Steps/mm: Don't assume!

For the last couple of weeks I've been on a crusade to calibrate my printer so that it gives perfectly dimensioned parts. This has led me down a rabbit hole of:

• Creating a new calibration piece that is independant of extrusion errors
• Rebuilding my diagonal rods
• Shimming my rod mounts so that all pairs are within 0.05mm difference between top and bottom
• Using PLA instead of ABS
• Getting all towers to be parallel to within 0.02mm (difference between top of tower span and bottom of tower span)
• Getting all towers to be exactly 90 degrees to the build plate (within 0.1 degree)

After doing all that, some directions (from the centre to the tower) were giving measurements that were 0.4% too small, and one was only 0.2% too small. I couldn't figure out what the heck could cause this issue.

Finally in desperation tonight, I set up my calipers to measure the distance travelled by the carriages when moving down in increments of 10 mm. Strangely enough, the distances they travelled weren't perfect. They were a long way off! The worst one was moving 99.6 mm when I asked it to move 100 mm. I find this astonishing that a GT2 belt could be off by 0.4%, but there you go. I also found it very difficult to measure this distance accurately. I ended up with my calipers in depth gauge mode, with the depth gauge resting on the bottom extrusion, and the top of the caliper being pushed down by the carriage. This was reasonably repeatable.

So if you're chasing some weird dimensional issues on your delta, and you're sure that your build is pretty good, then check how far your carriages move. The issue might be your belts.

Now I feel Like I have to measure my belts just to be sure I dont have the same problem.

Thanks, Is good to know.



Edit: grammar corrections

Edited 1 time(s). Last edit at 02/09/2016 05:31PM by Velric.

I recently started looking into my printer's calibration. I've seen a lot of web sites that say to use the mathematical steps/mm values calculated from motor ustaps/rev, pulley teeth, and belt pitch. My resulting calibration prints were always off by about -0.5%, so I cranked up the steps/mm by about 0.5% and get perfectly sized prints. I've been using the machine that way for at least a year. I started looking into why it was off by that much and thought it might be my cheapo caliper, or the extruder calibration, or belt stretch, or pulley diameter.

The weird part was, I have a belt driven X axis and a ball screw driven Y axis and both were off by -0.5%. So I checked the caliper against some machining standards, I checked the extruder calibration by printing a single walled part and measuring to make sure that when I specify a 0.5mm wall that what I get. The only other thing that was common to both axes was the ABS I was printing with. I tried printing with TPU and found it was off by -0.25% in both axes.

So it looks like material shrinkage is the culprit. How do you deal with it? You can do what I had previously done and just increase the steps/mm by 0.5%, which is fine if you only print with ABS, or you can take a more generalized approach and scale the print depending on the material you're going to use. You don't want to scale CAD drawings- they should reflect the target size/reality, so you can scale the print when you slice or at print time using host software or in the printer's firmware. I did some digging and found that Cura will allow arbitrary scale factors to be applied before generating gcode. Slic3r will only scale in whole number % increments. RepRap firmware (and no others I am aware of) supports scaling prints using the M579 command.

I brought this up at the smoothieware support google group and they didn't seem interested in it. I checked into Slic3r at Github and found that others have requested the same changes for the same reason before me. I have decided to go back to increasing the steps/mm by 0.5% in my printer configuration because when i make parts that need tight dimensional accuracy I am usually printing with ABS. When I'm using other filaments it is usually for things that don't require such accuracy, so a little error won't hurt. I think it would be very nice if Slic3r added a shrink factor to the Filament page where you specify shrinkage for whatever filament you use and then slic3r would scale the gcode appropriately.

---

Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]

For steps i believe its best to move in Z.
That way the carriages are only moving up and there stepping.
IF you move in X or Y and your rods are wrong in the firmware or any of the angles are wrong your get a different amount.
Same reason your change your rod size in firmware to make parts bigger or smaller.
So you cant just say that it is the steps.

Also when you print out a part, the filament melts and things come out different sizes.
But if your just moving the hotend and checking it, then it should be on point.

Also if your X and Y are different then you have uneven rods.