velocity mode. That is, set a small (say 100 mV) DAC output in each direction and check for symmetrical RPM each way
then try larger values. Also even at very low speeds (100 mV should be ~ 20-30 RPM) the drives local velocity feedback should resist load changes and maintain
the same speed.

To be clear, you were right in your assumption that I was initially trying to measure the 7i77 pwm output. When I got that, I wanted to get the servo amp output as well but realized it was not possible.

I will try what you suggest. New servo amplifiers are expensive but I'd like to get the most out of this machine.

Update: The servo amplifiers are Bosch Z15-1-240 which are velocity drives, based on what little info I've found on them.

I'm going to do the same test on my 4th axis and see what it looks like. My 4th axis is driven by a more modern AMC 30A8T amplifier.

Even my 4th axis, under similar test conditions and driven by its modern servo amp, shows 3 ms of delay between signal application and motion. PCW, what do you think an acceptable number would be?

Thanks for bearing with this thread even as it seems to be getting dragged out.

The acceptable number depends on your required accuracy vs acceleration
for example with a 10 mS delay and 10IPS/S acceleration, there will be an unfixable ~.5 mill error at the start and stop of acceleration.

There are always some following error spikes with second order trajectory generators (like LinuxCNCs), because the trajectory generator command has a acceleration component that is a step function (which has infinitely high frequency components) The actual drive always takes some time to reach the requested acceleration (time to develop the motor current (dIdT= V/L) needed for the requested acceleration, at low speeds, this basically adds an integrator with a few ms time constart)

I do wonder if a little FF3 would help. FF3 would be a 1 servo period long pulse at the beginning and end of the acceleration phase (Jerk feed forward) This would have to be added to the PID comp