Another bevel-beside wheel swerve drive. This one uses a delrin groove ball bearing. GBX-134 (not posted on CD but available in our offseason CAD release) weighed under 5.6lbs, but used an 0.3lb ball bearing and was optimized for manual machining and waterjetting (both of which we can do). GBX-136 has been optimized for waterjet and CNC work, but can be made with only a CNC mill and manual lathe.
It uses a 30:2 (15:1) 2-lead worm gear reduction and a 72:14 spur gear reduction to turn at just over 200rpm, but the 14t gear can be reduced to turn it slower. I felt safe with that higher speed as the worm gear cannot be backdriven in this setup. I calculated an efficiency of 86% for the worm and 98% for the spur for a total of 84% efficiency. This saved some weight as well.
The weight is my best yet at a hair under 5.2lbs. It is designed to be mounted to the top of a 2x1 or 1x1, but can also be mounted below. To allow both styles it has only a single plate for mounting. A 28"x28" chassis with this swerve drive mounted on 1/16" wall 2x1 weighs under 28lbs with a 6mm plywood bellypan.
The main thing I’m worried about is the worm gear failing under shock loads, but I can always replace it with a 4-lead worm to allow backdriving.
I never would have thought to use a worm gear to turn a swerve module. Very creative. Correct me if I’m wrong (I’m more of an electrical guy) but wouldn’t the worm gear reduce the potential speed to turn the module? If you end up assembling it, I’d love to see how it turns out.
I was worried about the efficiency loss so I used the chart and equations on this page to calculate the efficiency loss from the worm gear, and got around 86% efficiency. A 4-lead worm gets around 93% efficiency. I have not measured data on the common BaneBots planetary reductions or Versaplanetaries, but assuming they have 2 planetary stages, it should be around 92% tops (98% per spur gear), although I have not had great experiences efficiency-wise with either.
We would be making GBX-134 instead of this one due to easier manufacture, if any. We don’t own a CNC; this was just a lightening exercise for me.
EDIT: You can find both the CAD for GBX-134 and GBX-136 here at our GrabCAD CAD release. Look in the folders labeled “GBX-134” and “GBX-136”. If you want to avoid downloading the entire Parts library, just PM me and I will send the assembly to you without that.
The gear on the CIM is an 11t 8mm bore gear linked to a 1/2" hex bore 18t gear from vexpro. I have a cut Vex 8mm to 1/2" hex adapter on it to adapt it to an 8mm shaft for the 15t bevel gear down below. The CIM shaft goes into an 8mm bearing in the center of the turning gear to align it properly. The end effect is that the bevel gear is offset from the center, ad makes for an ultra-compact design.
The snap ring is just to retain that 8mm shaft for the bevel gear/ hex adapter.
Where is the the speed encoder and how do you know the position of the the rotation if you are using what looks to be CUI capacitive encoder on there worm gear? Could be missing some thing. Other wise looks great mechanically.
I’m using a CUI incremental encoder on the worm, but I forgot to embed a zeroing magnet into the rotation gear; rest assured it’s there now.
The speed encoder is the quadrature gear tooth sensor used in the Versaplanetary encoder mod (the Allegro ATS605LSG). The 11t pinion is steel and can be “read” by it.
Nice and light; too bad it’ll never see the light of day. I’m looking at GrabCAD right now, and I have some questions.
Is it worth it to machine a rim onto the custom wheel if rivets on the tread already hold it on?
How is the bevel gear attached to the wheel?
For the 4 little spacers/bearings on a bolt that help hold the 72t turning gear onto the main plate, do they do anything the delrin groove ball bearing doesn’t do? I guess they could just be helpers, but I can imagine that tiny bearing breaking if it has to take the whole weight of the robot.
On a side note, I have no idea why some of your bolts have blue threads.
The wheel is 3D printed and pressed onto the bevel gear, although you could probably throw it on a CNC or lathe now that I think about it. It has I think a 0.014" interference fit (similar to colson wheels).
It has a rim, because if you want to get away without one you need a lot more rivets, and you have to put more care in attaching the tread. We found out about that on our 2012 robot after the tread wore out.
The delrin groove bearing takes all the load in normal operation. The only time the small bearings are called upon is when the module lifts off the ground and the module is free-spinning. I might add a bit more vertical clearance to that due to the possibility of odd loading.
