4.1 Calibration Initial Fitness

Originally, I had appended the issue I was seeing in the 4.1 Calibration Issues and Questions thread, but @bar asked that we start new threads for different issues.

I am attempting to calibrate a Maslow 4.1 in a Horizontal setup. I have drop-in anchors at roughly 12’x8’. I have measured the 4 side lengths and the corner-to-corner lengths. This gives me the following coordinates.

  • bl: (0, 0) mm
  • br: (3635.38, 0) mm
  • tl: (20.79, 2447.25) mm
  • tr: (3658.93, 2433.91) mm

I cannot make it through the initial part of the calibration process. The Maslow will record Waypoint 0-6 and then attempts to calculate the fitness of these points. It seems to get as high as 0.37 but eventually over time it will continually return values in the 0.24-0.25 range. It doesn’t matter how long I let the browser try new calculations. The fitness never makes it to 0.45.

Everything else on the machine seems to work well. The assembly went smooth. I purchased a 4.0 machine in the original Kickstarter and never assembled it. I started assembly last week and used the 4.1 replacement parts during my initial build. All the encoders seem to read correctly. If I perform extend all followed by retract all, my largest offset is usually 0.04mm. My belts are extending to 2150mm which gives a couple of extra inches on the extend all.

I am attaching the latest serial output from my last attempt at calibration. Can anyone help figure out what is going wrong that is preventing this from calibrating? I know that @bar has been working on a calibration video, which I am hoping comes out this week so that I can attempt this with a follow along guide.

Maslow-serial.log (51.3 KB)

1 Like

A couple of things stand out to me:

From the log I can get your measurements:

CLBM:[{bl:2045.40, br:2099.78, tr:2313.16, tl:2301.12},{bl:2192.73, br:1971.75, tr:2200.15, tl:2412.66},{bl:2266.99, br:2046.11, tr:2106.11, tl:2338.21},{bl:2174.65, br:2168.40, tr:2194.25, tl:2215.47},{bl:2049.89, br:2293.30, tr:2320.18, tl:2094.61},{bl:1966.39, br:2231.56, tr:2401.76, tl:2172.57},]

Plugging those into the simulator to see if anything looks off, it looks like it’s just your first measurement which is throwing things off. If I remove the first measurement then everything looks good:

I think that the number one thing to do is to hang tight until I finish filming the walk-through video of the calibration process. I think that if you follow along with the steps that I usually do it’s going to fix that issue with the first measurement.

Re the video where the machine turned off. I didn’t realize that you had the 4.1 upgrade installed. I think I got confused about who had what setup in that other thread because there was a lot going on.

The first thing that comes to mind is this issue:

Does that match the behavior you saw?

Thanks @bar. I will hold off until the calibration video gets completed. No point in giving myself more gray hairs!

I think the second part of your post about the machine turning off is not for me. I have not had any issues with control board power.

1 Like

Right, I kinda got everyone in that thread mixed up

I got a successful calibration on the 2nd attempt. Even the first attempt was way better than any previous attempt.

On the first attempt, it made it through the initial 5 point calibration and then the 3x3 but got stuck on the 5x5. EVERY previous calibration had failed on this initial 5 point calibration.

On the second attempt, it made it all the way through and saved everything the to maslow.yaml file, which I immediately backed up!!

I think the only step that was different in the video vs what I had been doing is…
Before starting calibration, pull the sled so that the Tl and Tr belts are tight

I’m pretty excited to get started on a few test cuts!! Thanks again to @bar and @dlang for your help. I’m sure I’ll have more questions, but I’m feeling good about the progress so far!

1 Like

Excellent work!

A video can be so helpful for catching those little things which turn out to be important that I didn’t even know mattered :grinning_face:

You had pointed out when viewing the original 6 belt positions that from my previous calibration that the “Waypoint 0” was keeping the algorithm from finding a good fitness.

I guess this makes sense but also seems like a flaw by assuming that this point will be at the position where the Tl and Tr belts are tight. Is this leftover from the vertical frame design where the sled is hanging during that first Waypoint measurement? Is there a way to work around this so that the user does not have to pull the sled “down” before starting calibration?

My final fitness was 1.182.
Comparing the values from the calibration to the values that I hand measured, the Bl->Br and Tl->Tr are both around 3mm off my measurement. However, both the Bl->Tl and Br->Tr are about 15mm off my measurement. I am tempted to run the calibration again to see if I end up with a different result.

Calculating distances between the new points:

  • bl-br: 3638.6 mm (143.25") - Original measurement: 143.125"
  • tl-tr: 3657.2 mm (143.98") - Original measurement: 143.875"
  • bl-tl: 2434.4 mm (95.84") - Original measurement: 96.375"
  • br-tr: 2420.2 mm (95.28") - Original measurement: 95.875"
  • bl-tr: 4366.4 mm (171.9") - Original measurement: 172.375"
  • br-tl: 4385.7 mm (172.67") - Original measurement: 173.0625"

Comparing to your tape measurements, the differences are:

  • bl-br: +0.125" (3.18 mm)
  • tl-tr: +0.105" (2.77 mm)
  • bl-tl: -0.535" (-13.53 mm)
  • br-tr: -0.595" (-15.03 mm)
  • bl-tr: -0.475" (-11.93 mm)
  • br-tl: -0.3925" (-10.09 mm)
1 Like

That’s exactly where it came from. I don’t have a good workaround off the top of my head, but I will think on it and see if I can think of a way to make it handle that initial condition better.