0.70 Calibration - Calibration Fitness Too Low and slack belts

1

I tried a few calibration runs with 0.70 tonight and wanted to share the results.

Summary: Calibration process worked well but failed to find a suitable fitness. The problem seemed to get worse on the second run. When on the outer 9x9 square, it seems there is an issue where 1 of the 4 belts will tighten, another belt will tighten, then the first belt will slack.

I did an initial calibration run using 9x9 with 1800x1600mm calibration area (my machine is 2240x2130mm). It proceeded through all the stages but on the last (9x9) stage I noticed that 1 of the 4 belts was consistently loose when taking a measurement. That belt was originally tight but developed slack as the other belt was tightened. I recorded a short video that shows the behavior. I didn’t see this on the 3x3, 5x5, or 7x7 steps. (Could it be that the software assumes it will stay tight but with the long lengths when the 2nd belt tightens it moves the sled enough to put slack back in the first belt?). The first 3 steps all had Fitness values >0.5 and save the calibration settings. But on the 9x9 the fitness was too low.

I think repeated the calibration process from the beginning again. On the second attempt, from the beginning the machine seemed to keep all 4 belts much tighter when moving (I assumed because we had improved info from the first calibration run). The 3x3 completed with an OK fitness (but significantly less than the fitness for the 3x3 on the first run). On the 5x5 step, the fitness was too low and it stopped there.

I was hoping that sequential calibration runs would improve the results but they instead seemed to become worse. I’m not at all sure whether to keep trying additional runs give the result is getting worse.

Open to all ideas about what my next step should be…

Attaching the logs from both the first and second calibration run and the video.


0.70 Calibration Run 2.txt (6.0 KB)
0.70 Calbration Run 1.txt (15.4 KB)

2

My son and I had nearly this exact issue today calibrating. We attempted a total of 5-6 times.

First attempt, horizontal, one of the arms did not spool in. Error code kicked out.

Second attempt, horizontal, we raised the calibration current value to 2300. We are one of the machines that required a 2300 value for retraction to work at all. We made it to the 7x7 phase, but on the left and right sides of the workspace, as indicated here, the top “inside” belt loosened significantly, hanging pretty limp during the tightening phase. Fitness too low warning stopped us. There were many over 4000 warnings during this run.

Third attempt, horizontal, lowered calibration current value to 1800. Got to the 7x7 phase. There were a few over 4000 warnings. Loose upper-inside belts. Fitness too low warning stopped.

Fourth attempt, horizontal, lowered calibration current value to 1500. Got to the 7x7 phase. Don’t think there were any warnings. Loose upper-inside belts. Fitness too low warning stopped.

Fifth/sixth/seventh, horizontal, then vertical, suddenly, there was an error where after retracting, extending, and then pressing “calibrate,” two belts would tighten, the others would not. Eventually, I raised the current value to 1800 and finally to 2000 to get it to even move. Finally it did, but fitness too low because of inconsistency in step 2 of 9. lower right arm would consistently not retract.

Eight attempt, vertical, raised calibration current value to 2000, oiled lower right arm. Appeared to complete first 9, with VERY long fitness calculation, then up to the 7x7. At conclusion of 7x7, fitness too low warning stopped.

Ugh…

First image shows the amount of slack in my inside arm on attempts 2-4.
Second image shows the calibration error where two belts would not retract at all on attempts
5-7
Final image shows that even on attempt 8, the inside belt is a little loose.

Attempt 8.txt (149.6 KB)

IMG_8927
IMG_89275712Ă—4284 7.02 MB

IMG_8928
IMG_89284284Ă—5712 5.7 MB

IMG_8929
IMG_89293024Ă—4032 2.49 MB

1 Like
3

I should have included in my first post that I also increased the calibration current value from 1500 to 2000 on the second attempt hoping that would improve the tension (it did not).

1 Like
4

Super helpful. I’ve seen this behavior too, specifically for me it seems to happen just on the bottom left corner point. I think that it has to do with the sequence that we are tightening the belts in. We will come up with a software fix, but short term reducing the overall size of the grid should help. That’s probably why you only saw it on the final 9x9 grid. It seems to happen in cases where three belts are all pulling one way while just one is pulling back the other way.

Yes, I think that this is spot on.

I also saw this with @boxomatic yesterday. I’m not sure why that is happening, but it’s on the top of my list to look into.

A lot of the things that you are seeing sound like software bugs so if you want a break I should have an update soon-ish which could help.

If you are up for trying again I would reduce the size of the grid (the mm size, not the number of points) and try with different starting values.

5

I think that the most important thing is that we need to make sure that all the belts pull tight while taking a measurement.

When you are adjusting the current settings, which of these are you changing? (Or both)

image
image1356Ă—234 17.3 KB

6

I have retract current at 2300
And calibration current at 2000 (as of last night)

All the changes I mentioned were to calibration current.

1 Like
7

I think that we should try to work on getting that number down, that’s the highest that Ive seen and it seems like it could be leading to other issues. Are all four arms the same or is there one or two that are stickier?

8

Yesterday it appeared it was one. But historically, hard to tell.

Recall that we’ve completely disassembled/reassembled/greased all four arms (multiple times)

1 Like
9

From a previous challenge

1 Like
10

Does it seem like there is anything different about that one arm?

11

No, though i did notice that after last night, there were some gaps between arms in the stack as they were all pulling in different directions.

We can disassemble them all and re-grease them again.

It seems to be responding appropriately to encoding. I.e. moving around the 3x3, 5x5 and 7x7 as I would expect. Aside from the looseness.

12

My hunch is that the machine is feeling the extra resistance and it’s thinking that the belt is tight when it isn’t yet.

Before taking it all the way apart again which is a hassle, it might be worth setting your “frame dimensions” really large so that when you press extend all you can pull that belt all the way out and then retract it all the way back in a couple times. That can help to re-pack the belt on the spool and help it to spin more freely.

1 Like
13

I was changing the Retraction Force value in the Setup → Config screen.

image
image1310Ă—996 61.4 KB

1 Like
14

it’s been noted recently that there is some slop in the motor positioning, and
that can affect how tight the motor/idler fit is. you may want to try (at least
on the one arm) back out the motor bolts and shift the motor away from the idler
then tighten them again.

David Lang

15

That one will change just the retraction force used when pressing “Retract All”, I’m thinking about adding an option to set the other one (the force used during calibration) to that same menu.

The new calibration system seems more sensitive to that variable than the old system was because the old system would have so much extra slack that it would build up a lot of momentum before the belt would pull tight.

1 Like
16

Bar wrote:

My hunch is that the machine is feeling the extra resistance and it’s thinking that the belt is tight when it isn’t yet.

Before taking it all the way apart again which is a hassle, it might be worth setting your “frame dimensions” really large so that when you press extend all you can pull that belt all the way out and then retract it all the way back in a couple times. That can help to re-pack the belt on the spool and help it to spin more freely.

I was initially thinking it was the extra resistance of the belt against the
idler. I’m now wondering if it’s the idler and motor being too close together.

This is after a couple people mentioned that they improved the fit of the idler
against the motor by shifting the motor within the slop allowed by the bolts.

but I agree, unspooling and respooling is much easier if that gets the arm to
work better with lower current limits.

David Lang

1 Like
17

I’ll give this a shot next time we have a few hours!

1 Like
18

Major changes after moving the motors.

See video:

7 Likes
19

I moved retraction current back down to 1300 and it works like a charm!

I have run into a few separate issues but I will post them on their own!

1 Like
20

This is BRILLIANT! What a discovery!

I will test too and then update the assembly guide. Great work!

1 Like