After two years of testing and development, team 3737 is pleased to release our latest 3D printed swerve drive. Our goal was to enable us (and any team) to manufacture a relatively low cost but competition capable swerve drive using a quality FDM 3D printer. To this end, we have provided with the CAD model, step-by-step assembly instructions, full BOM, complete print settings and a speed calculator for various gearing and drive motors. The swerve assembly is designed to bolt to the underside of a 3030 T-slot chassis (see 3737’s robot release of BB Mantis). This lightweight but strong chassis design also provides immense versatility with infinitely variable bolting positions on all faces.
The base design uses NEO drive and NEO550 steer motors. However, all current FRCFIRST Robotics Competition legal brushless drive motors are possible (serrated shaft prints are provided).
Note that this assembly uses the ‘standard’ 4” x 1.5” Aluminum billet wheel. However, we have separately developed a fully interchangeable, FDM 3D printable, TPU ‘suspension tread’ wheel assembly (see end of this post for link).
Some highlights are:
Fully 3D printed design in PETG (excl. bearings, most gears and fasteners)
Competition tested
Relatively low cost (vs metal framed swerve drive)
Various drive motors possible
Five available drive gear ratios
Drive speed gearing can be changed by 2 or 3 steps with removal of two screws
Nice work! I assume this is for development and testing only? Otherwise I’m not sure I would trust a 3D printed thrust bearing with plastic races to hold up a robot in competition.
Thanks! We have been using 1/4" Delrin balls in a PETG printed V grooves for 2 years (in separate swerve drives of course) during both on and off season competitions and have had no problems. This year, our drivers managed to get some ‘air-time’ off the charge station with no ill effects. We will be using this design on our competition bot in a few weeks (assuming the terrain is suitable for swerve).
I just did the calculation - in static mode, each contact point on each Delrin ball takes just 2.8oz for a full weight robot.
Nope, they’ve used these in competition. I was discussing them with one of their mentors during an off-season event this year, so I’ve seen the units directly on their robot. They really do stand up to hard use and are completely reliable.
Derek, this is really fantastic work your team has done to develop these. Congratulations!
These look great. Low part count and built robustly. Would be nice to see some kind of reinforcement option and steel bearing for longer term use but I think it’s great as is for a low budget option. EDIT: nevermind, I see you’ve run these in season already. Wow!
Glad to see the as5600 get some love. Do beware that the magnets that ship with those often aren’t diametric magnets at all. Also, did you need to pull R4 off the PCB before using it? I wasn’t able to get an analog output before doing that or it would be stuck in programming mode. You can use pliers to remove.
Thanks for the compliments!
Regarding life, we did a full year of on and off season competitions, plus a STEM event (Wings over Wayne) where we played every other match all day for three days. We are confident that they will last another year of demos to sponsors etc.
We found the same issue as you with the magnets! So, in the BOM we provide a source for the exact match, diametric type.
We have not pulled R4 (or the other resistor) but our programmers know the exact details on reading.
You do not have to remove the R4 resistor. That method works but you can do it an easier way. The GPO pin on the board which is misspelled, it is called the PGO pin according the AS5600 documentation. If you connect the GPO (or PGO) pin to 5 volts it puts the chip in analog mode.
IMPORTANT UPDATE: Absolute encoder wiring. The assembly instructions we provide that refer to the encoder wiring omit the link wire (preventing the encoder from working).
However, while building our current swerves, we decided that removing R4 was the most elegant solution. This has been tested and is working. (Grab the sides of R4 with fine tip cutters and oscillate a little to crack the resistor. Dab the two solder blobs with a soldering iron to neaten them and ensure no connection). In summary: Follow the instructions provided AND remove R4.
Each swerve is about $200 without motors and controllers.
That, for example, includes the absolute encoder which for our units are $10 for 5 from Amazon. The CANcoder, often used on other swerve modules, are $70 each.
The full BOM is in the link which gives descriptions, quantities, part numbers and suggested vendors for every part.
Well, we have just finished another year of on and off season competitions and have had no failures with the 3D printed swerve drives.
I can reveal that for 2025, we have made some slight modifications to the design to accept the new NEO vortex motor (but still using the NEO 550 for steering) and the assemblies are no longer handed (2x LH and 2x RH), meaning that you could take one full spare assembly and it will fit in any corner. We are also employing steel gears from the motor to the wheel.
Hi David, we originally started with two NEOs when we developed our first swerve module in Aluminum. However, for cost, compactness and weight, we settled on the NEO500 for steering. It does a stellar job, paired with the UP 4:1 and 3:1 gearboxes. We have had no failures.
However, an innovative single print could adapt the two mount points to accept a full size motor. Don’t forget that the NEO runs at 1/2 the speed of the NEO550 and with the extra power, you only need 1/2 the reduction. I suggest that 3:1 would be fine. Cheers.
