Issue with step servo setup - x-axis cutting skewd

Some ini files use the STEPGEN_MAXACCEL variable to set the hardware stepgens max acceleration limit but if you started with the hm2-stepper config then you probably have this line in your HAL file:

setp hm2_[HOSTMOT2](BOARD).0.stepgen.00.maxaccel 0

instead of

setp hm2_[HOSTMOT2](BOARD).0.stepgen.00.maxaccel [AXIS0] STEPGEN_MAXACCEL0

In which case you only have the MAX_ACCELERATION to set

Have you tried setting G64 PXX with the higher acceleration?

Unfortunately yes, with the increased acceleration it's still offsetting the cuts even with G64 P0.001

I've tried multiple cam software. Same issue.

Incremental, and absolute position mode (same result).

Slowed everything down to crawl (same result).

I layed down a flat ruler across the table and jogged the machine incrementally by 1" and it was consistently on each following inch marker as expected.

Can you single step through your program (torch off)?

Maybe rig up a felt tip pen

If you rotate the part 90 degrees in CAM (so the hole lead-ins start 90 degrees to where they are now) is the error in the same direction?

If its still in the same direction with a rotated part I would start to suspect the mechanics

Backlash suggestion got me thinking about the gear racks. We are going to tear down the x-axis drive system today, pull it completely apart and inspect for wear. It has a center mounted servo/double belt reduction which drives a center drive shaft that distributes power to spur gears on each side of the machine. I'm hoping we find some hobbed over gears or something that would make sense of this. If not, then I guess we will try to measure the backlash that's present in the system. Where do we enter this for the system to account for it?

I ran a number of tests last night including cutting flanges that are specifically shaped for one direction to test skew when I rotated them 90 degrees and the error was always in the x-direction. So if I cut two parts 1 rotated 90 degrees from the other, the error was completely different for each part (all towards x).

I will try the G61 mode today.

I stepped through a program last night and marked all of the piece points with a pen, but somehow I was getting tired and didn't think to run it again to double check the positioning. doh

Well we tore it down to find that the tension on each side of the gear racks wasn't evenly set - one was loose and one was tight. We tried digging through the old komatsu manuals but fell short of a torque setting so we tried blindly just tightening down the loose one, which chattered like a monster and gave us the most incredible offset ever... Then we backed them all off, pulled out the digital torque wrench and started with 20 ft*lb which chattered a little, but ended up settling on 30ft*lb of torque which is relatively quiet.

I had them take 6-8 backlash readings with different jog speeds and different locations on the machine, but we got ~0.003 backlash on y and ~0.0055 on x. These are average values. We checked the .ini and hal files and they do not have backlash settings in them so I'm going to add try adding them tonight and retesting.

So..

[axis 0]
backlash = 0.0055

?

We applied the correct backlash compensation to each axis but the results are still offset.