BTW, internal ring gears can’t be hobbed. You use a gear shaped cutter that reciprocates along the gear axis. The nice ones generate the tooth form by rotating the gear and cutter as it reciprocates. Less fancy is a cutter that cuts the tooth form directly, but that type requires more cutters to deal with a range of gear diameters.
Yes, wire EDM should do the profile. You could pre-cut a blank with all the external features and a rough bore.
If you use nylon or MDF for the gears you should consider beefing up the thickness to deal with the lower elastic modulus/more flexing and/or adding flanges for the same purpose. Glass filling will help that issue…
1 Like
Looks like a great design. It’s very similiar to what 4143 and 5442 have been developing for quite a few years. Both of our teams are using a spur gear differential design this year (and in 2020 but we couldn’t compete). I know of one team who competed with a differential swerve in 2019. There probably have been others.
I can manufacture the ring gears out of AL on a live tool cnc lathe in one operation. I’ve been planning to offer the gear sets as a COTS item for interested teams to experiment with, but I wanted to get the design through at least one competition season. That should be this season.
I will warn you that our current batteries barely support this design. Telling 8 brushless motors to go takes more juice than we have available. You would be better off with a traditional swerve, but what fun is that?
Here are some of the relevant threads:
2 Likes
The hard part isn’t the theory of the code. Calculating a differential in software is not very difficult. The challenging thing is the level of precision you need in the output of your control loop. This requires both excellent mechanicals and extremely responsive control loops. It’s no joke.
The issue is mostly the complete lack of tolerance for error. So your drive is geared roughly 5:1, roughly 1000 rpm free speed either in the driving wheel or when fully rotating. Let’s say you’re trying to go at 500 rpm straight forward. One side goes at 502 rpm and the other goes at 498 rpm. Your differential of 4 rpm is now steering the module - doesn’t sound like a lot, but 24 degrees a second is plenty enough to give you fits. Now it isn’t actually that dire due to turning scrub - friction will give you some natural resistance to turning but you can see the order of magnitude of control/precision you need to simply drive in a straight line. Now try controlling four modules at once with a variety of complex driving+turning routines that will change the loading on each module dynamically, introducing greater disturbance into the loop.
None of this is an unsolvable problem by any means - I’m sure at some point it will be done. But it’s something to be aware of and to prepare for.
4 Likes
Thanks for linking the threads, it’s super useful to see a whole Diff Swerve robot not just the modules.
You say you succesfully made the ring gears out of Aluminium, I assume these did last a competition? Also, i’ve been made aware of the troubles tuning PID for control, did you use any special techniques or just alot of work into getting it right?
We are aware of current limitations from the battery, our team will likely continue to make very light robots but we will be keeping an eye on it.
You can seal the middle differential section and put good grease in it. We haven’t seen any wear on those gears through multiple competitions and lots of practice. The outer ring gear and bevel gears use carpet fuzz and dirt for lubrication. I have swapped out one set of ring gears because the outer ring was getting loud. We probably didn’t grease it enough.
The pid tuning is only hard if you make it hard. As long as the first reduction is large enough, its not hard to control at all. We’ve tested 8:84 through 14:84 first stage reduction successfully. Simple p terms work fine for rotation control. Open loop (just f) can work in teleop for drive control.
FUN just released a behind the bumpers video on our team. In it, you can see a 4 ball auto running with path following. I think we missed 1 of the 4 but not because of the drivetrain.
1 Like
What grease did you use for the ring gears? Red tacky grease?
Our current module is designed for a 14:70 steering reduction. I knew that is pretty high, but it meant we could use the bottom half of our last swerve to reduce unknown variables and re-use parts. I’ll pass your tuning advice onto our software guys, and its encouraging to know that it isn’t too dissimilar to what you’ve pulled off.
1 Like
This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.