During practice today, the sun gear inside one of the ring gear stages was completely shredded. We believe it is because the torque of moving the arm back and forth rapidly caused the gear to crack and fall apart. The prototype design was a 900:1 reduction using VersaPlanetary ring gears (10:9:10:1 total), and the arm was attached directly to the output shaft (though it was supported on the other side by an axle and bearing attached to the robot frame).
I’ve seen other teams use a chain/sprocket or a belt system to transfer power from the motor/shaft to the arm, but I don’t see how this could solve our issue. Rapidly turning the motor in the opposite direction while the arm is moving would still need to be transferred through the belt or chain, and through the gearboxes.
Would increasing the reduction help? Even at 900:1, the arm falls under its own weight, and I don’t think that backdriving a motor through the gears is good for any component involved.
The current solution is to use the dead axle method (which will relieve stress on the gearbox because the arm is not mounted directly to it), with two belts connecting the arm to the output shaft on both sides of the arm.
Any advice or considerations would be appreciated. I know the method is not perfect, but we do not have much in the way of resources and try to make do as best we can.
The higher ratio stages have TINY sun gears that are guaranteed to sheer under higher torque loads.
Have you/your team done any kind of engineering analysis to figure out the necessary gearing, or just “I know it needs to be high”? I highly recommend calculating the torque requirements then working backwards using a tool such as https://reca.lc/
Those tiny little sun gears are not well suited for any major loads, you want a 3:1 or so on the output side.
Now, in terms of overall design, getting the load off of that gearbox is for sure a priority, what you have initially is operating way outside the design window of that gearbox. The dead axle with the chain/belt drive is a much better call (chain would probably be preferable here as it is less likely to slip, making sure you have some way to adjust it with an idler or a cam on a sliding bearing block is also going to really help here)
Please refer to the load guide mentioned above for these gearboxes. Your best bet would be to make a reduction within that load guide, and then reduce it further with soemthing like gears, belts and pulleys, or sprockets
I did not realize there was a load table, and am rather amused at the state of affairs now. We basically did say “I know it needs to be high” and aimed for something reasonable. If we have the supplies/space/time to get a chain and sprocket system going, I think that would work best for us. At the very least, I know not to have a 10:1 be our output stage connected directly to the arm.
A counterbalance with a bungy (bungee is more familiar to me) is easy to install, as we have those in the shop, so we will take y’all up on that suggestion.
It might be a little late to start doing the calculations at this point, but I highly recommend going through the exercise of analyzing the mechanics of a design. If you know total mass and center-of-mass (relative to the pivot point), you can calculate torque. Once you know torque, you can calculate back to a reasonable speed range (i.e. probably don’t want 1000 deg/sec, but probably want more than 2 deg/sec). ReCalc will also factor in motor choice, current limits, efficiency loss, etc. Plug-n-chug is a perfectly valid approach here as well, just to get you in the right range.
I suspect you’d be fine with something in the 200:1 to 400:1 range for total reduction, though I don’t know the design of your arm. Try to do 2:1 or 3:1 in a chain reduction to reduce the stages needed for the planetary gearbox. With those numbers, your planetary gearbox would need something between 66:1 and 200:1 (granted, non of that range is within the acceptable load range for a NEO or Falcon500 per the published guide, but it’s a lot closer than you currently are)
I’ve been doing this a while, and this jumped out at me. Instant source of issue.
It does three things. First, it reduces the shock load (chain and belt have a wee bit more stretch, plus whatever you use to tension the chain can help). Second, it reduces what you need in the gearbox itself, which allows you to use some tougher stages. Third, having a decently large sprocket can help spread out the torque on the arm end a little bit.
The bold part scares me a bit. It’s doable, but now you’re adding a decently-sized shaft, with load on both ends. Not ideal. You don’t get any extra torque from this unless you’re adding a second motor.
Just for grins:
What motor setup are you guys using for this, currently? A 900:1 reduction seems insanely high to me, almost like you’re running a 775Pro for the entire arm load. (Note: I don’t recommend this.)
Also, approximate mass and length would help. Fletch is 100% on the right track here, some math required.
My team, 3512, found ourselves exactly where you are in 2019. We built a carriage arm that used a VP live axle setup with a small gear reduction of the VP to the arm pivot axle. This setup absolutely sucked and broke in every way imaginable:
shredded VP gears and housings
twisted hex shaft, yes including steel hex shaft
sheared hex shafts
cracked 20DP hex gears, including steel gears
more slop than what was reasonable for software to manage
We tried very hard to learn from the lessons of 2019 and use them in designing our 2023 carriage arm. I laid out some some arm best practices in this thread last month (read the whole thread, lots of good advice):
The biggest 3 pieces of advice I can can give you are:
Absolutely do not use VersaPlanetary gearboxes for high torque arms
Use a dead axle setup with a round aluminum tube
Use tensioned chain for your final reduction, the chain will help absorb shock loads and has substantially higher tooth engagement
Here’s our 2023 arm which incorporates most of the advice I laid out in the shared link above:
Slight self promotion, but what motor are you using?
I would recommend using a chain reduction on top as it helps with shock loads. We are running 2 neos on a 72:1 on our arm and have no issues, but if you run like a single 775 you might run into problems.
We know the spring print looks canted and that was due to putting a washer as a spacer because the first print was a little off. New print with no washer solved it.
Edit: do not try to put springs that strong on a PLA print. The material used for those print is CF-12 whish is one of the strongest if not the strongest printable material.
They will take you eye out. Ideally they should have some sort of guard or at minimum the attachments have a very high factor of safety with hazard training for the people around the robot.
For the OP: The Banebot sport gear boxes sold by Andymark gear boxes I recommend for arms. As others have said, don’t mount the gear box directly to the arm. Chains or belts will adsorb some of the shock loading. Current limit the motor. With that much reduction you have the power to break things. If you can add something that slips when over loaded.