Skew Compensation

[BackwardEcho]

How much space is left to work with on the controller board? I am wondering how feasible it would be to add Skew Compensation. I haven't been using my mpmd for a while since I've come to accept how bad the skew was on it. I plan to come back to it eventually to reprint and reinstall my mods (which I will print from another printer) to hopefully correct any hardware issues I thought I avoided, but there will be only so much I can do.

[BackwardEcho]

@aegean-odyssey pssssst, I have a question

[aegean-odyssey]

The skew compensation described in the M852 gcode seems to be better suited to printers with cartesian kinematics. The errors caused by a delta printer's actual vs stated geometry settings probably won't be fixed by such a skew compensation.

To your question: I doubt if there's enough flash space to actually implement the command, but more importantly, I think the extra (floating point) math operations to compute paths would just bog down the machine -- it's all ready pushing it's limits, particularly when the usb port is active.

Back to getting more accurate prints: the firmware does allow one to set every dimension involved with the delta printer's kinematics by using the M665 and M666 gcodes. I have used PurpleSensation's "Delta Dimensional Calibration Tool" from Thingiverse to dial in the parameters with pretty good results. After printing and measuring the calibration standard, the measurements go into a spreadsheet and correction values for the gcodes are computed. I then created a (gcode) file with the corrections and placed it in the folder with the other utilities on the micro-sd card. In the case of my printer, setting these values (and saving them in the flash, M500) was a worthwhile exercise -- it improved the "out-of-roundness" of circles and holes on the printer when printing away from the center of the build plate.

[BackwardEcho]

Well dang. I went down the skew rabbit hole for my corexy printer and I am currently using Vector 3D's Calilantern with olistrik's GoSkew post-processor as said corexy does not support skew compensation in the firmware so I need to post process the generated gcode. Since the mpmd is running marlin I'd figured I'd at least ask if it was possible. Sounds a bit more complicated that I thought, which makes it more of a miracle you were able to add the extra fan output for probably only a handful of people that can fully use it. Do you think a post processing script, like the one linked above, will work properly with delta kinematics? As far as I can tell, all that happens are the X and Y values of G0 and G1 moves are altered slightly to compensate.

Once I am done upgrading / reinstalling parts on my corexy, I'll visit my mpmd again. As of now the plan is to: fix corexy printer skew, redesign / translate the parts into a proper CAD from Blender, reprint and install, recalibrate for skew again since the motion would have been disassembled and reassembled with new unskewed parts. After all that, I'll reprint my mpmd parts as I think any skewing issue was compounded since I didn't catch how bad my parts were at first. I will keep that Dimensional Calibration Tool on hand for future use.

[mulcmu]

This site has some details on how different build tolerances/imperfections impact the nozzle travel/position. The XY error for rod and radius errors are in a radial direction but the end stops have a more linear shape. The errors for each imperfection add together to create a non-uniform result. I made some similar plots for 1 arm too long or too short and the tower angles out of position. The boim 3d plots are hard to compare details. My plots have xy error arrows with the color denoting the z-error. The bottom figure also has a cross section of the bed along each tower direction. The top down plots of the z error looks linear but the cross section view shows it isn't quite flat at the extremes.

The auto calibration routines optimize some of the calibration parameters to get the z-errors to zero. There are lots of potential calibration parameters that will get the z errors pretty good at the probed points. My suspicion is there is only one optimal calibration where the z position remains flat (or parallel to bed) with accurate xy movements over the whole bed. Which would be the delta calibration rabbit hole...finding a repeatable and reliable way to determine the optimal calibration.