# Discussion Regarding Triangular Calibration

#41

I need to make a correction about what I said earlier with allowing rotational radius to adjust. In reviewing the code, even though I allowed it to be adjusted, at the same time I set it back to the original value… oops… if I allow it to adjust, it drops down to ~133 in order to minimize the error to 0.0001. This is a bad thing to do in my opinion, so I calculated what I believe the real rotational radius should be for the ring kit (see this thread).

#42

I think one should think about using a fine point pen for this type of study…

#43

I’ve got a turned block from @marm but I need to sand it down some to fit well into the router body, it’s too tight.

Also, rather have something I can install in the router so I keep the weight and center of gravity the same. I found a pen that fits, but \$40 for the thing seems ridiculous.

#44

I measured motor spacing from outside edge of gearbox to outside edge of gearbox and subtracted 40.4 mm (what is reported to be the compensation amount to get center of sprocket to center of sprocket measurement). The measurement was made with the sled attached and as high up and in the middle on the workarea I could get it. I ended up with 3602.6 mm, which I used for chain compensation and was the motor spacing value used during the calibration cuts.

With rotational radius (139.82) frozen but motor spacing allowed to adjust , I get an RMS error of 0.312
Cut 1-4 errors: -0.495, 0.092, 0.3691, 0.0
Motor spacing ended up at 3607.818

With rotational radius (139.82) and motor spacing (3602.6) frozen, I get an RMS error of 2.023
Cut 1-4 errors: -0.702, -3.6897, 1.0953, 1.0

3607.8 - 3602.6 = 5.2 mm and is pretty significant. Hard to imagine my measurement is off by that much or a tape measure would be that inaccurate especially since I’ve measured the same value multiple different times. It’s also a bit difficult to imagine that much inaccuracy would come from the sprockets not being perfectly at 12 o’clock (they were at least pretty darn close).

So, adjusting only the Y offset and chainSag just doesn’t work well… at least when chain compensation is used… though I suspect the same would be true if they weren’t used.

#45

fair enough. maybe just a golf/ikea pencil inserted into a collet?

#46

I plan on just cutting a straight edge on one side of the sled and mark with a pencil. Still get center to center measurements. My waste boards are done too. Lol

#47

did you measure chain slop for each chain and enter it?

#48

Yes. All test were done with the following tolerances:

left chain tolerance: 0.3479
right chain tolerance: 0.3605

#49

@madgrizzle when I went to use my block it was too tight so what I ended up doing was putting a 1/2" long drill bit in the block and turning it in my drill press and sanding it down. If you do not have a drill press you could use a hand drill. I found the easiest was to use a power sander and the drill press.

#50

Good idea… I have a drill press… life got in the way for much work on the camera system unfortunately

#51

Might there be a way to calibrate the maslow by manually jogging the sled around your sheet to specific areas on your work area and then telling it where it is?

#52

That could, in theory, work. It has the same potential issues as the suggestion I made, that is several different changes might have essentially the same effect, so it would be hard to separate them.

#53

#54

Yes, the issue would be how accurately can you measure the locations, and how
accurately can you jog the machine.

currently we make small cuts and with a tape measure, measure the distance to an
accuracy of ~1/2mm

Is this new approach going to be any more accurate?

#55

Still a work in progress. Basically, I can minimize errors for the cuts 2, 3 and 4 but the error for cut 1 always seems too high (2-3x higher). I’m trying to figure where, if anywhere, to incorporate chain compensation into the calibration process. As long as the cuts are made with chain compensation turned on, solutions to the calibration end up with either too low of a rotational radius or too high if a distance between motors… ideally, these shouldn’t change from their actual measurements… But they do so something is not correct. If I don’t let them adjust, then cut 1 error never really drops low enough

#56

tarting to consider selling my kit. I spent days out there trying to calibrate.

Wondering now if its only good to make signs, logos… are there any hopes/advancements in getting the calibration process nailed down?

#57

It seems to me that the current method of calibration actually requires multiple iterations (multiple runs) until you arrive at a minimum error… The initial part of the routine calculates how long the chain should be to the four cut points based upon:

• distance between motors
• height of motors above workspace
• chain sag correction factor

Distance between motors and rotational radius are relatively easy to measure and could be set to be fixed (currently, only the distance between motors is fixed). The rotational radius, height of motors above workspace is entered as a guess and chain sag correction is set to 0 (initially). So when the routine calculates the “desired” chain lengths, it does so with at least three uncertain variables (rotational radius, height of motors above workspace and chain sag correction factor)

When the routine runs through its iterative process, it’s trying to arrive at a solution that best matches the desired chain lengths. If your height above motors calculation and rotational radius are fairly close, the calibration process probably isn’t significantly affected by it. But the default chain sag correction is 0 and is used to calculate the “desired” chain length… Therefore, you are calibrating to achieve what is likely a bad number.

So, if you run it once, you might be in the ball park because the chain sag number is higher. Eventually, as you repeat this process, you’ll get closer and closer to the real number. Maybe the second attempt is all it takes to get you close enough.

I wish I understood how chain sag compensation is calculated…

I wonder instead of calculating these imperfect “desired” chain lengths whether we can just quantify the error as a difference in actual cut length verus the desired cut length(which are known values.)

#58

I think some people are already there with calibration. Question is why are others not. I suspect the actual chain has something to do with it. I just got a new shiny set of chains to try out.

#59

The thing that frustrates me is that the machine seems perfectly accurate! Its just consistently wrong! If its 5mm off it can reproduce that exact same (wrong) result over and over! It seems there should be a way to simply say hey maslow just make Y (for example) 5mm bigger and be done with it! No need to recalculate all the numbers.

Like and axis compensation factor setting perhaps that doesn’t effect anything else. Why can’t we do this? Please for the love of god maslow add this (seemingly, probably not) simple setting! Is it possible that everyone is over thinking it and trying for a perfect solution instead of a solution that just works?

#60

I don’t think it’s easy to do. I notice that I’m off more upwards than I am downwards… so if I go up 12 inches, I might be at 11 14/16 but if I go down 12 inches I might be at 11 15/16… so a simple offset wouldn’t work. One idea I had was to do an error matrix at points around the work area and interpolate values at points between. Use this to offset the target coordinates in the triangular kinematics routine… it could work, just not sure how to arrive at the matrix. The difficulty is measuring the error at the individual matix points