Hmm that’s not much of a bump. Its an improvement, but not a huge one.
If you just want to get up and running it might be worth playing with the excel spreadsheet which seems to work in cases where our math has a tough time.
OK, so the race condition I worried about having a browser in my office opened while I was out running calibration on the tablet is a real thing. I had to unplug the unit as both were telling it what to do.
It was doing really really well, got through all the rings and returned to home. but… the tablet was still computing 3 fitness scores, and when I got back to my office, it had also computed them and was looping seemingly.
computer:
fresh-better-0.69.3-computer.txt (31.0 KB)
tablet output:
fresh-better-0.69.3-tablet.txt (36.2 KB)
booted maslow again and this is the set of measurements that got saved:
I think for the communication back and forth between the firmware and web browser we have to put in something that can be sure that the UI that started the calibration is the only one that can give the calibration new values, or some other mechanism to prevent this kind of confusion between connections.
In a better world, I wish I could just have it always talk to my computer browser to do computations and the tablet to do actions such as extend/retract/calibrate, etc. Also while I’m wishing I wish the messages would broadcast to all connected devices and if there is more than one, then they could synchronize but that is probably not worth implementing…
1 Like
I did download it and pasted the CLBM data in. I’m not sure if this is correct, it doesn’t look like it to me.
The third run of a 9x9 grid size resulted in a 0.52. So 3 runs back to back 0.47 → 0.49 → 0.52
Log#3.txt (210.0 KB)
2 Likes
That log looks really good to me. I’m not sure what is triggering the “not safe to move thing” that seems like a bug to me.
If you do “Take Slack” instead of calibration are you able to move around with the jog buttons?
Yeah, I can jog it around with the 0.52 result. The belts do get more slack than I would like as you get towards the end of the plywood. IE, if you are all the way left the two belts on the right side are not taught. To the point you couldn’t “strum” them.
1 Like
Hi Frank,
Thius spreadsheet was originally intended to be a one-time-use test to see if better solutions to this physical problem actually existed. It wasn’t intended to be a user-friendly tool.
In particular, if you have a different number of calibration points, then the spreadsheet needs to be heavily modified to work. You can’t just paste in the data and have it spit out an answer. It also needs to have Solver run to get the solution. In addition, I don’t know if it will work right in Google Sheets.
Later today I will modify the spreadsheet and get an answer back to you.
I’ve been working on the math behind a solution for solving for the frame dimensions and its hard. I have however learned a lot about the nature of the problem and the general purpose solutions I know of that will work for this are quite a bit of work to implement. I am looking at some solutions that might take advantage of the physical constraints inherent in the problem to see if a reasonable solution can be had.
1 Like
Thank you, sorry I don’t have excel and this math is a tad above my knowledge level 
1 Like
So I modified the spreadsheet and ran Frank’s CLBM numbers, comparing it to the results I saw in the log file. Note that my calculations of fitness are lower than Bar’s, as I don’t understand exaclt how he is doing it.
The file is attached, but the summary is:
| M4 | Excel | Delta | |
|---|---|---|---|
| Maslow_tlX | 3.30 | 4.16 | 0.86 |
| Maslow_tlY | 2802.60 | 2795.62 | -6.98 |
| Maslow_trX | 3628.00 | 3635.70 | 7.70 |
| Maslow_trY | 2844.00 | 2834.85 | -9.15 |
| Maslow_blX | 0 | 0.00 | 0.00 |
| Maslow_blY | 0 | 0.00 | 0.00 |
| Maslow_brX | 3646.60 | 3653.07 | 6.47 |
| Maslow_brY | 0 | 0.00 | 0.00 |
| SSE = | 1650.72 | 1503.78 | -146.94 |
| Avg. Dist = | 3.0807 | 2.9697 | -0.11 |
| Max Dist = | 7.6559 | 7.3092 | -0.35 |
| ‘Fitness’ = | 0.3246 | 0.3367 | 0.01 |
You will notice that the results have improved a little, but not greatly. I tried it a number of different ways and always got the same result. It may be that this is as good as it gets for this set of data.
Now that I’ve got the spreadsheet set up for a 9x9, does anyone have any other 9x9 calibration efforts you want me to second guess? I would need the CLBM values and the log file.
20240425 Frank Wiebenga.xlsx (99.9 KB)
1 Like
Hi, I run today evening the calibration on the new firmware 0.70. I got really nice values and a fitness I remember correctly close to 0.9 (I think 0.88) on a 7x7 grid with 1260 x 550 mm. The values also matched really nice to my measured values.
After calibration I cut another Maslow Logo and a circle at a feedrate of 1500mm/min. The maslow logo took about 5 minutes. Very very quick compared to 3D-printing! I show you a picture! What feed rate is recommended?
5 Likes
Is that like one pass? I’m doing a whole sheet at the moment, but went with 40in/min and 0.125" step down and the estimate was around 2.5-3 hrs.
Looking good! I usually do 1800mm/min, but I’m no expert 
Two passes with 6mm depth in 9mm plywood. The tabs were 4mm high so in fact only 1mm in the workpiece. The router was set at the lowest speed. This time there were more chips then only dust. It is really important to cut fast enough that you lift up chips and not only rub the bit against the wood.
3 Likes
3 Likes
I finally got some real luck with v0.69.3! I’ve run 4 calibrations and had success with 2 of them. My best result has been a fully calibrated 9x9 grid with a fitness greater than .9! I have all the details and I’ll include them when I have time. I did have to increase the current threshold like I did for .69.1 but with more of the expected successful results this time. I think I also found some other issues that I’ll save for another thread.
1 Like
Yippieh 
1 Like
What current threshold did you increase to?
Here is 0.70.1 which has lots of quality of life improvements for issues small bugs and also speeds up computing the anchor point locations 10x faster.
-Fixes the issue with the random “Can’t move safely” after calibration has finished
-Fixes the issue with the heartbeat getting sent in the middle of the data from calibration process
-Makes the e-stop error messages more clear
-Gets rid of some extra info being printed when you connect to the machine
-Makes the calibration algorithm more aggressive about moving towards the goal so it gets there much faster
Keep that feedback coming!
index.html.gz (156.1 KB)
maslow.yaml (3.2 KB)
firmware.bin (1.6 MB)
After FW updates, must i recalibrate?
No, absolutely not. Although if you update your maslow.yaml file you will need to copy your anchor point locations over from the old one.
No need to update maslow.yaml here though
Edit: You should never ever need to calibrate again unless you want to. In theory it’s a one time thing unless you make a new frame.
Here is 0.70.2 which continues speeding up the calibration process by increasing the speed of moving between measurements by 1.5x and the speed of taking measurements by 2x.
index.html.gz (156.1 KB)
maslow.yaml (3.2 KB)
firmware.bin (1.6 MB)
1 Like