Duet V0.85 Auto Calibration

timonemycat

Hey community,

I've got a Rostock Max V2 with a Duet V0.85 motherboard, and I'm using Tricklaser carbon arms and FSR sensors underneath a thick sheet of melamine for a rigid setup for autocalibration. I've checked to make sure my build is accurate; bed is perpendicular to the towers, the rod lengths are exactly the same between pairs, my carriages are solid.

In spite of all of this, I haven't been able to get the deviation under .134-.136. Every time I've calibrated gets me somewhere around there. What is a good deviation to be shooting for on a successful calibration?

I've gone through and updated the h parameters in the bed.g file to account for variances, which are very small, let me know if these are too small to have significance. They don't seem to help the deviation at all. The h values I've gotten are:
-0.03533333333
0.003333333333
-0.074
-0.03333333333
-0.07933333333
-0.044
-0.07
-0.027
0.02233333333
0.01233333333
0.002
0.01833333333

These are for the standard 18 points on my bed. These signify that the bed is pretty rigid correct? What other areas should I be checking for build accuracy? I've followed David's post at https://duet3d.dozuki.com/Wiki/Calibrating_a_delta_printer, and haven't been able to spot an error, and haven't been able to improve the deviation. Any help would be greatly appreciated!

Thanks!

jmjcoke

This was helpful when I setup FSRs before switching to smart effector.
http://www.sublimelayers.com/2016/01/fsr-auto-calibration-setup-for-duet.html

timonemycat

That's exactly the page I followed, Michael has been invaluable to me and the community. I followed it to the T, and I'm still getting fairly bad results out of it. Could it be endstop inconsistency? How do you like the smart effector over FSRs?

dc42

Two suggestions:

Measure the trigger height of the FSRs at each probe point, to check that it is consistent. It's not uncommon to find variations in the trigger height of FSRs, especially near the edges of the bed. Edit: looks like you have already done that.
Mount a circular spirit level on the effector to show up any varying tilt.

timonemycat

I removed the head and placed a bubble level on the top, and noticed a bunch of variation at the points specified in bed.g. Results are easier described with the video: https://youtu.be/VbzuqFZhGNA.

The tilt made me think it was an issue with the X-tower because of the deviation opposite it, so I checked the carriage and head effector more closely, and found that the threaded shafts that I'm using to attach my tricklaser carbon rods, (the version with ball links, not the injection molded ball links) are being bent by the way that the Rostock Max V2 carriages and effector attachment points are made. alt text
I'd think that bend is a problem.

I've gone about replacing these components with the injection molded ball link versions to remove this incompatibility in the design.