M4: Calibration fail observations, large portrait vertical

Troubleshooting
1

I ran a calibration today and realized where I went wrong and have some observations based on this frame. I have a 8’ W x 12` H at 15% slope. I used 600,600 offset and 2438.000 x 3658.000 for the anchor point coordinates.

Here are the results: Fitness 0.0384… 9700 cycles. WARNING FITNESS TOO LOW…

Observations:

  1. on the top and bottom and right the sled leaned over the spoilboard a bit, (especially on the right, which I anticipated and had a 1x3 up to prevent it from getting caught up in the frame…
  2. when moving from left to right (both on top and bottom of pattern), the maslow did not let up slack on the bottom left belt so it pulled tightly before it moved over, and there were a number of “over voltage” warnings/errors during some of these.
  3. the maslow was very tipped. I wonder if the anchor points need to be raised up because it seemed to tip upward a LOT during (especially the top half) of the calibrations.

Is there any data I can gather post mortem? I was not able to capture the output with copy/paste as I can’t seem to get the ipad to select all the data in the terminal window under the maslow controls. I took a lot of pictures of it, which I will post some stills.

I also noticed that the bottom right was getting caught on the dust hood and as I’m looking at it it looks reversed from what it ought to be, or I put the arms on in the wrong order. (bottom-right is on the bottom)

Advice? should I increase the offsets and try again?

2

I wouldn’t expect it to be tipping, that seems like something is up. Is it possible to capture a picture of that?

The first thing that I would try is to do Retract All → Extend All → Retract All and confirm that all the reported numbers are close to zero. That will tell us that there isn’t a magnet slipping which seems to be reasonably common.

If that looks good, I would try a really small pattern like 2x2 with large offsets close to the center of the sheet, like maybe 100mm square.

(I think we should switch from using the offsets to using the size of the calibration grid so I could say “2x2 points on a 100mmx100mm grid”)

1 Like
3

Here are a couple of shots. it looked more drastic in person :slight_smile: also, I think the other instances were because the sled was over the edge of the spoilboard at least a litte.

20240401_174754
20240401_174754622×829 183 KB

20240401_174756
20240401_174756622×829 176 KB

1 Like
4

I will try tomorrow to do as you suggest and report back.

1 Like
5

I would expect it to maybe tip a bit at times when all four belts aren’t tight, but then when the belts all get tight I would expect it to be level.

6

closeup of belt hitting dust hood. (I printed the 1.5" stl for this). Seems like the relief is on the wrong side or I put the arms in wrong?

20240401_175347
20240401_175347622×829 66.2 KB

7

after watching it go through this calibration, I think a tall vertical frame may not work as well as other orientations, at least for calibration. on the lower side, I don’t think it got very good measurements, as it moved from side to side quite a bit during the bottom belt tightening phase as the geometry of the belts seems to make this worse, the lower it goes. I’m not sure I’m phrasing that right, but maybe you get what I’m saying…

1 Like
8

Good catch! You want the lowest arm to be the one on the other side.

That might explain some of the tipping too because the arms are ordered to balance the forces.

9

So I snuck out during lunch and grabbed my maslow and re-assmbled the arms in the correct orientation. The installation gif is good, and I’m not sure how I did it in the wrong order before.

I did retract all / extend all / retract all and got:

[MSG:INFO: Retracting all belts]
[MSG:INFO: 3 pulled tight with offset -0.011]
[MSG:INFO: 2 pulled tight with offset 0.000]
[MSG:INFO: 0 pulled tight with offset -0.043]
[MSG:INFO: 1 pulled tight with offset -8.814]
[MSG:INFO: Extending all belts]
[MSG:INFO: All belts extended to center position]
[MSG:INFO: Retracting all belts]
[MSG:INFO: 0 pulled tight with offset 0.021]
[MSG:INFO: 3 pulled tight with offset 0.150]
[MSG:INFO: 1 pulled tight with offset 0.043]
[MSG:INFO: 2 pulled tight with offset 0.054]

I’ll try a smaller calibrate tonight… lunch is over :slight_smile:

Also had a side thought… I wonder if it would be useful to have a way to test your arms before getting things all assembled? it would be a “nice to have” for sure. taking things apart is not exactly a fast process.

Also brings up another question I had… Do you think it would be ok to run calibration without the linear rail supports/handles on? I’d think so as they probably are not getting any stress during calibration, with Z all the way down. Just going to save me some time as I will wait for my new arm/belt before doing any final calibration, since my top-right and bottom right belts are a little chewed up toward the inside of the spool from the tangle a few days ago.

1 Like
10

This one seems mildly concerning, but not the end of the world. Just something to keep an eye on.

It didn’t happen again or anything.

I 100% agree. What would we want it to do?

Wouldn’t the router fall over?

11

I was thinking of something like

  1. connect the board to power - may need to put the board on something like some cardboard?
  2. plug in one arm (motor and encoder) in a known port
  3. run extend (may need to ask user for “how much” to extend
  4. run retract
1 Like
12

no, I don’t think so… the supports meaning just these parts:

image
image622×829 49.9 KB

1 Like
13

Oh yes you can absolutely take those off :grinning:

14

be able to connect a single arm, no steppers to the controller then
extend/retract the belt the full length.

David Lang

2 Likes
15

I think that sounds like a great idea.

Where would we put it in the UI?

16

If needed, a separate tab, but could be a pulldown menu item

It’s not something you normally do, this is ‘expert mode’ type stuff (as
discussed elsewhere in the problems of writing software for non-experts to use)

David Lang

2 Likes
17

Give me some rough drawing of how it should look and where it should go in the UI and I can work on it

18

will draw it up later, but I’m thinking a series of buttons in a grid

top-left top-right bottom left bottom right topZ bottomZ
extend
retract

where extend continues as long as the belt is extending (so it would go just
slightly past max

I don’t know if the steppers can move independently, if not, just one set of Z
buttons (extend would be up, retract would be down)

each button would be a toggle, press to start, press again to stop

possibly with a distance readout for each of the 5-6 (if so, a button to zero
the readout)

ideally some would be able to be in extend mode, while others are retracting,
but that’s an advanced feature :slight_smile:

David Lang

1 Like
19

was just playing around with this, launched from settings/calibration settings (new button):

image
image787×734 32.1 KB

could also make your popup tabbed? I’m not crazy about that nor dialogs in general as they are easy to accidentally close

yea, adding another pane for the buttons David mentioned would be pretty easy too, maybe off of “test” button?

20

ok, btw, here is my calibration result from using 2x2 and insetting it quite a bit:
maslow-serial.txt (47.2 KB)

Excerpt:

Fitness: 0.37670880626693426 in 99000 cycles
Fitness: 0.37581898391581564 in 99100 cycles
WARNING FITNESS TOO LOW. DO NOT USE THESE CALIBRATION VALUES!
Calibration complete 
Calibration values:
Maslow_tlX: 18.4
Maslow_tlY: 3479.5
Maslow_trX: 2549.3
Maslow_trY: 3487.1
Maslow_blX: 0.0
Maslow_blY: 0.0
Maslow_brX: 2540.7
Maslow_brY: 0.0

I did video the entire process too if its useful. I didn’t see anything that concerned me.

1 Like