TPU90A Grippy Tire CAD published (Finally)

Thanks. I just remember people saying there are significant variations from one brand of 95A TPU to the next, and so I was curious which brand you’re using to get the “insane” grip.

We recently did a short test on Saturday and found the grip to be pretty good. What we were really happy with was the bump test we did, driving over bumps. Our competition bot from this past year was way smoother going over bumps. We went over a section of 1/2×2" stock and a 2x1. On the 1/2" the robot didn’t even notice and the 2x1 more so but way smoother than robots did in 2020. We don’t have a cable bump from this year to test but will probably build a wood one this week to test. We used Overture translucent yellow 95a

1 Like

I will caution that the extra grippiness of this makes your robot more sensitive to imperfect encoder zeroing. Before if you were off by a small amount resulting in a little toe in or out you would be fine but now with the extra traction it can cause some robot rotation when translating even when none is commanded.


This was the 1/2" bump test with just the side bumper removed to see the wheel

Edit: Most of video was slow motion to see the compression so the sound and the speed kinda look off


Here is a link for going over a 2x1

And here is super slow motion over the 1/2"


There is something I noticed that I don’t think will be as big of an issue after watching some of the slow motion videos. At speed the centrifical force keeps the wheel more round and less squished so I think the issue about odometry isn’t going to be as big of an issue.

Edit: I am horrible at spelling and tend to fat finger letters when typing

It might make sense to design an alignment jig if it’s now that sensitive. I know we’re in a different boat than most teams as we’re planning on using NEOs with with redux steering encoders directly connected to the steering spark max, but we plan on zeroing every module to face the corner via the zero button on the encoder and having our code offsets effectively just become part of a module pose. A 3D printed jig is on our to do list but we haven’t started designing it yet.

1 Like

Starting them on a 45° angle from straight seems like a pain to me. We just used a 4ft level to align the wheels straight. Don’t need a 4ft level, only a 30" or so, but 4ft is what we had hanging on the wall in the shop


It’s generally easier to just make the wheels face the same direction (forward), that way in code you won’t need any offsets. Forward will be 0 degrees. You can do this with a long 1x1 that gets clamped to the wheels to align them (across 2 modules).

The benefit i see to zeroing each module identically like that is each module is swappable without code changes , code would know where the model is on the bot by can ids and set the appropriate offset. just set can ids and go. kind of a cool idea but not sure its worth it.


I would say the clamp bar method takes somewhere between seconds and a minute, because all you need is clamps and a bar. But that’s fair that you could skip that with a module zero. I think that it would have a lot more error though due to jig tolerance.

We use neo’s and neo 550’s on our swerve. Motor controllers are not mounted on the modules. We calibrate with the wheel gear in the same position and the wheels at a 45 degree from straight. This way any module can be placed in any position with out any software changes. Module replacement is fast. Unplug 2 power leads and 3 encoder leads. remove 4 bolts. Remove. Place new module in and reverse procedure. Each wheel position offset is a parameter in code.

1 Like

Have you measured the error/tolerance on the jig?

We calibrate the swerve module off the robot. We either calibrate by eye (parallax) or there are index marks machined and printed into parts. We use an absolute encoder that is like a potentiometer. 12 bit and ball bearings 0 to 5 volt. The calibration is done by rotating the pot and then locking it in with the nut. It does take some skill to do but once done it is good unless the module is disassembled. Our software adapts to varying cross over point. Watches for hi and low values and adjusts accordingly.The encoder has very good full scale accuracy. We have no problem going straight. We pay attention to mechanical slop. Slop can be a bigger problem than calibration.

Anyone have good stl files of these they would like to share?

Ours were printed on a Bambu X1 carbon using the generic tpu settings and they came out great. The files should be at the start of this thread. We printed the V.40 version

What brand tpu are you using. I printed one in overture 95a and just a test by hand the black tread is alot gripper on smooth surfaces. Will have to test it under load.

We found smooth polycarbonate performance to be worse than standard tread, but carpet to feel better with that kind of a “hand pressure test”.

Are you in pushing matches on polycarbonate or carpet? :wink:

Obviously the need for grippy tires on polycarb depends on the year. My concern is something like this year getting on the ramps and maintaining accuracy in auto.

We are using the overture 95A tpu. On a side note we are planning to swap into the leaf spring version to see how those compare.