Missing Steps - Cant Print SpreadCycle StealthChop tuning help

carcamerarig

@fulg I’m sure @dc42 said it needed to be a bit longer like 130ms 150ms suggested.

Argo

Low jerk and medium to high acceleration is basically the standard setup for cartesian and Core XY printers. Also input shaper do not like high jerk values to function properly. Klipper limits "Jerk" (they use a different name for it) to 5 mm/s when using input shapers. This should not cause any trouble.

StealthChop initialization or calibration.
This is what I put into my config.g (bottom part):

; StealthCop calibration
M17
G4 P150
G1 X1 Y1 Z1
G4 S2
M84 ; disable motors

I'm using 1.8 LDO motors with my Mini 5+. No bonging, no issues. Prints 24/7 at the moment.

@carcamerarig has the "bang" issue with both 0.9 LDOs and 1.8 LDOs and @fulg has the bang issues with 0.9 motors but has not tested it with 1.8 LDOs so far.

With 0.9 LDOs I needed more current and could not print as fast but never had any "bang" issues during movement.
Imo the issue with vibrations and noise that TMC2209 + 0.9 motors are having is a different chapter / issue.

The question I'm beginning to ask myself: Can a hardware issue on the Mini5+ cause the "bang" issue?
I can reproduce those little hiccups (bangs) by enabling hybrid mode (V parameter) and set the threshold setting at a too high speed (like 100+ mm/s). The switch from StealthChop to SpreadCycle is then causing a "bang" sound which then also often leads to layer shifts.
BUT I checked the config of @carcamerarig (which is basically mine) which does not use hybrid mode (V0) so we can rule that out.

So maybe it's a hardware issue / problem with the board itself?

fulg

@carcamerarig I had the problem with M566 X600 Y600 ... and only reduced it to troubleshoot this issue by reducing instant direction changes, which I thought were causing the problem. I can go ahead and reduce acceleration too.

carcamerarig

@argo I think the point that my machine is only 6 months old, had less than 3 rolls of filament through it and performed perfectly before the upgrade. I only changed x and y motors and the board, nothing else changed and the issue was immediate. With the exception of repacking the misumi X carriage bearings I’ve stripped everything else down checked and rebuilt, repacking the Y bearing blocks and it changed nothing for the better. Then one or more of the large hard bangs broke my Y motor holder which I replaced.

ctilley79

@argo the bang issue is the more pressing concern. It’s enough to cause layer shifts. Even when locked in spreadcycle by specifying mode 2 instead of 3, OP still gets the slight banging on both 0.9, and to a lesser degree 1.8 motors.

carcamerarig

@droftarts

Rev 0.5 board

SputnikOC3d

@argo said in Missing Steps - Cant Print SpreadCycle StealthChop tuning help:

I'm using 1.8 LDO motors with my Mini 5+. No bonging, no issues. Prints 24/7 at the moment.

... but locked into StealthChop 100% of the time - at reduced travel speed moves and reduced print speeds below what the OP's machine was designed to operate at.

Important to be concise.

The mini 5+ duet and the .9* motors were supposed to be an upgrade in performance [ speed and print resolution ] over the Einsy and 1.8* LDO steppers.

carcamerarig

@sputnikoc3d
@sputnikoc3d
I ordered an SKR board yesterday with 2209 steppers to run klipper on, I feel uneasy about all the radio silence if I’m honest... I’m really hoping it gets resolved over the next week or two after which I think I’ll return the mini.

Argo

@sputnikoc3d

Not really reduced. I have a cartesian / bed slinger printer. I could do travel moves up to 200 mm/s but I don't because at those bed speeds I add additional cooling and also it's quite noisy. But doable without losing steps. We do btw almost have the same printer. He has Caribou, I have a Bear.

TMC2209 are a huge upgrade compared to TMC2130 (EINSY). Much less VFA and comparing the print results of 1.8 motors to 0.9 motors are (to me) only visible when making photos with a macro lens. You can also increase your steps to 200/mm and interpolate from 32 microstepping which the TMC2130 can't.

The Mini 5+ can handle 0.9 motors up to 120 mm/s but with much more noise during my tests. That issue I also have with Marlin 2. So it's a general problem with TMC2209 in my opinion.

The problem with the "bangs" and missing steps is a different chapter though and should not happen with 1.8 motors nor with 0.9 motors.

fulg

Hello @droftarts, one interesting data point we seem to have found, the problem is apparently much much more likely to happen if you use bed mesh compensation. Even @oc_geek who was quoted earlier as not having isssues, enabled a bed mesh and instantly started having the same layer shifts issue.

I looked at the start of the thread and @carcamerarig is also using a bed mesh (as shown in his M122 dump). So am I.

So perhaps that is the key to reproduce this issue. Personally I use a 11x11 mesh to correct a 0.1mm variation in my bed (not a lot but more than enough to ruin prints!)...

carcamerarig

@fulg I turned off my mesh bed leveling for a few prints and it’s was a lot worse.

droftarts

Apologies for the lack of updates, mainly because there have been no updates. Thanks for additional information. I’m planning on spending all Thursday looking at this issue (during school holidays I’m only available one day a week), and hopefully @carcamerarig will be joining me for the fun.

@Argo is low jerk high acceleration really the norm? Prusa3D disagrees with you:
https://github.com/prusa3d/Prusa-Firmware/blob/eb4cf1a77f57b0825b72c5209d48fceaa8c0ec58/Firmware/Configuration.h#L424
https://github.com/prusa3d/Prusa-Firmware/blob/eb4cf1a77f57b0825b72c5209d48fceaa8c0ec58/Firmware/variants/1_75mm_MK3S-EINSy10a-E3Dv6full.h#L100
Segmented curves will tend to print slow with low jerk.
Also your stealthchop tuning isn’t right, see page 35 and 36 here https://www.trinamic.com/fileadmin/assets/Products/ICs_Documents/TMC2209_Datasheet_V103.pdf#page35
(If you need more info or an example, let me know and I’ll write it up when I’m at my desk and not on mobile.)

Ian

Argo

@droftarts

Default Prusa slicer profile uses 8 mm/s jerk. But yes, still rather high.

I first stepped over low jerk recommendation in the Duet gcode dictionary.
M593 suggets "High X and Y jerk values reduce the effectiveness of DAA; therefore you should set the X and Y jerk limits only as high as necessary to allow curves to be printed smoothly."

So I did some testing and found 4 mm/s (or 240 mm/min) works very well without having issues with curves.

Also the Klipper input shaper documentation suggests to use low jerk and high acceleration
"If square_corner_velocity parameter was changed, revert it back to 5.0. It is not advised to increase it when using the input shaper because it can cause more smoothing in parts - it is better to use higher acceleration value instead."
https://www.klipper3d.org/Resonance_Compensation.html

What would you suggest to improve the StealthChop tuning procedure?

carcamerarig already told me about your plans. Very nice offer from you I hope you'll find a solution for his issues.

oc_geek

@fulg ...for some reason my account login had issues and i wasn't able to login neither to change password...back now

On my end (and i'm on the THIRD Duet 3 Mini 5+...)

i'm running Stand Alone
Have OMC 0.9 motors (but extruder)
Tried all releases including betas

Always had and have the bangs during prints; was able to "print" but had layer shits on a bunch (3-4) prints (which i attributed at that time to some other cause)
Since a few days i tried to enable mesh. Total disaster ...i started to get 25mm+ layer shifts

I really hope these issue get fixed and the Duet team takes it on cse i think there are severe issues at the moment to be solved... which make the card almsot unusable...

It is good that there is discussion and people contribute but the way this thread is turned... it is like users are doing debugging or finding caveats and work arounds to let things do not break apart.... i honestly think is not the way.... i counted few replies from the Duet team... ok take more time...analyze....test ....but come up with something or people will start moving / looking elsewhere (me included)

dc42

@fulg said in Missing Steps - Cant Print SpreadCycle StealthChop tuning help:

One thing that was suggested to me (outside of this forum) is that I am not doing M17 and G4 P100 before moving motors the first time, reading the docs this seems to say that the required stealthChop/spreadCycle calibration of the TMC2209 will not occur. I would have expected RRF to do this implicitly, but perhaps it does and I am misunderstanding.

RRF does not do this automatically. However, if you are switching to spreadCycle mode then it should not be necessary.

dc42

All who are experiencing the issue of occasional banging/layer shifting during a print, after it has happened please post:

A brief description of what happened, including the axis or axes affected if known
A M122 report
A M569 P# report for each relevant driver number #
Your config.g file

carcamerarig

@dc42 mine are all contained in this thread from the various modes in both standalone and sbc.

Argo

@carcamerarig said in Missing Steps - Cant Print SpreadCycle StealthChop tuning help:

@dc42 mine are all contained in this thread from the various modes in both standalone and sbc.

So he does not have to search for it as the thread has quite exploded:
https://forum.duet3d.com/post/223138

droftarts

@oc_geek thanks for the report. Did the other boards have the same problem?

What firmware versions is everyone using? Did you notice the problem in earlier firmware versions, eg 3.2 betas?

Ian

oc_geek

@droftarts yes they did
on my side i've been in 3.2.2, 3.3 beta 1 and 3.3beta2

I'll post the files that dc42 indicated later (can't now)

Thanks for your support and looking into it. We all appreciate

edit: As a head up... a simple mesh enable (corrections are within +/- 0.05) and movement from e.g. X150 Y290 Z5 to X150 Y5 Z150 causes a tremendous shift in Z (it landed at 60 instead of 150...) it shall not be difficult to reproduce... thanks again

edit: @dc42 as requested including dumps