I’ve read some of the replies to my first Butterfly Drive, and I designed this version to fix the problems on the first one.
Things I changed:
-made the wheel pods to be in-tube, so that I can add another bearing to the end of the driven shaft so it doesn’t become a “giant lever arm” when hit on the side with tractions down.
-pistons now push down rather than being connected by a clevis. This means that the pistons will not be affected as much on impact
-a compression spring on the other side of the shaft pushes the pod up when the piston retracts. If the robot happens to lose pneumatics, then at least we can still drive
-roller chain runs outside the tubing to avoid conflicts with rivets, which was a problem with the first version that had chain running inside the tube
-for front and back rails, I ran into some mounting problems because of the pods. If you looked at the CAD, there are custom “barrel nuts” which hold the rails together by screws and a threaded aluminum rod
Specs:
low speed: 7.31fps
high speed: 16.09fps
weight: ~39-40 lbs.
These speeds are made possible by the different wheel radiuses of the omni and the colson; a shifting gearbox is not needed unless you want more speed options, but I really don’t see a point in that.
It looks good, but just a few questions and things to check if you haven’t already:
I understand that your speed is 16 FPS with omnis and 7 FPS with Colsons, meaning that you will run mostly on omnis and use the Colsons for pushing matches and perhaps to kick off sprints; is this correct?
The weakest part of the side tubes will be where you cut the holes for the omnis, with the locations of the Colsons being next. Be sure when you mount bumpers that they mount to the chassis at locations other than these to avoid a bent frame.
The push-only, spring return is one way to avoid transferring lateral forces to the piston; we have found another for our strafe wheel actuator: the open portion of our clevis pin is significantly longer than the clevis is wide, meaning that small deviations will not apply a lateral force to our piston.
Be sure that the Colson ends of the modules, when faced with a lateral force, will contact the side walls of the tube away from the rotating axle, and preferably either between the wheels or above the green spacer/barrel nut where the tube is strongest.
You probably realize this, but note that your CoG is well off-center. Be sure to balance the load with the battery, control board, and/or manipulator placement. The CoG should ideally be between the two strafe wheels.
What angle of rotation of the strafe module is required for the wheels to engage the carpet? I’ve seen modules that went almost vertical, but these look like they only need to rotate a few degrees. One of these days I’m going to have to have a good session at a dry-erase board to figure out what the proper angles and proportions are for one of these modules.
Yeah, the gearbox and the sprocket reductions in the module were built so that I have those pushing match speeds. As for kick off sprints, I’m not so sure what that is, but if it is something like those drop-down omnis on 118’s 2014 robot, then most likely I probably don’t have that on this one.
The holes with the omnis… I kind of see that happening. An impact directly on the bearing that occupies it might cause the tube to cave in over time
Interesting. Honestly have nothing to say about that
So the modules need to be somewhat the same width of the inside width of the boxtubing? I guess that makes sense. I might have solved the shaft bending, but the plates might take damage after one of those traction side impacts, so doing so would reduce the bend
Yeah, those gearboxes would shift the CoG to that side, which I guess might make strafing some what strange. I placed them there because anywhere else would be a pain with the strafing module there, and so I can reduce how many sprockets I need. The battery, as well as additional components that would go on any FRC robot will definitely help in this situation.
Angle of rotation, I don’t have exact numbers at the moment I’m writing this, but it’s really easy to calculate using trig: I think it’s the inverse sine of the ratio of the distance from the driving shaft to the ground, over the distance from the driving shaft to the omni wheel shaft.
Did you add a foot gusset to the front rails you used the barrelnut on?
I made a custom barrel nut: take an aluminum rod, thread both ends, and drill a hole through the round face, and tap for 10-32. Screw them into place, and theoretically it should hold just fine in addition to the rivets from the belly pan
I’ve never heard of a foot gusset before, but to me it sounds like one that has an L shape and can mount to the side of the boxtubing. I could do that, but probably not with those modules in the way.
If your modules are strong enough to not have to touch the sides, that’s great. I was just suggesting that when/if they make contact, it should be at a point of strength on both pieces, and should not cause the axle to rub, which would cause extra wear and loss of drive force.
I don’t have any CAD software on my computer, so all I can do here is guess from the perspective render.
The module plates are 6061-T6, 1/4" in thick aluminum with no pocketing. I think the most I would expect would be some bending after one regional, or even no damage at all. I haven’t seen such a material in high-stress situations as much in FRC, so I’m really not sure about this
With 1/4" plates, I would not be so much worried about permanently bending those plates as much as rubbing the axle or bearing against the inside of the tube, especially with T6. However, note that tempering usually causes little change to the elastic modulus (that is, the spring constant for a given geometry). Unless you are going for a really tight tolerance in the tube, you’re not likely to get contact under likely sustained forces such as scrub and shove on the FRC field that would cause this contact. However, consider what would hit what and where they would meet if you were rammed from the side, or if you ran into a wall or another robot with a corner of your robot.
Today was the day (though I used a sheet of paper rather than a board). I’ll follow up later (probably over the weekend) with a diagram, but the bottom line is that if the module has a short pendulum arm (vertical distance between the drive axle and the CoG), and ignoring bearing friction, wheel slippage can be avoided by ensuring that
cos(θ-φ) > r cos(φ)/μR
where
θ is the angle between the drive axle and wheel axle then the wheel contacts the carpet, measured down relative to horizontal
φ is the pressure angle of the gears
r is the pitch radius of the wheel gear
R is the radius of the wheel
μ is the coefficient of friction
The optimal angle for floor engagement is therefore the pressure angle of the gears, though small variations from this should not cause too much of a problem.
I also see that doing this with belt or chain would not work; the key point is that the drive gear pushes **down **on the wheel gear to provide the basis for the normal force. In order to apply enough torque to the module, there would have to be a significant gearing **up **within the module.
I’m having trouble trying to derive this so I’d love to see your analysis.
Edit: As far as the design itself goes, I have a couple of comments. First, I’d highly recommend that you extend the hex shaft that goes through the vex clamping gearbox so that it supports the other side of the module as well. This will not only improve serviceability by making it far easier to access the e-clip but also make the module more stable against twisting. It isn’t critical for the module to be solid against twisting because it won’t see much of any load in those directions, but there’s no downside to improving it and, at the very least, it will make the meshing between the 48T gears nicer. Finally, if you could put a standoff in the bottom corners of the side rails it would significantly improve the rigidity where you have the giant cutouts, though they still make me uneasy.
Is the axis of the barrel nut vertical? The barrel nut would have to be made fairly accurately for it to work well (length, ends square, holes for screws all in the same plane). When you start putting stress on your chassis, will the single screw holding the side tube to the barrel nut cause that part of the side tube to deform? It really isn’t a tube anymore in that area. Is it possible to add an L-shaped gusset to the top surfaces where the side tubes meet the front and rear tubes? That would spread the stress out much better than the 3 screws in the barrel nut and save you having to make the barrel nut.
Since you have a large bottom pan providing rigidity in it’s plane, you can probably pocket the top and bottom surfaces of the two side tubes quite extensively to reduce weight. The inside surfaces of the side tubes and the front and rear tubes can probably also be pocketed. You can probably also pocket the middle tube quite extensively too.
Do you have a small gap designed in between each of the two plates and the inner surfaces of the side tubes? Can you add a piece of low friction material on the outside surfaces of the plates? Nylon screws/bolts are injection molded, I think, so their dimensions are pretty consistent. You could add some tapped holes to the drive module plates and install some nylon screws or bolts so that their heads will rub on the inside surfaces of the side tubes when the module is deflected. The screws/bolts can be cut flush with the inside surface of the module so they don’t interfere with the wheels. A few threads will be more than enough to retain the screw/bolt in the module.
What retains the bearings holding the ends of the shafts with the omni-wheels on the outside surface of the side tubes?
How will your drive modules be built? What is the order of assembly? It looks like you will have to insert the hex shaft after you insert the module plates into the bottom of the side tubes with the omni-wheel and sprocket for driving the traction wheel “held in place by a third hand”. Our modules from 2014 were like this and were really frustrating to work with, requiring two people working in very tight quarters to get them in or out. I cannot remember the details but 148’s drive modules could be pre-assembled and formed a self contained unit. They could be swapped in about 2 minutes. I think they used a dead-axle.
You may want to delete the pocketing on your slide-drive module with the two omni-wheels. You may also want to beef up the “ears” that the two green standoffs attach to and add fillets where they join the main part of those plates. What keeps the two plates from “sliding” relative to each other when the slide-drive is pushing your robot sideways? It does not look like the main shaft extends through both plates. The two standoffs will tend to twist the ears they are screwed onto. Can this whole assembly be made from a piece of tubing with the wheel bearing holes offset so that the wheels stick out the bottom of the tube? It could be much more rigid.
It may also be helpful to extend the main shaft from the VexPro Clamping block through to the front side of the module and add a support that is attached to the bottom pan. The VexPro Clamping blocks are plastic and will deform more than metal, allowing your shaft to wobble from side to side a little.
Will the slide-drive module twist the main central shaft so that the front end moves towards one or the other side tube? This is the same issue as with the drive modules in your original design.
Be careful when installing the motor on the VexPro clamping block. One student ruined one or two of them by overtightening the screws. Since cap screws have a relatively small head, the clamping forces are very concentrated and the screw was pulled through the plastic, deforming it. The screw head ended up touching the mounting face of the motor. The hole in the plastic became a press-fit for the screw head and could no longer retain the motor properly.
I did make an error in my earlier analysis. I assumed that the force applied to the wheel by the module at the hub was horizontal, as it was there as a reaction force to the wheel pushing the robot. I reviewed this assumption, and figured out that it did not make sense. I now realize that (ignoring a couple of minor effects, such as bearing friction, pendulum effect, and wheel inertia as compared to the robot inertia), the force must be applied along the line between the drive axle and the wheel axle. This updates the answer to:
tanθ >= r/( μ R (cos Ф + 1))
For a 4" Vex omni wheel and a Vex 60 tooth gear (I guessed), this comes out to a requirement that θ be no less than 19.1°. If the gear is larger relative to the wheel or the coefficient of friction is less or the pitch angle is larger, the angle will increase. Also, I did make a number of simplifying assumptions, but the bottom line appears to be that unless the module rotates to a nearly vertical orientation, the greater the rotation angle the less the likelihood of slippage.
I’m putting some finishing touches on a white paper (about four pages) that I’ll post tonight or tomorrow.
