Loss of Y position on fast jogging.

[Geoff_S]

So, it was an interesting exercise ! Overall, we've confirmed there is no issue with UCCNC software nor the UC300ETH board, and the problem is due to noise from VFD drive.

As Andrew mentioned, we put the scope on the step signals from the UC300ETH (ie going to the opto-isolated breakout board), and also the signals output from the breakout board, while jogging the machine manually. Immediately it was obvious that there was some funny signals - small amounts of noise on the UC300ETH outputs, but more on the breakout outputs.

Top signal (yellow) from UC300ETH. Bottom signal (blue) from breakout board. Notice strange noise before main pulse

Even worse - now have a complete false pulse before the correct step pulse

After a bit of pondering, I checked whether there was any noise with all axes not moving - and yes there was - a lot !! Noise (about +/- 1V peak to peak) on the UC300ETH outputs, being converted by the breakout board to full 5V pulses at 43.5 kHz !

Noise with all axes stopped.

Andrew suggested we disconnect the analog signal to the spindle VFD and sure enough, all noise went away, leaving a nice clean step signal.

All noise gone...

Nice clean step signals

Now it made sense why things got worse when the kernel speed was reduced. We had a fixed frequency of corrupted step pulses due to noise at 43.5 kHz, which becomes more significant when compared to the true step pulses as kernel speed was dropped from 400 kHz down to 200 or 100 kHz.

Thanks all for the hints and suggestions for troubleshooting. I think Andrew and I have both learnt a bit from this little problem Although I was a little surprised that the opto-isolated breakout board amplified the noise, rather than using something like schmitt triggers to filter it out and avoid this sort of trouble.

[ger21]

Good news that you have it figured out. Hopefully you can start making molds soon.

[cncdrive]

Geoff,

Good to see that you found and fixed the issue.

Yes, I agree that a schmitt trigger would also fix this issue, because it would filter out the 1V amplitude noise signal around the positive or 0V power supply rail.
But it is even better that you fixed the issue at it's source.

[AvB]

The only difficulty out of this is it brought us back to the problem of getting the spindle speed working properly, which is why we resorted to hooking it up using the UC300 analogue output. I think we've got it close (using step/dir) but it's a real fine trial and error thing. At a 200khz kernel speed I think we've got the spindle steps/rev at 63 or so (can't remember exactly) but quite small changes to this number vary the minimum/ maximum speed and the response to the % change. Have other people got suggestions on how to get this right (other than trial and error) or do most people connect up a resistor/ capacitor network and use PWM??

[Battwell]

Does your vfd support modbus?
Glad you found the problem

[cncdrive]

Normally VFDs using isolated analog input circuitry. I've only seen one in my entire life which did not isolate the analog input and the interesting thing is that it was a high-end VFD not a cheap one.
And even cheap chinese ones have isolated analog inputs.
If the analog input is not isolated then it can cause issues if the noise is getting back to the UC300ETH or to any other electronics which provides the 0-10V for the VFD, so is why mostly all VFDs isolate it.
So, the question is what VFD are you using and is it really not isolating the analog input?

And the UC300ETH analog output is a PWM plus an RC network after it and an OPAMP in voltage follower configuration.
If the VFD is not isolating then the solution is to use a high speed optocoupler on a digital output of the UC300ETH and then low pass filter it with RC filter.
Remember that for a PWM you need a high speed optocoupler, because however the PWM is only a few hundreds hertz normally, but you must throughput that signal without much distorsion, so basicly the optocoupler should have an analog cut frequency at around the PWM frequency. I would use a HCPL2631 or similar one which can throughput 10Mbits digital signals and the switching time is only in the hundred nanoseconds range. But look up it's datasheet for more proper and detailed info on it.

By the way our UCBB breakout board is suitable for the task, it has HCPL2631 optocouplers on all outputs.

[Geoff_S]

The VFD is a Fuling brand. I've looked before but not found a manual for the particular model Andrew has, but manuals for similar Fuling VFD's do not mention isolation at all, so no idea if it has isolated inputs or not. Perhaps the noise we had was on the ground anyway ?

Thanks for the info on the UC300ETH outputs. We have the spindle speed working even though it's not great, but I might make an opto isolator for the PWM output of the UC300ETH some time in the future.

[dezsoe]

Could you make a photo of the VFD's data plate? Maybe somebody has similar.

[shad]

A long time ago we had a similar problems on the plasma tables with HF oscillator for initial arc setup. Like China LGK100/160 or Hyprtherm MAXPRO200 or HPR130.
Full optoisolation between table and CNC solves this problem.
And do not forget about GOOD grounding of the CNC machine.

[AvB]

Here are pictures of the info on the VFD.

General advice ... at this point, given that we may have to do a workaround to get the spindle set up well, is it wiser to switch to the UCCBB? What I mean is, are there other benefits with it compared to the old C23 board? On one hand if all is working OK, maybe we shouldn't fiddle any further and save the $ for another time - but if the benefits are clear, perhaps it's smarter to change over?

It's interesting that my old C23 board has ports on the back that match to the Warp 9 ESS so you just plug it directly onto the board. And they promote its compatibility in the manual. I'm not an electronics whiz but I couldn't help wondering why the UC300ETH doesn't similarly plug directly onto the UCCBB, to avoid the need for ribbon cables which could pick up EMI?