I was hoping for a little more detail here if possible. When the machine is measuring the points, I see the following…
Two belts let out slack
The remaining 2 belts adjust to move the sled.
At the end of these 2 steps, 1 belt is tight, 1 is a tiny bit slack and the other 2 are noticeably slack
The belt opposite the tight belt starts to wind in and when it tightens, it pulls the sled about 1/4"-1/2"
The other noticeably slack belt starts to wind in and when it tightens, it also pulls the sled about 1/4" to 1/2"
It surprises me that the calibration would does these Waypoint measurements this way. I would think you would need to move the sled with 2 belts and then slowly wind in the last 2 belts in a manner that it doesn’t pull the sled. Once the sled has moved like that, how could you even know what your trying to measure.
If belt 3 and 4 (the noticeably slack belts) were alternatively pulled in a few millimeters at a time, would this reduce the movement of the sled at the waypoint measurement? My assumption is that with the Vertical calibration, gravity is tensioning the tl and tr belts and tightening the other two does not cause the same movement.
As for the need to refresh the browser, I have posted in other places that I cannot get the machine to start calibration without doing this. I accept that this might be a cause for the problems I am seeing with calibration, but I don’t see any other alternative. EVERY time I try to calibrate, the Maslow will go into the Belts Extended state but will not start Calibration without a refresh. The Maslow output will simply print “Calibrate” to the screen each time the button is pressed but after that, nothing will happen. I have repeated the calibration at least a dozen times and everytime it is this same result.
I was hoping for a little more detail here if possible. When the machine is measuring the points, I see the following…
Two belts let out slack
The remaining 2 belts adjust to move the sled.
At the end of these 2 steps, 1 belt is tight, 1 is a tiny bit slack and the other 2 are noticeably slack
The belt opposite the tight belt starts to wind in and when it tightens, it pulls the sled about 1/4"-1/2"
The other noticeably slack belt starts to wind in and when it tightens, it also pulls the sled about 1/4" to 1/2"
It is supposed to tighten all 4 belts. If you have it in vertical mode, it
assumes that gravity keeps the top two belts tight, if you have it in horizontal
mode, it should go around and tighten each of the 4 belts (the order will
depend on what quadrent of the machine you are in)
If the belts do not get tight before going on to the next point, that gives bad
measurements which will lead to bad calculations.
It surprises me that the calibration would does these Waypoint measurements
this way. I would think you would need to move the sled with 2 belts and then
slowly wind in the last 2 belts in a manner that it doesn’t pull the sled.
Once the sled has moved like that, how could you even know what your trying to
measure.
If belt 3 and 4 (the noticeably slack belts) were alternatively pulled in a
few millimeters at a time, would this reduce the movement of the sled at the
waypoint measurement? My assumption is that with the Vertical calibration,
gravity is tensioning the tl and tr belts and tightening the other two does
not cause the same movement.
As for the need to refresh the browser, I have posted in other places that I
cannot get the machine to start calibration without doing this. I accept that
this might be a cause for the problems I am seeing with calibration, but I
don’t see any other alternative. EVERY time I try to calibrate, the Maslow
will go into the Belts Extended state but will not start Calibration without a
refresh. The Maslow output will simply print “Calibrate” to the screen each
time the button is pressed but after that, nothing will happen. I have
repeated the calibration at least a dozen times and everytime it is this same
result.
I have a similar problem, although not 100% of the time. When my machine does what you are describing, the only way for me to get it to do anything correctly is to reboot the machine via fluidnc and start the process over (which is really frustrating and time consuming when you’ve had to do that many times in a row - there needs to be a high speed retract/extend option so the start process doesn’t take 5+ mins every time). As @dlang mentioned, whenever I’ve tried doing something after I get those blank calibrate messages by reloading my browser, I’ve just had bad results.
I have tried both windows and mac, and various browsers on both (safari, chrome, firefox). I haven’t had repeated/consistent success with any option.
During the calibration waypoint check, all 4 belts do tighten before it moves to the next point.
My observation is that when the 3rd and 4th point are tightening the sled gets pulled to a new position each time a belt tightens. I don’t know how you could determine what your measuring since the sled keeps moving. I am trying to understand if this is the expected outcome or if the calibration is not running correctly
I would think you would want to use two belts (tl and tr in vertical vs any 2 belts in horizontal) to move the sled to a location and then slowly tighten the other belts without moving the sled anymore. By jerking the sled 1/4"-1/2" twice while trying to measure, I would think any calculation would then be completely thrown off.
During the calibration waypoint check, all 4 belts do tighten before it moves to the next point.
My observation is that when the 3rd and 4th point are tightening the sled gets pulled to a new position each time a belt tightens. I don’t know how you could determine what your measuring since the sled keeps moving. I am trying to understand if this is the expected outcome or if the calibration is not running correctly
you don’t know where you are on the frame, all you are doing is gathering data
points where all 4 belts are tight, it’s only after you know where the anchors
are that you can figure out where you were when you measured that point.
I would think you would want to use two belts (tl and tr in vertical vs any 2
belts in horizontal) to move the sled to a location and then slowly tighten
the other belts without moving the sled anymore. By jerking the sled
1/4"-1/2" twice while trying to measure, I would think any calculation would
then be completely thrown off.
@bar needs to weigh in on this, but I don’t think the position where you are
matters, just the fact that you have all belts tight at whatever point that is.
After reading @dlang and @boslaw comments, I decided to change computer and browser. I am now uses a Win 11 laptop using the Chrome broswer.
I ran the calibration starting with a power cycle followed by retract and then extend. The calibration started without refreshing the page.
However, with a calibration size of 1500 x 1000 with a 9x9 grid, the calibration measured 5 waypoints and then started doing the fitness calculation. Shouldn’t this be measuring more points than this? Is there something else in my setup that would cause this. If I get time this week, I might look through the code on GitHub to see if I can get a better understanding of the calibration steps and post calibration calculations.
Again I attaching the serial output in case it helps anyone. Maslow-serial.log (9.2 KB)
My calibration process does a similar thing where it will often start calibrating before all waypoints have completed. On the other end, after seeming to complete calibration, it will sometimes go back and start more waypoints at ever increasing numbers beyond what it should. There’s no rhyme or reason to it.
It starts with an initial 5, then measures fitness, then does a 3x3, measures fitness, then 5x5, then 7x7 and finally 9x9, measuring fitness in between all. I suggest for the initial 5 on a horizontal setup you manually pull out the slack belts, otherwise they can tangle on the spool and it becomes a nightmare. I had to try a variety of different calibration grid sizes and calibration forces until i could reliably get it to calibrate across multiple days. I found a higher calibration force around 1800 seemed to help with this
For the record i also have the unresponsive machine issue where you send commands and it does nothing and have to refresh browser or sometimes power cycle the maslow. This is coming up more and more often
The only thing i do different is manually manage the slack belts in the first stage, if they get a loop in them it will result in a bad measurement and a failed calibration. Otherwise its the same. As i mentioned above though, i had to try heaps of different calibration grid sizes and calibration forces before i could get it to work reliably, i imagine the nuances of everyones different setups might make it difficult. Ive never had it power off in the middle of a calibration like that though. If it were me id just power cycle it and start the process again
Yeah, which I would have done if it wasn’t my 1000th time through this process. I’ve spent full 8 hour days already retrying stuff. I’m at the point now where I can only handle 1 or 2 tries at a time before I have to pack it up and walk away. I have built lots of things, from 3d printers to pinball machines to computers to whole houses. None have given me this kind of trouble where there is no solution other than to keep trying random things and hope they work.
I suggest for the initial 5 on a horizontal setup you manually pull out the
slack belts, otherwise they can tangle on the spool and it becomes a
nightmare.
note that this is just managing the slack until the maslow pulls it tight, not
putting enough force on the belts to affect the measurements.
I think that we have too many folks with too many different issues going on in the same thread. Let’s try to keep this one focused on @caddad’s issues and start new threads for other issues.
I’m working on a calibration walk through video for this week’s weekly update, I think that the best thing that anyone could do right now is to set things down and wait for the instructions to be updated. Sometimes just changing things randomly can make things worse not better.
A LOT of the things that I’m seeing described here are totally normal behaviors that are working right, I think that a video showing what’s expected would help a lot to make things more clear.
This looks totally normal to me. The green light on the PCB being off here means that it’s a power supply issue. Is there any chance that the outlet that it’s plugged into could be an issue?
This is an issue with one of the encoders not being connected. Spraying the connectors with compressed air or contact cleaner could help.
@bar I did just rebuild one arm to replace a mangled belt. Might be the BR one (not sure of what order i reinstalled), but I also got the TL having errors.
The soldered connector pins extend past the board a significant amount. So much so that they might come into contact with the bearings below them and short.
I trimmed mine and then put a piece of electrical tape over the pins.
Oh, good to know. However, my first calibration attempts (before I mangled a belt) didn’t have this issue. I didn’t unmount the motor or board during the belt replacement.
