Oval Holes

[handles]

Thanks Jazz, that is very helpful.

I retrammed and recalibrated the X and Y/A axes, and could find no discernible backlash. It is basically an Avid replica with their Rack and Pinion drives. Perhaps my measuring tools are just not accurate enough, but I managed to get an acceptable result by adding 0.15 mm backlash compensation to the X with a compensatory acceleration of 20% extra (600mm/s/s/).
I set the max Velocity at 5000mm/m and Acceleration at 500mm/s/s on a small item profile.

The maximum before lost steps is actually 21000mm/m at 1000mm/s/s acceleration, but I would have to bolt the table to the floor to run at anything like those speeds with a gantry of around 80 or 90kg total. Seriously powerful motors those Lichuan Nema 23/24s. Thanks for the recommend on those btw.

The OP on that forum with the Out Of Round problem was confusing the terminology, as he linked to NEMA 34 3 Phase 12NM Closed Loop Stepper Servo Motor Driver Kit he said he was using, not true Servos. His solution was:

"I was right about it not being backlash or anything loose in the system. I went through all the settings on the drives and I found a setting for speed smoothness.. This was set at 8 on the X-axis and 2 on the Y-axis.
The consequence of this difference between the two settings was the 0.5mm OOR on all circles. When adjusted to 8 there was a pretty aggressive speed smoothing going on.
I adjusted both drives to 2 and behold, perfectly round holes and even sized diagonals in squares."

I ran the debug software on the drives and all settings are identical, just with the exception of the Current Loop Proportional and Integral Gain, which I see is set by the drive automatically according to current fluctuations I take it. The Position Loop Proportional Gain, although I take your point re servos v closed loop steppers, is set way above the default, at 2500 v default 600. The Speed Loop Feedforward is way lower than the default at 35 v 120 and Position Tracking Error Threshold is at 4000 v 400 default.

I surmise that I could play with adjusting the Speed Loop Feedforward or other settings but I don't have enough knowledge on this, and how they relate to IJK commands on steppers v servos, so I will leave that for now, after all I am not milling parts for the space shuttle.

[JAZZCNC]

Thanks Jazz, that is very helpful.

I retrammed and recalibrated the X and Y/A axes, and could find no discernible backlash. It is basically an Avid replica with their Rack and Pinion drives. Perhaps my measuring tools are just not accurate enough, but I managed to get an acceptable result by adding 0.15 mm backlash compensation to the X with a compensatory acceleration of 20% extra (600mm/s/s/).
I set the max Velocity at 5000mm/m and Acceleration at 500mm/s/s on a small item profile.

The maximum before lost steps is actually 21000mm/m at 1000mm/s/s acceleration, but I would have to bolt the table to the floor to run at anything like those speeds with a gantry of around 80 or 90kg total. Seriously powerful motors those Lichuan Nema 23/24s. Thanks for the recommend on those btw.

Well first off 600mm/s/s is low, I'd at least double it, this will help straight away with reducing ovals if it's an IJK issue. However, R&P can be a pain in the arse regards getting the gantry square if the linearity of the racks isn't spot on, esp on a wide machine like an 8 x 4. So are you sure the gantry is spot on square.?

[handles]

Well first off 600mm/s/s is low, I'd at least double it, this will help straight away with reducing ovals if it's an IJK issue. However, R&P can be a pain in the arse regards getting the gantry square if the linearity of the racks isn't spot on, esp on a wide machine like an 8 x 4. So are you sure the gantry is spot on square.?

Thanks, I will adjust the acceleration and run a few more tests. I didn't realise that this was on the low side.

I spent quite a bit of time making sure the racks were mounted square on the EN3B steel angle I used, meshing their teeth together to make sure they were symmetrical and flush to each other along their length, and then mounted them on the ally extrusion running a dial indicator off the linear bearings to ensure they were as parallel as possible to the linear rails. Similar care was taken building the sideplates and gantry itself.
I am no more than a tenth or so of a millimetre out on the hypotenuse of a 600/800/1000 mm right angle triangle. I will do another measurement tomorrow to make sure it is absolutely exact. I have a home switch both on the Y and A for self squaring but it was pretty square from the get go and I haven't had to pull it into line with the homes. Maybe it is just a little more tweaking that is needed.

Thanks for the help!