Mach3 XML and Stepconfig issues.

Well, you do what works. The stepgen type 0 produces a pulse that is always the same length (set by stepgen.N.steplen). Stepgen type 2 produces a pulses whose length equals 1/2 the frequency. We still don't know for sure what the pulse that Mach3 generates looks like. (Until you scope it.)

The last few posts between you and Andy seemed to imply that you were having trouble setting up and getting the stepgen type 2 phase-A signal to work. That is why I posted the hal file I used to show that signal in my last post.

But according to your Mach3 set up screen it should be the type 0 with a step length of 5000 (or maybe its inverse) That is why I suggested going back to that with the settings I suggested in my previous post. (Nice and confusing huh? If you under stand all this your probably smarter than me.)

"But according to your Mach3 set up screen it should be the type 0 with a step length of 5000 (or maybe its inverse) "

And that did not work either way inversed or not. Quad pulses no better either. Still as a bad two-stroke.

So I have to take down my scope and look how the signal from Mach looks like.

At this point I think that is the only way to be certain what exactly is going on. Use the "real" scope to see just what Mach is doing. Then set up Linuxcnc to replicate that. Then use the "real" scope to confirm that Linuxcnc is actually doing what you configured it to. If Linuxcnc isn't doing what it is configured to then something isn't configured right, there is a big latency problem, or you've found a rather nasty bug.

I wonder if maybe there might possibly be something goofy going on with your chargepump causing this output or maybe all of them.to pulse. Is there a way that you can check to see if the outputs on your breakout board stay on steady when it is set to a constant true value while Linuxcnc is powering the chargepump?

Something is fishy that is for sure.


Mach3 out = 1µs sweep = 5µs pulse as set in Mach3


LCNC delivers with 1ms sweep a pulse of 2.75ms or something like that.



Input is set to 1V/div

Above is with 250000 steplen so it figures, but shorter and spindle will not start.
Will go down after dinner and test stepping down until there is no spindle starting.

Tested with 2500 and nothing

Obviously I am missing something here

Testing in 0,0,0
Now I have set steplen to 5000 and I can start spindle BUT it takes for ever about 35 to push the plus sign before spindle starts. Scope says 30mikro s and not 5.
Now also I have smooth spindle rotation.

How to get spindle starting earlier and why not working in MDI?

Position scale did the trick but still no MDI

Doh, homing axis now MDI works