Trying to optimize base period, etc.

nerdsite wrote: if I manually increase the base period in the BOB.ini file from 62,000 to something greater I get "following errors" in running the axis test file.

With a stepper system the "following error" means that the system simply hasn't been able to make enough steps to move the motor far enough.

If the base period is 50uS then is can only make 20,000 steps per second. If the base period is longer then it can make less.

What's the PC? 5.1uS isn't particularly good latency. Have you tried any optimisation?

I thought 5100ns was pretty good jitter (just according to what I read on this site). I've only used isocpus=1 for a dual-core G620 cpu. I've done nothing else other than using an external graphics card. If I get that down to, say 3000, will that make a big difference? Thanks.

Also, does the max speed override the program g-code speed settings? My son is the g-code cnc expert, not me. I'm guessing the speed is too fast causing the following errors.

It seems to work if (1) Microsteps are just 2, or (2) Microsteps are (4) AND speeds is 30in/min (.5/s). Is that pretty much my limits or is there something else that I should be tweaking? Accel is 5in/s^2 (don't really know what's realistic). With the 2ustep config I get a base period of 100,000ns. Not sure if this is large, small or average. Just trying to understand real-world configs. Thanks.

Have you read this?
linuxcnc.org/docs/2.5/html/motion/tweaking_steppers.html

does the max speed override the program g-code speed settings?

Yes.

Rick G

nerdsite wrote: I thought 5100ns was pretty good jitter (just according to what I read on this site).

Sorry, my mistake. Yes, 5100 nS is excellent. I was out by a factor of 10.

So, with a 5100nS latency but realtime delays on a 60,000nS base thread, something funny is going on. Maybe an infrequent issue that didn't happen during the latency test?

Have you tried running the latency-histogram?

Yes; however, that doesn't mean I understand it all fully. It seems if I put microsteps, then the calcs change. Do I then need to adjust the step lengths or something for microsteps? If I kicked down the % speed it seemed to not get the following errors @ 4 usteps. However, all seemed to run fine @ 2 microsteps.

nerdsite wrote: Yes; however, that doesn't mean I understand it all fully. It seems if I put microsteps, then the calcs change.

Yes, because at 4 micro-steps it takes 4 x as many steps per second to move at the same speed as it does at full-stepping.

The motor is probably 200 steps per rev. A 4x microstepping it is 800 steps per rev, at 16 microsteps it would be 3200 steps per rev.

Yes, and it doesn't seem to be the reason (unless there's something weird w/ linuxcnc running).

You could try ignoring the warning and then see if anything bad happens.

I found that a CamView tab in Axis ruined my otherwise-good latency, so there are definitely things that can cause problems that don't show up in the test.