Kevin,

How does Mach use the index pulse during threading cycles? What kind of speed variation can it compensate for? Does it compensate during the cut or only as it is starting each pass?

See Threading on the Lathe-Mach3 Turn write-up in Members Doc's section 2 as it explains in detail.
The more consistant the speed control the better. 30-40 rpm variation is not good. Speed is monitored and an adjustment made based on the previous pass. The monitored speed is very accurate and more so than you may  be able to measure. The DRO should
not change by more than 1 when not threading.

One problem is if the vfd is changing the rpm and Mach is compensating for the rpm then the two are fighting each other.
Thus screwed up threading.
I would suggest you try manualy setting the rpm and see what happens with the scribbing tests.

RICH

Screwed up threading 

I set the speed on the VFD manually, and it stayed steady within 1 or 2 rpm. Gave me much better results on the scribing test. I checked the PWM signal from the SS with my volt meter, and that appeared to stay constant. The analog output voltage from Homann's DigiSpeed board seems to vary up and down a few millivolts  though. Not sure if this is the issue. I need to sit down and do a little math and see if millivolt variations are enough to affect my speed by that much.

Don't think it's the DigiSpeed board that's causing my trouble. I'm seeing output fluctuations of maybe .001-.002 volts. This shouldn't affect my spindle speed by more than 1rpm, if even that much.

Possibly something with how the drive reacts to analog input commands. I'll be contacting tech support to see if I can work something out there.

I know my drive has a ModBus interface. Is it possible to command VFD speed and direction from Mach via ModBus?

One issue I'm seeing, although slightly unrelated, is the linearity between the S command and Mach's PWM signal. Although I'm sure using a voltmeter to test that signal isn't perfect, at 2Khz base frequency, I should be able to read a fairly clean voltage represented by the PWM signal. And it isn't very linear. Which shows up in the fact that at low RPMS, S actual reads higher than S commanded, and at higher RPMS, S actual reads lower than S actual.

 Did you ever get resolution to this issue? I'm experiencing similar issues.

SG