Ultimaker Mark2 - The Smart Dual-Extrusion Upgrade

[RudydG 62]

This is a copy/paste of the conversation going on in the beta-tester forum of the Mark2.

We are moving the discussion here

----

Houston, I think we have a problem

I made a very simple model to test the "leaning to the left" behavior.

Test 1:

Single extrusion

Good print, it stays perfectly aligned.

Test 2:

Dual extrusion

It's fucked.

Here, we can see that somewhere in the first quarter of the print, it failed to pick the second print extruder.

I simply put the second print head away from the printer, so there was no interaction between magnets : it doesn't change anything. Magnets are not responsible for what we observe

Comparison between test 1 (left) and test 2.

Something is wrong, do you guys have ideas ?

Could you please try to replicate this problem ?

Here is the project file, Cura 2.6: https://1drv.ms/u/s!AgmRtaUlyi-AkZhYSI2hRGizd3o3Cw

I had it now with several different prints, and different slicers (Cura 2.5 an S3D)

In my case, it is skewed to the right, and to the back.

Would it be some kind of rounding factor in the firmware, related to the calibration? And a kind of zero with the toolchange?

I believe i have it from the beginning, but until now, i had not noticed it, printing abstract things where it was not noticable, and i knew there was something wrong, just could not detect it. Thought is was a wrong calibration.

[Melka 57]

@foehnsturm

This new printer has been announced on kickstarter:

They disclose that they were inspired by your magnetic tool solution, is it infringing a licence ?

[neotko 1,417]

Indeed @melka

On a side note I think that like @rudyg it might be a rounding error.

How does the T0 T1 work on @tinkergnome ? Is there any 'clean' from decimal points on the process from t0 to t1? I wish I could run the firmware to check but all my machines don't have the oled um2 system

[gr5 2,167]

@rudydg and @melka - My first thought is you have to tighten your pulleys some more.  I've seen this lean many times.  When it goes to pick up the other head the forces are higher and maybe one of the pulleys slips a bit.  It even makes sense that it leans left if your tool changer is in the front right corner.

Rudy's leans the other way - is your head in the left corner rudy?

My second thought is you are doing so much force when doing the tool change that it loses a step occasionally.  The X and Y steppers are 5 steps (not microsteps - full steps) per mm.  But if it lost 1 step on every other layer (every tool change) or even then it would lean 1mm for every 10 layers (or maybe 1mm for every 20 layers if it only loses a step when picking up and not dropping off the second head?).  That seems like more than you are seeing?  Or is that about right?  Of course the more it leans the farther it gets from the proper tool change location and so that starts to fail also (doesn't pick up the second head).  

Or it could indeed be a software issue but this seems unlikely.  But maybe.  When you change tools does the firmware just add the offest to incoming gcodes like it should?  Or does it actually change the internal position of where it thinks the head is?  The latter method could slowly add up to cause lean errors.

Anyway I strongly strongly recommend you tighten the hell out of all the set screws in the pulleys. They need to be REALLY tight. The hex wrench should twist a little and if you use an L shaped wrench your fingers should hurt after. And most important to get the 2 short belts - the one on the motor and the one nearest the motor.

Edited June 8, 2017 by Guest

[Melka 57]

I checked the pulleys and they were all really tight.

Could you guys try to replicate the problem we see by printing this cura project ?

https://1drv.ms/u/s!AgmRtaUlyi-AkZhYSI2hRGizd3o3Cw

You don't have to put materials in the second printhead and you can even put it out of the printer. Just be sure that you tape it so it doesn't burn shit.

Should take around 1h45min to print.

If you can replicate the problem, then we know it's something widespread.

[gr5 2,167]

@neotko mentions that foehnsturm himself has seen this leaning parts issue. It is still probably a mechanical issue but maybe not pulleys. Maybe lost steps due to high forces. The docking magnets are quite strong and overcoming their force in order to undock is one possible issue. But some people have the head hitting part of the dock (or the side blocks hitting the dock) and that can cause lost steps in the other direction.

If your tower leans *towards* the dock then try putting a layer of blue tape between the magnets to reduce the magnetic force a tiny bit. If they lean *away* from the dock then look for things hitting during the docking maneuvers.

[neotko 1,417]

No @gr5 latest test from @melka show that that happens even without the second head installed, so nothing is touched, not friction, nothing. It really looks like a slicer or firmware issue.

[RudydG 62]

My thoughts;

-my pulleys+belts are rock steady, no doubt about that.

-if it was a mechanical problem (hitting the 2nd head agaist the holder, then it would dissapear once it has shifted a couple of thents.

And it is so regular, that i really doubt it would not be software.

My 2nd head is at the the right side, like in the building explanation.

[RudydG 62]

Maybe something to compare?

My2nd extruder Offsets:

X 18.93

Y -23.61

Z 2.06

[ultiarjan 1,223]

Could you guys try to replicate the problem we see by printing this cura project ?

 

@tinkergnome

just did the print, code generated in Cura 2.5, I don't see any leaning. Also measured v.s. the left panel, no leaning. Will do the test based on the 2.6 profile now.

[RudydG 62]

I checked the pulleys and they were all really tight.

Could you guys try to replicate the problem we see by printing this cura project ?

https://1drv.ms/u/s!AgmRtaUlyi-AkZhYSI2hRGizd3o3Cw

You don't have to put materials in the second printhead and you can even put it out of the printer. Just be sure that you tape it so it doesn't burn shit.

Should take around 1h45min to print.

If you can replicate the problem, then we know it's something widespread.

 

printing right now!

[MarcoUbtz 15]

Could this error be caused not so much by firmware or cura but by a wrong alignment in the slicer when it was imported? Slightly rotated around the X and Y axis so it was not more orthogonal than the buildplate.

Had someone take a look at the gcode which gave these strange results?

[ultiarjan 1,223]

and here's the Cura2.6 version, clearly leaning, same machine.

I'm too tired now to compare the gcodes, but here they are if someone @tinkergnome wants to have a look. I could not load the 2.6 project file in 2.5 so not all settings used are the same.....

https://www.dropbox.com/s/s5aj9vswv5jemgp/C25-C26-leaning.zip?dl=0

[Melka 57]

YES ! Thanks @ultiarjan

I'm not crazy ! \0/

Edited June 8, 2017 by Guest

[ultiarjan 1,223]

oh and to be complete, firmware used is 17.02.2 , the tinker version.

[RudydG 62]

Cura 2.5

-Sliced for 2 extruders

-removed head 2

-started print

-obeject skewed to the rear (straight in the L-R direction)

Redo now with same Gcode, and second head in the machine + filament

Wait and see in the morning.

Already sure it is not a mechanical thing, otherwise it would not be skewed to the rear with the second head outside of the printer.

[RudydG 62]

And another update:

Cura 2.5

-Sliced for 2 extruders

-print with the 2 heads

-started print

-object skewed to the rear (straight in the L-R direction)-exactly the same as with only 1 head.

I can assure now, it is not a mechanical thing.

I had it with S3D, Cura 2.5

Will try now, with other offset settings, see if things change. WIll ignore the calibration right now and set the values to a round setting. (i have a feeling it has something to do with this)

Edited June 8, 2017 by Guest

[Melka 57]

Shit, so it's not because of Cura 2.6 :-/

Could it be because of acceleration parameters that are a bit too high and missing some step while using a second printhead because of power management ?

[conny_g 251]

Regarding the offset per layer, here's a calculation I did on basis of Melka's print in the developer's thread:

-------------------

That's 486 layers. 3.5mm/486=0,007201.

Hey, but when dual printing Cura prints two layers before it does a tool change!

0,0072x2=0,0144 and that's a tiny bit more than one microstep.

486/2*0.0125=3.04, that would be the offset/leaning with the microstep.

Are you sure it's 3.5mm and not 3.0mm?

Anyway, it looks like a toolchange offsets one microstep in X and Y. -1 in X and +1 in Y. (Weird that it's not both the same direction)

Now the question is if it's a mechanical issue or a software issue.

I think it's a software issue, it's too steady for mechanical and even the 2nd head was not present at all.

-------------------

[conny_g 251]

Would one of you want to try to add this at the beginning of the gcode file:

M218 T0 X0 Y0

or

M218 T0 X0.0125 Y-0.0125

(with correction of the leaning direction however this is happening for you)

This sets the head offset in the firmware.

The head 2 offset is calibrated explicitely, but maybe it helps to zero or to correct head 1.

[RudydG 62]

Update,

I changed the extruder offset values to 18.90 and -23.60

Print is perfectly straight now.

If someone else could do the same test,we could narrow down the possibilities.

About the extruder offset, if you calibrate trough the normal procedure, 1 step is 0.4mm. If my theory is correct, it could be solved by setting the offset to a round value before doing a calibration. (and the influence would be invisible)

Will try a new calibration tonight, being sure that i have a round figure as result. and then a redo of this print.

[MarcoUbtz 15]

I just got a look at the gcodes, the only differences I have found are a constant -/+ 0.001

Tonight I will run these gcode to see if they give me a mistake too.

[tinkergnome 927]

If my theory is correct, it could be solved by setting the offset to a round value before doing a calibration. (and the influence would be invisible)

Let me recapitulate:

1. The extruder offsets are added to the current position on a tool change.
   
2. We are guessing that the leaning has something to do with rounding errors of the Marlin planner (which "translates" the moving path to the number of steps).
   
3. the current suggestion is: rounding of the extruder offsets to a multiple of one microstep (1/369mm for the UM2+ feeder) and check, if this makes a difference
   

correct me if i'm wrong...

[neotko 1,417]

correct me if i'm wrong...

I think the main thing to check atm is like @ultiarjan tested, on Cura 2.5 it doesn't happen but Cura 2.6 does ?

[tinkergnome 927]

I think the main thing to check atm is like @ultiarjan tested, on Cura 2.5 it doesn't happen but Cura 2.6 does ?

Well..., it's stated above that it (probably) does not depend on the Cura version, because it happens with Cura 2.5 too - and even with S3D. The difference must be elsewhere...

That's why I ruled this out at the moment.