My team and I were considering purchasing the new rotation sensors to replace the three optical shaft encoders we use for odometry, but they are a hefty investment at 40 dollars a piece. We have heard mixed reviews about these sensors, and were wondering if anyone had a definitive verdict on whether these are actually better than and worth investing in to replace optical shaft encoders.
Any help is much appreciated! Thank you!
Our team does not yet have the new rotation sensors, but I will put in my two cents having partially created our teams odometry code. The current shaft encoders have 1 degree resolution. With a 2.75" omni wheel connected to each encoder, that gives a 0.024" measurement resolution. Using the rotation sensor spec of 0.088 degree resolution, an odometry system with these should give a resolution of 0.002" with the same omni-wheels.
While this is a major improvement on paper, I really don’t think that this translates to major gains in real robot precision. There are other sources of error in odometry that cause much greater errors than the encoder resolution. On top of that, the movements we make on the field are quite large compared to the sensor resolution, usually on the scale of inches. We usually don’t try to move 0.05" in autonomous! Having just said all that, the new rotation sensor looks much smaller than the shaft encoder, possibly making mounting easier.
i’ve heard mixed reviews about them as well. ill probably get some just to try them for myself, but at $40 for 1 rotation sensor, compared to $20 for 2 optical encoders, it really ought to be noticeably better than the old ones for such a crazy price. the complaints i’ve heard about their lack of many mounting options and higher friction than the old ones is… disconcerting… especially considering the fact that a magnetic sensor like this is supposed to be frictionless…
I agree that the increase in resolution won’t make the accuracy of movements better, but it will probably help more with preventing drift.
As good as odometry is, you can still get a certain amount of drift in the system because of sensor limitations. It shouldn’t be enough to throw off an auton, but it will add up and might become a problem in Programming Skills.
I don’t personally have any experience with the rotation sensors, but I am planning on getting a couple soon to test out.
Another thing to consider is that you can use a high strength shaft with the new sensor, which can eliminate the several degrees of slop between a low strength shaft and an LS to HS adapter in the wheel. Granted that slop can also be removed by roughing up the surface of the shaft or jamming in a small plastic fragment to take up the slack.
The lack of mounting solutions is a major design flaw as I have found the need to use zipties as a second point of contact because someone thought it was a good idea to just put one hole all the way through it and give them the ability to rotate around that hole. The extra friction is noticeable but it can be counteracted with more tension downwards on the tracking wheels, Despite the, for lack of a better word, interesting design features, they work great as sensors. They are very accurate and the ability to remember its position makes them very versatile.
I have a few of the rotation sensors on hand and they have noticeably more friction than shaft encoders. Not enough to cause significant issues, but enough to have to consider as another variable.
I would say if you’re going to use them for tracking wheels, use a larger diameter wheel (larger than 2.75") so the friction from the sensor becomes more negligible. This is because the torque causing the wheel to turn (i.e. the friction between the ground and the rollers) is increased as torque is a function of distance from the center of rotation (and also force, but there’s not much you can do to increase the friction force drastically). You really don’t want your tracking wheels to slip.
I would honestly recommend people not to buy the rotation sensor as the difference in precision really isn’t enough to be much better than the 3-wire encoders.
We’ve already tested the sensor out, and I can confidently say that there’s way too much friction for it to be used as good tracking wheels.
So far, I would say that using certain sensor for tracking would need to be decided on a case-by-case basis, as the amount of friction in each individual sensor seems to be very subjective.
I just got mine in today, and all 3 have roughly the same amount of friction.
I don’t even have to push down on the wheel, and it glides effortlessly on the tiles.
I might have just gotten very lucky, but the ones I have work just fine.
Freespin test (mind you this is somewhat of an arbitrary test because of how many variables are at play) –– relatively low friction for the purposes of a tracking wheel
