Archive for October, 2011


Revisiting RazEr: Finally getting rid of those concrete wheels

Oct 14, 2011 in Project Build Reports, RazEr rEVolution

Let’s return to a scooter which is not made of wood.

For the longest time, the only thing that’s prevented RazEr rEVolution from being an actual useful vehicle has been the rock-hard wheels I selected for some reason last year. Both were surplus finds and very cheap, and the rear wheel especially was selected at the time for its low tread profile. Being industrial casters, though, they were hard, a fact that I didn’t really think about being a possibility before I got them. The front wheel is 95A Durometer (pretty damn hard) and the rear wheel was 90D Durometer (might as well be plastic).

So basically every time I rode RazEr around, there was a loud rumbling sound on top of the earsplitting 4kHz PWM sound, and afterwards I could not feel my ankles. RazEr rides were limited to indoor, smooth environments. This isn’t to say I didn’t try riding it back and forth across campus a few times, but the numbness in my joints afterwards threatened early arthritis and I was constantly afraid of shaking apart the components inside the motor and destroying the melontroller. Ultimately, though, the vehicle prevailed over my willingness to ride it…until I accidentally detonated the controller in Singapore. With RazEr out of service anyway, I got a little more motivation to replace the wheels. The other portion of the motivation came from discovering how well Straight RazEr handled with the 5″ 65A durometer Colson Caster wheels (a staple of combat robotics in the past). Straight Razer rode very well as just a regular scooter, but it just had too much horsepower to safely handle day to day.

I never thought you could machine a Colson wheel all the way out to the tire, however. My impression of them was that the plastic ended very close to where the rubber began. Therefore, coring them out would not end well as the ring lost integrity while being machined. But while visiting the GT Invention Studio, I watched Colson wheels being machined for A Certain Other Hub Motor Scooter (no, not that one…. the other one). It was indeed possible -the Colsons had a plastic ribbing that the tire was molded over, so even the clamping pressure of a lathe chuck wasn’t enough to deform them for light and moderate cutting passes.

Well, could have had me fooled. It was time to change tires.

There was a problem, though. The 5″ Colson wheel can only be bored out to about 3.3 inches, and the Dual Non-Interleaved Razermotor had that big aluminum outer case which measured at 3.5 inches. This would have required alot of machining of rubber, which was suboptimal.

However, the original Razermotor V1 (dear god that’s an old post) had an outer can diameter of 3.25 inches, the maximum diameter a 125mm scooter wheel can be bored to. Its successor, Razermotor V2, had the same dimensions. Conveniently enough, a 5″ Colson can be bored to 3.25 inches with no issue. So the solution was to make a longer can version of Razermotor V2 to fit the existing stator of the DNIR.

This is a 5″ black Colson wheel. It is in fact the 100th build picture of RazEr rEVolution, continuing my streak of photographing utterly boring subjects for the centenary build picture. To bore this wheel out, I needed a lathe big enough to grip the entire outside of the tire.

Luckily, one existed in the form of a recent addition to the vehicle teams shop: a late model 19″ LeBlond Regal . This thing is the size of a car, and before I hopped on it to start messing around with the controls, I have in fact not seen it ever run. But run it does, and it’s probably now my favorite heavy metal mayhem machine on campus (no less because I have access to it 24/7 and it has all the nice features… minus a DRO). The installed chuck is a 10 inch type which opens wide enough to shove the whole Colson in.

Even more luckily, the only accessory it came with was a boring bar. There were other tool blocks to fit the toolpost, but I would have to “borrow” the tooling of the other machines to use them.

A few minutes later, a centerless Colson. The cores of these things are made of polypropylene, so they’re exceptionally easy to machine.

I ordered my usual Giant Steel Pipe from Speedy Metals to build the new rotor.

This is where the horsepower and mass of this thing really shined. By this time, I had already assembled some basic tooling, and was having fun making 3/8″ inch wide bright blue steel curls that were several feet long. Technically that’s a bad thing to come off a lathe during operation, but it’s a welcome change from the slightly anemic Old Mercedes (I haven’t thought of a snappy nickname for this one yet, but nothing short of Leviathan or Behemoth or Kraken or something would suffice).

I also spent a little while reading about how the Servo Shift worked. This was LeBlond’s “thing”, and it was pretty much an automatic transmission for a lathe driven by hydraulic cylinders and controlled by electrically actuated valves with switches as feedback (Turning the big dial selects which set of switches the gears have to move in place to trigger). The spindle literally jiggles back and forth while the shift is occurring (in a manual transmission lathe, you jiggle the spindle yourself to mesh the gears together). This machine completes the cycle in a second or two, which apparently means the Servo Shift is working exceptionally well – it sounds like an utter pain in the ass if it ever breaks.

I love overly complex mechanical solutions to a simple problem. The rest of the variable-speed machines here use belt-based CVTs, which are far less parts but less heavy metal.

Okay, enough diddling. Time to make a motor can.

The protoform can emerges. This is actually the second attempt. On try number 1, I was still getting used to where the threading controls were and how they functioned specifically to the machine, so after almost finishing the part I accidentally started the carriage incorrectly and instantly plowed off all of my threads.


The thread is a 3.25″-24, which is not in any standard catalog of threaded objects. It just means I have to make my own matching nut again.

I borrowed a Nice Carbide Threading Tool from the Edgerton Center Student Shop, and its niceness shows. These were cleaned up a bit with a pass of Scotch Brite.

Unfortunately, hugelathe didn’t have a parting tool yet, and other tools couldn’t be used on it again because of toolpost incompatibility. I therefore had to transfer the part over to another machine to perform the cutoff. In this case, it’s the second largest machine in the shop, a Clausing 4900 type.

After the cutoff was complete, I transferred it back in order to finish the stepped bore for the endcaps. Tossing a part between machines like this is risky for concentricity unless the workholding implements are exceptionally well-made and precise. The old American machinery did not disappoint.

The finished product.

The next day, I returned to remake the endcaps. These now mount directly inside the steel rotor, so they’re smaller in diameter.

I finally took apart the DNIR to recover the stator and center shaft. I’m glad to see that even after being stuffed into a rock tumbler posing as a wheel, the stator, windings, and sensors are all intact. However, the winding potting epoxy, which I used a cellulose filler (read: paper fluff) for, seems to have charred from the heat. That was quite interesting to see – at first I thought the windings were burnt, but it’s just the paper fluff not handling high temperatures very well.

With the “protoform” can and endcaps done, it was time to drill the radial mounting holes using my trusty indexing fixture.

I still have to glue the magnets inside the new can, but after that, I should be able to just put everything back together and have it work. This takes care of the back wheel of RazEr – the front wheel is less of a priority, but given that I have another Colson, I will probably go ahead and finish that too.

Oh, yeah, and fix the controller. Hmm, this might take longer than anticipated…


Oct 09, 2011 in Project Build Reports, Stuff


The Inaugural Beyond Unboxing: Hobbyking 1-8-Type 80A Car Controller

Oct 07, 2011 in Beyond Unboxing, Reference Posts

I’m going to formally open another category since I’ve noticed that I do something like this quite often. The Beyond Unboxing category is strictly reserved for those posts where I strip something, usually a cheesy robot or vehicle part, apart down to the component level and examine it in detail to find ways to abuse it, or just have a good laugh. I’ll be back-marking some pots as “Beyond Unboxing” which contain details, but this will be the first in-depth examination of anything.

The first topic at hand is yet another option for controlling your small electric vehicle drivetrain. Motor controllers are almost always the hardest component to understand in a vehicle for new builders. The hardware itself may be the difficult part to make (okay, unless you’re truly crazy and making your own motor controllers), but a motor controller is usually that black box component you buy or swindle out of someone, plug in, and don’t really care about how it works in detail.

There are plenty of options available for the small vehicle builder now [citation needed], the easiest to obtain of which is the Kelly KBS line. The only problem? They’re enormous. Compared to a R/C type controller, the power density of the average EV controller is much smaller. This is for good reason, since EVs aren’t guaranteed a stiff breeze blowing past them all the time, so the controller has to deal with more heat buildup. But for ultra-compact vehicles like the RazErs (both domestic and internationalinstitutional), the KBS isn’t an option: It’s just too huge.  The thing is thicker than my entire frame on the original RazEr. Past this, we have to turn to R/C airplane controllers, which are nice and flat, but hard to deal with when your load isn’t a propeller.

That’s why I think the HK 1/8 Scale series of car controllers (the ones in the nice case with no exposed metal – there are several different lines) have alot of promise. Actually, not just promise. They have been proven already on two vehicles: Amy’s doohickey and the Razor Wind. The best thing about them besides being small is cheap. Hobbyking wins the best amps-per-money award again:  The 150A is $60-65, but the 120A and 80A are only $35-40. Both of those vehicle designs used the 150A variant (HK150A) which is advertised as being 2 to 6S lithium battery (7.4 to 22.2v nominal) capable. The other ‘sizes’ are 80A and 120A, but are advertised as only being 4S capable. That’s not useful in a vehicle really, at least one of decent performance.

Now, I tend to believe in the Law of Chinese Packaging Inertia. If the cheap chinese thing outwardly looks the same, it’s probably the same. The 80A (HK-1-8-80A) and 120A look really similar to the 150, but they’re smaller in their pictures. Were they also smaller in person? The 150A has 3 input capacitors but the rest have 4. If they really were the same, could the 80 and 120A ones in fact handle 6S too?

Curiosity drove me to buy one and chop it up.

This is the box. As far as I can tell, it is static from model to model, since I also have seen the boxes for the 150A.

I’ll skip the dramatic unboxing process and skip right to the beef. This is the thing – it’s about 1.75″ square and has wires coming out of it. This is a good sign, but not if you’re the bomb squad.

R/C modellers complain about the size and weight of this thing – as someone used to seeing house bricks and small appliances mounted on vehicles, I contend they have no idea what they’re complaining about. Regardless, suspicion #1 has been confirmed: It is the same size as the others. I guess HK just cropped the picture closer for the 150 or something.

Starting the pornographic stripdown. The top shell comes off with 4 screws, which also anchor the fan.

And the whole assembly pops out of the bottom case. It’s two boards, one power board and one signal and gate drive board….just like the 150. Hmm, suspicion 2 is on the verge of confirmation. Let’s look at the power board in closer detail.

So this is where the 80 and 150A clearly differ. On the 150 amp controller, the power board is composed of no less than 36 Super-SO8 package FETs. This board uses 18 IRLR8726 type FETs, 3 per leg of the 3 phase bridge. They’re about 6 milliohms at 10 volts gate drive, so the combined parallel on-resistance of each leg is, exactly as spec’d, 2 milliohms. Suspcion #2 is therefore partially debunked: They’re not really the same thing inside. What I’ve seen before is that a 100 amp controller with 5 FETs has one FET taken out and is sold as a 80 amp controller. The same kind of ‘binning’ occurs in CPU and other silicon manufacturing, except the cheap hobby version is on purpose and for marketing purposes only. However, these ESCs therefore do use the same architecture but with different power boards, at least.

The FETs are 30 volt rated parts. The bus capacitors are 35v parts. This is clearly not what’s limiting the controller from handling above 15 volts, so a closer analysis of the signal board is required.

The voltage limiting factor for cheap controllers with low cell counts is the gate drive power supply and the logic power supply. Often, controllers with 2-4S (or up to 12 nicad cells) rating derive their gate drive power directly from the battery. Because modern power semiconductors usually have Vgs (gate voltage) ratings of 20 volts, this is just fine. However, it means they literally cannot be run above 15v nominal since a fully charged 14.4v battery is usually closer to 16 or 17 volts. Any higher, such as during braking or slowdown, could destroy the semiconductors through the gate pin. Usually some kind of hard limiter like a zener diode is supplied to avoid this.

The issue is that if this controller does indeed derive its gate drive voltage from the battery input, then it’s practically useless for vehicles. Sure, you could spend the time and try and jump 15v to the gate drive supply and cut it off from the rest, but in my opinion there’s a point of diminishing returns for hacking and chopping a commercial board – especially an inexpensive one.

The logic power supply is also important. Typically these 2-4S controllers have a single 7805 type linear regulator feeding the logic and possibly the 5v receiver circuit too. The chip will not be able to handle voltages much above 20 volts just due to heating. When a single-7805 logic power supply is found in a 6S (22-24v) rated controller, it is typically only driving the logic and the controller has no BEC. This keeps the current and therefore heating acceptable. The HK “200 amp” 6S airplane controller is a good example of this, and is worth its own writeup because it’s so ridiculous.

After some staring under a microscope, it was determined that the BEC had its own switching regulator: a MP2565 switching regulator controller. The giveaway to this was the big inductor core at the bottom right of the signal board. The logic was definitely not derived from this, which was unfortunate. That regulator can operate up to 50 volts, but the 6V BEC precluded the logic from being run of a normal 5 volt regulator (which needs 1.5 to 2 volts of overhead to start working). Of course, there’s the chance that the logic is 3.3 volts, in which case it would work.

Since the BEC power supply was eliminated as the source of gate drive and logic voltage, I wanted to look at the gate drive itself (which, by the way, is handled by 3 IRS2003 half-bridge drivers similar to the IR2184s I am fond of)

I scoped the gate of one FET while the controller was running in order to check the voltage. If it changed as I changed the power supply, there’s no question: I’m hosed. The test began at 12 volts and I swung it between 10 and 14.

However, it definitely did not change. The gate drive voltage held at about 9 volts the whole time. That means there was clearly another power supply somewhere I missed, and this power supply is most likely where that 7805 on the board (top center  DPAK component) is drawing power from.

Closer examination of the board revealed a small 6-pin IC, a tiny inductor, and an unknown 3 pin regulator. Clever guesswork with Shane revealed that the tiny SOT-23-6 part to be a LM2681, a switched-capacitor voltage converter.

Mystery solved.  One possibility is that this voltage converter runs in doubling mode for low voltages and is bypassed as the voltage increases past 9 or 10V. I’m not sure where the transition occurs – the datasheet lists a maximum of 11.6 volts only, so there might be something else at play. At that voltage, the input just spills over the diode, incurring one Vforward loss in the diode of about 0.6 to 0.4 volts, and regulator loss (1.5 to 2 volts) before ending up as the 9v power supply. The supply held constant as the power supply voltage was decreased past 7 volts, but the logic shut the controller down (low voltage cutoff for battery protection) after that.

The mystery 3 pin regulator is then either a linear regulator or a BJT rigged as a linear regulator. The 9 volt output is also piped to the 7805 for easy downconversion for the microcontrollers.

So there really seems to be nothing stopping this thing from running on 6S. All of the important components appear to handle it – even if the voltage converter failed, the 7805 and gate drive regulator can handle 24 volts but at increased heat stress.

Therefore I only did what came natural: turn up the voltage until something becomes unhappy.

It was fine.

In fact, it was so fine with 6S (22.2v) that it even beeped 6 times. The firmware for 6 lithium cells is on the chip.

As I increased the voltage past 26 volts or so, it became unhappy. But not smoke-pouring unhappy – in fact there is a soft cutoff. The controller simply stops running the motor and start blinking furiously. The “undervoltage” LED morse code is also a “overvoltage” sign. This also occurred when the Turnigy SK3 test motor was braked quickly, causing a power surge back into the supply (A battery would have low enough impedance for this to not be a problem). Momentary surges exceeded 30 volts without smoke, but I bet it won’t live long near that mark.

Because the controller was able to scream for mercy (blink for mercy) before ultimate destruction, I spared it any higher power supply voltage tests.

So it seems like this is a pretty solid controller and can definitely handle 6S. I’ll bet a Singaporean dime (I have those) that Hobbyking “releases” a 6S version soon and charges $60 for it instead of $35.

The controller logic is the same as the 150A one (considering the signal board is most likely identical). The basic operation is:

  • Apply throttle, motor goes.
  • Release throttle, motor coasts down to lower speed.
  • Releast throttle to neutral, motor coasts to stop with some active braking (the “drag brake”)
  • Apply negative throttle (stick back, trigger forward, knob left, whatever), motor starts reversing after at least 2 seconds of neutral have passed and the speed has reached zero. While braking, motor makes a cute chirp of a distinctly lower frequency than when driving.
  • Apply positive throttle and motor instantly brakes to a stop and turns in original direction.

So the reverse isn’t symmetric (bad for robots) but uses the same input and requires vehicle stopping beforehand. This is good for vehicles, where you usually don’t shift into reverse while moving at highway speeds. I’m sure you COULD in theory, it just won’t end happily.

The conclusion is that the HK 1/8″ scale ESCs are nice things. If you have a vehicle electrical system that doesn’t mind running on 24 volts maximum, it’s the smallest AND cheapest option available. Things notably lacking: current control and limiting, cool side functions like contactor drivers and light controllers. You know what, I don’t really care, it’s 6 cubic inches.

I wonder what I could possibly use this on.



Return to Singapore: The SUTD Open House

Oct 04, 2011 in Events, MIT & Boston

Hello everyone, I spent the last week in Singapore on a visit to MIT.

Wait, what?

It’s kind of like MIT.


It even has MIT colors.

That’s right, the SUTD is now open, and the crack team of MIT faculty and students involved in the International Design Cent(re?) collaborative effort was shipped over (students in economy class, mind you) for the inaugural Open House event to demonstrate and exhibit the latest research efforts be total scooter-mounted assclowns.

The temporary campus is located right next to Singapore Polytechnic and a little north of the National University of Singapore, because piling universities next to eachother is awesome. The campus formerly belonged to a technical school. Singapore being a tropical country where the weather is always 90 (okay, 32 Celsius) and 110% humid, the buildings are very airy and feature such things like tropical gardens. Why can’t we have tropical gardens? I’m starting one in Lobby 10, right now.

During the week before, billboards and bus stops in the area started featuring SUTD posters advertising the event. This one shown is of the Leveraged Freedom wheelchair designed through the MIT Mobility Lab, led by Amos Winter.

Oh, yeah, not only were posters on the bus stops, they were even on the buses. Full livery for several buses featured Amos and the giant ratchet wrench wheelchair. Why don’t we get entire T buses painted in scooter and ballcopter colors?!

How to know you’re going to get a classy education: SUTDBot.

How to know your university is going to be super legitimate and renowned: your orientation guides are Jack Sparrow, Cleopatra, Jasmine, and Pocahantas.

The MIT Scooter Demo Fleet made another appearance: RazEr, along with Kitmotor and random odds and ends. Besides Pneu Scooter, Shane also had his tinyquadrotor along, and the gymnasium space was perfect for getting some flights in.

The small group of current SUTD students also had an EV faction, and they took a formerly street legal 200-watt lead-acid electric bicycle and turned it into a not-very-legal lithium-powered chopper. And painted it glossy black. The electric bike was originally from MKP Bikes.

MKP came to the rescue late on Saturday as I decided to change the PWM frequency on Melontroller 2 to 8 kHz instead of the earsplitting 4kHz. The latter was done a long time ago using code copied and pasted from the Arduino forum, before I R’dTFM the ATMEGA328 manual and found out which registers to flip in order to get more PWM frequencies. In order to do this, I borrowed a USB-to-Serial converter from another student group display.

Unfortunately, it was most definitely not a FTDI cable. The pin difference meant I was shorting my logic power supply the whole time, and for one reason or another, I damaged the gate drive stage. What’s weird is that the Arduino survived the affair.

The solution was to buy a cheap and shady electric bike controller from MKP, which was a short bus ride away. Then plop it right on the back of the deck and keep rolling for the rest of the weekend like nothing ever happened. You know, just like last year.

With Melontroller 2 a proven design that works great when I don’t kill it out of stupidity, I’m going to 1. be making another one very shortly to replace this one and 2. Redesigning the board to take care of some of the hardware bugs and Little Blue Wire hacks.

I totally took this picture. This is so awesome I might swap it as the title picture for RazEr’s static page. It’s otherwise a great “press shot” for posters and whatnot.

meanwhile, in singapore

Because we had a few days of diddle time before the Open House proper, we visited several area businesses to create institutional and educational relationships dick around. Now, several of the former did come about as a result of visiting, so it wasn’t totally for our amusement.

Firs stop: EV Hub / FSG Mobility Concepts, a local EV conversion house and alternative vehicle dealership that was featured on the Discovery Channel. We chatted vehicle design and marketing (the market structure being very different here in Singapore for automobiles) and also got to ride some super slick personal EVs.

I take it back. It’s not slick, it looks like a malformed reverse dong. Made of carbon fiber.

One of the vehicles that FSG/EVHub sells is the YikeBike, a… you know what, I don’t even know. It’s an industrial designer student’s term project come to life. Great for the industrial designer sure, but it rides and handles like ass (or a malformed reverse dong thereunder). The drivetrain is an electric brushless system, but it’s not a ring motor like I expected, but an internal rack and pinion drive. It makes a horrible racket and is impossible to freewheel. The handlebars are backwards and in an unconventional place, and you’re up so high that you’re constantly shifting weight on them. It definitely takes a few minutes of practice to even begin riding one, and probably many hours to feel comfortable on it.

Why deal with all that when you could have tiny electric scooters?

In this picture: Better than the YikeBike

We also visited the holy Mecca of ekranoplans, WigetWorks. I didn’t get around to it the last time I was in Singapore, but….


But it ultimate amounted to nothing. We were denied entrance due to intellectual property concerns. I guess they don’t like random surprise fanboys? Maybe if I brought Chuckranoplan, it would have been better…

Either way, what on earth are you guys protecting anyway? It’s not like your design wasn’t purchased from someone else 10 years ago and has remained unchanged since then… oh wait.