Maslow Home Maslow Community Garden

Optical Calibration Demo and Three Hours Working on a Bug


Understood. My floor in my shop is definitely not level. My ceiling is a bit more level than the floor so I have mine hung from the top. But it’s all a bit out of level.

Your idea about the oscillations is an interesting one. Seems like the best hypothesis we have. Though we would expect .25mm of oscillation right? Instead of ~1mm?


There are also other things that could potentially affect it. I think the ring kit itself might be the biggest culprit. A few weeks back I did a number of straight cuts and there was very slight waviness to them. They sanded right down, so it wasn’t a big deal. @dlang reported seeing the same at some Makerfaire(?) with @bar.

I might reassemble the wooden linkage kit onto my old sled… I’m not too sure how accurate the center was, but we can at least see if the waviness persists.


Watching this calibration run really closely I think this is the case. Frequently when moving to the next square I’m seeing the target point land high on one square, then low on the next, then high on the next etc:


I wonder if this effect is from the small vertical movements that happen during homing. When the target point lands high in the square, the routine slowly lets out some chain bit by bit until the sled makes it ‘close enough to the center’. But what I’m noticing is sometimes the sled doesn’t move down immediately, but instead the chains get a little bit of slack in them. I think what happens is when the sled then starts moving again, the slack is taken up and then the target point lands low on the next square. You can kinda see this in the videos I posted – you can see the chain bounce around a bit, even with the sped up video.

If this is the case, more weight and a lower friction sled could probably help. We could consider mitigating this by always lowering the sled below the target square and then slowly raising it until it hits the target center point. Though it would definitely slow down calibration.


This is an ongoing problem, a lower friction sled, more weight (although watch
the motor limits), or angle closer to vertical would help.

As far as your mapping goes, always approaching from belon will be far more
accurate (but raises the question of how well your measurements match real use)

David Lang


I’ll try to at least match the weight of the router first. I just ordered 7lbs of BBs to put in my 3d printed camera mount to match the router weight. My frame is at a ~15deg angle. Any personal opinion about UHMW PE tape vs HDPE board?

It’s a good question. Yet to be seen how well this matches real use. But I think the best we can do for now is get the measurements as repeatable as possible.


Hey @madgrizzle, my paper test pattern is getting obliterated. Should I take the plunge and get a full size vinyl banner version? Are you happy with yours? Or do you have a different idea you’d like someone to try?


That would affect chain sag anyway.

Also, moving towards the center or away from the center will matter. Moving away
from the center involves letting out chain from the far motor and can have a
similar sticking problem.

David Lang


what about this?


Would it make sense to do both? The process could start from the bottom and step up to find the center, then start from the top and step down to find the center, then average the results. It would be super slow, but maybe more precise.


this place has some really cool films in addition to -


Looks like a great lightweight board to put a pattern on! Are you thinking you’d use that as a backing for the vinyl banner, or are you thinking of a way to apply the test pattern directly to that board?


Ah i see, maybe cheap backing board, then paper pattern, then a protective film? That could definitely work.


My thought is get a vinyl cutter to make the pattern and “stick it on” a sheet of PVC then maybe a film - If the PTFE is clear they have.

Thank you


no, when moving away from the motors, you only have the force of gravity working
for you (and that modified by the chain angles), when moving towards the motors,
you have far more force available.

Note, I still do not believe that this will result in a usable map to correct
out errors, but I think it’s a good way to measure accuracy with different
calibration settings and algorithms.

David Lang


I’m very happy with the banner. There appears to be other places that will do the banner cheaper, but I know what I got from Vista Print will work. The texture on the banner didn’t affect the performance of the image recognition and it seems reasonably accurate (now that we incorporated scaling). Regularity of the pattern and the contrast are really the important things and what I got seems to meet that… It appears durable as well.


I’ve seen it as well and it has been a concern of mine, but I figure that it really doesn’t matter because if too much chain is let out, when it finally overcomes friction and drops, it won’t measure to be within 0.125 mm error… so it will start taking up chain again until it does get into range.


The challenge then becomes sticking on vinyl with <0.5mm accuracy right?

I think I’m noticing an in-between state sometimes. Chain is let out, and the sled moves down a little bit, but not to the point that all the added slack in the chain is taken up. Then when the sled starts moving laterally again it falls a little bit and ends up hitting the next square a little low.


I wonder how the vibration of the router affects this in practice. My sense is that vibration would help shake loose any sticking friction and help keep the chain tension more consistent. I don’t really have a scientific way to reason about this though.

You still have the code in there to measure the center point without homing right? Does that have the same periodicity? I have some ideas on how to improve the optical centering to work with sled rotation (probably by collecting a bunch of points and using RANSAC) so that it could maybe work to find the sled center. Then we have another option if small moves for homing are problematic.


Yeah, I the same thought. But, I have a piece of HDPE to try to reduce fricton, just haven’t attached it to the sled yet.

It wouldn’t be hard to add a toggle between homing-in and just measuring, but then we have to be precise in measuring the rotation of the square and size of the square… vs. whether or not we are at the center of the square.


We can add a vertical offset to HomingPosY and shoot low and then have it move up to see if this affects things. Can try tonight.