What must be adjusted to correct out of round circles

I think this is the key, the zero point is not the end of the link, it’s the middle of the last link. Take a look at the left side of the first diagram in the post (ignore the fact its a 7-tooth sprocket). When you set a tooth to 12 o’clock and press “define zero”, it uses that point as zero. When you place the first link on top of the tooth, the middle of the link is what coincides with the zero point. So when you measure rotational radius, you measure from the center of the router bit to the center of the first link of the chain. I just did another diagram, to scale, that suggests the rotational radius is 139.1 mm. The real answer is somewhere around there.

I couldn’t agree more.

I get the same thing.

I wish I had an answer. I cut out a large rectangle (24x36) the other day and besides being two short (calibration issue), it came out straight. I try not to lose sight of the saying “perfect is the enemy of good”. I wonder if that shouldn’t just be an unofficial motto of Maslow.

3 Likes

OK, I live around the good vs. perfect thing…

Did you advocate an additional run through the triangular cuts? Will this do any good?

I do not think I want to mess with the chain calibration, as most reports suggest worse results.

And finally, Let’s make a t-shirt with “139.1”. It will be the password to our secret society!

Thanks again!

2 Likes

first off, see how bad the problem is. The wiki used errors large enough to be
obvious in the diagrams, it may be close enough for your work.

David Lang

1 Like

Working on a file now to test some dims for cabinets I am planning. Will report on the results.

Thx!

4 Likes

This is what I am thinking more and more might be biggest source of our problem using chain stretch calibration. I suspect the wear is mostly limited to the portion of the chain that runs over the sprocket and even that is not linearly distributed because some portion of that portion runs over the sprocket more than the others. Also, I suspect that end of my chain (that connects to the sled) is also suffering from wear because I’ve gotten wrapped around my sprocket too many times. I don’t know that this can readily be solved if the case. I have a brand new set of chains to try out and will see how they perform.

2 Likes

we can setup a test for this. link both chains together with the master link,
pull them tight to measure the length, then advance a bit and repeat, see how
the lengths vary.

If you think the end of your chain is a problem, reverse the chain so the other
end is at the sled.

David Lang

1 Like

So, back in early April I started this topic, and I am pleased to report the today I produced a round circle in the center of my Maslow!

Now, for the rest of the story…

First, the circle was part of a test file I created based on some cabinets I hope to create with the help of the Maslow… Here is a photo of the completed test:

The rectangle is made from a file that will be the back wall of a 32" wide x 42" high x 15" deep wall cabinet carcus. The photo is oriented the way it was cut, but the bottom is on the right. The squares and circle were added for test measurements. The test was cut from 1/4" ply with a 1/4" bit. The final will be on 1/2" ply.

Based on the results of my latest Calibration Benchmark, I did not expect the circle to be round, as the 3 long vertical benchmark measurements were short, with the center one off the most, 3mm. All the horizontals were close or dead on. But the circle should have been 17 7/8" in diameter (set the cutting path to “on the line” instead of “outside the line” by mistake). Instead, it was 17 11/16", 3/16" small on all around.

Most everything else on the cut was also slightly small…
File Cut
Full Panel 42w x 31.31h 41 15/16w top, 41 7/8w bottom x 31.19 h
Dados/
Vert Pockets 3/4w 23/32w
5" Squares 5 x 5 all dims short by range of 1/64 to 1/16

Measuring the diagonals of the entire panel, there was 1/16" difference between the two was 1/32" over 52", virtually square. All the edges seem straight.

The full panel is off 1/16 wide x about 1/8 high, while the circle is consistantly 3/16" small. Makes no senes to me.

If I use this cal to make the cabinets, other than widening the dados, as long as all the parts are cut with the same orientation, I think it will work.

I know, easier on a table saw, but hey, this is not a table saw forum!
Also, sorry about mostly English with a few Metrics mixed in. The old brain is harder to re-calibrate than a Maslow.

8 Likes

Update!

After my Aug 28th post, I wanted to do a couple of signs, thinking that they would be relatively easy, and also to show my wife that the Maslow can actually do something other than endless calibration. Well, when the machine is doing pocket cuts. I noticed the sled doing a lot of “bobbing” up and down, rocking as the machine made sudden movements up and down the workpiece. This caused uneven cutting depths. @dlang had commented from a photo I posted that the ring was too high on the sled.

I felt that if the sled was balanced, this would not be a big deal. Of course, David was right. Because of the way I had added my beam, I needed to mount the motors forward on the beam so that the chain cleared the mounting brackets. and to get the chains parallel I needed to raise the ring. So, this issue needed to be addressed. In my mind there was no practical way to change the beam mounting, so I “packed out” the board mounting area to raise the workpiece, adding 2x4’s on all vertical and horizontal board mounting areas. This raised the workpiece 1.5", and made my Maslow even more solid (and heavy!)

With all the turmoil over how folks are going to get kits, parts, etc, I have been getting a few key items so my baby can keep running (spare motor, Router parts). @Bee’s Blue Smoke Herder seemed like a good way to go for back up electronics

so I ordered the Option 2 Maslow Brain (Sheild and Arduino Board, along with enclosure and cooling fan). It arrived just at the time of the frame upgrade (Brandon is finalizing the enclosure and fan, so that is yet to come). The board is too cool to be a back up, so it was installed, V1.21 firm and software was enabled, the ring was lowered, and on to calibration.

To get to the August 28th Calibration, I used the Maslow chain measuring tool to ensure that my new chains (for the new 12’ beam) matched in length (they do). But my steel rule measurement was 13mm longer than the machine’s. I felt that if I got the machine’s number to the correct result, all would be good. So I adjusted the Chain Pitch from 6.35 to 6.37 (getting there was half the fun) and got the machine to agree with my ruler. After a couple of “Triangulation Test” cuts, the following was reported by Maslow, which led to the Aug. 26th result:
Distance between Motors 3605.72
Vertical Offset 578.6
Rotational Radius 136
Chain Sag 42.649137
and the following Benchmark result

Fast forward to the last couple of days…
Using the Manual Calibration process, the ablve values were entered, except for the Rotational Radius, which I changed to 139.1 (the consensus “correct” value). The results were pretty bad…

Ran the Maslow chain measurement, along with confirming my steel rule number, and found an additional correction was needed, which took my chain pitch adjustment to 6.385. Went through the regular calibration process, and got the following numbers after a couple of Triangulation Tests…
Distance between Motors 3606.09
Vertical Offset 556.8
Rotational Radius 126 (Yikes!)
Chain Sag 43.186239
and the following Benchmark result

Deep in the weeds, I retreated and re-read @madgrizzle’s piece on Maslow Precision and Accuracy

This is a MUST READ to get some kind of understanding about how the system works. Also follow

to get an picture about how the calibration software works, and an innovative new approach to the process (a work in process).

From the Accuracy article, “If the controller thinks the motors are further apart than they are, then it will feed out too much chain and will make the vertical distance of your cuts lower, wider, and actually shorter than it should”. For sure, my long vertical measurements have always been too short, and when I made my Paulk Workbench (early version of machine)

the long horizontal cut at the top did just what the simulator shows, bow down.

And from the “Holey” article, that the Maslow calibration uses the Rotational Radius value to correct other problems. The fact that my Benchmark long verticals are short, and the Rotational Radius is radically adjusted smaller, leads me to believe that the Maslow is sending out too much chain. Measuring 10 links of my chain in several places, it was clear that my chain links are pretty close to the correct size of 6.35mm.

So, the chain pitch was reset to 6.35, and I ignored the discrepancy between the machine and steel rule measurement, used “hard” measurements in the manual calibration routine for Distance between Motors and Rotational Radius, and the latest Vertical Offset and Chain Sag values…
Distance between Motors 3606.09
Vertical Offset 541.338
Rotational Radius 139.1
Chain Sag 43.186239
Set the chain length, ran the Triangulation Test, and got the following:
Distance between Motors 3606.09
Vertical Offset 579.4
Rotational Radius 145.4 (again, YIKES, but…)
Chain Sag 10.826962 (smallest number ever seen by me)

Ran the Benchmark Test, and got the following


Which is probably the best result I have ever gotten, especially the long verticals. I ran an 18" circle, but my test board is not real flat, so it did not cut everywhere… But where I can measure it, it is the prescribed 18" diameter.

So, at least for now, I am happy!

9 Likes

Fantastic post that is a wiki in it self. Thank you for the clear summery!

3 Likes

Nice write-up. The high rotational radius is compensating for something. I’d have to guess it’s chain tolerance, but surprised (and happy for you) about the benchmark scores.

the fact that setting the chain link size works but setting chain tolernce
doesn’t says we have something very wrong.

David Lang

The firmware I’m testing changes how chain tolerance is used. It just modifies the chainStraight calculation in triangular inverse. I think this is simpler and hopefully results in improvement.

And, yes, something is wrong since both chain pitch and chain tolerance modify the exact same value in the firmware (distPerRot)

1 Like

My question is, does anyone get the correct chain measurement from Maslow’s chain measuremt tool, compared to a phisical measurement

2 Likes

right now, no.

If you think about it, the chain has to have some clearance on each pin to be
able to rotate. If that clearance is a mere 0.001" (0.025mm), across ~3m you
have 470+ links, this results in an error of over 11mm.

stock chain is rated as having an error of +0 to +0.15% (and it’s typically
replace when wear makes it have an error of +0.3% or so)

We know we need to account for this error (this is what the chain tolerance
values are for), but there is something in the current code that is not using
them appropriately.

We also know the chain sag calculation is invalid and the calibration
corrections are bad.

This is something we have a bunch of people working on right now.

David Lang

1 Like

Seems like there is an issue of offset. Is this related to using the “second link” as is recommended due to safety? Possibly, the Maslow measurement of “distance between motors” may be missing a link. I am trying to remember, but I believe there were other threads where the “distance between motors” was measured as ~6 mm short. In your latest (best) calibration, you got a Rotational Radius of 145.4, which is incredibly close to the length of one chain link larger than what it is supposed to be, 139.1. @Dustcloud, could you confirm the difference between the steel-rule measurement and Maslow measurement for “distance between motors”?

Can someone double-check if there is some mistake (one chain-link off) in the number of links that are assumed to be wrapped around the sprockets when measuring the Distance between Motors?

1 Like

It removes two links of distance (2*6.35mm)

1 Like

Well, when I first started the calibration process with the new 12’ beam, I measured the motor spacing with the rule, and got 3606 mm, whereas the machine measured 3593 (13mm difference!) I then made corrections to the chain link value, from 6.35 to 6.37, which got the machine to something close to the rule measurement. At that point the calibration was pretty good. Then I rebuilt the frame to move the workpiece closer to the sprocket plane by 1.5 “, lowered the ring to get the chains back in parallel, and removed about 2 lbs from the sled to re-balance it. Thus another calibration adventure, where the machine chain measurement with the 6.37 chain link value came up short again (3601). This was corrected with a new link value of 6.385 , and one of the worst cal results I have seen (benchmark chart above), which included a rotation radius result of like 125. Rather than trying to force the machine to do a correct top beam measurement, I went to the manual cal, using the steel rule measured motor spacing, assumed to be correct 139.1 rotational radius, and the latest calculated chain sag and motor offset. Maslow increased the rotational radius to 145 to deal with the chain now being too short, and I ended up with a pretty decent Benchmark. Although I am happy to have what I think is a working Maslow, I am not sure that what I did is repeatable. For sure right now I just want to make some stuff.

2 Likes

Interesting, this is the third time I have seen messing with the rotational radius and how it increased accuracy.

See here: Fine Horizontal, Short Vertical - #3 by xrok1

1 Like

To be clear, I did not mess with it… the values I entered were what I think to be correct, at least the rotational radius of 139.1, and the motor spacing of 3606. The chain sag and motor offset were the latest values I got from the previous triangular test. After running another triangular test, the Maslow reported rotational radius was 145, which yielded a good benchmark.

1 Like

Good Work @Dustcloud. I appreciate the work yo put in to “tuning” and sharing your results.

Thank you

1 Like