3.5.0rc1: Input shaping causes layer shifts!?

dc42

@NeoDue @adrian I've found a likely cause of RRF trying to apply input shaping to moves when it was not possible, which is likely to lead to jerky motion. So please can you try the new firmware at https://www.dropbox.com/scl/fo/p0136wx04h8xf6ejwdnn9/h?rlkey=efrfwyb6o5tqid11gustz3uvy&dl=0.

PS - a check I added has just found an anomaly that I need to fix; so you may wish to wait for a new build from me tomorrow.

adrian

@dc42
I’ll try and test tomorrow with the upcoming version

dc42

@adrian @NeoDue and anyone else having layer shift issues in 3.5.0-rc3 when IS is enabled, please try the new main board binaries at https://www.dropbox.com/scl/fo/p0136wx04h8xf6ejwdnn9/h?rlkey=efrfwyb6o5tqid11gustz3uvy&dl=0.

This version fixes a probable cause of layer shifts in previous versions. It also produces debug output if it detects a malformed movement segment. To avoid having to connect a USB terminal to see any debug messages, you can use the new M111 B parameter. If you are running on any Duet 3 board then I suggest M111 B4096. If you are running on Duet 2 or Maestro then you won't have enough free RAM for that, but you can try M111 B256 and accept that the messages will be truncated.

Please report any debug messages you see here, as well as whether you get any later shifts (and a M122 report at the end of the print if you do).

NeoDue

@dc42 Perfect! And I was lucky and have some time right now. Just a quick question regarding M111: if I use "P1", I get flooded with messages such as

Received 365 bytes
New conn on socket 0 for local port 80
Found responder
Received 378 bytes
New conn on socket 0 for local port 80
Found responder

How can I turn that off?

dc42

@NeoDue don't use P1, use P0. Or don't use P at all. The debug messages I was talking about are sent regardless of the debug enable status.

NeoDue

@dc42 This means "M111 S1 P0 B4096" or "M111 S1 B4096"? Okay, thanks!

edit: 1st "dry" test (print w/o filament, measure x and y offset) ist running right now!

NeoDue

... and once again @dc42 has fulfilled the picture of an electronics wizard that I have of him:

electronics wizard created with hotpot.ai_art-generator.png
(no, I cannot draw at all, this was generated with hotpot.ai art-generator, and I could not get that stupid KI to make the PCB larger)

Three prints up to over 6mm without filament and one with filament, all with Input shaping enabled - and not a single "bang" or layer shift! For the first time ever!

Congratulations, Sir!

As far as I can tell, it seems you have solved the issue!! THANK YOU!

Now I guess I can finally throw all these away

(By the way, the larger blob is still there, however not in the first affected layer but in the other two for some reason. But now the printer simply runs over it, so this is a minor annoyance now and nothing more. I will check if including the retract/unretract command into the print file instead of letting it the Duet do has an effect or such. In case I find something, I will open a new thread for that)

Edit: @adrian @oliof Please also write here if the current version solves the issue for you, then I would mark this as solved

adrian

@NeoDue Testing... Fingers crossed

Update: 35% finished, so far no shifting... I have few more hours to go to complete the print.
Update 2: 46% and going strong...

Checked this morning: finished correctly

Working for me

adrian

@dc42
Newest version works with IS on. It was quieter during infills, as well. Thanks for all your work

NeoDue

@adrian Great! Then @oliof is the last one missing.

balajiramani

@NeoDue I too, previously had layer shifts. With the newer firmware, there are no layer shifts and it works great!

Thank you, @dc42!!

oliof

@NeoDue I did a quick test today and the new version has no issues for me (but that's not worth a lot since I fixed my previous issues with increased current).

NeoDue

@oliof well, maybe you can reduce currents now again

But thanks for the feedback! I will mark the thread as solved then.

Once more @dc42: thank you very, very much for all that detective work you did!

dc42

My thanks to all of you who provided data to help me solve this issue.

balajiramani

I tried a different model and I see the layer shift with the newer firmware.

When I switched back to 3.4.6, It prints fine. So, it looks like the problem still exists.

NeoDue

@balajiramani you need to help me there... where is the layer shift in your picture? I only see some ringing...

balajiramani

@NeoDue

I am printing a dactyl manuform keyboard case. The one on the right was printed with 3.4.6 and the one on the right was printed with 3.5.0-rc+.

The walls are supposed to be straight but are tilted.

Exerqtor

@balajiramani have you tried these binaries?

NeoDue

@balajiramani Ah, that is rather some calculation error that seems to cause leaning if I interpret your picture correctly, but not a layer shift, and most probably something different than the issue in this thread. It might be worthwhile creating a new thread for that.

Edit: ... or resurrect his thread here where you already had posted a while ago. That one is about your leaning issue: https://forum.duet3d.com/topic/33886/y-slope-with-3-5-0-rc1-not-seen-with-3-4-6

balajiramani

@NeoDue Good point. Let me move this issue to the other thread. Thanks for pointing that out.