threading - spindle encoder synchonization issues

Its actually the opposite, a servo motor can maintain a precise speed but a VFD cannot...

If you only have an index, accurate threading is relying on perfect speed control for a
complete spindle rotation, since the control must guess the spindle angular  position between
index pulses. Having an encoder allows the control to "gear" X motion to spindle rotation
so even without precise spindle speed control, accurate threads can be cut.

yes, more accuracy

I'm having threading issues where the machine cuts a slightly different thread each time. This has always been the case since I retrofitted a late 70's CNC lathe 8 years ago. Recently I've decided I've had enough and want it running "perfectly" which is why I'm finally digging into the issue.

Initially I thought the issue was due to noise on my encoder lines. I finally tracked down the source of the noise I had and it was EMI from my cheap chinesium servo drives. I've thoroughly isolated those now and have very clean encoder signals so I really doubt it's that.

• I'm running LinuxCNC 2.7.4
• The OS is: Debian 7.10 (Wheezy) and the kernel is 3.2.0-4-rt-686-pae
• The encoder is a US Digital HI-1024-IE-D with a differential sort of backpack that connects to the single ended encoder and drives differential signals to my mesa 7i76 breakout board.
• The 7i76 is connected to a mesa 7i92
• I keep doing test runs on a thread and every pass is slightly different. It's as if it's not getting correct the index pulse and I suspected that to be the case which is why I put so much effort into getting rid of any noise. However this does not appear to be the case.
• When I turn the spindle manually 10 turns I expect to see 4096*10 or 40960 counts in motion.spindle_revs, but it seems to undercount by about 5-10 counts per revolution. I've swapped in a different encoder and I have basically the same issue. What is strange is that if I rotate the spindle back and forth +- 45 degrees or so a few times and go back to the original position I'll have accumulated a couple hundred encoder counts. What I think I've discovered is that the system undercounts in the forward spindle direction, but seems to count properly in reverse.

Attached is my .hal file.

Any ideas?

grossman83 wrote:

What I think I've discovered is that the system undercounts in the forward spindle direction, but seems to count properly in reverse.

Do you have a scope to monitor A and B encoder lines?
-
I had something similar a few years back with a china encoder where the pulses were correct in one direction, but would get thinner with speed on the other direction.
This was a new out of box encoder, the same make/model encoder worked properly after replacing it.

If it does not count reliably, it suggests a problem with
the quadrature.

The count should be absolutely consistent.

Tommylights suggestion to use an oscilloscope is good.

Also if you can run the spindle at (very) low constant speed you can look
at the encoder A/B signals with halscope.

The signals to watch are:

hm2_7i92.0.encoder.00.input-a
hm2_7i92.0.encoder.00.input-b
hm2_7i92.0.encoder.00.input-index

( assuming the spindle encoder is encoder 00 )

1024 count encoder or 4096 encoder ? if it's a 4096 encoder, that's likely having resolution issues,. i ran into a similar issue with my lathe at 1000rpm, i wound up stepping down from a 4096 to a 1024 encoder. also check the max speed rating of the encoder, make sure it can handle the speed