Is there a kinematics calibration guide?

The cut on the left is from the maslow, the cut on the right is from a calibrated cnc table. width is spot on, however, hight is off almost an inch. My maslow seemed to be fare more accurate with stock hardware than it is with Kinematics. Is there a Kinematics calibration guide somewhere I could try to follow? I’m about ready to ditch kinematics altogether. I was within a 1mm with stock hardware with slightly wobbly edges, now i’m way out of wack. Any guide to follow would be great. Thanks in advance!

3 Likes

I assume by “kinematics” you mean triangular kinematics right?
I’m pretty sure I’m preaching to the choir here but for the sake of anyone else reading this who might not know: The stock Maslow uses quadrilateral kinematics, which simply means that there are four points and four “sides” (from motor A to motor B to sled bracket A to sled bracket B). Quadrilateral kinematics tries to calculate the position of the router bit which lies below the short side of the quadrilateral shape.
The linkage kits effectively converge those two sled bracket points into a singe point located (mathematically) at the router bit itself. So with a linkage kit you now have triangular kinematics (only three sides/points). This reduces the processing power required and increases accuracy by quite a bit.

You should never use triangular kinematics (in GC) without some sort of triangulation add-on (like linkages or a ring etc.) Likewise you should never use quadrilateral kinematics with a triangulation add-on installed on your sled. Doing either of those will lead to weirdness in your cuts.

Back to your actual question, did you cut that left piece using the linkage in the picture (along with triangular kinematics)? The rotational radius should be set to the distance between your router bit and the chain attachment point.

I do not think that the error you are seeing is due to linkage slop but here are some other things to consider when using a linkage setup:

With a top mounted linkage there are potential problems that can compound if you aren’t careful. I would suggest mounting the linkage on the sled with the router installed (for full weight) and run the sled to the top center of the work piece then measure the distance between the vertical bars of the linkage at their tops and at the chain attachment points. These two bars should be parallel (and vertical), if your numbers are different then your bars are splaying under load and this can cause accuracy problems (note that this is not a constant inaccuracy because the amount of splay in the linkage is dependent on how much force is on the chains which changes as you move across the work area). If you do run into that problem then you’ll have to do some more sleuthing to find out where the splay is coming from (it could be joints, or mounting points).

Another option that is a little more work would be to make yourself a 45˚ linkage (or similar) and see if you still have the same results. With the 45˚ design all of the bars that attach to the sled are under tension so assuming your bars are the same as each other, any error in the joints only adds to increasing the rotational radius, which can be accounted for when calibrating. With the top mount design the top two bars are under compression and the bottom two are under tension so you can run into compounding errors.

I’m sorry for the long response that is likely filled with information you already knew…

3 Likes

the potential errors from the top mount linkage would not add up to nearly an inch in height. At worst those errors are going to be tiny fractions of an inch.

If we assume that there is 0.02" of error in every joint, the worst case error is only 0.04" (pulling straight up if you are pulling straight horizontally, some of the errors cancel out and the error is only 0.01")

That would not account for being off by an inch or more

What version of GC and firmware are you using? have you done a calibration after upgrading to 1.0? please check that the motor distance reported matches reality with a tape measure (if something is flexing as the measurement is taken, that will cause errors)

You are absolutely correct! I had actually clarified that when I was writing the post but for some reason I deleted that part when I was editing before I posted and I forgot to put it back! That’s my bad, I just edited my post for clarity.

To be clear, I do not think that the 1" error is caused by linkage slop (I do not even think it could be caused by linkage slop). I’m sorry for the confusion, it was 1AM and I was using my tired brain to write.

Out of curiosity, what orientation was the part on the workspace when cut on the Maslow? Horizontal, vertical, etc? Which triangular linkage kit are you using, and what rotation disk radius setting do you have?

1 Like

the picture shows a top mount kit (the one I make), but it’s not clear if that was used to make the cut or not (the way the arms are show isn’t a way that they could be used on the machine)

you have both kits (and actually the worst available version of the top mount kit), could you run a test of the accuracy of each of them?

just cutting squares or a grid across the work area, flip the board and repeat with the other kit. Then measure to see if there is a measurable difference between the two.

I’ll send you money for the wood and a set of calipers if needed.

play with the simulator a bit, even the worst case error of 0.04" would translate into less than 0.08" of error near the top center (where the errors cancel out rather than add)

as we are talking about linkage kit errors, I’ll point out that if the wood links flex even a couple hundredths of an inch under tension, that will also result in errors that vary across the work area.

but in any case, I think that chain sag is a larger factor inaccuracy (and chain stretch could easily be larger) than linkage caused errors (of either type)

I completely agree, any possible slop in the linkage kit would be far overshadowed by chain sag, etc.

I would suspect a possible issue related to one of:

• Motor horizontal spacing
• Calibrated chain length

It’s odd that the sides appear to be straight, and the single dimension appears roughly accurate.

If it’s cut near the horizontal center of the work area, you won’t see as much distortion of the horizontal distance compared to the vertical distance if the motor spacing or rotation radius are incorrect.

Good point. The question would then be, was the part cut in the vertical orientation? If so, that may hint towards calibration issues.

Before the update to V1.00, I was experiencing a similar issue to @Onelonedork’s problem. I was able to get it within 3mm tolerance but couldn’t get it any tighter than that. After the update, I was able to get it to ~±1mm in a couple of tests. I still plan on trying to dial it in more than that.

I have done calibration routines ad nauseam with the top mount linkage system. Full disclosure, I don’t use the auto calibration process. I think my machine is not physically long enough, because everytime I’ve tried to use it I end up going in the wrong direction. I’ve come up with a manual calibration process that involves cutting squares, measuring the results, and tweaking the motor distance and the rotation radius. Those two values have the greatest impact on machine calibration. The simulator is a great tool for testing which calibration values are messing you up. I’m guessing you might be having an issue with the rotation radius, since you had the machine working well with the stock kinematics. It’s hard to say, however, without being at the machine.

Compounding errors are quite easy to calibrate around and get center of the machine cutting at least as accurately as ±1mm. I’ve seen some distortion to the far left and right sides of the bed, but that also could be because I have a short machine. My efforts to test compounding errors further (in the manner @dlang has suggested) are moving at a glacial pace. Especially because my shop hasn’t been above 20 degrees for a few weeks now. The last thing I cut was the HDPE pad for the bottom of the sled, so I’m getting close to putting together the test sled. Here’s hoping things warm up for a week or so that I can get more definite numbers.

1 Like

The part was cut vertically, i’ll post my settings shortly as well as some photos. I’ve cut the part across in multiple spots across the 4x8 space. Comes out the same everywhere, always short by about an inch. I’ll start over with calibrations (round 10).

when you do this round of calibration, put a tape measure between the motors
before you start, and again while the chain is pulled tight, see if there is a
difference between the to measurements.

I’m thinking calibration process issues here.
I’m using a top-mount kit like the one pictured. I just did a full calibration using v1.02. I checked that the motor spacing (3365.37mm) under load matched the motor spacing at rest. The rotation disk radius after the calibration (116.815mm) is the same as the previous time I did the calibration. This calibration gave 1905.2 on the first cut. This value for rotation radius seems odd, as David’s esitmate for the value was 133.35mm or 136.525mm. I’d like to see a diagram that shows just what is being measured here. At any rate, the calibration test passed, so the machine should be accurate.
As a test, I ran 20" up from the center position and made a mark, then 40" down and made another mark, then back to the center position. It returned to the center position precisely. The upper mark, though, was 19 15/64" above the center, and the lower mark was 19 47/64" down.
Moving 37.5" to the left and right to make the same pattern yielded similar though slightly different results. The bottom marks lined up across the sheet, the left and right center marks were 5/64 above the center mark, the left and right top marks were 19/64" above the center top mark. I pulled the sled down and vewrified thatn then center of rotation for the

I’m thinking calibration process issues here.

I’m using a top-mount kit like the one pictured.
I just did a full calibration using v1.02. I checked that the motor spacing (3365.37mm) under load matched the motor spacing at rest.

That is a little wider than a stock frame (just over 11’), but not unreasonable.

whenyou say the spacing under load matches the spacing at rest, do you mean that
the distance measured here matches what you get from a tape measure, and that
there is no difference between under load and at rest ith the tape measure.

The rotation disk radius after the calibration (116.815mm) is the same as the previous time I did the calibration. This calibration gave 1905.2 on the first cut.
This value for rotation radius seems odd, as David’s esitmate for the value as 133.35mm or 136.525mm.

that is odd, 116.815mm is 4.6" and the horizontal links in the kit are 5"

This is an impossible value for this kit and means that something else was
measured incorrectly, and the ‘calibration’ is trying to correct for the error
(see more details below)

I’d like to see a diagram that shows just what is being measured here.

The rotation radiuks is the distance between the end of the chain (more
precisely the center of the hole for the pin at the end of the chain)and the
center of the bit. With this kit, this is 5" (the length of the horizontal arms)
plus 1/4" (the length of the master link that’s added to the chain)

but it’s not actually being measured, what’s happening with the current
calibration routine is that it makes cuts that should match if the yoffset
height and motor spacing are correct, and then ‘adjusts’ the calculated rotation
radius to counter whatever error you find. If the motor spacing isn’t correct,
then the rotation radius that’s calculated won’t be correct (I suspect that if
the yoffset value, the height of the motors above the work area, is wrong, that
will also throw off the calculations)

At any rate, the calibration test passed, so the machine should be accurate.

yes and not, it’s introducing error in one measurement to counter errors from
other measurements. The fact that it ended up with an impossibly small rotation
radius for this kit means that something else is off.

As a test, I ran 20" up from the center position and made a mark, then 40"
down and made another mark, then back to the center position. It returned to
the center position precisely. The upper mark, though, was 19 15/64" above the
center, and the lower mark was 19 47/64" down. Moving 37.5" to the left and
right to make the same pattern yielded similar though slightly different
results. The bottom marks lined up across the sheet, the left and right center
marks were 5/64 above the center mark, the left and right top marks were
19/64" above the center top mark. I pulled the sled down and vewrified thatn
then center of rotation for the

and this confirms that something is off.

Yes

Hard to see why that would be; it should only affect the vertical location of 0,0 on the workarea.

and that is the reason for the post

Hard to see why that would be; it should only affect the vertical location of 0,0 on the workarea.

not quite, you are trying to calculate a triangle

the horizontal part is half the distance between the motors

the vertical leg is the Y location + yoffset

and the angled leg is the chain + the rotation radius

so if yoffset is off to far, it will affect the calculated rotation radius
calculation

So I finally had time to run through calibration a few more times. The cuts are actually getting worse. After two more trips through calibration, I’m not more than an inch and half shorter than stock Maslow.

I also measured between motors with chain and only ended up 1mm different from my tape measure which I calk up to the looseness of the metal tip. Any other ideas?

1 Like

I’ve had disappointing results from the automatic triangular calibration, too. Much better when I entered the value by hand. For the linkage shown in your picture (like mine as well) 127mm is the appropriate value. Still not perfect, but not 1" off over 6" distance .

1 Like

It’s gotten worse, which means my changes have made a difference. These last couple of chain calibrations, my chain has also been putting the center point of the work piece a foot or so low.