4x4x5 Differential Swerve Module

The Differential swerve described here draws from the many talented designers on CD. My objective was to take the concept a bit further by reducing the number of parts, while getting the performance of two NEOs.
I also wanted it to be the least space claim inside of a 1”x2” tube frame and under 5” tall.
And just fun, it fits inside a 7” ball with room to spare.
Looking for some feedback.

Wheel Diameter: 2.5 in
Differential Gearing
Drive Ratio: 3.62 (17ft/sec)
Twist ratio: 8.11 (695rpm unloaded)
Ground clearance: 1 inch
Space Claim: 4.95” Tall and 4”x4” area inside of tube frame
(8) Fabricated components
(1) One-piece yoke (3D printable)
(2) Internal/External Ring Gears (3D printable)
(2) Straight Spur Gears (3D printable)
(1) 2.5in Wheel with Dual integral Face Gears (3D printable)
(1) Mounting Plate (Aluminum)
(1) Encoder anti-rotation arm (3D printable)
No standoffs
Motion Control:
(1) Mini Magnetic encoder mounted on axis

  1. diffy swerve is usually not that good of a call
  2. your “twist ratio” is 695 RPM free, unless it’s different than normal swerve (which I don’t think it is) that’s way to fast. the fastest you really ever need is 360 RPM, but even then I’d aim for like 100-200
  3. if you want this to be viable i would go with less 3d printed (esspesaly for 2020)
  4. 1 inch of clearance isn’t a lot, 2020 had a 1 inch bump around the shield gen that would be impossible at this hight.
1 Like


Do you have CAD available? I’d be interested in tearing it down further.

Definitely concerned about the 3D printed face gears but I’d be happy to be wrong.

I’d also try to minimize cantilever on the NEO shafts. Motors don’t like side loads very much. Supporting it on both ends is the move, though if you can’t fit that in the space you have, running it through a bearing close to the gear works.

Kind of a fast drive and steer ratio, though certainly workable (within the bounds of diffy swerve that we’ve seen, as it’s still a fairly unexplored field)

I’d probably spend the extra few parts and rigidly attach the encoder to the module.

How did you design the face gears?

1 Like

This is really cool!

What I’m most worried about is the potential for large moment loads (like from running into the many obstacles on most fields) damaging the module. Seeing as you have a boss already on the bottom face, you could get another thin-section bearing on the bottom of the differential module. The housing for that bearing could potentially be a single 3D printed part which also acts to protect the gears from the carpet and floor gunk which inevitably gets all over the place.


Id be interested in taking a closer look at this in CAD as well!


I’m just curious, from the cut away it looks like the top big gear meshes with the top small gear on the inside right? and that meshes with the wheels bevel gear. in addition, the lower small gear ALSO meshes with the wheel as well as the lower gear, so wouldn’t ALL gears turn if only one motor activates? or am I missing some detail

The right inner spur gear meshes with the top internal gear and the right crown gear.
The left inner spur gear meshes with the bottom internal gear and the left crown gear.

If both motors turn in the opposite directions, all the gears spin and the wheel moves forward but the module does not turn.

If both motors turn in the same direction, the inner spur gears attempt to move the wheel in opposite directions, causing the entire inside of the module to lock up and the module turns while the wheel stays still.

1 Like

Here is a link to the Excel spreadsheet you can purchase for designing face gears.


It exports a cad file of one surface of the tooth profile which you have to mirror, create a root surface, cut that profile from a gear blank, then do a circular array with the number of teeth. Kind of a pain but the face gear profile gives a lot of design flexibility.

You got it! Great explanation.

1 Like

This is earlier version uses top and bottom stationary plates with a 3.5" x bearing with very low cantilever. The lower static plate could incorporate outboard bearings supporting the motor shafts. It also includes a rigid encoder mount.

With a 3" wheel: Twist axis free speed is 487 rpm and a driving speed of 13.5 ft/sec.

The trade off vs my initial objective is increased size and more parts.

Feedback has been great, keep it coming.


Using the “footprint inside the frame” convention, the dimensions would be 5.7x5.7x4.9.
The ground clearance is 1.3". This design can incorporate a conical ramp feature, extending from the bottom static plate down to as low as desired, and still leaving room for the wheel to stick out.

This is a really excellent design, I didn’t notice the use of crown gears before. I’m a fan of the 3" wheel. With this type of setup I’m sure you can make the twist axis speed faster, as you are driving it with one stage from 2 NEOs anyway. I believe previous differential swerves have azimuth (as it’s usually called) free speeds up to 1,000 RPM without significant issues.
Of course, nobody has fielded a differential swerve at the highest levels of play yet, so it’s hard to say at what point things get hairy.


Here the CAD for the Las Vegas Swerve Rev-4. There are a number of dormant parts that I used to try different arrangements.

Las Vegas Swerve -4a.zip (35.0 MB)

I do have a working 3D printed model that I am using as a fidget.
Loads of fun.

I will publish the CAD for the 4x4x5 Differential Swerve after its cleaned up.

Hope this provides enough to keep the differential swerve enthusiasts thinking of the next swerve.


it really was! I guess I was thinking too traditional and didn’t stop to think about how the motors would turn! Its an amazing design that I would LOVE to see in motion, do you have plans on actually making one full metal anytime soon?

I would like to build and test one, but we have not had access to our robot lab for quite some time and therefore the 3d printers. I will post a video of the Las Vegas Swerve module to give you an idea how it works. I am planning on incorporating a conical ramp feature, extending from the mounting plate down, to redirect forces generated by field bump impact. I will post the CAD before the end of the year.

1 Like

Here is a video of the 3d printed model of the Las Vegas Swerve. It is the predecessor of the 4x4x5 Differential Swerve. It demonstrates the motion of the gears when driving and twisting.

Las Vegas Swerve Drive and Twist.zip (29.5 MB)


I took a shot at addressing the issues outlined in previous posts, and II think this architecture hits them all. I took some screenshots of the Las Vegas Swerve with a 360 degree fixed ramp (BLUE). It is an 8-inch spherical section that can be integrated into the static lower-plate. The wheel protrudes through a hole on the bottom, allowing 360 deg rotation in azimuth. Initial contact angle with the ramp is at 36 degrees. In LVS design, the cantilever distance from the mid-plane of the azimuth bearing, to the centerline of wheel axle is only 0.060inch, so basically near zero cantilever. The pictures below show the motion of impact with and without the BLUE ramp ring. As always comments are welcome.


The ramps do help. We managed to break components last year in testing before we added ramps to our drives. The original design we had was also round, but in the end we made them square to make them easier to mount on our machine.

The ramps were just drilled to match the bolt holes on the drive and the ramps were mounted to the bottom of the frame and the dirves to the top.

Here is one before the bolt holes were put in. In testing the kids hit the barriers from all directions as well as rotating over them. We were lucky enough to run in a week one competition, and had no issues with damage from the barriers.


This is an elegant implementation. What was the clearance from the ramp bottom to the floor, the angle and how far above the 1-inch obstacle did the ramp start?