Everything's better under the C (a quick poke at the frame flex value)

[RandomDave]

@dlang

that is correct. the initial math did a triangle from the anchor to the center
of the router at the level of the arm, and I showed that with a reasonable Z
offset and 600mm on the xy plane, the fact that the anchow end and the arm don’t
flex like that resulted in an error of ~5mm when all the Z difference was in the
belt length rather than anchor to bit length (I may be off slightly on the exact
numbers, but it was significant enough to fix)

Yeah, that makes sense

[Dano]

I think we would want 4 separate belt extension values, similar to z height offsets. I can think of two examples.

1. Using materials like cabling or chains thatay e aren’t perfectly matching.
2. I could see where those with vertical frames might want extensions on the top two belts for easier mounts and dismounting . But possibly don’t use any with the bottom two.

Dano

[themanro]

This might be more on this topic, for some reason I’m constantly getting about 4 (mm?) difference. Looking at the values you guys are getting, my seem off, but I can’t figure out why. I don’t believe it’s frame related. Should I figure out how to get those values to match closer or it’s irrelevant?

[MSG:INFO: Flex measurement: TLBR: 0.892 TRBL: 4.119]

[dlang]

Roman Rusinov wrote:

This might be more on this topic, for some reason I’m constantly getting about
4 (mm?) difference. Looking at the values you guys are getting, my seem off,
but I can’t figure out why. I don’t believe it’s frame related. Should I
figure out how to get those values to match closer or it’s irrelevant?

[MSG:INFO: Flex measurement: TLBR: 0.892 TRBL: 4.119]

my question would be where the flex is actually happening.

If you can get a video of things (ideally closeup videos of each arm and of the
sled itself)

is it the anchors moving??
is it the cage/router/sled moving/tilting?
is it belt stretch?

until we know where the movement is actually taking place, it’s going to be
impossible to eliminate it.

David Lang

[themanro]

I can see the sled moving a bit while measuring the flex, perhaps belts tightening up more. I will try to get that on film. Is there a way just to run the frame flex measurement command without the entire calibration?

[RandomDave]

There are a few points to understand with the Frame Flex calculation (it’s not really a finished feature, but you can get interesting info out of it):

• The values returned are the TL+BR and TR+BL (belt lengths at +500 Calibration Force) minus (belt lengths at Calibration Force).
  • So if you calibrate at 900 it’s the length the belts/frame/Maslow stretch between 900 and 1500
• It’s the first thing in the calibration sequence, so you can theoretically stop calibration after it happens.
• But if you want to get an interesting data out of it, I’d suggest letting it run the first initial set of waypoints and doing the first fitness calc, repeating for a few different Calibration Forces
  • You want the belt data from the first point if you’re going to graph it.
  • I occasionally saw a bit of weird behaviour if I stopped during the fitness calcs.
• This isn’t quick - when I have time i’d like to rustle up a PR to do something better, but for now this is likely to take you an hour or two with having to restart the calibration from scratch each time.

[dlang]

If you have a usable machine to start with, make sure you save a copy of the
maslow.yaml file with your good calibration in it.

David Lang

[RandomDave]

Something that occurred to me when pondering what I could do with my system flex calcs.

I whacked the data into linear regression to get myself actual fitted lines. I suspect these would be cleaner with better data capture than the way I did it, but it was interesting that they do appear to have quite different slopes, more than i’d realised:

It occurred to me, if I use the actual, laser measured distances between the anchors as C, I could get the scaling factor at any tension I wanted.

I could then use that to scale the belt lengths from a calibration run, and run fitting on those to see what happened. Interestingly, it did come back with very close to the laser measured lengths (within a mm or so).

Kind of interesting but not very useful? I did wonder if it would allow me to plug my laser-measured values into the yaml files and set a scaling based on it, but annoyingly that doesn’t really work because the current scaling is X and Y, not TLBR + TRBL scaling

[bar]

@RandomDave

I could then use that to scale the belt lengths from a calibration run, and run fitting on those to see what happened. Interestingly, it did come back with very close to the laser measured lengths (within a mm or so).

I think that this is a super valuable check that we’re not crazy and that belt stretch + flex is the thing we need to be correcting for

[RandomDave]

@bar

I think that this is a super valuable check that we’re not crazy and that belt stretch + flex is the thing we need to be correcting for

Yeah, for sure! My take on it is we might need to look at it on a per-arm basis or per-diagonal though, mine certainly seems to be different for the two diagonals.

I want to try again with a Force 1200 calibration at some point and see how my theories hold - it’s all a matter of finding the time