Archive for September, 2010


Random Site Updates, Once Again

Sep 29, 2010 in Stuff

Every once in a while, I remember that other parts of my site exist besides the build posts. I’m going around and cleaning up the post categories and making sub-pages for more projects. I’ve also reorganized the project categories themselves, to better reflect what is in them.

You’ll notice the “Not Robots” section is gone. I’ve split off the vehicles into their own index page since I’ve built so many things that aren’t robots.

Pretty soon, I should have a more updated About page too. After all, if I keep directing people to it, I should actually consider putting something there….

Update: Hey, guess what’s live now? The About page.

RazEr rEVolution: Dual Non-Interleaved RazErMotor

Sep 28, 2010 in Project Build Reports, RazEr rEVolution

Spoiler time!

No, you can’t ride it yet. Yes, I know it has ground clearance issues. Those will be resolved.

And yes, I know my laptop screen is cracked. That’s like a whole new set of motor controllers for Clocker right there.

After finishing most of the framework, I turned my attention back to the motor. I finished the can and magnet ring before forgetting about the project’s existence for a while. Left to do at the time were both endcaps and the centerpiece of the whole assembly (literally), the stator mount & shaft.

Oh, and the random auxiliary features on the can.

I’m going to hand-wave the endcap construction a bit, since I just randomly came into MITERS one day and did it. Its construction is identical, dimensions notwithstanding, to the RazErBlades motors.

The cross holes for both piece were made on my haute usinage indexing fixture. It turns out that all my designs which use features in multiples of 3 are actually less time-efficient to make since I can’t use integer turns of the handle. But if it divides evenly into 40, then I just have to count full revolutions. Yet another serendipitous laziness-increasing discovery!

After finishing the last machined features, I shoved the magnet ring into the can and pressed in the other endcap (not visible above)

I normally don’t do this much hand-waving, but again, this center shaft was done while I randomly showed up to MITERS on a whim without my camera. The stock material is 1.25″ aluminum turned down to 28mm to press in the stator, 25mm to seat the bearings, and 0.75″ for the spacer nub. The center is tapped 5/16″-18.

Right now, those 3 stators are just lined up on the shaft. For space efficiency, I need to remove the epoxy insulation coating from one side of the outside stators and both sides of the center one.

Fortunately, that’s why these things are laminated. A flick with a knife (and some plier work) later, the stators are ready for installation.

(Okay, so the real reason they’re laminated is for reduction of eddy currents, but I like to think they work like Pringles potato chips or those funky layered biscuit things!)

As usual, these stators were yanked from copier motors. Specifically, the Xerox Phaser 4400, with Xerox P/N 127K35701.

This isn’t a picture of the shaft being pressed through the stators. Nope – I’m actually just gluing them together at the moment. I spread a layer of thin CA glue between each, then threw the whole thing under a 20 ton hydraulic press (not using the whole 20 tons, I mean) and let it sit for a few minutes to cure.

For your lack of viewing pleasure, here’s a few layers of the skin on my thumb. It turns out that spamming something with CA means the excess CA will just sit, uncured, waiting for something to come in contact with it.

Now that the stator is ready, I put the wire access features into the shaft with the indexer.

I’m trying something a bit different on this design. Normally, I just shove the stator onto the shaft with Loctite and call it good. This time, I actually machined a 1/8″ round groove into the aluminum shaft. A 1/8″ roll pin locks the stator (using its own round half groove) with the shaft. This retainment method has the benefit of being reversible. If I ever wanted to take the motor fully apart, change out stators, or something else, I can actually do it without destroying everything!

The fully assembled core.

This thing is a 48mm long and 68mm in diameter. For reference, that’s a whole Turnigy Short Melon. I might as well have just yanked the stator from that thing.

Hey, if you want to build your own DIR, that’s a good place to start.

A test fit confirms that no, I didn’t machine everything wrong. This thing cogs like a beast, even with the 12/14 magnet layout.

Like on the skates, the wheel is retained using a little locking ring. So I had to make another one for this motor.

Problem: My right-hand internal threading tool was lost in a terrific tip crash a few weeks ago.

Solution: Use the left hand tool while the machine is running backwards and feed from the inside out.

It took a few tries for me to get the multiple chirality changes right, but…

One 3.5″-24 threaded wheel locking ring!

Now that’s a nonstandard thread if I ever saw one…

Test fitting the wheel. Did I ever mention how ridiculously wide this thing is?

On a whim, I tossed the ring back on the mill and machined in these two little spanner wrench grooves. One of the things which I had neglected to do (out of ambivalence and laziness, probably) for both RazErMotor V3 and all the skate motors was actually make tightening tools (or fixtures) for the wheels.

One day, I’ll actually make a little spanner wrench or something so I can actually use these grooves.

Now doesn’t that look epic?

The eventual wires will exit the motor and lie within the little indentation under the (temporary) shaft bolt.

structural editing

I went back and countersunk the structural top plate’s holes so I could use my flathead 4-40 screws.

But much more importantly, I added these additional 1″ long 10-24 screws to the folding joint attachment area. They are actually legitimately threaded into the aluminum below. Imagine that – me, legitimately drilling and tapping a hole.

After reviewing the results of my 5AM engineering on the frame, I realized that while the M6 x 1 machine screws will definitely support my weight and survive the impulses of things such as, say, landing a curbhop, the tiny 4-40 T-nut slots will not. Not even come close. It amounts to trying to jump start cars with 22 gauge breadboarding wire.

Just to check if I had any manual fab skills left at all, I freehanded the hole on the drill press like I used to. I’m proud to say that I can still use a drill press.

begin menial labor

I began winding the motor Friday evening (before leaving for Maker Faire early Saturday).  Now, I had said before that I would be winding the motor fully interleaved, that is:

  • A-b-C-a-B-c
  • -a-B-c-A-b-C

But that causes a mess of wire terminations interleaved with each other too. I’m electing to do it instead as

  • AabBCc——
  • ——aABbcC

That’s right – just splitting the dLRK down the middle. So really what this is is the Dual Non-Interleaved RazErMotor, but DNIR just doesn’t have the same ring to it.

Above is the first finished motor-half, with 40 (+/- like 5?) turns per tooth. I have yet to terminate or wind the second motor-half, so I should probably get on that soon…

RazEr rEVolution: Hey, I forgot I had this…

Sep 28, 2010 in Project Build Reports, RazEr rEVolution

The last rEVolution post was on August 9th. Some might recognize this as about when I entered the panic phase of robot construction for Dragon*Con (that event which I still haven’t written a summary of). Since then, the parts of rEVolution have been taking up space on my electric vehicle shelf.

A few weeks ago, I decided to trip down to Maker Faire NYC to see what this whole “making stuff” scene is all about, since I don’t know how to make things. Originally, I was just going to go… but the temptation was too great to not try and bring something. Classwork and other preoccupations meant that I really only had two days last week to blitz rEVolution together.

Long story short, I didn’t get very far, and just went. Crap, yet another event I need to recap.

Long story long, here’s the results of that blitzing!

Step 1: Throw everything at the waterjet.

All the major structural components are made of 1/4″ aluminum that I picked up off eBay (per normal procedure of buying the first thing that said “aluminum plate”). The black top plate is the nonstructural deck cover and was cut from left over 1/16″ black Garolite, the same sheet I used for Cold Arbor’s top and bottom panels. The one immediately under that is the structural top plate, made of 1/8″ aluminum.

The new frame next to the current RazEr. Yes, it’s big.

I went ahead and compensated for the thickness of regular oversize aluminum plate when designing the slotwork here. It turns out that aluminum plate, unless specified tight-tolerance ($$$$), is usually 0.01 to 0.02 inches thicker than the nominal thickness. I generally design on-size, which has resulted in more sanding than should be necessary in the past.

All the slots on this frame are .005″ larger in every dimension and the tabs are .01″ shorter. Most of the frame was therefore pushed together pretty easily.

With the new motor can in place. Isn’t that just ridiculous looking?

A side shot, showing the text cutouts and the deck installed.

wheel? more like squee-l.

The next task was to hollow out the massive aluminum spare semi truck wheel that is to become rEVolution’s rear tire. For this task, I elected to use the heavy \m/etal machine of the auto shop, since it could more confidently grip the inside diameter of the rim (which was the exact right diameter to be unchuckable with our small 6″ chuck in either configuration).

The process itself was surprisingly painless. I bored the wheel all the way through the center hub such that it just fell out, then turned it around and cleaned up the leftovers.

And the wheel fits around the motor can like so.

Let’s go back to the frame. This piece is the anchor point for the steering neck and folding joint. This always seems to be the trickiest part of scooter building unless you have a premade solution already, because it’s one of the highest stressed parts and also generally sprouts from the base at an angle.

I’m going to retain the stock Razor A3 folding joint, so the bolt pattern fits the base and the thread matches the stock screws (M6 x 1)


Nothing I build is complete without at least one t-nut, and this build is absolutely full of them. These are 4-40 square nuts, which seem to actually be quite rare in small packs. McMaster didn’t carry them at all, and I could only find packs of five thousand.

That’s okay, it’s only $50 for the pack. I decided to spring for some because I would never, ever need to buy them ever again.

I’ve built up too many pictures to reasonably write one long post, so I’m going to split this here. In the next episode of RazEr rEVolution, the inner workings of the Dual Interleaved RazErMotor.

Melon Scooter: Yeah, So It’s Done.

Sep 21, 2010 in Melon-scooter, Project Build Reports

Well, everyone probably saw this one coming: the post where I make up for the lack of mechanical issues and dearth of unconventional machining problems with seemingly neverending tales of electronics woes. The bad news is that I did not find out how to control the Melon with just a relay and a diode. I’m sure it can be done, but further investigation is needed in this study’s focal area.

The good news is that it finally works. I mean, it works as well as all my scooters so far do, which is to say it’s probably on the verge of exploding.

Now for the things that didn’t work.

hexbridge shield

As described in its own post, the Hexbridge shield is an attempt at creating a system for Arduinos to control more power than they deserve to. With twelve total IRFB3207 FETs straight out of International Rectifier’s private corporate Hell, the board could process more than a kilowatt of power.

The six half bridges are nominally divided into two independent 3 phase bridges by the 2 PWM pins of the Arduino it occupies. Now, while one rail of IRFB3207s alone can probably handle the Melon motor with proper current control, the board itself has no current sensing provisions. I elected to activate the other 3 phase bridge and put it in parallel with the first for that extra burst current overhead, since I was aiming to just make a simple open loop BLDC commutator. The way to accomplish this is to disable one of the PWM pins completely (high-Z) and then tie the selector pins together in either software or hardware.

This mode is the “secret mode” of the Hexbridge, and comes with risks such as EPIC CROSS-PHASE SHOOT THROUGH if the paralleled phases are not in absolute sync with each other, such as if the selector pins aren’t ganged together properly.

What was extra cool about the Hexbridge shield was the fact that it fit exactly between the two \m/etalpaxxx. Space efficiency!

I yoinked a few feet of silicone wire from the Media Lab and hooked up the main power wiring for the system. As usual, with things I wire, it’s a mix of gauges, strands in parallel, and absurdism. Yes, that’s double 12 gauge coming out of the shield…. fed, of course by a single Deans connector at the other end.

Hey, I figure too much copper is not a bad thing from a system resistance perspective. In fact, the shady master key switch probably contributes more resistance to the system than anything else.

Because the motor does not come with sensors, I had to craft a sensor board out of some thin tube plastic. This is a departure from how I usually do sensors – which is stuffed inside the motor’s stator slots – because I didn’t particularly feel like taking apart a motor that big.

As such, I could take advantage of the fact that the sensors can be located physically close together to detect the proper magnet transitions.  The formula for sensor spacing in degrees is 360°/P/3¹, where P is the number of pole pairs. So for the Melon, it was just 360 / 7 (since it’s a LRK wind) / 3, or about 17 degrees. Across a roughly linear projection, the sensors had to be about 0.48″ apart. This board was superglued to the exterior of the stationary motor mounting base.

The cheap hall sensors I used (p/n ATS177) exhibited some very weird behavior when I read the 3 lines out with the scope. One of them seemed to be “preloaded” and did not exhibit the 50/50 duty cycle I would have expected from a Hall sensor subject to a constant frequency alternating magnetic field composed of equally sized magnets. It meant that one motor state would overlap slightly with another, which at best could cause the motor to just not start (and at worst cause current spikes).

Here’s the basic wiring hooked up to do a commutation test. I loaded the same Hexbridge-and-Etek code on this assembly, making sure to lock out one PWM and hardwire the two PWM pins together before powering the system on and trying out the throttle. After digging through a handful of sensor:phase combinations, the motor finally started up. It was a little rocky sounding, but it did work.

Now, those of you who pay attention to words I put in bold and the subtle details of my sentence grammar will probably guess what happened.

After 7 seconds of operation, there was a large spark, and the power side of the Hexbridge was history.

The most plausible explanation, as it turned out, was that I neglected to tie the ABC selector pins to the UVW ones – in either hardware or software. I made them all high impedance instead, in software. So not only does that mean one half of the whole thing was never actually switching, the floating FETs themselves might have been subject to dv/dt-induced spurious turn-on.



I was determined to get this thing running. Maybe too determined. After the Hexbridge exploded, I jumped to my backup plan of using a KBS series Kelly controller, of which we have aplenty in the Media Lab. At this point, I didn’t even care about monster horsepower. I just wanted it to move, because I wanted some transportation (that didn’t involve me actually doing physical work, I mean. Heaven forbid I use my legs to propel myself…)

Kelly controllers have a penchant for randomly exploding, but I must say this one did an admirable job given the adverse circumstances it had to tolerate. It’s a 20 amp rated controller with peak 50 amp ratings and here I am asking it to run a motor which will easily knock down 150 to 200 amps on a bad day.

It did an excellent job of keeping its own current so low that the scooter was able to move maybe 3 or 4 miles per hour, all the while sounding like a moped. The rockiness of the sensors really exhibited itself here.

In the end, nothing happened. I took the Kelly out of the system while it was still undamaged and useful to our Media Lab-y purposes.

oh, damn it all

Fuck it. I’m going back to sensorless.

After a while of stewing, I bought a spare Turnigy 100A controller off Shane (thus funding some of the Pneu Scooter, which I look forward to) and threw it on there with the servo signal interface board.

The new layout with the Turnigy ESC hanging awkwardly off to the side. Instead of stuffing it into the cavity, I elected to drill some mounting holes in the deck and put it in a path where it could get some real airflow, since heat is the killer of power-dense controllers like those.

The R/C meter tucks in between the batteries, and my secret ninja servo converter hides under that, waiting to strike.

And guess what? It works. It’s a power system I originally assembled in 2007, while still in high school. It’s nothing more sophisticated and would blow up the moment I floored the thing from a standstill (so I’m always careful on the throttle until the vehicle is already moving a few MPH). Hell, it even uses the exact same cheap R/C power meter (model-wise… not the physical same, mind you).

I said this thing was just a quick dumb-but-awesome build. It’s Fankart with powered wheels and the ability to navigate city streets. Maybe there will be some running video, but Melonscooter has become my staple commuter vehicle now.


But I’m not stopping there.

Putting together such a system isn’t particularly contributory towards making myself a better engineer (or person). I haven’t laid out the board yet, but I am absolutely determined to tame the Melon. And I intend to do this by making a current-controlled BLDC motor driver board that

  • will operate with a 48 volt electrical system
  • at up to 50 real amps continuously and 150 to 200 amps peak (limited)
  • have current sensing on the DC bus and two motor phases
  • be fully regenerative
  • be two stacked boards about the footprint of a common business card
  • be modular, with one signal processing board and one brute force driving/current feedback board
  • and possibly do crazy insane batshit things with motor control.

Oh, it’ll run on Arduino.

It’s a long shot, and will probably cause melonscooter to be the LOLrioKart of 2010-2011 (not that LOLrioKart isn’t already the LOLrioKart of 2010). But it will be the only component of Melonscooter that makes it a worthwhile project, so I’m going to go for it.


Melon Scooter: Yet Another Dumb Electric Scooter Build

Sep 14, 2010 in Melon-scooter, Project Build Reports

I know that I have an entire robot event to report back on. In fact, the robots have even had time to arrive safely back to MIT.

I know that I already have an electric scooter build that is in process and which features a design paradigm not found on most scooters extant today. You know, one actually worth building.

… but some times I just want to build something dumb but awesome. After all, it’s not like that’s never happened before to me. Most people in academia would probably argue that’s all I do anyway.

So let me tell the tale of the Melon Scooter, which has been occupying my time for the past few days.