Skewed print any ideas?

So printing a cylindrical object and it leans towards the -x direction. I am printing it on an all metal xl kossel, which so far has been faultless. I was using 1.17rc1 reprap firmware for duet and assumed it was a firmware bug as this is new firmware, but I've tried with the new grid compensation on and off and gone back to 1.17 dev8 and 1.16, same issue so it's not the firmware.

Towers are 90degrees to the bed and obviously the bed is level as a result. The arms are Hayden Huntley magnetic type, the linear motion is MGN-12H rails, the bed is tooling plate.

It autocalibrates to 0.03mm deviation and until 2 days ago nothing has been skewed like this, the only thing I changed was the firmware. Its not the object as I've tried a few different cylinders. I printed a tall narrow tower which did not show the same tilting. The M122 output on RRF does not show any skipped steps or processor overload. Pulleys aren't loose, belts are tight, arms are all properly attached.





Stumped....

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

A thought that strikes me is a year or so ago (In fact probably nearer to 18months) a friend of mine had a similar issue on thin wall objects at certain layer heights which turned out to be a rounding error in Marlin maybe worth trying a different layer size or wall thickness and re-slice it and try again.

Just a thought?

Doug

ps still working on CNC cut effector and carriages?

Hi Doug, it could be might try a different slicer.

Not at present as it's working very well. But I'd try some if your metal ones if they are available.

Edited 1 time(s). Last edit at 12/17/2016 01:01PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Make sure you're following whatever process the firmware calls for to clear the bed height calculation. That certainly sounds like a bad bed probing is saved somewhere and messing with things. I'm far from familiar with that process specifically, but I recall a Gcode existing to recall a saved heightmap file, make sure that's not floating around your config.h somewhere, alternatively I'd delete the bedheight.h or whatever its called entirely and see if that clears it up.

That's a good suggestion but my bed.g file has a gcode (M561) to clear any existing bed transform before autocalibration and grid levelling. I might try a solid cylinder and a couple of others with different wall thickness and put them in different locations around the bed to see if it varies depending on x,y coordinate..

Edited 1 time(s). Last edit at 12/17/2016 02:05PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Like a souvenir vendor's table by the leaning tower of Pisa.

So wall thickness, overall diameter and position on the bed make no difference.

Next to see if its print speed or seam position making a difference. I usually use nearest or aligned seams, and I have been printing these fairly fast, maybe a little inertia is throwing the effector out to the left as the seam is at the right on these cylinders and it has had chance to acquire some speed by that point, still wouldn't explain why the right side of the cylinder is slanted as well.

Edited 2 time(s). Last edit at 12/17/2016 03:35PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

yeah that's a weird one indeed. Lemme now how the test goes. have you done any other prints since updating?

Ruled out mechanical issues? loose bolts, leaning towers, loose carriages/rods/effector tilt?
It'd be a weird coincidence coinciding with the firmware update, but stranger things have happened. Outside of not clearing a bum bed probing I can't think of what 1.17 has that 1.16 didn't that would screw your calibration by so much without changing bed.h and config.h

Edited 1 time(s). Last edit at 12/17/2016 03:42PM by clearlynotstefan.

Yes to mechanical factors as I said above. I updated it yesterday afternoon, I was using previously the beta version of the firmware I'm now using, so I don't think its that. I printed some stuff for a customer, a tv bracket thing, a circular base with some grooves in it, and two cylinders which came out skewed like this. I;ve since printed about 10 more trying various things, moved on to 20mm cylinders to save time/filament as they are just as skewed. Prior to that I printed some airsoft gun parts for someone else who commented they were some of the best 3d prints he's seen.

Edit - I just measured the distance between the towers, with a 300mm digital caliper there is 0.1mm difference between the upper and lower of the X-Z pair and they are 1000mm towers, the rest are less than 0.05mm difference, the frame is not the issue.

Edited 3 time(s). Last edit at 12/17/2016 03:46PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Weird, I guess wait till DC42 comes in with his duet wisdom. I'd try deleting the bed height file anyway, just in case the glitch is in the gcode not clearing that, but beyond that I've got nothing. Sorry.

Thanks for having a think about it, its definitely a stone cold whodunnit not a smoking gun this one. Just tried random seam position but it made no difference. I'll try printing slowly next. Then try turning on the new feature that fades out the bed compensation after x layers, see if that makes a difference. Its lucky no one ever wants anything printed in gold abs, or I'd be wasting a lot of useful filament.

I did post on duet forum but David couldn't immediately see a reason why it should be skewed, he asked me to check M122 to see if there are underuns/overuns/high maxreps but the answers there were 0/0/20 which is all pretty normal.

Edited 1 time(s). Last edit at 12/17/2016 04:05PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Try switching your motor plugs in different orders and see if the skew changes direction? Roll back to 1.16 and see if it fixes itself? Chance of motor failure itself? I'm just spitballing here, I'd be losing my mind in your position so trying to help.

edit: saw you rolled back. The fact that didn't fix it is especially troubling. Did you post your bed/configh files for dc? maybe something screwy in there? Also is the lean in a particular direction relative to the towers? towards one/away from one etc.

Edited 1 time(s). Last edit at 12/17/2016 04:50PM by clearlynotstefan.

Appreciated. I will test the motors and some other basic stuff like seeing if they all move 100mm when they are asked to, even at 300mm/s etc... but I don't think its anything like that I'd expect a print to be all over the place if a motor were failing, but Ill test it.

Not that bothered, there is a cause and I am quite enjoying tracking it down, I'll be enjoying it even more when I fix it.

Plus I have 3 other printers so its not a big deal, they'll get some attention instead of this one for a change.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

See edit above, sry about that. Also, do you use H parameters in your bed.h? More just a curiosity of mine, although I could see if something changed firmware/mechanically and those weren't valid anymore causing problems.

I dont use them no my effector doesn't tilt much if at all, and the IR sensor is very close to the nozzle. As for direction it appears to be towards -x not towards/away from a tower in particular which is again, very odd.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Damn I am so invested in this conundrum. Rotate the XYZ plugs and see if it leans in the same direction, then I'm out of ideas lol.

So now I checked the motors they'll move exactly 200mm when asked to do so at 300mm/s. Enabling the grid compensation fade out makes no difference, I wasn't expecting it to really as the bed is really flat.

I suppose one thing I could try if I really cannot find the cause, RRF does allow for correcting non-orthogonal axes in software, it would just be a case of very precisely calculating the amount and precise direction of lean and applying such a correction, it would be a very unsatisfying workaround though, and I am not sure if it works for deltas.

Edited 2 time(s). Last edit at 12/17/2016 06:56PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

This problem just keeps getting stranger and stranger.



After 50mm or so the part starts printing straight again.
Grid compensation was set to stop after 10mm.

I might have to sacrifice an animal to the 3d printing god to solve this one. The axis compensation option is out then as it would work for 50mm of z then start skewing the print after that point.

Edited 1 time(s). Last edit at 12/17/2016 07:45PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

DC should come yell at me for giving advice I'm not qualified to give, so I'll couch it in this. If I were you and truly stuck and with no more ideas, I'd pull my memory card, reset+erase and reflash the 1.16 onto the cleanly wiped duet, then go through bed/config/etc with a fine toothed comb for weirdness and try again lol.

Apart from the lean, are you losing height, possibly more on one side than the other?

How much does it shift sideways per layer?

Have you checked that M122 is reporting zero step errors after the print?

Have you tried reducing acceleration and jerk?

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Any chance of some larger photos? I'm having trouble making out the detail in those ones.

Also if you grab the hot nozzle with a pair of pliers, can you move it around? The only thing I can think of is that your nozzle/hot end is loose, and is being pushed around by the cable umbilical. That's the only thing I can think of that would explain the change in slant at a certain height.

If it helps any, I've spent about 20 hours this week trying to get a certain part printed, and had all sorts of crazy issues. Now I've discovered that there's effector tilt because the guy who built the machine wasn't worried by an eccentric hole in a rod end. *sigh*

Definitely the most plausible explanation so far.
The hot end assembly is not as tight in the effector as I'd like it to be, the only weak point on Hayden's effector is the hot end clamp is a bit flimsy. I will try to get it tighter/redesign it.

Those thumbnails are clickable for a bigger image but here's another if it helps.



This version was after wiping the duetwifi and reinstalling all the software, so I think we can rule out software.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Hi David,

Yes the height is correct its the same on both sides of that tower shown above, it came out 100.55mm for a 100mm tower.

The sideways shift is very uniform. It slopes towards -x (not the x tower) until 50mm and there is a 2mm discrepancy in the sloping part, which seem awfully round numbers to me. Its 0.01449mm off per layer.

M122 gives me 0 underuns, 0 overruns, max reps 6.

The tower printed at 60mm/s but with my default accel and jerk at 3000, 1200 respectively it barely got up to speed printing it. I tried printing a cylinder at 120mm/s with those settings and at 60mm/s exactly the same cylinder was printed.

Genuinely stumped.
Based on Nebbian's suggestion I am printing a strengthened effector.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

It was the step errors count from M122 that I asked for, not the figures that you already provided in an earlier post.

Three more questions:

1. Is the bed perpendicular to the towers?

2. When you auto calibrate, how many factors do you specify?

3. If you run M666 with no parameters, what values does it give for the bed tilt correction?

---

Large delta printer [miscsolutions.wordpress.com], E3D tool changer, Robotdigg SCARA printer, Crane Quad and Ormerod

Disclosure: I design Duet electronics and work on RepRapFirmware, [duet3d.com].

Okay, step errors 0.
1. The bed is perpendicular to the towers, I adjusted it manually yesterday to attempt to diagnose this problem. My G30 S-1 at the base of each tower was x1.296 y1.262 z1.134 Bed is 330mm.
2. 7 Factors and 16 points generated by your bed.g generator.
3. M666 Endstop adjustments X1.45 Y-1.34 Z-0.12, tilt X0.00% Y0.00%

Edited 1 time(s). Last edit at 12/18/2016 05:05PM by DjDemonD.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Quote
DjDemonD
Definitely the most plausible explanation so far.
The hot end assembly is not as tight in the effector as I'd like it to be, the only weak point on Hayden's effector is the hot end clamp is a bit flimsy. I will try to get it tighter/redesign it.

Those thumbnails are clickable for a bigger image but here's another if it helps.

Thanks DJ, I tried clicking on the thumbnails previously but only got the small version again.

If it was me I'd just get a cable tie, or a small washer or something and jam that hotend heatsink to something else on the effector. You could slip some thin aluminium shim (beer can ) inside the clamp. It doesn't matter what you connect it to, or even if it's vertical, it just has to stay in the one place relative to the effector. That way you are testing your existing effector. But then again I am a bit of a caveman in the workshop

I had issues with my hotend wandering around on my Sintron delta so I know the pain. I ended up making a clamp out of a large washer and a couple of screws. That thing is NOT moving now.

My beefed up version of the clamp for haydn's effector, I was able to tighten mine completely and still wiggle the effector, it basically wasn't thick enough in the middle, I bright that part out a little and beefed up the screw holes so they didn't tear as easily

[www.thingiverse.com]

Edited 1 time(s). Last edit at 12/18/2016 11:13PM by clearlynotstefan.

Thanks, I beefed up the clamp and added some spars to the centre of the effector, I was printing it on another printer and we had a mains power "wobble" and it spaghettified. Try again tonight.

---

Simon Khoury

Co-founder of [www.precisionpiezo.co.uk] Accurate, repeatable, versatile Z-Probes
Published:Inventions

Quote

M666 Endstop adjustments X1.45 Y-1.34 Z-0.12,

To me it looks like your M666 endstop difference is huge. Are they really that far apart?