Rostock V3 Bed issue

[Th0mpy]

Those are very consistent, which is good. Interestingly, I'm seeing a pattern with several of us having very similar "level" issues. I also ran Z height calibration a number of times in a row and it was dead on each time.

I am experiencing exactly the same thing you are also along the Y tower direction. I want to do this test: auto calibrate and test print to verify the tilt and then manually calibrate and compare the results.

Do you suppose it has something to do with the position of the accelerometer on the circuit board in relation to the arms?

[mhackney]

Too early to tell but an interesting question. The manual calibration & comparison might shed some light.

[mhackney]

Turns out it should be easy to test the "which side the accelerometer" is on theory. It is easy to simply rotate the entire HE280 120° simply by unclipping the arm joints and moving it. I am setting that up now. It will be interesting to see if the out of level stays the same or follows the rotation.

[Th0mpy]

I had a similar idea, swap two sets of the arms. I am attempting that same thing right now. We're heating up getting ready to print, I've already run the calibration. Stay tuned.

[mhackney]

Swapping arms is not as good a test as actually rotating the HE280 by 120°C - at least to test if the position of the accelerometer is a contributor.

[Th0mpy]

I did just exactly what you suggested, turned the hotend -120 degrees (fan is towards the Y tower now). No change, still the same issue.

My original thought was to swap the arms to remove the arms from being the culprit, I may still try that. So far the accelerometer does not seem to be the issue, based on orientation anyway.

What did you find?

[mhackney]

Ok, good. So you still have the same tilt along the same direction as you did before the rotation?

[mhackney]

Oh, and I found that rotation had no effect. I still have tilt in the same direction as before.

[Th0mpy]

Yes, that's correct.

[mhackney]

I still need to do the manual calibration. It will have to wait until tomorrow.

[Th0mpy]

I was thinking along the same lines. I did attempt to swap the arms, and guess what, no change. Clearly something to do with the autocalibration and likely not mechanical (imho). I found myself a beer and am going to try the manual calibration on the endstops. I'll see where that gets me.

[mhackney]

Friends don't let friends calibrate while drinking

[Th0mpy]

Well, it didn't help anyway... I am having an absolute NIGHTMARE trying to calibrate this damn thing now. While trying to manually calibrate I found that not only does the Y tower have the issue, but the other two have issues as well. The hotend will be at the right height while probing near all of the towers, but as you go straight out from the tower, the X and Z axis go up while the Y axis goes down. I've tried to adjust the radius up and down and just have no luck.

Now, one thing i just realized. I was previously using Repetier Firmware 92.9, and as I recall, went to that version because I couldn't get things calibrated when I originally built the printer. I might throw that back in tomorrow and see if I can get back to some success printing again.

...and before you ask, when I loaded 92.2 from SeeMe for the new hotend, I did wipe the eeprom.

Any other ideas?

[geneb]

The only branch of the firmware that supports the HE280 probe is the SeeMeCNC branch.

g.

[Th0mpy]

Sure, I understand that. I would need to work to port those functions over. The problem I have, however, is that I did not have this issue on a later version of Repetier.

I am heading out soon for one last fishing trip, maybe something will come to me while sitting on the boat.

Just to confirm my findings last night (early this morning so I had to double check after getting a little sleep). The area in front of each tower, and center, have the same height. The areas between ZX are low while ZY and YX are high. What am I missing?!

[mhackney]

Missing nothing, this is what I'm observing and working to try to understand if its mechanical or firmware. It is suspicious that the 4 or 5 filks who have seen this all seem to be in the Y to Z-X direction.

[Th0mpy]

What if we shifted the tower angles by 120 within the eeprom, instead of (x,y,z) 90, 210, 330 we do 210, 330, 90. If it's firmware it'd move the low point; right?

[mhackney]

I haven't wrapped my head around what I'd expect to happen but it might be worth the experiment to try.

[Th0mpy]

Ok, just tried it... If I essentially turn the towers 120 as I said the low point STILL exists in the same place. That tells me that I've got a hardware problem; wouldn't you say? If it were software I'd expect the low point to have moved 120 as well.

I've got a 4 hour trip ahead of me and I'll continue to mull it over.

Maybe there is some basic flaw with how the towers are leveled to each other and the bed?

[mhackney]

Where are you fishing?

[mhackney]

I'm going to pull out my machinists squares and check things. I don't know what I would expect the results of your test to be, I need to think about that. The point of calibration is to compensate for small errors in the build. What we are seeing appears to be a leveling issue and "level" should be easy!

The thing that stayed consistent when you rotated the towers in firmware were the probing points themselves. If you notice, the delta radius probing is done at the center and base of the Z tower. That stayed the same in your test. The next thing I'll try is rotating the probing points themselves - this can also be done in firmware easily.

[Polygonhell]

Ok, just tried it... If I essentially turn the towers 120 as I said the low point STILL exists in the same place. That tells me that I've got a hardware problem; wouldn't you say? If it were software I'd expect the low point to have moved 120 as well.

I've got a 4 hour trip ahead of me and I'll continue to mull it over.

Maybe there is some basic flaw with how the towers are leveled to each other and the bed?

Not necessarily, the towers are still physically located in the same place and still labeled the same XYZ in the firmware.
If this is a software issue and I had to guess what it was likely to be, it would be an error in offsetting XYZ after computing the 3 offsets as part of the tower calibration, the only thing that affect that is physically swapping wires or changing the pin mapping in the firmware.
I haven't looked at the modified firmware in detail, The easy way to do tower calibration is to measure points below the towers, or at some known offset from them, and use a scaled Step offset foe each tower. However I would assume that they then either nominate a tower as the 0 offset tower or generate the mean of the 3 offsets and displace the calculated offsets by that, this last part is what I would suspect in the software.
You could get more complicated than this and measure more points, best fit a plane, and compute the plane intersections at the towers, but I suspect it would actually increase the error, rather than reducing it since your introducing other errors.
If it's a software or an obvious mechanical issue, after doing a tower calibration, manually checking the offsets should show a clear error at one of the towers.
If it's software physically swapping the tower wiring should swap the tower affected.

[mhackney]

I'm moving all of my commentary over to my build thread to keep it all in one place.

Thanks Polygonhell, I did a test with rotating the probe points. Basically, the delta radius calculation relies on 2 points - the center and the X3 Y3 point configured in firmware - which by convention is at the base of the Z tower. I swapped points 2 and 3 to force the probing for delta radius to use the Y tower. The results were unchanged, the bed was still tilted in the Y to X-Z direction. I'll try your recommendations here in a bit.

[Th0mpy]

Where are you fishing?

Northern Wisconsin. Day 1, so far so good!

[morgandc]

I have reprogrammed the probe points to match yours, during the next recalibration I still had "head tilt". I also ran a 1st layer test after and am still having issues with bed leveling. I will continue to dig. The good news is any testing/builds in the center of the bed work well