Propositio Tertia: Boston sidewalks suck ass.

It was certainly the most fun impulsive build I’ve ever completed. What began as a desire to mount a giant brushless motor on something, anything, turned into a foray into the world of exotic power systems, electric vehicles, battery technologies, and electronics engineering. After I finally learned how to control the motor’s power, it became a great local transportation device. However, after a month of operation, I’ve decided to pull the plug on Snuffles.

Actually, it sort of pulled its own plug, but I was the root cause. The pot-to-R/C interface board pretty much completely shattered over the course of several trips around Boston. This just shows that I needed more support on the board itself. The most important thing, though, is that if you want extreme performance, electric will triumph over internal combustion. This was known to me for a long time, but only by building Snuffles did I experience firsthand the power curves of a giant electric motor.

Too giant, actually. The 3.5″ diameter brushless airplane motor was utter überkill for such a small chassis. I couldn’t stay on the thing the first few times. Gradually, I learned that I could only stay on if I give it a really good kick start and then slowly ease the throttle in. More practical performance can be had with a motor significantly smaller – perhaps even in the 40mm range. The 70×50 motor of Snuffles would be slightly powerful even for a bike.

The successful execution of what amounted to a prototype and proof of concept therefore begets refinement and improvement. Recently, I have exercised the power of Googlepedia to perform more in-depth research on in-wheel electric motors – that is, slamming all the drive components into the structure of the wheel itself. This opens up space that would other be used to house drive components and lets you… well, use it for something other than drive components. Like cushier seats. I had used integrated motors before with Pop Quiz, which had a motor built into the blade. It worked well for the amount of quality non-control I put into it.

And so this is where the next chapter of my obsession with giant brushless motors will start. Now that I have membership at the MIT Hobby Shop and am working with the Media Lab, I have access to just about every weird manufacturing process available. Over the past few days, I experimented with designing wheelmotors for Snuffles. The goal of the design was to pack the motor into the confines of a 125mm scooter wheel such that minimal modification of the rest of the structure is needed (the only mods being, for example, drilling extra wiring access holes.) The real “inspiration” for this madCADing was my… uhhh, ravaging of several large copy machines (Example picture from earlier in the year). For those of you who don’t know, large copiers tend to have rotating-can brushless stepper drive motors.

I drew up two motors in Inventor. One is a conventional ironcore “outrunner” motor as built by R/C hobbyists around the world, using parts from the copier. The other is a more radical axial flux ironless motor which is used in high performance, high efficiency motors and generators. For narrow applications such as the scooter wheel, an axial-flux type motor seems to make more sense for space efficiency. However, I was just as successful in shoving the copier stator into the wheel.

Here are some pics.

The outside of the axial motor. Actually, they both look sort of like this. To mount the motors, I have to drill out at least one hole in the wheel mounting locations on the scooter frame to accomodate running wires outside (since you can’t really get power to something like this without wires). A hollowed-out screw will be used so the shaft can be fastened while providing access to the wires.

Here’s a shot of the inside! The coil shapes are obvious stand-ins, since I have yet to figure out how to draw an odd-shaped coil in Inventor. Nor do I intend to, since that’s over the top. The coils stay stationary, and a hole is drilled crosswise in the dead shaft to give wire access.

Two of these magnet rings are used, one on each side of the windings, to give the best concentration of magnetic flux. This part rotates, attached to the wheel sidewalls. I had to “composite” smaller magnets together into bigger ones, since not everyone makes 3″ diameter flat motors with 14 magnets (read: I can’t buy the right magnets stock)

The inside of the conventional motor. In the end, only a weird combination of metric and english unit parts made everything align and space correctly – otherwise, things were always too far apart, too close together, or some other anti-three-bears condition. Four magnets group together into one bigger magnet, and there are again 14 of those.

To understand why I’m using 14 magnets, one must understand the concept of LRKs.

Here’s the outside of the conventional motor with one side plate removed. In reality, there will be little amber copper windings wrapped around the stator arms. Or big ones… I hope to either use giant magnet wire to push as much current as I can through it, or use a multitude of tiny wire strands in parallel. The latter achieves better volumetric packing.

When will I build it all? Nobody knows. Unfortunately, I have these stupid things called classes.

However, I hope to integrate this with the research I am helping the Media Lab with – which involves, of all things, light electric vehicles. So watch this space!

Snuffles failsafed in the middle of the Harvard Bridge today, and I almost took a faceplant into the concrete. I didn’t, but my custom signal generator MUST be replaced before I’ll ride it again. The unsmooth Bostoncaster sidewalks are shaking the electronics apart.

I was crossing a street while on the way to, of all places, the bike shop to get a big U-lock so I can actually start using the thing as a campus and town vehicle. While cruising across the street, the ESC suddenly reset.

Now, brushless controllers play a setup tone through the motor, using it as a speaker. This means the motor suddenly applies power in random directions, maybe more than one direction at once. Either way, the rear wheel locks up and I almost fly off. After a few power resets and poking, the ESC wouldn’t even play the setup tone, meaning there was no control signal at all, but the main power relay still latched, so it is getting battery power.

Fortunately, Snuffles is small enough to kick scoot and limp back to home base. I’ll find out the problem later.

One, giant outrunners are insanely powerful.

Two, this one is far too powerful for a small frame like the ElectricX2.

Three, I need a running start before jumping on or else the motor will torque so hard as to wheelie backwards and throw me off. That or I need to lean across the handlebars, Motocross style, so the most leverage possible is exerted againt the motor. This, however, looks really retarded.

Snuffles, so far, is completely uncontrollable and is a monster. This is a good thing, as it was the original design goal, but I have yet to tame this monster. I have a few days to practice, though!

I’m not sure if I’ll ever understand what the heck it’s doing, and just take its little flashing lights for face value.

It also comes with two fuses, but with no fuse holder or any other visible sign of a fuse. It also seems to be ultrasonic-welded together, with no screws, so if I ever have to replace a fuse, I seem to need a sledgehammer.

Here’s hoping it works and never breaks (yeah right). I will not hesitate in sledgehammering it open at the first sign of trouble.

1. Tuning
2. Battery armor
3. COLD CATHODES!

Anyway, no Major Build Disasters were logged, which is fortunate. It had its share of problems, however, including misfitting components, random switch shorts, and way too many hot soldering iron burns.

Some final build pics!

One of the hallmarks of my well-known power tool abuse portfolio is “saw-milling”. This involves using a power miter saw, my favorite universal tool, and using it to cut sideways. This involves pressing the piece against a solid backing and making sure it cannot flip upwards (which results in probable bloody injury), then carefully slicing off layer by layer by sliding the piece past the saw teeth. High tooth count compound cut blades do this readily, and I regularly use this method to cut small slots and such. Here, the right side motor cutout is being “machined”.

It fits pretty well, however. The sides are 45 degree approximations of a full circle, which is good enough. This doesn’t support the motor in any way, as the outer can rotates, but functions as a anti-scrape device.

Both plates installed. The right plate also functions as a motor guard, preventing me from whacking it right into something, which would be rather bad.

Of *course* my remaining aluminum is a quarter inch too short to be an undercarriage plate. What, did anyone actually expect it to be right on?

A sheet metal assembly might be created to deal with this.

The two major electronic bits mounted.

Wiring up the back end. Some of the parts are a little tight, but they just barely manage. I also mounted the control board a bit far forward, but not enough to risk stepping on it. Wires are conveniently tied to the many standoffs present.

Enjoy this shot of the water heater. The “switchbox” mounted at the top houses the big power relay and the switch to actuate it. Remember, both red is actually battery negative!

This was the site of probably the most retarded electrical problem I had ever seen. During testing, the battery power somehow got to the ESC while bypassing the switch. It wasn’t a dead short – things kept working just fine if the relay turned on. After testing and unsoldering everything, I more or less got the problem down to the long run of 12 gauge zip wire. Maybe it was a manufacturing defect, but after slitting the wire down the middle with a razor, separating the red and black, things worked just fine.

Therefore, the only valid conclusion is that cutting solves all of life’s problems.

No, that’s NOT ME. Quit asking.

The Splitting of the Paths; or, rather, of the control and power cables. Since the boards are on opposite sides, I routed them early.

Okay, here it is. The Sharper Image Electric X²/2, codenamed “Snuffles” by a lovely lady friend of mine. So… uhh, Snuffles just needs a few more bits and piece to function at 100%, but is fully working right now.

That’s it! With a week left, I’ll be focusing on other things… like…

Let me snap out of build-stuff mode first.

Pics!

We begin today with a picture of my disaster area of a work bench, as well as a handy creative use of the hand drill and the drill press at once. There is a 1″ wood spade bit in the chuck, and a piece of aluminum in the vise. The spade bit does not go through the aluminum. Solution? Brute force machining using the much more powerful corded drill, transmitting power to the spindle through a 15/16″ socket. I figured that with the drill cranking at full force the spade will eventually melt through the aluminum and be on its way.

This was not what happened. Unfortunately, the spade bit (which should never be used on metal anyway, especially not like this!) just blunted out completely and polished the aluminum at a depth of about 1/16″.

I left this full size (3000 px wide, 3 megs, blame my carelessness with the quality slider) if you want to investigate the metallurgical structures of aluminum. I think I almost snapped the table right off the drill press by slamming the lever so hard.

Therefore, I had to resort to the OTHER brute force machining. Drilling holes in a circle, then bridging those holes with a Dremel drywall cutting bit. This worked very well.

Here is the throttle body with the potentiometer mounted. It will be clamped to the handlebar. Why couldn’t I use a stock, well-built, easily purchaseable throttle? Because the handlebar is a weird diameter nobody makes accessories for.

Complete throttle body with aluminum thumb lever thing, which will probably be dipped in something soft. I couldn’t believe that none of the random mechanical tidbits I stockpile had a single usable torsion spring. In the end, I faked it out with a tension spring wrapped around the resistor shaft. The spring anchor screw was dug out from my hardware trap, and is an odd standoff-screw at the most convenient height possible. This screw also acts as a travel limiter.

I found that using a 10K pot was better than a 5K as perscribed in the original servo tester circuit since it made the useful travel range smaller; that is, more usable with a thumb lever. The bottom travel limit is the potentiometer body itself, and corresponds to 0% power on the controller. The top end travel limiter is the screw, and it stops exactly where “full throttle” on the controller is. The controller automatically shuts down if it gets past this angle, causing the electric brake to kick in and me to probably fly off. Therefore, the stop is right before this limit.

Mounted on the right side handle bar. It’s nice when things just work.

Working on the switchbox. I had to yank the big 10 gauge cables through the holes in the box with pliers, because the holes weren’t quite big enough. Originally, the holes were for LEDs – a red one for “low battery” and a green one for…well, something other than low.

I had to break out the huge 15 inch long 100 watt iron to solder the big cables to the big relay terminals.

Power meter tape-welded to a bent metal plate screwed to the completed switchbox, which has been mounted back in its original position. It’s a nice fit. The angle could be a bit shallower, but I won’t take it off and adjust it again for fear of breaking things.

Notice the white wire next to the thick red and black ones. This is actually a four-conductor cable salvaged from a computer mouse. Conveniently, it was six feet long, and had all four conductors I needed – +12v for the meter, and three wires for the throttle. There’s so much convenience on this thing it’s not even funny. Actually, it is.

One more interesting feature is that the red and black wires are both on the battery negative side of the whole electrical system. Why? Because I had six feet of red and black zip wire, not six feet of black and black. Interesting things will happen if I wire this red cable to positive, such as dead shorting through the relay and making a huge smoke cloud. Watch for this.

Dummy motor says everything is working as it should!

Now the only thing left is to get another slab of 2.5″ aluminum and some more servo leads. Then it could be done. By this weekend, even.

Watch for a MAJOR BUILD DISASTER in the near future! All my projects have Major Build Disasters around now, and I don’t think this is an exception!

And also some mechanical work.

But before that, here’s a shot of the beginnings of the instrument pod. I reused the orignal controller casing because by some strange cosmic alignment, the guts of a 80 amp relay slips perfectly into the space previously occupied by the stock motor controller. This also means I saved one more long wiring run to get to the power meter, which is now right next to it. The meter itself, with no mounting holes, will probably be tape-welded to the controller box.

And by tape-welding I mean gray 3M outdoor mounting tape. It can hold up the universe.

And here it is. The 100 amp, 12-lithium cell (44v) brushless controller, along with a handy programming card. I tried listening to BESC morse code once before and vowed never to do it again, and this programmer lets me select parameters and tune the controller as needed.

Testing the rig on the table, using a small outrunner. This was AFTER almost losing the controller to reverse-polarity shorting. How did that happen?

Notice the wires between the meter and controller are crossed, and obviously wrong in color, and not matching the label. But that’s actually the correct way – before I switched them, of course. The first test of the controller, using a small lithium polymer battery (Not the giant SLA seen here) resulted in the controller dead-shorting the battery. This was strange. Next, it tried to dead-short the 12v SLA (thank goodness small test cables saved it) and a 12 volt power supply before releasing a little poof of smoke.

Uh oh. I fired off a support email to United Hobbies asking if it might be a DOA. With nameless stuff, I gave it an extra chance of that being the case. Then I looked at the bottom label a bit closer and discover something horribly wrong – the wires were switched. Now, me being used to red = positive and black = negative, ignored this once-over inspection, so it very well could have been my own fault, but it could have happened to anyone.

So I carefully reversed the wires and tried again. It would struggle a bit, beep the motor, but would heat up intensely and not start the motor. Figuring it was toast anyway, I cracked it open and noticed only one bank of transistors was cooked. The controller uses all (48) of the same transistor, so I carefully plucked the tiny, surface mount chips from the other legs and replaced the burnt ones.

And that was how Charles saved the day. Sort of. The controller works, but might as well be a 60 or 70 amp one, with a few transistors missing. But at least it’s something. That will not spare UH, however, as I still intend to prod them about this…slight mix-up.

After calibrating and making sure nothing else would smoke on power, I moved the test to the floor and tested all systems. The result is that all systems are functional – including the awesome controller controller! I was surprised such a contraption could fool an advanced electronic device. 12 volts wasn’t too exciting on the motor…

… but the full 28 cells was. I was worried about rotor balance from the start, but the increased voltage makes it much worse. I had to hold the thing down to prevent it from vibrating across the floor.

Written off as increased appeal to the opposite gender.

Here are some mechanical bits. This is the left side “armor plate” and mounting surface for the electronics. This was my only slab of 2.5″ wide aluminum, so I’ll need to get more before work can continue on the right side. They will be secured to the scooter frame by standoffs and bolts.

The scooter does not fit in my drill press. Nor most drill presses, actually, so I did alot of the work by hand drill and careful eyeballing. It worked pretty well. The aluminum used in the frame is a soft and rather gummy one, and it threaded like absolute misery. Fortunately I don’t have to do too many holes on it. Wide shot is not for close-ups.

I’m a terrible photographer, but this isn’t an art site. The ESC, with mounting ears, will be bolted to the frame. The CCFL power supply will most likely be tape-welded, as it is flat and light. Conveniently enough, the motor wires and controller wires are just long enough to meet and be soldered together in the middle, and be restrained to a standoff.

One side plate installed. Standoffs are a bit too long for my taste, so I might get shorter ones. The other side plate will have a big circle thing cut out of it to account for the motor. They will be joined at the back by a UHMW skid plate, for slipperyness. Looks like wheelie-bar mounting isn’t too promising, so it might be left out. ESC and CCFL driver goes on the left side, controller-controller and power distribution on the right. Main switch and power meter by the handlebars, and the throttle will be designed and implemented as needed (Or I can just hold the little dial in my hand and fiddle with it while cruising…)

Home stretch! One more week! I really should be doing more productive things, but this is more fun.

Things that are done:

1. Powertrain mechanics
2. Battery pack
3. Power-converter-signal-generator-thing

Things that are not done:

1. Throttle assembly
2. Electronics mounts / side panels
3. Instrumentation
4. Wiring
5. Testing
6. Breaking
7. Crying
8. ..and Aaron, I didn’t forget this time: Cutting self and ranting on MySpace!

Pics from the past few days:

Hey, it’s my R/C signal faker. The 100A controller requires a 1-2 millisecond duration pulse every few dozen milliseconds. This is not exactly something a potentiometer can do by itself, so it requires additional circuitry. The schematic used to build the prototype was transferred over to a 2-year-old reused-thrice burnt-flux-covered piece of Radio Shack perfboard. It’s akin to using the same sheet of toilet tissue three times, once on each side and once by creative origami.

DC-DC converter! I realized that the little tiny 5 volt regulator chip would most likely detonate, and in the process detonate everything behind it, if I layed the full battery voltage through it. A bit of asking around got me these 9-36v in, 12v out converters that conveniently mount on a PC board.

Part 2 of the important electronics is this RC power analyzer. Devices like this are dropped in between a battery and its load, and read information such as volts, amps, amp-hours, watt-hours, watts, and other combinations of SI units. It will form the instrumentation of this contraption since I’ll be needing all my dead reckoning skills to prevent actually becoming dead.

Conveniently named the “Watt’s Up” meter, it is an even more convenient possibly-illegitimate cousin to the Astroflight “Whattmeter”. With a name like that, Astro will never notice.

DC-DC converter and associated equipment mounted on another piece of old perfboard. Now I’m digging napkins out of the trash can. The two are connected on the underside, and soldering this whole assembly was a precarious balancing act to make sure the tiny wires didn’t snap.

The large pic is labeled with the “pinout”, if it can really be called that. I stole a bunch of 0.1″ headers from circuit boards lying around, and regular R/C connectors slip over them in standard fashion. Only I know which way they go.

Which means this whole thing is about to detonate.

Here’s the “ghettoboarding” on the underside. This is what happens when you don’t have PC-board making equipment and freelance it using plucked strands of a 12 gauge wire. The DC-DC converter sticks out past its own board a bit, and so I glued the two boards together using it.

When I first switched everything on, there was no smoke, but there was only a shaky 7 volts coming out of the 12 volt converter. As a result, the 5 volt regulator only could push out 4 or so. I wasn’t sure why. A bit of reading on DC-DC converters found out that this thing needed capacitance on the output. A

Equate it to using a bucket to move water to a different level. The capacitor acts like a tank to smooth out the water flow on the other side. I plucked one from the nearest switching power supply board and the problems went away. Testing was done on 18 volts, but 36 shouldn’t be much different.

So I came to the conclusion that tying all my expensive electronics to the back wheel forks wasn’t going to hold up to any sort of engineering (or practical) scrutiny. Therefore, this pile of standoffs and cap screws will form the anchors for aluminum side plates, onto which everything will mount. I don’t have a concept sketch yet, but imagine big plates parallel to the wheel forks on either side of the wheel. This has the added bonus of being a starting base for a wheelie bar, and protecting the giant motor from hitting the ground, which *might* be bad.

Anyway, Hong Kong Express mail better hurry up with my controller. Who knows, the thing might take its maiden flight into the wall within the week…

Oh, and…

This must be done.

The next best thing after flangeless pulleys is a hacked-up belt guide/tensioner. Here, I used one of the stock motor cover screw holes to attach an aluminum lever with a shoulder-screw-bushing-washer-collar assembly on one side and another screw on the other side, onto which a deceptively strong tension spring is attached. This pulls directly down, with the other end of the spring attached to another screw. The whole thing keeps the belt from slamming into the frame as well as adds more tension (as if it needs more tension).

This was completely unanticipated and was a midnight engineering hack, which means it looks rather uncouth but functions just fine.

Another view of the tensioner assembly. As fate would have it, turning the motor in the forward direction causes the belt to jam against the frame, grinding the whole thing to a stop, whereas going backwards just bumps the belt against the wheel and lets it be a natural “flange”. Maybe I should just always move backwards!

No matter now, as the tensioner takes care of it.

The beginnings of my controller controller. I was not about to mount a 75mhz transmitter and receiver on this thing, so I have to come up with a way to fake the 1-2ms control pulse that R/C motor controllers take. A quick visit to Google and I found a workable schematic that involves a 555 timer chip, a bunch of random parts, and a 5K variable resistor, which I have a handful of.

When finished, the controller controller will probably have a DC-DC converter attached to it to drive accessories. This is contingent on everything else working.