Loss of Y position on fast jogging.

[cncdrive]

No, it's not possible to adjust the pulse width. The pulse with is always the 1/kernel frequency divided by 2, because a pulse must contain a high and low state, and the duty cycle is 50%, so the half of the pulse is high and the half is low.
If the pulse was longer than that then it would not fit in the kernel frequency.

[AvB]

Justin (machine maker from Vermec) came to check it out yesterday and verified the position loss with a dial gauge, and fairly quickly came to the same conclusion as you guys, that it has to be a communication problem. He gave me some tips on things for Geoff to test when he brings the oscilloscope etc.

I was showing Justin the various things we had tried and the results. One thing was that he commented that kernel speeds of 200khz etc were very high and that his machines on Mach3 were set up with 25khz. I hadn't tried settings that low so I tried switching to 25khz, but found that at that kernel speed the axis velocity was very low. It wouldn't accept axis speeds of eg 3500 mm/min and it moved very slowly, which Justin couldn't understand.

[cncdrive]

Kernel speed is the maximum pulse rate in other words.
For example 25kHz is 25000Hz which means that then the controller will output a maximum of 25000 pulses per second.
The stepper or servo drive moves the motor one position forward for every pulses.
So, ofcourse if the kernel frequency is lower then that will limit your axis velocity if the stepper drive has a high microstepping or the servo drive has a high resolution encoder and so it requires lots of pulses to move one motor turn.

[AvB]

Yes that's what I understood but we used Mach3 at 25khz, with this same machine, operating at normal speeds. Or is the 25khz speed in Mach3 a different thing?

P.S. I also want to add my thanks to people for answering my questions and trying to help with intelligent feedback. I realize that I'm asking a lot here but I am very thankful for the opportunity to get the feedback because it is helping me to learn a lot more about how the machine works and as well as the software etc. Thanks very much and please feel free to give any more feedback as I'm learning heaps! Thanks.

[ger21]

With Mach3, you set the kernel speed in the ESS plugin. The 25Khz in the Mach3 screen is not used.
At 25Khz, you have 25,000 steps/sec, or 1,500,000 steps /min
You have 1640 steps/unit.
1,500,000/1640 = a max speed of 914mm/min at 25Khz.
100Khz would allow a max speed of 3650mm/min.
I think you are set to 3500? If so, then anything above 100Khz should work.
Do you have Gecko G320 or G320X drives? They have a min pulse with of 2.5-3.5uS, depending on model? If my math is right, at 200Kz, the pulse width is 2.5uS??

[AvB]

Ah, OK, thanks Gerry. My Gecko drives are G320X. So the pulse width is calculated as one half of one cycle in the frequency/ pulse stream?? In that case 2.5 microsec would be right. Interestingly, the machine runs fine and the error is less at 400 khz ... which I would have thought would put the pulse width at 1.25 microsec ie lower than the minimum for the Gecko drive?

[ger21]

The G320x also has a max pulse rate of 300Khz.

I've had a similar experience with a Leadshine AM882 stepper drive. It has a published max pulse rate of 200Khz, but it actually will run faster with UCCNC set to 400Khz.

[Geoff_S]

Thanks for the tips Gerry. There is an opto-isolated breakout board (a C23 from cnc4pc) between the UC300ETH board and the gecko drives. It quotes a maximum supported frequency of 1 MHz but also states "The recommended pulse width for the inputs and outputs is 2us" which equates to 250 kHz. At this stage we're planning to do some tests with an oscilloscope on Friday. Hopefully have some useful results after that.

[Battwell]

if you write a program with all g1 moves and a feedrate- is it losing position?
ie no G0 moves

[AvB]

Success we think! Geoff has spent some time hooking up the oscilloscope to the inputs (from UC300ETH to the C23 breakout board) and also from the C23 to the Gecko drive on the Y axis. He found a regular interference patch with a peak of about 1V, at a 43khz frequency on the input side from the UC300ETH - which was triggering the board to recognize it as a real pulse, which is magnified.

We wondered where it could be coming from. I had last night had a nagging thought that it could be related to how we have wired up the spindle. When we initially installed the UC300 and were trying to get Mach3 to work with the UC300 to test it, before switching over to UCCNC, I couldn't get the spindle speed to work properly and we took the easy option of using the UC300's analogue analogue voltage port to supply variable voltage directly to the spindle VFD. So when we saw the rogue frequency waves today and started thinking "how are we going to find where this is coming from???" We initially just took a simple investigative step of disconnecting the connector
from the UC300 analogue port to the spindle drive, and immediately the interference went away! And the test results (position loss) are fine.

Geoff says that since we've bypassed the opto-isolators with this alternative spindle drive method, we've allowed a 43 khz noise coming from the VFD to go straight into the UC300 via the analogue port and be fed into the pulse trains.

Funny to think that if we'd just thought to try disconnecting that earlier we could have crudely diagnosed the problem without Geoff having to do complex brain surgery to find it!

Geoff has some screenshots of the traces which he hopes to post later.