Missing Steps in circular interpolation

[kman1967]

I have an Avid CNC machine using a Centroid Acorn controller. Machine is using 2 motors on one axis for the Y. All NEMA34 steppers.

Any linear moves are fine. When I use any circular interpolation I lose steps in what appears to be a crossing over a quadrant (you can actually see it make some weird jump). I if have multiple cuts on the Z axis by the time I am at the bottom the part edges look like a staircase with each step off by the same amount.
I did some tests and starting from a 0 point I move out on the X axis to a point and then do a G03 to make a circle then I go back to 0. Each time I do that the spindle is a specific distance away from 0 in the same direction and accumulates each time. If I do the same but with G02 the error goes in the opposites direction.
I am using Gecko G210X drives. Originally I had it configured for 200 stps/rev and the noise and shaking was crazy. I am now at 2000stps/rev and I thought it fixed the issue but it is still there. Any parts that use circular interpolation in the G Code come out messed.

[centroidsupport]

Please post a current report

[kman1967]

report attached

[kman1967]

Added a pic of what is happening...

[centroidsupport]

The first thing I notice is that you have the accel set to .5 and the paired axis set to 1. I suggest that you set them both to the same value and see if that helps.

It looks like you are losing steps.

What are you using to generate your g code?

If you cut a simple circular pocket and/or circular frame in Intercon, do you get the same results?

[CNCMaryland]

Just to add.

360(degrees) / 1.8(degrees)(assuming a 1.8 degree stepper) = 200 steps per rev multiplied by the micro stepping of 10 (I believe that all Gecko's are set for 10x microstepping. = 200 x 10 = 2000. That is the formula. You need to verify the stepper degrees, and the microstepping of the drive.

Per Gecko Drive:
The step pulse multiplier board, the G901X, has four resolution options: Full step (200 PPR), half step (400 PPR), five microstep (1000 PPR), and ten microstep (2000 PPR. An adjustable low speed smoothness trimpot compensates for motor nonlinearity at the low end while midband resonance compensation ensures reliable torque output through the midband dropout region.

[cnckeith]

report attached

not related to the big step in the arc, but definitely related to machining performance... report shows overall turns ratio around 1 turn per inch.. this will not produce good machining results.

is this a rack and pinion machine? photos please.

also again unrelated to the big step in the arc. the cncpc cpu has a single core benchmark of 1200 that is below the minimum recommended single core performance and will not run the g code smoothing feature well.

now related to the big step. looks like backlash to me. measure the backlash of the X and Y axes and then adjust the machine mechanicals to minimize it.

[kman1967]

In response to
'centroid support'
- I actually have the Y axis output tied to 2 drives for the Y, not Y and N. I originally had N paired with Y but removed it as I thought this may be causing some issue. So N is not being used.
- G code is generated from Autodesk Inventor CAM using the Centroid post processor settings.
- Yes I have manual cut circles and they do the same thing.

'CNCmaryland'
- The software is set at 2000 counts/rev. I have set all the drives the same using the jumpers on the drive.

'cnckeith'
- not sure I can do anything about that. Yes this is a rack and pinion. It is an Avid Pro 4x8' machine pic attached.
- I have checked backlash and cannot seem to measure any appreciable amount

[ShawnM]

Are the drives wired to the Acorn headers or do you have them wired to the DB25 via a breakout board? What wiring diagram did you follow? You could also be losing steps on linear moves it’s just a lot harder to notice.

[centroidsupport]

What kind of drives and motors? Are they identical?