Saturday, August 17, 2013

Catching Up... First Drive, First Fail.

It's been a busy summer... the good news is that the first drive was done back in July, and I was really happy with the performance of the chassis and drive system.

The rover had no issues with the lumpy, weed-infested patch that pretends to be our backyard. I wasn't sure it would work, since the tires are very soft, and small obstacles tend to push into the tires, instead of the tires rolling over the bumps.

Putting down some lumber ends as ramps, it had no issues with low angled climbs. Again, I wasn't 100% sure it would work, as basement testing proved that smooth angled surfaces are a challenge for really knobby tires like these. I can see why NASA opted for the wheel grip patterns on their solid wheels.

The bad news is the steering servos suck. At $17 each, and a rated 83 oz-in (which is pretty low), they can't turn the wheels correctly with the rovers weight on them.

Here is a short video:

That bit at the end really ticked me off.

Although I have spare wheels and hubs, I didn't want to chew through my supply just for the sake of additional testing. That meant an end to testing until I can get some solution to adding more torque.

A few of the ideas I've had:

- tiller arm designs, which could get a 1.6x improvement on torque.
- bigger servos, that might still fit in the servoblocks. That might get a 4x or 5x improvement.
- geared servo's, but those are quite expensive (after shipping and customs it would add $500 to the rover). They could get 8x - 12x more torque than I have now.
- dc motors as servos.

Right now I'm leaning to the dc motor route. I have plenty of Atmel processors and L293D's to act as the servo controller, and I've breadboarded them with four small geared dc motors I have. I don't have a torque figure for them, but at 24v (and fairly low rpm) I don't think I could stall them by hand with pliers on the output shaft. I don't really want to try, either.

The hardest part is getting the position feedback done right. I could add gears to feed a rotary potentiometer to the mix, but the price goes back up to $100 a wheel. I still have to think about how to add bearings (once I remove the servos), so I know it's not going to be a free ride.

I did consider investing the $400 in a DIY 3D printer (I have a few steppers as well) so I could print my own gears, but that's a pretty big distraction from this project.

My interim solution is to use linear slider pots, which should (with a little plastic construction) be able to feed position back to the processors. Because the arrangement of angle to position wouldn't be rotary the processor can also do the math to get the angle based on linear position. I should be able to rig something up that doesn't cost an arm and a leg.