3.5.0rc1 Input Shaping Layer Shifts not addressed in 3.5.0rc2

bulbs

Re: 3.5.0rc1: Input shaping causes layer shifts!?

I've had layer shifts related to input shaper(M122 output had numerous input shaping failed errors, turning off IS fixed the issue) and I have been waiting for a new release candidate so I can use input shaping; however I can't see anything in the rc2 changelog to suggest this issue specifically has been fixed. Is there any value in updating to rc2 and potentially ruining a print or is it clear that it's not yet been addressed? Should I just go back to 3.4.6 and try input shaping on the stable release?

I don't want to waste filament trying to reproduce the issue to get the M122 output that's necessary to make my own bug report, at least not without some more info so I can try something new to fix the layer shifts.

gloomyandy

@bulbs The input shaping reports in the M122 output do not indicate an error, they are for information/debug purposes only.

There have been lots of changes since RC1 (including at least one fix to input shaping: https://github.com/Duet3D/RepRapFirmware/commit/bc204a3b92b8ac97f7f4239709d2b9355fc049d7), so I would say that you should probably retest with rc2. I don't think there have been any other reports of layer shifts with rc1 and input shaping, so if there is a problem you may be the only person that has run into it and so if the problem exists in rc2 it is even more important to identify it.

0

dc42 committed to Duet3D/RepRapFirmware

Fixed processing of M593 L parameter, increased version to 3.5.0-rc.2

dc42

@bulbs yes I've fixed an issue in RC1 that could case layer shifts. It's the same issue that caused one of the filament monitor issues.

NeoDue

@gloomyandy If I may, I also posted this issue (if there was a previous topic, I did not see it - please feel free to merge the threads): https://forum.duet3d.com/topic/33926/3-5-0rc1-input-shaping-causes-layer-shifts/

I will check after christmas if the change from dc42 in RC2 helps in my case.
(edit: seems RC2 was withdrawn - at least I do not see it on Github?)

dc42

@NeoDue rc2 hasn't been released yet. We expect to release it tomorrow.

tas

@dc42 I tried 3.5.0rc2 on my Duet2Wifi standalone.

The part printed perfectly up to 5mm then began a very slight shift (0.1mm per layer maybe) towards the origin in the Y axis only.

There were no collisions of the printhead and part and no errors generated. The part finished printing ok except for the shift.

Input shaping was enabled with
M593 P"zvdd" F45.7 S0.0 L100;

As well as pressure advance with
M572 D0 S0.0825

Jerk set to 300 in X and Y.

Printing speed was 50mm/s max on a heavily modified Ender 5 Pro, direct drive LGX Lite with E3D V6 and E3D 0.9 steppers.

Hopefully this info will help isolate the problem.

oliof

@tas M593 L parameter changed in RC2, and may need adjustment.

gloomyandy

@oliof Yes it looks like L is now a value between 0.01 and1.0 which is used to control how much the speed and acceleration of a move can be adjusted to allow input shaping to be applied. The default setting is 0.25.

The docs do not seem to have been updated yet though (one for @droftarts ?).

tas

@oliof And that is why I should always read about the changes. I will do some more testing today if I get time. Thanks.

tas

@oliof @gloomyandy
I changed L parameter value to the default of 0.25 and the problem is completely gone. No Y shifting at all.

Many thanks, that was an easy fix.

tas

@tas Well I posted too soon. The first part on the bed printed perfectly.

The second part had a single shift at about 7.5mm layer height and another at about 10mm. Each time it shifted it moved towards the Y origin by 0.3mm this time. A larger shift than before but much less often.

I will try turning off Input Shaping to confirm that is actually the problem.

tas

@oliof @gloomyandy
I tried the same gcode with Input Shaping disabled and there were no issues with the print so something is up.

gloomyandy

@tas What motor current are you using on your Y stepper? I wonder if increasing it would allow your print to work with IS enabled. It may be that if the current is "marginal" using IS is pushing things over the edge.

tas

@gloomyandy
That is a good point.

I like my motors to run cool. I run the X axis with the E3D 'Compact but powerful motor' (40mm) and Y axis with the E3D 'High Torque Motor' (48mm). The rated current is 1.68A max for both so 85% is 1.428A and I am only at half of that.

M906 X700 Y700 Z800 E300 I30 ;

If you would like a test done let me know but it won't be tonight.

I didn't have issues with Y shifting until this rc2 release though.

gloomyandy

@tas I would say it is worth running a test with a higher current to the steppers, but perhaps @dc42 would like to comment. The way that IS is applied was changed in rc1 and so it is likely that it is now being used on a greater set of moves.

Exerqtor

@tas Like @gloomyandy said, just for troublesbooting purposes try setting the current high. Try with 1400milliamps just for one print and check if it makes a difference.

tas

@Exerqtor @gloomyandy

M906 X1400 Y1400 Z800 E300 I30;

M593 P"zvdd" F45.7 S0.0 L0.25;

Test print running now.

tas

@Exerqtor @gloomyandy

The test print worked perfectly.

So the Y shifts were due to a configuration error on my part. The motors do need more current to meet the demands of input shaping.

The motors got a bit hotter than I would like by the end of the print but most people would call them warm so that is interesting for me. I will bump up the current in my configs.

Thank you both for the suggestions.

dc42

@tas with input shaping enabled, do the motors sound normal? I get a slight clunking sound when some types of IS are enabled on one of my machines. However, as the IS code is being rewritten I decided not to investigate this before releasing RC2.

You might like to run M122 after doing one of these prints. It has a new field that lists the maximum number of overdue steps encountered, and that may be related.

tas

@dc42 I can't say I noticed a noise difference but I will do some testing today and report back.