I have noticed that the maslow doesn’t always have all arms respond to initial movements.
For example, when I am releasing the tension a lot of times, only three out of four arms will respond and I will have to click the button again to get the fourth arm to respond. The same thing with retracting belts, sometimes there are one or two arms that don’t respond on the first click.
This has got me wondering about my calibration. I really have no idea what is going on during calibration. From what it looks like, there’s usually one or two arms that is slack, but it does not seem to follow a pattern, so this makes me think that maybe my calibration is off or so low because at some of the different points some arms are not pulling in the tension when they should or not releasing when they should.
Can anyone describe to me what the arm should be doing at each calibration point?
That makes some sense because what is happening when you release tension is that it’s giving the arms the option to extend if there is some force pulling on them, but it’s not actually driving the motors to extend the belts. If one (or more) arms doesn’t have any tension on it it won’t release.
This I think has more to do with the retraction force setting. Bumping up the amount of force that the machine uses when retracting the belts should fix that:
That helps a little bit but my setup is horizontal.
I just need confirmation that when the Maslow is measuring, all four belts should be pulled tightly.
I think my problem is that one of them was not being pulled. Tightly in my retraction forces were getting up close to 2000.
I’ve already torn down the machine another time and am polishing the spool and trying to realign the motor as I saw in another forum post to try and make sure there’s no resistance in my retraction forces are reasonable.
It was never stated in the video, but from what I can see in horizontal mode, there is one belt that should be released in tension and then pulled tight at each measurement point.
Bar I was initially frustrated with your reply but I think your wisdom has prevailed.
I noticed that my machine does not release its top left motor during the last two or three measurements of the 3x3 grid.
I think this is what is throwing off my fitness and causing it to calculate forever.
The initial fitness score during the first part of calibration where two of the belts go slack got up to 0.7 or 0.8 - so it seems pretty promising, but I think that because the top left belt does not release, it is causing some kind of kind regression in the fitness calibration as it moves to the 3x3 grid - it just seems to get worse unlike yours in the video.
I have the calibration and retraction values at 1300.
I may try higher but I fear I need to take apart that arm and get it nice and smooth.
If you only ran a 3x3 grid i’d double check what dimensions it generated when it calibrated (if you don’t have the serial log, check what’s in the Maslow.yaml file). It should be similar aspect ratio to measured dimensions, and likely slightly smaller.
In terms of what the calibration actually does (for explanation):
It puls-tight-and-measures the belt lengths at multiple points, and uses that set of belt lengths to generate an estimate/model of the dimensions of the frame. The fitness is a measure of how well the set of belt length points match the current estimate/model.
There is a chicken and egg element to this in that you need to have the frame size estimate/model calculated to move the maslow round with tight-but-not-over-tight belts, but you need tight-but-not-over-tight belts to move it around to get the data from different points.
That’s got around by working incrementally - the first 5(?) points it slacks the belts that are not being tightened for each move (so if it’s travelling right to the next point, slacks the left belts), then re-tightens them at each point. Edit: this is in horizontal mode, I think vertical it always slacks the bottom belts but not certain). After those first 5(?) points it runs an initial fitness calc, and uses this while it moves the Maslow round to collect most of the points.
So you shouldn’t expect all the belts to be always perfectly tight doing the movements within the calibration, but they should all tighten up at each waypoint.
The frame size that it generated was the first thing that I checked.
The generated file is -0.8, 3125, 4592, 3111, 4590
The frame size that I measured was 0, 3124, 4584, 3124, 4584.
They seem to be pretty close.
Thanks for the explanation about the calibration process, but the most useful part of it for me actually was trying to discern when a belt should be slack and when a belt should be tight. My belts were having difficulty releasing and I think they still do to some extent, when I hit the release all button I usually only get one or two to release and I have to click it a few times.
But what I find out is that the very first measurement phase as bar describes in the video. Two belts are tight and two are slack. Then after that first measurement three belts will be tight and one will be slack. At each waypoint one belt should release tension and then retention. I was noticing that I had one arm that was struggling to release tension and that was the one that I had to pull out and smooth down.
My retraction and calibration current are both set to 1100. Do either of these settings affect the machine’s ability to release tension?
Nope! Those both relate to how hard it will pull in.
The releasing tension mechanism relies on detecting when the belt is pulled on (by seeing the encoder register outward belt movement) and then the motor will spool out more belt to let it keep extending