There is for sure something really weird going on here. These points should be pretty much in a grid (other than the first 6 which are more random).
The things I would look out for is 1) Are all four belts getting tight at each measurement location and 2) Is the machine moving smoothly between the points. Generally does it look like the timelapse here: https://youtu.be/7X5S4LkapX0?si=N1kveXxRuyjBo7W2&t=293
I’ll pay extra attention to the 4 belts and get back to you as soon as i ran it
@bar It’s interesting that these frame dimensions are similar to my frame, and those are the two corners I have an issue with too.
Although the belt was salvagable, the spool was not. It was damaged beyond repair when the belt tore off. Waiting for my new spool.
To be continued…
Is the belt almost completely wound off the spool in the bottom left (and top right) corners? If so, I would suggest adding an extension on each belt. I used square aluminium tubing just large enough that I could terminate the belt ends in. I drilled two holes in each end, to allow me to drop a bolt through the belt end and terminated the other end on the wall with a 6mm bolt through a couple of angle brackets. It is important each extension is exactly the same length or you can’t get the Find Anchors to work.
@bar How does the software compensate for the different number of revolutions required as belt spool goes from full to empty?
Then the same would have to apply for the other 2 corners, assuming all 4 belts have the same length?
Yes you have to do all 4 belts, they need to be the same.
I did another find anchors run with my spools all sanded, and the newly fitted belt. Unfortunately the results are still the same:
What i noticed during the process was:
Does any of this make any sense?
Looking at the top Photo is the TR anchor point vertical. It doesn’t look like it in the photo, but that could be camera distortion. If it’s not vertical is the belt jerking as it rotates around that bolt?
That’s distortion. It’s just as vertical as the door frame
Have you still got a copy of the yaml file from 1.17? If you load Anchor point values into your current yaml file can you move the Maslow reliably? For the people who are having difficulty with Find Anchors, it appears to be BL and TR anchor points.
Thanks ian_ab.
My old yaml file was based on a wooden frame, now I’m working from the shop floor, so these are totally new anchor points.
I’ve manually entered the anchor points and entered the data from the anchor point calculator (see above) which produces reasonable results, but not accurate enough.
I’m working on a 1700x700 mm frame for a wooden boat. The vertical dimension is 2mm off (which could possibly be corrected with the Y-scale factor of 1.003), but it’s also slightly skewed. The cut frame matches up different with the 1:1 drawing if I flip it.
The whole purpose of this exercise is to end up with a symmetrical boat :).
So at the moment I don’t feel confident enough to cut the real thing.
I was just thinking about this, because I think I can hear the motors struggling at those 2 corners. Could it be that the TR belt is too tight, causing the other 3 motors to work harder, and then suddenly loose tension?
Have you considered going back to 1.17, running Find anchors on the new setup and checking the accuracy with the older version of software. Then load v1.24 again and see if accuracy varies, using the 1.17 Anchor point yaml file.
No i have not. I already know my anchor locations. I just wanted to find/confirm them with the new algorithm
Just to chime in @martijn; I also have issues with calibration and excessive belt tightness post version 1.17. I’ve now seen several users having similar issues post this version. For your reference: V1.21 broke a spool and my heart
Sorry to hear that Lasse. Bar was kind enough to send me a replacement spool so I’m up and running again. Right now I’m tweaking the X and Y scaling. Almost there…
By the way, I’m also based in the Netherlands 
