Lost Steps: Square vs Pulse


Peter Boreland
 

I have recently been testing a new controller running  nema 23 motors. I discovered 4" spikes in the guide response. At first I thought that this might be a mechanical issue, but later came to realized that they come from the controller.  



Above, I am showing the result using step_wave_form  SQUARE.with the number of micro steps set to 32. Reducing the number of micro steps to 16, the spikes where fewer, but larger. 

I then went on to set step_wave_form  PULSE.  At 16 micro steps I saw no evidence of any spikes. At 32 I'm not so sure that there were none.  I saw no guiding degradation at 16 microsteps (0.56" step resolution), so I have decided to stay with this arrangement..

What might be going on here?

Peter


Peter Boreland
 

The controller I'm using is the MaxESP3 with TMC5160 drivers


John Walker
 

This guy did a test on a few different drivers.  None of them are perfectly linear and some hare worse than others.
https://hackaday.com/2016/08/29/how-accurate-is-microstepping-really/


Peter Boreland
 

Thankyou, John. I recall now seeing this on Youtube sometime back. I recall in a forum posting someone mentioning that pulse may be more accurate than square based on the timing method within the processor. 

The thing is, this is a serious problem, and I would imagine others have this problem but may not have noticed, especially if they are using a belt drive where the spikes would appear much smaller due to the gear ration of the pulleys. I do not know if the issue is load dependent. At first, I thought the motors were not able to deliver sufficient torque to the worm (directly coupled in my case).  To sort this out I would have to reduce the load and test guide, which I do not feel like doing right now!

In pulse mode with 16 microsteps, the spikes were not visible at all in the guide log. This is encouraging.  I'm hoping Howard will weigh in on this.

Peter


 

It looks like it gets stuck 16 microsteps at a time before catching up (4.5" = 16 microsteps for a 0.9 degree stepper at 32 microsteps).  Too much stiction?  While you might not easily notice this with a belt drive, the stiction torque is reduced proportionally to its ratio so it might not happen.  I just got Nema23 steppers meant for direct rive too and hope I will not see the same!  I will be following this thread.


Peter Boreland
 

Henk,

I thought stiction too, but the problem resolved when I went to pulse.  I.e switched back and forth between the two modes.  Square = spikes, Pulse = no spikes. 

The guide chart I posted was with 32 microsteps, 0.9 degree stepper, and spike  looks to be 4.5. Do you concur? Your response was a bit ambiguous?.

Yes it will be good if you can confirm this in your testing. I'll email my config.h file.  What is your timeframe? I have product going out the door quite soon. 

I happy also to share some logs with you. I got awesome guiding behavior, even with spikes.

Peter


 

On Mon, Sep 19, 2022 at 03:59 PM, Peter Boreland wrote:
The guide chart I posted was with 32 microsteps, 0.9 degree stepper, and spike  looks to be 4.5. Do you concur? Your response was a bit ambiguous?.
Here's my reasoning, correct me if I am wrong:  You are running at 32 microsteps so one microstep is 0.9*3600/(32*360)=0.28125", and 4.5" is 16 times that.  That's why it seemed to me like it got stuck for 16 microsteps then caught up again, for each spike.  The 3600 in the numerator is degrees to arc seconds, and the 360 in the denominator is the G11 worm reduction factor.

Yes it will be good if you can confirm this in your testing. I'll email my config.h file.  What is your timeframe? I have product going out the door quite soon. 
I first have to make the harness, and those Nema23s are a lot heavier than the Nema17s so my use of metal epoxy may fail.  Once I have it ready, I will let you know but first I want to see how the Nema17s perform in the one-piece worm configuration.


Peter Boreland
 

I think I resolved the problem.  I switched from square to pulse.  I also decreased the microstepping from 32 to 16.  I've had two full nights of imaging, and no spikes. Isaw no degradation in guiding.  In fact, I got as low as 0.26" total rms per this screen shot.



Peter