0.68 calibration failure and belt chewing

Troubleshooting
1

Ugh. I setup a new 4950x3150 “frame” using the 3d printed anchors glued to a concrete floor , with a 2400x1200 sheet in the middle as spoil board.

First calibration run had a terrible fitness score but completed the procedure without issue.


Serial Messages
[GC:G0 G54 G17 G21 G90 G94 M5 M9 T0 F0 S0]
$/axes/x/max_travel_mm=2438.400
$/axes/x/homing/mpos_mm=0.000
$/axes/x/homing/positive_direction=true
$/axes/y/max_travel_mm=1219.200
$/axes/y/homing/mpos_mm=0.000
$/axes/y/homing/positive_direction=true
[MSG:INFO: Channel auto report interval set to 50 ms]
[GC:G0 G54 G17 G21 G90 G94 M5 M9 T0 F0 S0]
[MSG:ERR: Calibration data not acknowledged by computer, resending]
CLBM:[{bl:2325.28,   br:2363.43,   tr:2360.48,   tl:2362.99},{bl:2174.49,   br:2176.45,   tr:2562.44,   tl:2590.23},{bl:1513.88,   br:2866.65,   tr:3173.03,   tl:2075.02},{bl:1646.96,   br:2973.33,   tr:3079.84,   tl:1870.12},{bl:1831.93,   br:3080.19,   tr:2969.67,   tl:1682.54},{bl:2035.38,   br:3205.85,   tr:2880.00,   tl:1518.80},{bl:2276.53,   br:2894.26,   tr:2524.56,   tl:1839.87},{bl:2135.45,   br:2750.98,   tr:2596.45,   tl:1978.17},{bl:1973.34,   br:2627.08,   tr:2717.81,   tl:2134.31},{bl:1835.69,   br:2525.36,   tr:2857.22,   tl:2308.25},{bl:2146.71,   br:2176.56,   tr:2594.59,   tl:2596.48},{bl:2255.49,   br:2292.59,   tr:2433.95,   tl:2436.25},{bl:2397.83,   br:2432.91,   tr:2292.84,   tl:2295.30},{bl:2553.44,   br:2589.30,   tr:2175.53,   tl:2178.12},{bl:2855.05,   br:2318.80,   tr:1836.66,   tl:2523.38},{bl:2749.51,   br:2137.36,   tr:1945.09,   tl:2635.23},{bl:2625.50,   br:1975.31,   tr:2107.26,   tl:2753.38},{bl:2523.63,   br:1837.85,   tr:2284.49,   tl:2890.37},{bl:2852.21,   br:1516.56,   tr:2072.29,   tl:3206.03},{bl:2939.00,   br:1680.53,   tr:1867.13,   tl:3081.68},{bl:3045.87,   br:1865.14,   tr:1679.25,   tl:2971.56},{bl:3171.45,   br:2065.07,   tr:1515.07,   tl:2882.06},]
Computing... This may take several minutes
Fitness: 0.013719239291010129 in 100
Fitness: 0.11208588594460811 in 200
Fitness: 0.1265253804978789 in 300
Fitness: 0.12662950327120728 in 400
Fitness: 0.12689730827713833 in 500
Fitness: 0.12710002768365702 in 600
Fitness: 0.12718271128078473 in 700
Fitness: 0.12726715459111998 in 800
Fitness: 0.12735304003614076 in 900
Fitness: 0.1274477048611932 in 1000
WARNING FITNESS TOO LOW. DO NOT USE THESE CALIBRATION VALUES!
Calibration complete 
Calibration values:
Maslow_tlX: -25.1
Maslow_tlY: 2195.7
Maslow_trX: 4145.4
Maslow_trY: 2186.0
Maslow_blX: 0.0
Maslow_blY: 0.0
Maslow_brX: 4157.6
Maslow_brY: 0.0

I retracted, extended, and calibrated again, this time filming in case there was something obvious happening preventing a reasonable fitness score.

Bottom right belt let out too much slack and ate itself.
Could this be an issue with using a large frame size? anyone got any other ideas?

2

I’ve had multiple jams with the bottom right belt. It does let out WAY TOO much slack during the calibration process on the initial moves.

1 Like
3

Agree. Bottom Right is the worst slack producer of the 4 arm units.

143dcf2549ccaebbb1c4d4b552762c0e0e84ad29-1
143dcf2549ccaebbb1c4d4b552762c0e0e84ad29-12000×1126 279 KB

Here is my motor that not only flipped the belt but jammed up completely to have to take the machine apart again.

1 Like
4

While it’s moving between locations I definitely recommend tending to those belts which are slack like this:

I’d like to work on making that more self sufficient, but I’m not totally sure how to. Its a bit of a chicken and the egg problem where until the machine knows the anchor point locations it can’t tension the belts properly, but it needs to use the belts during the calibration process.

I’m positive that someone will come up with a clever solution, but I’m not sure what it is yet.

5

Why not have the user extend 2 belts to the opposite corners, say upper right and lower left. When you pull tight, you then have the diagonal distance and can center the machine on that line. Next, let the user extend the remaining 2 belts to their anchors. Another pull tight and you can at least center the machine between all 4 anchors. Granted, you still don’t know the exact coordinates of the anchors. Would that at least help with a rough estimate (assuming the points are close to rectangular)?

1 Like
6

That would get us a rough estimate, but once we have a rough estimate what would we do with it?

7

Could it help to limit the amount of slack that has to be fed out when moving to a new calibration point? And, if the diagonals are the same, we know that we’re dealing with a true rectangle.

8

Unfortunately I’m not so sure :slightly_frowning_face: I could be totally wrong, but it seems to me that unless we have a good understanding of where the anchor points are we can’t really move around.

I wonder if we could do a spiral so we would basically take a couple of measurements near the center and then use those to find a decent calibration near the center…then we could expand out a little bit and add the next set of points to the mix creating a calibration which would be accurate over a slightly larger area…and repeating until we have the accurate calibration which works everywhere?

2 Likes
9

I think that would be a good idea.

David Lang

2 Likes
10

My bottom right also let out a ton of slack and chewed it’s belt up. I also noticed that when it was trying to calibrate low on the spoil board the too long belts wanted to catch on the frame. I’m mounted on a wall and operating vertically. I propped some scrap ply in front of the frame plates and that made tending the belts easier, so if you haven’t calibrated yet and you’re thinking vertically on the “stock frame” that might be a good idea for you too.

PXL_20240412_020448649
PXL_20240412_0204486491920×1440 183 KB

PXL_20240412_020440980.MP
PXL_20240412_020440980.MP1920×1440 185 KB

1 Like
11

Hi Bar,

There is an other way of knowing where the anchorpoints are. I mentioned it before in another post, but maybe that was not clear enough, so let me try again;)
If the machine can measure each side of the frame rectangle, and at least one diagonal, the anchorpoints are known.
By using the machine as a measuring device, and using just two opposit belts, this can be accomplished.

Arjen

2 Likes
12

@arjenschoneveld- You and I appear to have the same idea… Possible alternate approach(es) to M4 Calibration

1 Like
13

It’s possible, but I’m pretty skeptical that we could make it work reliably in practice. I think that getting those measurements precisely is going to be tough, and anything which requires a lot of steps for the user to do is going to confuse people. I would really like to focus on getting to a system where we can extend the belts, press calibrate and it’s good to go, no user input required. I don’t think that’s impossible at all.

I’m super open to the idea though and I would love to be proved wrong!

1 Like
14

And I would love you to be proved right, with a calibration sequence like you envisioned it.
Having said that, the other way needn’t have too many steps when we can park the M4 centered on an anchorpoint measuring 2 sides and the diagonal, and then park it on the diagonal opposite to measure 2 sides.
But I understand fully that you want to get your system working smoothly, so keep up the good work.

And thanks for the CAD files!

Arjen Schoneveld

2 Likes
15

I’m wondering why the bottom right belt seems to be the black sheep in many cases (mine as well).

2 Likes
16

I think that is an EXCELLENT question. It does seem to be a pattern.

17

During calibration the lower belts got very long and had a lot of slack at various points… so much so that they were getting caught on the frame, even with me watching and trying to prevent it. I wonder if them getting caught occasionally is confounding the calibration process by causing them to register as fully tensioned but being too long? but the lower right was loose a fair amount of the time even before it ate itself.

1 Like
18

Same issue on the vertical frame setup. The 2 lower belts already let out a lot of slack, and when the calibration starts and the machine moves down and then to the left, the lower belts let out even more slack. WAY TOO much. Its hard to keep them from getting tangled even when watching and tending them. And yes, they will get caught on the bottom of the frame if you are not careful during calibration.

1 Like
19

What I am not understanding is why so much belt slack is being fed out. You have
said that it already uses the same logic as ‘extend belts’ where it only feeds
out if the encoder shows movement, so why is more than a minimal amount of slack
fed out if nothing is pulling?

David Lang

20

When we use Extend Belts we can rely on the user to pull the belt out as fast as it’s being unrolled…ie if we spool the belt out faster the user will pull faster.

When the machine is moving between points on the calibration process that isn’t true anymore. The two motors pulling tight will move a set speed and then the belts that are getting pulled will also be pulled at a set speed.