My team has been coming up with some ideas for drivetrains and octocanum pods recently. One of our team members is worried about the backlash caused by the WCP/Vex gears that we plan on using and suggested using belts instead. We were not sure which would cause more backlash, gears or belts and pulleys, or if it would even matter. We have the ability to precision machine bearing holes, gearboxes and such to .001" or higher if needed. With all this in mind, would gears or pulleys/belts cause more backlash, and would it even be enough to be effect our driving in auto?
There are two things about Vex gears that lead to some backlash:
The hex is oversized by about .003-.005
Center-center distances in gearboxes are increased by .003
If you are machining your own gearboxes, you can CAD them at exact center distances to remove that source of backlash. If you do so, you will want to break in your gearboxes before using them to ensure everything is running smoothly. I usually design a .001-.002 center add, if only to artificially shift manufacturing tolerances to the + direction.
A way to get the backlash out caused by the hex interface would be to machine your own oversize hex shafts. I guess you could potentially add shims to your gear fits as well? I don’t think either are worth the trouble for most teams or applications.
All of this said, the fairly small amount of backlash in such setups is something a good autonomous mode should be able to handle. It’s not really a problem in most cases.
For FRC drivetrain applications, backlash should never be an issue, especially with your drive pods. That being said, there are many tricks in reducing backlash for both belts and gears. Our team loves using both, but they certainly have different uses and applications. For your drive pods I would say to use belts because they are a bit more forgiving than high DP gears. Depending on the pulley you use, you might not need a live axle and can strengthen your modules tremendously. Some teams like to oversize their axles for gears (something I’m pretty sure you guys won’t be able to do) and opt for higher DP gears with less spacing given the machining precision. For belts, you simply run with higher tension (more spacing) and will much less backlash. In the end, all machines have backlash (look at manual mills!) and you really won’t run into any issues in autonomous programming. If you use a gyro to turn, the precision on the encoders will beat the tolerances of the field anyways.
You tell us! Try and calculate your theoretical backlash using the attached calculations.
Every system is different, but I’d say that you’ll have some amount of backlash with your current setup using gears. Much, if not all of that backlash might be removed by using properly tensioned belts, but I doubt there’s enough to worry about it. There might be other spacing and packaging reasons to use belts, but I don’t think the backlash would be reason enough to switch.
What about the torque we would be putting on the belts? I feel like a high amount of torque could cause the belts to slip in a pushing match, for instance.
If you put encoders closer to the wheels, backlash is less of a problem. For most teams’ drivetrain purposes you don’t really need that much precision (maybe 1/8").
Belts will not slip under load if tensioned. Normally they’ll snap first, and then at a very high load. If you do see ratcheting, increase the tension and they’ll be fine.
One user (Oblarg) follows the teeth = 8*diameter rule, which says that to use 9mm wide 5mm pitch HTD belts in a drivetrain, you should have a number of teeth on the pulley equal to 8 times the diameter of the wheel in inches.
That means that for a 4" wheel, if you use a 32t pulley, you’ll be fine.
If you plan to use 15mm wide belts you can easily drop down to a 24t pulley with no problem.
The KOP drivetrain uses 15mm HTD belts for a good reason. When assembled properly, they don’t slip even under massive shock loads.
That said, I would probably go for gears in your pod design. I think that belts would allow you to reduce backlash more than Vex gears, but I don’t think they extra design/machining work for belts would be worth it. Keep in mind the KISS principle especially for your early designs. Design and build something as simple as possible and get it working well before you start optimizing ever corner of it. Gears will keep your particular design simpler but will probably have more backlash, but it is easier to overcome backlash than it is to overcome a complicated design.
Engineers don’t use the word ‘feel’ when evaluating an option.
What’s your gearbox torque? What’s your wheel friction? How much force will the belts see before your wheels slip? And what are the manufacturer recommendations and specifications?
Strictly speaking, FRC robots usually run belts well above their rated specs. Although it’s possible to calculate everything if you test for wear over time, it’s a lot easier to use empirical data. Most robots fall in the <150lb category with a CoF of between 1-1.3 on 4" wheels. If you go with what has succeeded for bots in that range (15mm wide belts on 24t pulleys, or 9mm wide belts on 36t) and have 4" wheels, then chances are you’ll be fine.
Yes! We’ve done both, and if you do both (custom over-sized hex, and slightly smaller c-c), you can get backlash down significantly. We only do that for manipulators where low backlash lets us get more precise control.
Belts will have lower backlash than gears, but you still need to worry about hex backlash. For very high precision joints (1/16" of backlash on a 3’ arm), we’ll run the last stage as a tensioned chain run to reduce backlash further. #35 chain has higher load capacity than belts.
For a drivetrain, I wouldn’t bother with all this fanciness. We don’t.
Has anyone used a gap filling adhesive to try and mitigate backlash on hex shafts?
Alternatively, has anyone compared backlash on hex shafts to keyed round axle? Worn keyways have tons of backlash, but a new keyway should be pretty minimal.
