Wednesday, August 8, 2012

New blog and old projects!

Alright, so while I'm over in Washington jumping out of planes and with very limited shop access, I decided my downtime would present the perfect opportunity with which to categorize some old projects, as well as things I'm currently working on. Hence the start of yet another generic Mechanical Engineering blog!

I figured I'd start with an old project of mine which I recently got working again: Derpcopter! This is one of the few of my projects lying in the "Actually Finished" category of stuff I built/am building.

Derpcopter was initially conceived as both a good way to make my first foray into CAD, and to satisfy my high school graduation requirement of a senior project.

So, the first Real Thing I ever CADed:

Yes, that's right. AutoCAD for a 3D assembly. It was all I knew at the time, and it got the job done. Notice how this design is currently lacking any motor mounts whatsoever, if I recall correctly I actually just made those in Visio (the only drawing software available on the computer hooked up to the laser cutter). I was initially intending for the lasercut parts to be acrylic, but ended up settling on polycarbonate because it's much stronger/less brittle. (Note: don't ever ever lasercut polycarb, unless you want all manner of delightful brain damage which I probably have by now).

The platform was based on the Arducopter build of the beginning of 2011, when the project was still very much in beta. Knowing approximately squat about electronics at the time (not that I can really say much has changed on that front), I just went with their default Ardupilot Mega hardware and Arducopter firmware, some 30A ESCs, APC 10x4.7 props, and these actually pretty decent motors.

I specced out 5/16" aluminum squaretube for the arms, 1/4" polycarb for the frame and 1/8" polycarb for the electronics mounts. It turns out 5/16" squaretube with 1/16" wall is commonly used in upscale railings (who knew?) so I picked up a few lengths for free from a contractor who no longer had use for them (not to mention they came pre-painted a sexy matte black). As for the polycarb, TAP Plastics has this great scrap bin in their stores where you can pick up excess sheets for practically free.

After making three or four passes on the 1/4" polycarb at max laser power with low speed and succeeding at nothing other than charring lines all over the place (and throwing around all manner of nasty fumes), I opted to swap all the structural components to 1/8" as well.

From that point, machining/assembly went pretty smoothly, and I was able to finish the wiring/electronics assembly with much digging through old forum posts to find any and all relevant info. If you're looking to build an Arducopter now you'll have a much more pleasant time of it, considering documentation by this point is far more substantial.

Finished build, weighing in at a whopping ~2.6 lbs with battery:

I uploaded the firmware, calibrated ESCs, mounted props and decided to see if Derpcopter would lift itself a bit. I applied what I thought was just a little bit of throttle, and immediately WHRRR-CRACK!
Derpcopter flung itself sideways straight into a wall, instantly shattering three of the props. Not only did I grossly underestimate the amount of thrust the over-specced motors I used could produce, I'd also failed to double check directionality of the props and one of them was backwards. Well, crap. I'd only bought one extra of each direction of propeller, so I was stuck waiting over a week for new props to ship (and yes, this time I bought like 10 spares).

Unfortunately I had to present the project before the new props actually came in, so I just spun it up a bit without props mounted to show my teachers that yes, it actually seemed to do what it was supposed to, which evidently was enough for me to pass with the project.

New props came, I put them on in the correct direction, flew around my neighborhood a bit, and then I must have given it a few too many hard landings because eventually Derpcopter became pretty much uncontrollable. I monitored direct sensor outputs from the Ardupilot on my PC and noticed that one of the accelerometers was spitting values completely all over the place. Derpcopter thought it was shaking like crazy, so the reason it was impossible to fly was that it was actually just trying to stabilize itself (even though it was already stable) and thus jerking all over the place. At this point I was nearing the end of senior year, and shelved the project in the "I'll fix it eventually" basement shelf.

Come the end of freshman year at MIT, I started spending some time over with these guys, who inspired me to dust off Derpcopter and bring it back to life. I noticed an inadvertent but quite convenient feature in my design:

It folds!
It turns out that converting Derpcopter into travel mode only requires taking out a total of 6 screws, yay! Anyway, I brought Derpcopter into my lab to work on, where a perceptive friend noticed that a smoothing cap on the APM board was toast. Whoa, that would explain bad accelerometer readings! Here's a picture with the rather hilarious looking not-at-all-surface-mount replacement cap:

Looks good! Now to upload the new firmware, and.... crap. It turns out that APM has upgraded their boards and are now based on ATmega2560s instead of 1280s. This means my old board doesn't have enough memory to support the modern Arducopter firmware. The solution? Upload firmware manually using Arduino, after having commented out parts that are irrelevant to my uses (there is a lot of bloat on the main Arducopter code, there are ridiculous amounts of features).

Somehow in the process of screwing with firmware I managed to get the plane edition of the code. My quadcopter spent several hours fully convinced that it was a plane before I finally got correct code to compile and upload. Planes like to go at mid throttle when the throttle stick of their controller is all the way down. See why a quadcopter that thinks it's a plane is a very bad thing? Fortunately I didn't have props mounted on it, so Derpcopter just spent a while happily spinning all motors when given no throttle input as I tried to figure out what it could possibly be doing.

I finally got Derpcopter running, fixed orientation issues, and adjusted PID values to get something flyable.

Yay! No one is dying! It still flies backwards (the taped arm should be the front, at the moment it's treated as the back), but I can live with that for now. When I get back to Boston I'll take some footage if it doing crazier stuff over Briggs Field, but for now you'll have to be satisfied with the first working test flight.

No comments:

Post a Comment