The gearbox: http://www.chiefdelphi.com/forums/showthread.php?threadid=158688
With the wheels that far apart, you are going to have a lot of scrub when turning them. Your turning gearbox is going to need a lot more torque than a one wheel option, or even two close-together wheels. I think I remember hearing that high turning scrub makes it harder to tune PID loops to control swerve direction, but I could be imagining that.
What is that bushing connecting the module to the robot? It looks like brass from the color. Is is a COTS part or something you are custom-machining? How many pieces is it? Is the top flange a bearing that got mis-colored or is there no bearing between the large bushing and the vertical drive shaft?
How thick is the horizontal plate at the top of your module? What size screws are you tapping into the side? Are you at all concerned about the plates parallelogramming?
What is the large turning gear made from? (i.e. is it some kind of custom plastic gear or is it just transparent in the render?) What is that set screw thing between the turning gear and the top plate?
Sorry for all the questions; it’s just a bit hard to figure everything out from just a render.
The wheels are about 4 3/8" apart center to center. I could probably make it about 1/2" closer if I shave off the back end of the bottom bevel gear or even more if I find bevel gears with a smaller footprint. Even as it is I think it would be fine, if just a little slow.
All of that stuff is aluminum bronze, and COTS from McMaster. The top ‘flange’ bit that you are referring to is just a thin thrust bearing that prevents the module falling out the bottom if the swerve is suspended. There is a bearing at the bottom of the vertical drive shaft and a sleeve bearing closer to the top.
Thats 3/8" thick 6061 plate. 1/4-20 Bolts tapped into the side. I am not so concerned with that, but if it becomes an issue a standoff can easily be added towards the bottom of the plate.
That’s just a gear from Vex with the center bored out to accept the hub underneath it. It’s transparent for the render.
Thanks for the feedback!
I would suggest running the Andymark 2:1 bevel set (am-2621).
As far as scrub goes, we have never had a problem turning our duel wheel swerve. This years swerve used an RS550 with 100:1 gear box and that was too much power. We needed to VOLTAGE limit them to 6 volts in order to get the PID stable and optimized. That sounds counter intuitive but trust me. Another local team used an RS775 PRO for the azimuth and could not them to behave (blown breakers, hot motors etc.) until they clamped the voltage. If weight wasn’t an issue, we would use the PG-71 (am-2971) from Andymark.
The biggest issue we had with the duel wheel design was how much the robot moves in error when sitting still and the swerve turns. For example, if you’re almost got the gear on the spring but you find you need to move to the right by 1". You try to go 1" to the right but the scrub from the wheels upsets the robot 2". This is the only issue we have found with the whole scrubbing thing.
Our swerve is in GrabCAD and there is a white paper here on CD. I would recommend looking at them if you haven’t yet.
Looks good. Stay calm and Swerve on
With multiple wheels that far apart, you’re going to run into problems with turning as previously mentioned. If you want to stick with this design, I would personally make a custom differential to drive the wheels.
I would suggest running the Andymark 2:1 bevel set (am-2621).
As far as scrub goes, we have never had a problem turning our duel wheel swerve. This years swerve used an RS550 with 100:1 gear box and that was too much power. We needed to VOLTAGE limit them to 6 volts in order to get the PID stable and optimized. That sounds counter intuitive but trust me. Another local team used an RS775 PRO for the azimuth and could not them to behave (blown breakers, hot motors etc.) until they clamped the voltage. If weight wasn’t an issue, we would use the PG-71 from Andymark.
The biggest issue we had with the duel wheel design was how much the robot moves in error when sitting still and the swerve turns. For example, if you’re almost got the gear on the spring but you find you need to move to the right by 1". You try to go 1" to the right but the scrub from the wheels upsets the robot 2". This is the only issue we have found with the whole scrubbing thing.
Our swerve is in GrabCAD and there is a white paper here on CD. I would recommend looking at them if you haven’t yet.
Looks good. Stay calm and Swerve on
Thanks! You guys were actually my inspiration, so I took some of the concepts that you guys had and tried to apply them to my design. I tried to stay away from the meat of the whitepaper so I wouldn’t be influenced by your specific design.
Right now I have a ~54:1 reduction on a 775 Pro for steering, but that can easily be increased during testing since its on a VP.
I will take a look at the 2:1 bevel gear set when I get a chance, but probably after I publish the first revision of the design.
We fell in love with the 2:1. I wish they made a 3:1 or even a 4:1 On our swerve we are always needing more gear which always ends up in the pulleys between the Motor and the swerve.
My gut feeling is the 775Pro is too much. If you put too much motor on your steering it becomes hard to tune. I had a hard time grasping this phenomena but our control guru proved it in spades. The 54:1 helps mask this problem but if you have issues tuning your azimuths, go to an RS550 with 100:1 or even a bag motor! Keep this in mind if there are issues.
Do you think the problem with tuning comes with increased power or increased speed? I would think that for two systems with the same speed but different powers, the higher power one would be easier to tune because it’s easier for the system to overcome wheel scrub at slower speeds. You can give the motor less voltage to get less speed as you approach the setpoint without losing so much torque that you can’t turn the wheels anymore. I don’t have any experience with tuning swerves, but that would be my guess from tuning PIDs for other systems. I’m interested to hear what you think.
I am going to have to completely redesign the shifting gearbox for this if I use the 2:1, so I definitely want to get this design out before I modify it. Right now I just have to figure out how to get azimuth position. I can stick a VP integrated encoder in the gearbox, but then I have to find the best way to zero position.
Your observation of less voltage/less torque is why simply limiting the voltage on a too hot system won’t fix the problem. Dropping the gear ratio also helps but ends the same way.
Well, I’m not a controls guru so I am not able to explain it. It has to do with the inertia of the rotor, how fast the rotor can be accelerated, how quickly the motor slows from the load of the system. Any more than that and it’s over my head. I’ll ask our control guru if he will explain. If he can, it won’t be until next week. I’ll see what I can do.
If the 1:1 is right for you system, then you’re good to go.
In the old days, we used a zero switch. Not any more.
The VP encoder has an absolute and an incremental output. To calibrate the zero position, we physically “zero” the wheels. We run a zero calibration program that reads the absolute encoders, generates an offset, remembers it. You do this once of whenever you think the cal is off.
When the main program runs, it can now figure out where the swerve is at. After this it uses the incremental output to control position.
I see. This design has a spur reduction outside of the VP gearbox, so it gets complicated. I think I am going to get rid of that and just do as you described.
You’re right. To use the absolute part of the VP encoder, it needs to be 1:1 after the VP. Sorry about that.
