Bimodal layer shift

[mrehorstdmd]

@spudaddict That first photo at the top of this thread looks like something is shifting by a mm or two. There has to be something loose. Can you detect any play at all in the bearing blocks on the linear guides? Are the pulley axles solidly mounted? Grab the heater block on the hot-end and try to wiggle it.

Try pushing on the bed and see if you can make it wiggle.

The frame has an open front in that design for some reason. Does it flex when you push the two top corners toward each other?

[JamesM]

Did my eyes deceive me or do you have a 1.8deg stepper on X and a 0.9deg stepper on Y looking at your config.g and the steps/mm??? I'm guessing by the heatsinks on the stepper drivers that it is a clone duet board since the standard duets don't come with them. If that's the case I would put some active cooling on that board since the clones generally have only 1/2 the amount of copper in the pcb compared to the real duet boards.

[spudaddict]

@mrehorstdmd Yes, this last round seems to have much less shift.

I can detect some rotating on the "Y" bridge, but takes some force. (pushing forward on the left end, pulling on the right at the same time - and visa versa) I tried doing that during a print with long lines (a large box) and it didn't seem to have an effect. The open front has bothered me since before I began the build. They will flex a little if the Y bridge is near the back, less flex when it is near the front.

I DO get some wiggling on the nozzle in the Y direction. Not sure how that can create a diagonal offset, but I plan on reprinting the head assembly today/tomorrow with stiffer supports. The X and Y MGN carriages seem solid.

Thank you for your continued suggestions. I will see what reducing the print head wiggle brings.

[spudaddict]

@JamesM lol... yes. Sharp eyes.

My only spare 2A stepper motor is for the 2nd filament driver(1.8deg). I swapped it into both the X and Y positions to see if there was a motor issue and need to pull it back out. I tried the M121 command near the end of the gray tube print and there were zero errors and zero hiccups - so likely a mechanical issue.

When the back cover is on, the 4x120 fans move a lot of air across the Duet board, but I have had the cover off since I started troubleshooting. I will put a fan on and under it next.

Thank you for your help and suggestions.

[Danal]

@mrehorstdmd said in Bimodal layer shift:

The frame has an open front in that design for some reason. Does it flex when you push the two top corners toward each other?

Good catch. That looks REALLY suspicious. Given the nature of the layer shift, specifically that a few layers are "this way" and a few more "that way", it would seem that something is flexing or stretching, ONLY when some part of the printer is at "one end" vs the other, AND that the slicer is doing a few layers with all "strokes" moving in one direction... and then a few more layers the opposite way.

And... all of that would fit VERY well with that open front moving, when the crossbar is not there to brace it.

Perhaps try bolting a temporary brace across that opening, and reprint?

[spudaddict]

@Danal Morning Danal. I am not ruling anything out at this point, but a couple considerations:

There are two cross bars near the bottom (not ideal) that are rigid - so any flexing would require the vertical 2040 extrusions to actually bend against the 40mm width. Im not sure there is anything that happens during normal operation of the printer that can generate that much force.

Also, the Y bridge works to hold the width steady since it cannot get longer/shorter as it travels front to back... as long as the MGN rail alignment is correct.

I am going to put a temp brace on as you suggested, but I suspect what I will get is confirmation that my MGN alignment is either right or wrong. If it is wrong and the top is now absolutely rigid - the Y bridge should bind up. Badly.

[Danal]

@spudaddict said in Bimodal layer shift:

@Danal Morning Danal. I am not ruling anything out at this point, but a couple considerations:

There are two cross bars near the bottom (not ideal) that are rigid - so any flexing would require the vertical 2040 extrusions to actually bend against the 40mm width. Im not sure there is anything that happens during normal operation of the printer that can generate that much force.

Also, the Y bridge works to hold the width steady since it cannot get longer/shorter as it travels front to back... as long as the MGN rail alignment is correct.

I am going to put a temp brace on as you suggested, but I suspect what I will get is confirmation that my MGN alignment is either right or wrong. If it is wrong and the top is now absolutely rigid - the Y bridge should bind up. Badly.

Yeah, it is a long shot...

But two things seem very likely, given the pattern of the offset:

1. The gcode is doing something in "this direction" for a few layers, and then "that direction" and keeps swapping. This would make absolutely no difference on a mechanically correct printer.

2. Something in the printer is allowing movement that gets forced "this way" when the code is printing in "this direction" and vice versa... AND it is allowing that in one corner and not the other. Corner? Edge?

It is item 2 that makes me look at things like a belt with a stretched spot near-ish one end of the belt, or the frame being more rigid "over here" vs "over there", or a rail that is loose at one end, or anything that can be different. I'd look at anything asymmetric first, and then look for loose things at one end. Like maybe a linear rail that's bolted to the frame very firmly at one end, and has somehow come loose at the other. GIven that it looks like a full mm or more of shift, probably not a subtle wear item, probably something that is very loose.

I just discovered last night that I can twist the carriage on my Jubilee tool changer. The "slider" on the rail is fine. The bolts between the carriage and that slider must be loose. Need to take some things apart find out. Explains some things I've been seeing.

So heave on everything. Twist, push, etc. I'm sure you already have... and I'm sure you've done it while the carriage is positioned near the corner that shows the worst symptoms. Power it off and play with things in that corner. Push hard.

I will say you've got a bizarre one going here... fun challenge for all those of us throwing ideas at you remotely. Not as fun for you!!

[JamesM]

Bracing the front is something I did to my BLV printer from the start. The open front is nice but I wanted it rigid. It is an old picture of mine since I have since done a RailcoreII bed with 3 leadscrews and steppers.

[spudaddict]

@Danal lol... troubleshooting was fun at first, but im getting to the point i just want to print.

Both items are very true. Working through the many great suggestions I have gotten so far. Will definitely post if I find the smoking gun.

[spudaddict]

@spudaddict OK, so smoking gun found. I went to swap out the other stepper motor and noticed three case screws (on the bottom) were missing. I replace the screws and two stripped out and two were loose. I pulled the T20 driver gear off, and there is grease speckled underneath.

So, I am guessing without the case screws the internal bearings got skewed and that the motor is likely unrecoverable.

I tapped the motor case and got 2 screws to hold - luckily they are diagonally opposite each other. Just for grins I started my "tube test" design as 3 of the 4 tubes consistently turn out terrible. Here is what I got:

Comparatively perfect minus some relatively minor tuning.

New stepper motor is on order and this is one to put in the troubleshooting books - if you ever get a repeatable bi-stable layer shift on a 45degree angle, check your stepper motor.

THANK YOU everyone for your input. I am implementing pretty much every suggestion I got as there are definitely other items that I can improve on. The improved belt alignment and stiffer front plane has been integrated into my new print head design. (and now I can even print it!!) Other improvements will continue. Thanks again - especially JamesM, mrehorstdmd, Danal who stuck it out to the end.