Something I’ve been working on over our postseason.
Features 6:1 reduction for 15.5 fps free, 121 RPM steer.
Steer feedback from the PG hall effect encoder and a microswitch. Drive feedback from an SRX mag encoder.
Cost is ~ $206 if you own the 42T pulley, mag encoder, CIM, and wheel. $286 otherwise.
Weighs a few ounces more than 7.178 lbs.
Continuing in our tradition of 3D printing things that we could probably find better solutions to but don’t, this design features 3D printed bevel gears and a fully 3D printed steer housing.
the structure keeping the pulleys at the right distance is shaped like a C. When the tensioned belt runs, it may bend the module or the motor inwards. Can you support the steering motor more? One of the great things about having your steering down at the bottom is that you can support the module there as well and avoid cantilevering the steer housing.
If you’re 3D printing a gear, maybe not make it so small. I’m not really sure though, since the real solution is to not use 3D printed gears for the last stage of your drivetrain, since usually the torque is high enough to probably break those gears.
Otherwise, it looks fine. 3D printing sounds risky, but hey if 192 has done it before I’m sure it’s possible. Definitely would be fun to see if it works. However, if this is your first swerve and you want something that just works, this risky 3D printed thing might not be the best for that.
Really interesting design, and I like the 3D printed pulleys.
Definitely try to avoid using 3D printed bevels, but if you’re set on it try to use something like 10 pitch bevels instead of what appears to be around 20 pitch now.
The module caster box looks super cool. Is that all a single piece?
I’ve had bad experiences with CIMcoders (yes I know, email AM and all that but some of the issues are easy to run into and outside the “normal” operating parameters). If you can, switch to Muchskull’s 3D printed version instead; it’s cheaper and uses 3D printing.
Using the PG71 and encoder like that will add a lot of encoder backlash to your steering. Using a shorter gearbox or something like 5817’s enveloping globoid gears will let you put the encoder on the end of the shaft to be close to the actual rotation of the module.
Are you planning on having a plate on the bottom of the module? It looks like you have mounting options for that on the PG71.
Oh man, this thread is right up my alley. I work in a 3D Printing lab, and swerves are an interest for me. So where to start?
Bevel Gears: Printing gears is totally doable, provided you take into account the post-processing required to eliminate the interlayer adhesion weakness inherent in FDM/FFF processes. Check out post-print epoxy treatments, or maybe look into doing Lost PLA casting out of Nylon. Still cheap, plenty strong.
Bottom Support: 192’s printed swerve worked because of one major design difference, bottom support. The material they used was Stratasys Ivory ABS, which is a relatively flexible blend of ABS. Why does the malleability matter? Because with printed parts, specifically out of FFF/FDM, you run into load-focusing. Imagine this in slow motion: You start the robot from a standstill, and demand full throttle with nearly instant acceleration. As the wheel begins to rotate, there’s a fairly large lever force being applied at the axle of the module. What then happens is the module deforms, flexing ever so slightly, until you find that one layer that was printed just slightly too slow. BANG, the layer snaps, and you lose a module. Swerve systems don’t work too well when you snap the modules off.
**
Think about the forces.** Here’s what I mean by that: Now that you’re printing, you’ve really got to understand how forces will be transmitted through your object. In this case, you’re creating a nice lever, focused at the top pivot shaft. If the parts are printed with layers perpendicular to that lever force, chances are good you’ll see an interlayer failure. If the parts are printed parallel to the lever, chances become decent that you’ll see an intralayer failure. Both can be mitigated by carefully printing, orienting, and load-sharing the design. 192’s printed swerve had four bolts (looked like #10?) running through the module. See how the module bits are printed perpendicular to the normal plane? This makes the layers work together to resist flex, and the bolts both add stiffness, AND assemble the part. Especially important is that little radius at the bottom of the module. By using laser cut acrylic as their rotational bearing, they support a lightweight module effectively.
Changes I’d make: Support your module from the bottom. Keep the printed pulleys, they’ll work just fine. Look into intended print orientation, and see if you can focus and share some of the load into cheap fasteners. Seriously consider using an absolute encoder 1:1 off the module for steering sense. It basically solves most issues you’ll run into with control and hardware compatibility, and is WELL worth the cost increase.
Good stuff!
Disclaimer: All 192 swerve knowledge came from pit conversations with them.
I was planning to run a couple churros down the steering motor to support that, but I guess I forgot about those before posting. I’ll definitely add a plate on the bottom.
This is definitely an experiment. Where’s the fun if it just works?
Cool! Thanks. I’m most interested in the bevel gears. Have you done testing in situations like this?
It’s split up into 4 pieces, to be bolted together: top plate, 2 side plates, and the bottom pulley. Each is designed to be printed flat.
I guess it’s hard to see in the render, but I’m using an SRX mag encoder mounted above the drive shaft, not a CIMcoder.
It’s a PG27, not a 71, but I assume that doesn’t change much? I am planning to add a plate to the bottom from a previous post.
If I switch to an absolute encoder, can I put it on a small pulley and have it idle in between the module and the turning pulley? It won’t have much tooth contact, but it’s not taking much load either.
@Mechvet thank you for your input! I’ll definitely add bottom support. I’ve got the side plates printing perpendicular to normal, but the top and bottom plates both “flat.” I’ll also think about adding some bolts through the middle of the module to take the load up. Same question to you about the encoder on the small pulley.
Correct, but now you need some kind of homing routine or sensor in order to ensure no loss of position, either at the beginning of the match, or due to any other issues.
Absolute encoders make a lot of potential problems go away. My selling point is that with an absolute encoder, the only tuning you need is an offset to establish where “home” or zero is. After that, module starting position doesn’t matter, and minimal time is lost homing/zeroing the system at the beginning of the match. You just have immediate feedback of exactly where you are.
Switched to a BAG on a 25:1 VP and increased the pulley ratio. Module now uses a block of delrin to sit on a bottom polycarbonate plate. Added the MA3 absolute encoder on the same steer ratio as the module.
Looking quality! New absolute encoder setup looks really good, and I like how your components are designed around being printed.
VP + BAG for turning looks great too. I like using RS-550s, but to each their own.
That being said, those gear teeth are still in danger of shearing. Using a larger pitch shouldn’t be too hard, especially if you offload some reduction to the CIM.
Thank you! A positive comment from one of the most prolific FRC gearbox designers really made my day.
After wrestling with Solidworks Toolbox for a bit, I switched the bevels to be 10 DP at 12:36 and changed the pulleys of the CIM to be 18:36, resulting in the same reduction in a stronger way and the same geometry.
@Mechvet Yes, very similar! A bit more compact though. Is this a concept or the swerve you ran this season?
Concept based on the lessons we learned running swerve this season. Once we compile all those lessons into a more digestible format, I’m sure we’ll be releasing a paper on both our thinking going in, and coming out of swerve.
Your designs are similar to what GRT 192 ran this year. The students developed a lower bearing and that added a lot of strength to their plastic swerve drive. Good luck
