Marlin: [FR] [Idea] [Proof of concept] Mechanical dithering on Z axis to improve accuracy

[AlessandroAU]

I noticed that the main cause of z artifacts is mechanical stiction caused by binding on the lead screw or linear rails. Playing with a dial caliper I noticed this can be resolved by giving the table the printer is sitting on a whack to jostle it into position.

So my idea would be to do something like this intentionally after each z move to allow the axis to settle into position.

I have done a proof on concept, I use (abuse?) the babystepping feature to oscillate the axis about it's intended position after each z move for a small amount of time and with a decaying amplitude. This allows the axis to 'average out' and settle more accurately on the intended position.

I expect this would help a lot of cheaper printers improve Z accuracy. On my machine it mostly helps with -Z motion.

I have made a video:
https://www.youtube.com/watch?v=cjPFJGFkVU0

Cheers,
Alessandro

[salacpavel]

Very interesting.

[h55ia]

Very good man, looking forward to trying this

[AlessandroAU]

Made a PR https://github.com/MarlinFirmware/Marlin/pull/15409

[AnHardt]

How about some more grease or weight instead?
Does this do any good under load?
Can't this be compensated by backlash compensation?

[AlessandroAU]

Sure, all possible fixes and on a 'good' printer this is not really needed, just like K advance is not really needed at low speeds or on direct drive. But the goal of this is make not so good printers print a little better with some tricks. Actually, you will find dithering is used in a lot of applications as an easy way to improve accuracy.

This isn't backlash either, this compensates for the non-deterministic stopping point you get under different static/moving frictional coefficients on the lead-screw where there is some elasticity in the system. (most printers to some degree) It also helps counteract the loss of torque when micro-stepping.

If you have a dial indicator you can experiment with nudging you printer after a z move to see if it will settle a bit closer to the intended position.

[h55ia]

Your PR got denied, better check that out please

[AlessandroAU]

Yeah need to rewrite against 2.0.x instead, I've done it but need to test before I PR.

[github-actions[bot]]

This issue is stale because it has been open 30 days with no activity. Remove stale label / comment or this will be closed in 5 days.

[thierryzoller]

@AlessandroAU any luck so far?
Has anyone the time to get to push this into 2.0 ?

[AlessandroAU]

I've put a version of this into the 2.0 branch, but since my printer has been out of action for awhile I've yet to test.

[thierryzoller]

How would we proceed to test this ?

[thierryzoller]

@AlessandroAU : Is it this one ? https://github.com/AlessandroAU/Marlin/commit/2049468a144a2c0158c1ea130e1deb1d663ea8a1

[AlessandroAU]

@th33xitus I think so, but I need to check my laptop to see if I have any uncommitted commits.

@h55ia That's awesome! no idea someone was using it. Can you share any results? Does it improve z-banding? did you try different settings etc?

[thierryzoller]

@AlessandroAU did you have the time to check? are these all changeS?

[MoellerDi]

@thierryzoller I have it working in my 2.0 fork.

See the following commits:
https://github.com/MoellerDi/Marlin/commit/4d4c25f6dc1074bd99f86d99c168d1ff9fb80710
https://github.com/MoellerDi/Marlin/commit/45cd50251e0fe33699873d8b0a07c77456c79984

I'm still testing, but as of now it seems be an improvement.

@AlessandroAU are you still planning to raise a PR for 2.x to get it merged some day?