Archive for July, 2011

 

The Fiery Demise of the Land-Bear-Shark

Jul 18, 2011 in Land-Bear-Shark, Project Build Reports

You already know this is going to be exciting. But first, a video.

In accordance with my goal of bringing one random vehicle to every Swapfest, I rolled LBS out of MITERS (this time on a handcart to prevent 50-feet-away-death syndrome) once again after rebuilding the motor controllers for the nth time over. It proceeded to run the entire day without issue. Due to its more tame nature compared to things like LOLrioKart or melon-scooter, I let a few friends zip around the seller booths and parking garage on it (some of which is seen in the compilation above).

But with the day reaching temperatures of 95 degrees or so, and after climbing a dirt mound for a while – with much stalling and low speed running – and then afterwards gunning it at full speed up the parking garage ramps, the Beast-it-trollers finally gave out.

The little molten metal balls and pyrolyzed epoxy are classic signs of thermal overload failure. I’m satisfied, though, that this is probably my first real legitimate controller failure. Not some design flaw, or gate drive explosion, or being plugged in backwards, but honest-to-goodness overload. Now, if these had just exploded right away, I would have probably called it quits and left the DIY controller market forever. Maybe now I can slowly climb back up the controller ladder once more…so the next version will return to half-bridge synchronous rectification? And then I can have an H-bridge again? Then maybe one day, back to brushless!

Though with all of the molehill climbing I did, I also seem to have toasted the left side drive motor. That’s what probably led to the controller failure – the motor distinctively has the “shorted windings” feel when turned by hand.

But that wasn’t the most exciting thing that happened!


OH GOD KILL IT WITH FIRE

I wheeled LBS back to home base and pushed it off the handtruck and that’s when a majority of Hell broke loose. For seemingly no reason, massive plumes of smoke began pouring out of the frame. I figured the motor controllers had somehow failed short, but that couldn’t happen – I had already removed the key switch. So the next logical conclusion was BATTERY FIRE. However, the smoke did not smell of burning lithium battery electrolyte – it was purely wire smoke. I began pulling the thing towards the nearest door anyway in case it ever decided to transition to lithium battery electrolyte. By this time, the rate of smoke production had almost tapered off, so I just started removing the skateboard top instead.

It seems that the entire balancing harness shorted and caught fire. My assessment is that because I transported LBS upside down on a handtruck, the unrestrained battery pack mashed the balancing connectors and abraded the wiring down to the point of shorting. It would only take one short between cells to cause massive heating because of the cells discharging, and subsequently it would spread to the wires around it by melting the insulation.

The heat was enough to destroy the kapton insulation wrapping.

Flipping the cell bank over to the side where the individual wires were consolidated reveals the “flame trail”. Looks like the lower 4 cells shorted on the two packs.

And the packs after some selective surgery to get rid of hanging bare wire leads (and a little bit of brushing). The cells themselves seem to be fine – which I would imagine, since it’s really hard to kill an A123 cell using a 22 gauge wire. All cell voltages were checked and they seemed to be consistent with a pack that was just used intensively with cell voltage between 3.3 and 3.4 volts.

I’ll probably just rebuild these packs using more carefully routed, heavier insulated wiring.

So with pretty much all aspects of the tank now out of commission – controllers, motor, and batteries, I think I’ll temporarily declare this thing “done” for now since a full restoration is now going to need alot of effort. Maybe it will be seen again when the snows come back in the winter…

A Quadrotor-shaped Sculpture, and the Continued Tragic Goals Reduction of the Land-Bear-Shark

Jul 17, 2011 in Emergency Quadrotor, Land-Bear-Shark, Project Build Reports

Hello Internets! If you want the latest about what’s going on with this design (since this post is kind of an awkward in-the-middle post where nothing is working yet), see the EQ category here.

-ChArLeSg, 19 July 2011

No, it doesn’t fly yet. Quit asking.

I basically spent the better part of a day putting together the entire frame and placing the components. Given that everything was rods-and-joiners, this process did not take long at all. The hard part was actually making sure things came out square (or 45 degrees in the case of that X). I also made a center coupler for the X that also mounts the flight controls. The coupler increases the torsional rigidity of the frame signifiantly by forcing the crossing rods to be a constant distance apart.

INSTANT QUADROTOR!!

Not really. In this picture, I’ve slipped the 4 fans on and aligned them pretty exactly according to the models. I cut the rods pretty exactly to length – tolerances of maybe +/- 1/16″, so overall the thing is very square.

The weight of the fans is causing a bit of upwards bow visible in the carbon fiber rod. That will be fixed once the ‘upper level’ is added.

Here are the cute little corner landing legs, still incomplete. Ultimately they’ll just have a section of carbon fiber tubing embedded in the hollow conical point, upon which will be impaled a nerf ball or similar squishy ball-shaped thing (no, not Ballcopter)

The second ‘level’ has been added, including the battery hangers, so now the frame is much stiffer. It’s still lacking in the torsional rigidity department (i.e. if I grapped each end with 2 fans and twisted), but that’s a given since the upper and lower rails aren’t really joined at all, and the CF tube is still somewhat flexible. Since this thing should never be picking itself up by 1 corner, I don’t think it’s problematic. I have, however, considered just cutting out a long balsa or lite-ply plate which bridges the long center gap between the tubes.

I really like my battery mounts here. They’re Velcro cinch straps from McMaster, and are perfect for things like hanging a not-too-heavy battery in a not-too-solid manner. Little flat 3d printed feet give the battery packs a surface to be aligned on, and also hold the nylon straps from falling off sideways.

Most important of all, they’re adjustable side to side and the rail itself is adjustable front-to-back. I’ll need it when I’m trying to balance this thing.

And here it is, with appended Arduino carrier board and a Razor 6DOF IMU appended on top of that. What’s missing? Well… everything else. The heavy power wiring needs to be added still, and then oh god the software.  The total weight in this picture is exactly 8 kilograms.

I probably just made some Aero/Astro folks cry there, but with all four fans hauling, I have 8 more kilograms to go.  (Okay, so realistically like 3 or 4!)

Can you spot the tiny quadrotor?

A visitor from Physical Sciences Inc. dropped by MITERS with that cute little thing in the center. It’s adorable…and exceptionally stable for something so small. I hope this contraption works…

failshark

So let’s see, what have I actually built compared to what I wanted the Land-Bear-Shark to be?

  1. Brushless drive, wireless dual-channel wrist-controlled, reversible, with electronic braking
  2. Brushed drive, wireless dual-channel wrist-controlled, reversible, with electronic braking
  3. Brushed drive, handheld radio controlled, forward-only, with electronic braking

I am proud (…?) to say that now it has been reduced to forward-only with coasting. And radio controlled…..barely.

After the last update, LBS kept blowing through motor controllers (sound familiar? Only all of my vehicles ever do that). For some reason, the signal-side and gate drive problems were never resolved, and the gate drive chips kept failing short. Their inputs simply became low impedance for some reason, causing the microcontroller pin voltages to no longer change sufficiently to indicate a shift in logic state. The outputs would also some times randomly die, but there was never a power semiconductor failure. That means I’m getting somewhere, right?

In any case, LBS was going to premier at the last Swapfest in June. Literally the morning of, I tried riding it to the event….and made it about 50 feet out of MITERS before the left side drivers gave out. Dragging that 70 pound brick back to the shop was unpleasant.  Then in late June, along with my partner in engineering malfeasance Amy, I replaced the drivers (again) for a joint “demo event” to staff in the Mechanical Engineering department. This time, I made it all the way onto campus and around some building. But once again, in the middle of blasting around in the hallways, the right side quit this time.

Derp. At least I had a pushcart to transport it back this time, because I was just going to leave it in someone’s office otherwise.

After that failure, LBS sort of sat on a stand at MITERS, taking up space until last week…

Clearly, I just needed a new start on the motor controllers. I didn’t care any more about regenerative braking, reversibility (standing on the thing while spinning in place is actually really hard and taxing on the motors), efficiency, brushlessness, or anything. I was just going to rig up a single-quadrant forward-and-coast-only controller. The most bone-simple thing you can build that can control any reasonable amount of power at all. This controller has been called Beast-it-troller because it’s time to just beast something.

This controller is a single low-side pull-down configuration. The reason there are “upper side” FETs is because they have been coerced into acting as diodes by bridging the gate and source leads. This acts as a flyback diode for the motor. The single low-side configuration is vulnerable to uncontrollable latchup failure.

Now, doesn’t that make you comfortable? Regardless, it’s dead simple and ran LOLrioKart for almost a year.

I designed it in about one hour and had the board files immediately sent to Advanced Circuits for fabrication. Their quick-turnaround basic service is reasonably cheap and great when you need a board…………….right about now.

I decided to totally gut the system wiring and start over, too. While I had all the components dismounted and unwired, I decided to do a full bench test on a current limited power supply. The code was knocked down to the barest functionality needed to go forward or do nothing. The dummy loads here are some adorable small Minertia servomotors picked up off a free stuff list.

A board designed in an hour probably has bugs and numbskull errors, and this was no different. Like every controller I’ve ever built (ever), it needed a Little Blue Wire hack. I neglected to actually route the MOSFET source pin (power side ground) to the gate drive IC’s gate return pin (signal side ground). While the grounds are in fact connected elsewhere, so it wasn’t the end of the world, having the gate drive return current come through a few extra feet of wire would have introduced noise into the system.

After that, it just kind of worked. It’s hard to get this type of controller wrong.

I also discovered that this thing has a very unique and convenient self-balancing service position.

The downside of the fowrard-or-bust control is that turning with a short radius is less predictable and more finicky, depending strongly on the available drag on the slower tread. It’s also almost impossible to turn in place unless I shift way back and keep most of my weight on the ball tail. However, larger radii and sweeping turns are no problem.

Guess what? July Swapfest is tomorrow!

Declaring Emergency (Quadrotor)

Jul 13, 2011 in Emergency Quadrotor, Project Build Reports

So I have been quietly working on quadrotor business for the past week and a half after Ballcoptering from 3000 miles away. All four of the thrust modules are done, due in part to Make-a-Bot logging something like 70 or more hours of cumulative operation during that time alone. There was also a little bit of work done shortly before I left, so this post will recap all of that. I’m fairly confident I can get this thing making an ungodly racket by the weekend (those fans are LOUD) if not a controlled and scientific one.

Speaking of science, I actually got the chance to put one of the tailcone-equipped fans onto the Fankart Rail for a real honest-to-robot-Jesus instrumentation session. I was curious to see how addition of the smooth cone surface would affect the thrust output of the fan – since without it, the motor ends bluntly. Given the exhaust velocity of ducted fans, I was expecting a difference of a few ounces at full throttle.

I think my expectations were in line. Using the same 10S lithium pack as last time, I got a maximum sustained 2.97 to 3.0 kgf of thrust. The pack dipped to 31.5 volts during this time. The results do closely match last times – 31.1v minimum and 2.85 kgf.  I didn’t get a RPM reading this time, however, and that extra 0.4 volts might well account for the difference.

But if that is not the case, the tailcone appears to net me back around 100 grams or so. Not that significant, but regardless it adds to my aero/astro cred, right?

It turned out that one of the stock 5mm adapters that I had reamed to 6mm was still way out of round. That’s how horrible they were to begin with. I didn’t have a small enough boring bar (that went deep enough, anyway) so I had to make do with a 6mm reamer. It probably just followed the path of least resistance anyway -  I didn’t mark the parts to distinguish them, but one of them was distinctly rushed.

To resolve this, I elected to just duplicate the most important dimensions on the stock part and make one myself from scratch. It worked well – while I couldn’t find a lathe which could turn metric threads (and did not have a 10mm x 1) die, 26 threads per inch is close enough to 25.4 (1mm pitch) that the prop nut was able to thread on enough.

I also managed to destroy one of them instantly when I forgot that I couldn’t thread at 600 RPM.

Oops. Luckily, I only needed one.

However, ultimately the effort was for nought – the one that I didn’t destroy was improperly reamed out to 0.238 instead of 0.236 (6mm). The other one was left at 15/64″ (0.234″) and it turned out to be a snug fit on the motor shaft. For some reason, I went for the loose oversize fit on the second because there wasn’t a 6mm reamer. That 0.002″ of radial play translates into OH GOD WHY DID I EVEN BOTHER at the end of the prop adapter once everything was tightened down.

So discovering that I can no longer Course 2 properly, I borrowed Shane’s prop adapter.  :<

After many hours of scraping 3d printer droppings, here are all 4 thrust pods completed!

The upper left controller is zip tied on since during my trip away, someone knocked the module off my shelf, breaking it into a few pieces, some of which was the Turnigy controller’s mounting flanges. I had spares of the wood parts, but still. Not smooth, bro. Not smooth at all.

At least they piled the wreckage very neatly back on my shelf and proceeded to not tell anyone.

I had to print a total of 32 of these little corner joiner things, in batches of 9. Pretty easy, right? Yeah, each batch took 4 hours and I had to make 3 more sets after discovering that some were done in the wrong orientation (where the lamellae of 3d printing are stressed in their weakest direction). Oh, and one set died mid-print, so that’s one more.

While I was staring idly at the printer, I whipped up these switch panels. They’ll attach in between the upper and lower set of frame rods. This thing will have a front and back power system, since otherwise I’d be facing very long wiring runs. This setup doesn’t necessarily weigh less (in fact, the Hella switches are about 4 ounces each), but it simplifies the wiring somewhat. Each battery feeds 2 controllers directly instead of all battery wires coming together and then branching off again. I also get the benefit of redundant 5v BECs, one from each battery. The small switches are designed for precharge circuitry, allowing the logic to be powered without main power application (and also not beating the controllers in the face with all 40-odd volts of the battery at once).

I’m actually going to try a very unconventional arrangement for the master power switch. Because the Hobbyking 10S batteries are actually 2 discrete 5S packs with a middle “bridge” connector, I’m going to wire these switches such that they open and close that middle bridge. This should, in theory, split the battery into those 5S packs with an isolated center, allowing me to charge them with a single 4-channel charger. I have yet to work out all the implications of this, but it SHOULD work.

So here’s what a batch of parts looks like fresh off MaB’s print bed. It’s covered in stringers, little noodles, and support lattices which are also extruded ABS, so some degree of knifework and sanding is required to clean the parts up afterwards. These are the joists for the center truss.

But this is what they look like after they’re cleaned up!

With these important pieces complete, I could start cutting carbon fiber tube to length. The right angle pieces also shown here are the (+37) Rails of Battery Mounting, which are actually made using square CF tubing.

I couldn’t assemble the entire frame, however. I didn’t design any means of landing this thing safely yet, and putting the frame together would have meant that the motors were sitting on the tailcones. Probably not good news, and definitely not good if I want a remotely aligned frame.

So I made a frameholding jig for the time being. Its only purpose right now is to create a place I can safely put the quadrotor down on.  This is not landing gear, though it’s of the right shape and not very heavy, so maybe it could be pressed into the task. This little thing is made from laser cut 1/4″ cheap plywood. The V grooves seat the carbon fiber frame tubes, and the distance between towers is also exactly the distance between the joists of the truss.

To come: Putting it all together!

I already received the ultrasonic distance sensors and the IMU, so after frame assembly I should be able to move quickly onto master power wiring. Otherwise, this thing is mostly plug-and-(program-and-test-and-)play.

Ballcopter!

Jul 09, 2011 in Ballcopter, Project Build Reports

Hello everyone, I am pleased to introduce Ballcopter.

It doesn’t work.

But it sure is cute, though. Look at how cute it is!

Ballcopter was inspired by a recent Japanese experimental ball shaped drone which made the rounds on the Internets about a month and some ago (though it appears to have been in development for a year or more). Here’s some more test videos:

Needless to say, I was instantly enamored with the spherical design. It’s so much simpler than a multirotor machine (like A Certain Emergency Quadrotor) and appears to be even more agile. And being Japanese, infinitely cuter. Having no details about how it was built besides just staring at videos, I sort of mentally reverse engineered it and designed one, but did not go about actually building it.

…until two weeks ago.

visiting the other coast

I was extended an invitation to visit Makani Power, an airborne wind turbine (AWT) research and development company based in Alameda, California, by MIT alums affiliated with the company. Lots of MIT graduates (and recently, MITERS alumni) end up in the San Francisco Bay area for some reason, and I found the environment at their shop/facility very much reminiscent of what I’m used to here. While I’m not someone who is enamored with wind power, I did take the opportunity to talk to the engineers and learn a bit more about aerodynamics and wing design, and also tried my hand at laying up carbon fiber.

Those two things can only result in more desire to build flying objects from me, so I’m not sure if the outcome can be considered positive or negative.

But I also built a ballcopter. I spent a total of 6 days at Makani, and while not discovering that carbon fiber fabric actually does deform and change shape (I had wondered how on earth smooth CF surfaces were made before), or looking at how composite airfoils were manufactured, I whipped up a design for Ballcopter, ordered parts (and had them overnighted to the facility), and used the shop to cut out the frame and finish it up. The parts cost was  about $150 including the whole overnighting thing, and the total time of build from start to finish was only about 60 hours.

And it kind of shows….

This is a picture of the first few hours of design. Again, there is absolutely no science to this at all – it’s a crude visual pirating of the vehicle shown in the test videos – I just sat down and started CADing. The frame was designed to be made using foamcore (stiff polystyrene foam overlaid with heavy paper), also known as posterboard, foamboard, etc..

I’ve added the control flaps here. The way I guessed that the JSDF drone worked was using the upper set of flaps to control direction (tilt and movement in XY, or I guess in this case XZ) and the bottom set as a rudder (spinning about the polar axis). Opposite diagonals of the upper flap set move with eachother, and all the rudder flaps displace the same relative angle. I was intending to just use a stiff tape as the hinge, which seems to work out for most small airborne implements. The servos are 9 gram miniservos, and the propeller mockup is standing in for a 10 x 4.7 inch prop.

The design progresses a little more with the addition of the… Tropic of Capricorn? The wide band adds structural rigidity to the outer shell and also acts as a crude duct to funnel air over the rudder flaps.  I have no clue what they actually do, but J-ballcopter had it too!

Because the timeframe was so short, I pretty much just went downstairs and heaved foamcore on the laser cutter and went for it.

Wait, foamcore on a laser cutter? Now, MIT’s rules have always said that cutting foamcore on a laser is an instant death sentence followed by permabanning from everything ever. You won’t even be allowed to use the shop in Hell, even though it’s already on fire and people are already dying around you. Foam tends to melt and burn, then drip and burn more, until a serious fire develops. I was not entirely convinced that the rules were founded on technical barriers, just peoples’ stupidity, so I elected to experiment with foamcore cutting settings on a foreign laser cutter.

And you know what? Like I suspected, it worked just fine. The key is to use many passes of a low power cut with very high head speeds so the laser never parks long enough on one spot to burn the foam. I ended up finding that 4 passes of 20% power (on a 150W system) and “30%” speed made for a very clean cut in 5.5mm foamcore, with the last pass just barely breaking through the paper backing on the far side, resulting in a good ‘puzzlebot’ sheet part. These settings are for a Universal Laser ILS1275, so I don’t know how that translates to Epilog or any other manufacturers.

Disclaimer: if you’re an MIT student, DON’T EVER CUT FOAMCORE ON THE LASER CUTTER!

Instant ballcopter. Just add hot glue! Notice that one side panel has been left off so I can actually service the thing.

And the parts arrive. I got all of these from Hobbypartz.com, which seems to be a U.S. variant of Hobbyking (you know, the one that isn’t Hobbyking’s U.S. office). I used to see them spammed all over the place, but hey, they’re legit. The lineup includes a shady 6 channel 2.4ghz radio (functionally identical to the 6 channel HK radio I bought for the robots last year), eight 9 gram miniservos (because they’re like 2 bucks), a 28mm outrunner with 25A controller, a 1Ah 3S lipo battery, and a few little linkage parts.

The one thing which I could not get because it was out of stock, and which I couldn’t find at the local hobby shops because they were useless, was a propeller hub for the motor. So I actualled machined something (!) and made one. It’s based around a short 1/4-20 bolt that was drilled out and had a set screw added. Stock locknuts and washers take care of the fastening of the propeller. Sure, it’s heavier than an aluminum prop collet, but could I find an aluminum prop collet?

A laser-cut buffer plate made of plywood secures the motor to the frame. The servos were also given little laser-cut mounting plates because let’s face it – foamcore is just not structural.

And here it is!

Notice the small carbon fiber rods spanning the the upper flaps in an attempt to get them to synchronize.

The word “attempt” is key across all aspects of Ballcopter here. It really didn’t work, and after a bit of head scratching (and “duh” moments), I concluded that…

  • Those flaps are just too rigged and horrible. I ended up using electrical tape for the hinge material because it was in front of me. Protip: Electrical tape stretches. Instead of rotating the flap about the hinge when the servos moved, it was more liable to just displacing the flap without much rotation.
  • The servo linkages were not well thought out at all. In fact, they weren’t thought out – just built on the fly. The linkage is a 45 degree bell crank of sorts. While all aircraft control surfaces seem to rely on the small angle approximation, this only works if there are not like 3 different small angles involved and one of them isn’t based around displacement from 45 degrees. This arrangement is clearly seen in the above photo. Basically the servos were more yanking on the flaps sideways after only a few degrees of rotation. The same goes for the rudder flaps, where 2 flaps were actuated by a servo between them.
  • As soon as the prop started and the vehicle took off, those sad little linkages and hinges just collapsed and the result is not very much flap displacement at all. The vehicle was therefore barely controllable, and tended to just run away in one direction (or run away while twirling if I applied any rudder control)

The JSDF drone is gyro-stabilized, too, and mine was totally and utterly open loop.

So over all a great execution of Cheap Chinese Imitation that would probably have worked if it wasn’t crudely reverse engineered and then even more crudely reassembled. I do want to try my hand at this again some time, though – direct drive flaps (no loose and bendy links), real hinging, reducing weight a little (there is really no need to carry a 1Ah battery on something this small), and adding some intelligent stabilization using IMUs.

It’s so cute that I have to try making it work. Before I left, I stripped down Ballcopter back into parts, and hopefully those parts are en route to me in Boston. The frame was left mostly untouched and would make a great lamp or something.

Onwards ボールコプター!!

More Thrust Vectoring

Jul 07, 2011 in Emergency Quadrotor, Project Build Reports

I’m back.

Wait, where did I go for like 2 weeks? I’ll explain that later, as well as detail what Shit Went Down™, but I was finally able to run the Emergency Quadrotor fan with thrust vectoring on a full 36 volts:

Yeah, I just kind of grabbed it. No thrust measurements except “wow, my arm is sore from holding onto that”, but I wouldn’t expect it to be more than 4 kilograms as recorded legitimately before.

This test confirmed the servos as being strong enough to muscle around the fans, even at full speed. I was concerned that gyroscopic forces would mean the servos would act too slowly to control the vehicle, but they are still responsive. There is visually discernible lag if I manually “step” the servo position from limit to limit, but it’s not enough to concern me: the fans should never be split-second actuating from lock to lock.

The test also (fortunately) showed that the structure is strong enough to hold back the full force of the fan. There’s no reason itwouldn’t be, but still.

Next step: Make 3 more. Poor Make-a-Bot.