Archive for March, 2009


I accidentally

Mar 24, 2009 in Project Build Reports, Project RazEr


I am now thoroughly convinced beyond any doubt that power electronics just plain hate me and want to see me suffer.

Last week, the 5v switching regulator on RazEr went out with no explanation. Fortunately, it just stopped switching, and didn’t jam 35 volts through all my control logic like last time.

Today, still depressed from the failure of LOLrioKart and wanting at least one working vehicle, I cracked it open and found out that the (poorly) reflow-soldered output capacitor fell off. After soldering it back on, and still have it be dysfunctional, I made a 30-to-5-volt linear regulator from a 78T05 chained to a 78T12.  One 7805 cannot handle the drop from 30 volts to 5, so the 7812 is an intermediate stage. This is an arrangement that’s worked for me before.

After installation, the headlight, servo driver, and ESC fire right up. I was able to reset my throttle endpoints as usual. Then I pushed on the handlebar throttle, and


(Oddly enough, the headlight and servo driver  stayed on through the entire ball of fire. That means my regulator works.)

Back to DC motors and relays…

The Short and Tragic Flight of the LOLrioKart

Mar 23, 2009 in LOLrio Kart, Project Build Reports

I did it.

Well, kind of.

The (physically) largest and most (spiritually) ambitious engineering project I have embarked on yet moved under its own power, finally, after nine months of on-and-off work, like some bizarre, monsterous electromechanical pregnancy. There’s still development to come, such as adding some form of braking as well as making sure the throttle is not held on by a strip of duct tape.

But alas, the giant is stillborn. You saw this coming, didn’t you?

Anyways, some pics of the completion.

The glue has dried.

Now it’s time to lay out the electronics!I had no particular layout in mind, so it was clamp-and-drill in a semi-oriented fashion. I did make sure that appropriate connection points were on the appropriate side of the board, however.

While using machine screws in wood seemed like folly at the time (and still does), I didn’t have wood screws of the correct length.

More goodies start appearing, mostly scrounged. I discovered some copper screw terminals designed for 4-gauge wire in a box full of Jolt-can-sized capacitors. The golden blocks are car audio power distribution blocks, spares from the car project. The three-terminal Wiring Blocks of Convenience™ were found in a box of giant industrial contactors and switches.

The large controller is a Kelly KBL72201, 200 amp with regenerative braking ability. The car used 4 of these with a custom control interface, and they give a pretty good bang for the buck (with the bang/buck ratio being a little under 2 for the highest-capacity ones, assuming bang is measured in amps and buck is measured in 2009 U.S. dollars).

Speaking of 4-gauge wire, here’s a few feet. I wrestled this stuff around for a few days getting the car wired up, and these are leftovers. Requisite handling tools included a sharp razor blade and pliers for stripping, a hacksaw for cutting, and a bench vise for crimping on connectors.

This time, I had an upgrade – instead of a hacksaw, I got to use a BANDSAW! Rejoice!

Here’s a battery interconnect. These are made of the aforementioned 4-gauge wire and two copper terminals.

Most of the SERIOUS FUCKING \M/ETAL routed.

I tried my best to make this assembly as modular as possible. Notice the three terminal blocks surrounding the controller. The bottom-most is a main power block, and the other side is intended to go to a master power switch, then back to the controller. The leftmost is the motor output – conveniently, three wires for three phases.

The tiny terminal strip at the top is for signal connections from the controller.

This way, nothing is hardwired to a particular component. I could have just as easily duct taped all the wires to eachother, but that’s not a very sustainable activity.

Signal wires brought out of the controller and onto the terminal strip. Luckily for me, Kelly Controls sells premade labeled & shrink-wrapped signal cables.

Short motor wires for minimal loss! I used a variety of giant connectors, some scrounged from spare parts bins and others yanked off existing giant wiring.

The magic moment arrives. With controls hooked up and switch installed, it was time to perform the first live power test!

But like a good CNC machine, I looked ahead in the program a bit and decided that maybe, just maybe, jacking the rear wheels off the ground would be a good idea, in case of some stupid error, such as a reverse-wired motor, the 200 pound kart goes balls to the wall backwards. Maybe literally, and possibly with someone else‘s balls.

I hit the power switch, and…

No, it didn’t blow up, but the motor instantly jumps to full speed.

In reverse. Glad I performed the last minute sanity check, eh?

A quick installation of a reverse switch (instead of guessing at what combination of power leads need to be turned around) and I was in business.

This is the best part about the whole first test run. The throttle lever was retained by duct tape and there was no braking to speak of. As in, none. Zero. The little band brakes weren’t yet hooked up, nor did I have another springypot handy for the controller to perform motor braking.

All that stopped me (LOL PUN) from certain disaster was drivetrain drag and air resistance. Nothing could _possibly_ go wrong like this.

So let’s bring it outside. I wanted to finish this test setup by sundown, but didn’t quite make it. The first (dark, shaky, blurry) test video is here.

All things considered, I think it was a great test run. The frame survived numerous potholes and high-speed curb landings. Nothing seemed to come loose or fall off.

The battery basket worked as designed. I was, furthermore, surprised by how well it handled at high speeds. You’d expect the center of gravity to be dangerously high, but it wasn’t – with about 100 pounds of batteries about 3″ off the ground, it’s actually exceptionally stable.

Stable enough for stuff like…

After 10 minutes of putzing around, we decided that it would better to test at a later time… you know, when cars weren’t coming and going regularly. I brought it back inside, then went away to get dinner.

When I returned, I went to flip the power switch, and everything exploded. There, that’s the kicker you all were waiting for!

A bright blue flash emanated from the rear of the controller, accompanied by firecracker sounds and a small cloud of black smoke. Needless to say, I flipped off the master switch as quickly as I flipped it on, but it was already too late.


Of course it was the electronic side of my project that magically works and then doesn’t work. The same thing happened with RazEr a few times.

This time, however, I have absolutely no idea why anything could have failed. It’s not like the controller was something I cobbled together in a night, hanging together with alligator clips. It’s not like I even attempted to modify or change anything in the hardware. I literally left it alone for half an hour.

Now we know why I’m not EE/CS. Because if something doesn’t work in the world of mechanical engineering, you can usually see why.

After flipping the switch back on just to see the last few FETs fizzle out, I started removing components and probing around. I did discover one suspect – the frame had continuity to battery negative, through the throttle potentiometer – I had neglected to notice *how* I was duct taping it to the metal frame, and accidentally the pot terminals. Nothing else had low impedance where it shouldn’t have. Hell, the entire thing is mounted on wood.

But would such a thing be enough to destroy the power amplifier? Questions remained after this, such as

  1. Why did it work at all for the test run, then?
  2. The throttle cable is 22 gauge wire. Is this really going to short enough battery current to blow out an entire row of FETs?
  3. If it can momentarily, why didn’t it blow when I held the switch to see what would happen?
  4. Can a fault on the logic side of a controller, like a ground loop, really cause the power side to fail?
  5. The explosion was practically instantenous. If it was a simple short, wouldn’t it have taken at least a few seconds before the FETs thermally overloaded, then failed?
  6. What can possibly cause a voltage surge on a 54 volt battery (62 volts fully charged) that would destroy a 72 volt controller?
  7. Why haven’t I switched to a brushed Etek and giant contactor yet?

Regardless, I’m bummed, because I’m out the mo$t importan£ ¢ompo₪€nt. Hell if I can get Kelly Controls to budge on a swap unit – it’s $100 to send it in for replacement.

Maybe it’s time for me to build something that doesn’t involve complex power electronics (or just start building the complex power electronics themselves?), because I can’t afford (literally) to keep having electrical demons delay all my projects.

Nontheless, here’s a (95%) final shot. Note the important addition, the single large revolving light at the front.  I have to have my priorities, after all, and this was far more urgent than, say, brakes.

But I didn’t get to test it.

Things to come? Real pedals! Both brake and throttle! Random decorations! Giant aluminum spoiler! Ground effect lighting!

And maybe a working electrical system.

LOLrioKart Update 13: Yeah, something like that.

Mar 18, 2009 in LOLrio Kart, Project Build Reports

This would actually be update 12 if I didn’t count the co-update with the battery meltdown. However, I did.

In the last LOLioKart episode, I had settled on a scheme for mounting the giant NiCd packs in the frame.

Generally, in life, it’s a bad idea to put all your eggs in one basket (or invest all your savings into a single financial avenue…), but in this case, it was the easiest and most compact. The “4-across” method kept all the packs next to eachother and the space above them available to mount whatever else – in this case, the retaining plate was to double as a fixturing surface for all the electronics and wiring.

So I whipped up this quick “basket” with compartments, each of which holds one pack with some spare volume that I can fill with shock-absorbant materials, important for a suspensionless vehicle that will eventually travel over streets whose most ubiquitous surface feature is the pothole.

I mean “whipped up” pretty literally, because several hours after finalizing the design, I committed it to aluminum.

Part of what I find appealing about working for the Media Lab is the degree of “have metal, will build” that it promotes in my building habits.

Maybe that’s actually a bad thing, since I’ll admit to becoming increasingly spoiled by quick access to neat things like waterjet and laser cutters, and have started basing all of my builds around them.

Anyway, shown is the waterjet-cut basket undergoing assembly using my favorite tools for such an activity.

One of my now favorite methods of “expedited construction” is erecting 3-dimensional objects using 2D (flat) cut plates with slots and tabs, which allow them to interlock and assemble into something more complex. This is similar to those old “wooden model dinosaur skeleton” toys, and “chinese puzzles”… In fact, the latter is what I always call this method of construction, since I don’t know if it actually has a name or not. It’s a bit tricky to get used to, and accounting for cut kerf and corner radius (especially important on waterjet cutters) is somewhat of a trial-and-error process, but I couldn’t have built NK5 in a day without it.

If you get ambitious, you can even put orthogonal screw-and-nut slots to make the whole assembly stronger as well as disassemble-able. I decided to forego that this time.

Uh oh. This is the result when I forget to pay attention to where I reference my mirror planes from.  The slots are offset by about 1/4″.

1/8″ carbide endmill to the rescue!

Maybe a year and a half ago, this would have been a serious project-delaying fuckup, since I’d have no way to move the slot over except by file and patience.

Actually, more than a year and a half ago, I couldn’t even imagine having waterjet-cut parts.

Or machined ones on-demand, for that matter. Ah, the joy of bot-poverty.

Whole thing squished together. All the tabs and slots are compression & interference fitted, which means this whole thing is pretty solid. Optimally I would have screws retaining the slats, since this kind of thin compression fit (a swage) is sort of weak against high vibrational loads.

We’ll see where it goes.

Also, I lied when I said that I committed the design to aluminum – the bottom plate is actually steel, since I was unable to locate a contiguous piece of aluminum plate of that size.

An initial test mounting was then in order. Here’s one battery pack sitting in its compartment, and the whole assembly dropped into the shopping cart bottom frame using the half-inch thick mounting ears.

Unfortunately, even the most precise eyeballing can be erroneous at times. I had no coordinate-measuring device large enough to cover the bottom of the cart and the batteries, so I had to eyeball dimensions with a tape measure. This worked well enough, but I was off by maybe a quarter inch in width.

Problem is, with the frame having a total front-to-rear taper of 10 degrees, even a quarter inch of error is quite a substantial in terms of longitudinal placement – and by that I mean about an inch and a half difference.  I had to cut out a small portion of the tube frame to position the basket where I had originally intended.

This does not substantially affect the structural integrity of the cart frame much, since the tubing is (pretty fuckin’) thick-walled.

I’m not seriously going to just hang all the batteries from a hook, am I? No, that’s too daring even for me. Instead, the mounting ears have corresponding bottom halves that clamp around the frame tubing. I was not able to account for the longitudinal taper when cutting these, but figured it shoud be insignificant enough over 1/2″ to plug my ears and go LALALALALALA.

Notice the small discontinuity in the middle of the mounting ears. This is the result of me sanding down a “tab” I added to the cutting path. Small parts on a waterjet tend to either 1) fall into the tank, never to be found again, or 2) float up and get caught between the stock and the cutting head, which then 3) converts some of the mass of your checkbook to a “crunch” sound. None of those options are particularly appealing, so the best way around it is to add a “tab” – a small cut discontinuity such that the part stays in the stock, but you can break off easily.

Here is one, semi-tightened. Notice that I also cut off part of the mounting ear past the cap screws. It covered up part of the two smaller holes that I was going to use to mount the +4 Plate of Battery Pack Retainment, because the placement error caused the mounting ears (which have that single slot for adjustable placement) to become seated too far inwards.

The belly of the beast, with all four corners clamped.

I discovered a roll of 1/8″ neoprene rubber int he back room. How convenient – it’s the perfect thickness to cushion and insulate the battery packs in their compartments.

I sectioned the roll and cut the chunk into small strips to line the compartments.

Hmm, looks like I need a chain tensioner after all. It will only be worse when the Plate of Battery Pack Retainment is added.

All four battery packs in place. Getting them into their compartments was a bit of a challenge, since I could not just drop them in from overhead or load from below. They had to be inserted carefully from the side, which was a balancing act for me.

LOLrioKart is now too heavy for me to heave around – so if I feel the need to move it around, it has to roll. I estimate the total weight at around 175 to 180 pounds at the moment.

Let me say right now that parallel-parking a shopping cart in a confined space like MITERS is quite a unique experience.

Now onto the retainment plate. These “nut blocks” screw into the quarter inch aluminum plate, and the retainment plate screws in from above using the large screw holes. The front mounts are smaller and made to original dimensions.

The rear ones are longer and offset because I found out that the original retainment plate design interfered with the motor mount, so I was forced to widen it.

The new plate with wider-spaced ears and motor cutout.

So, here’s the plan. This plate will be pressing directly over the batteries, so it’s not like I have the ability to simply drive a screw through. That would be bad – instead, I am making the plate (currently 1/4″ plywood) thicker by adding a buffer layer of 5/8″ plywood, and a layer of 1/8″ neoprene above that. This would allow the use of modestly-sized wood screws to mount components without having the screw tips stick out underneath. The added thickness would add stiffness to the whole assembly.

The neoprene is rubber-cemented to the thick plywood, which is in turn wood-glued to the thin plywood.

I threw the entire thing under about 150 pounds of cast iron milling accessories, parked over the hotspot in the room (MITERS is situated over the N52 *boiler room*, and so there are spots on the floor that are always warm) to let the glues properly adhere.

In the next episode, I find out if my glue has set or not, and continue working on mounting!The plan is to get LOLrioKart rudimentarily running on the full electrical system – probably with a “GO” option and no “STOP” for now. Then work will shift to decorations, improvements, and..uh, braking.

Spring break is appropaching fast, then after that, Campus Preview Weekend. How many pre-frosh can I send to the hospital?

(Don’t answer that.)

This is how not to charge large nickel-cadmium batteries.

Mar 13, 2009 in LOLrio Kart, Project Build Reports

The batteries I picked out (literally, from the bottom of a crate filled with cruft) for LOLrioKart have been problematic. They’re more or less in hibernation from being stored so long. While I’ve used my multicharger to try and tend to the batteries, it’s simple too low-power to really wake the batteries up.

Charging individual cells was an option, but not something I wanted to repeat 44 times, if not more, so I would try to ‘zap’ the cells to get them to a workable voltage, then charge the entire pack. But, by the time one pack completed (usually after 18+ hours at 6 amps), some cells would have zero-volted again in the other packs. This was obviously not a sustainable activity.

Plus, pegging a charger at maximum capacity for hours is unhealthy for the charger. Alas, after weeks of nearly constant work, my charger died with a little puff of smoke.

I’m not peeved. The fact that it lasted this long near the top end of its ratings means that the cheap stuff has gained some serious ground in terms of reliability. For those interested, here it is.

After that, I decided to pour a little more brute force into the equation. So now I present…

A 2,000 watt variac. A 50 amp bridge rectifier. 10,000uF of giant capacitors. And… well, that’s about it.

I even added little caster wheels so I wouldn’t have to pick it up and drag it around.

It’s a completely unregulated power supply that *sort of* functions as a constant-current supply when I hook it up to something. There’s no auto-cutoff, no temperature sense, no readouts (have to poke at it with a multimeter), and no SCRAM switch, besides a 40 amp circuit breaker (really 2 20-amp units paralleled). It vibrates at 60Hz ominously, and probably raises the level of the Homeland Security Terror Threat Level a notch just by virtue of existence.

And it works beautifully.  Throwing the switch on all 4 packs and giving it full crank  trips the local 120 volt circuit breaker (20a) within a few seconds. So, to keep things tame, I limit the battery charge current to 25 amps, a healthy 2 electrical horsepower (+a pissed-off pony) at 70 volts.

Note that this is a later pic, where I have added a current shunt  (upper left corner of the board) such that I can  get an amps reading without dying.

I actually lied in the title. NiCads charge optimally under constant-current conditions, and a great way to terminate the charge is the rate of temperature rise (dT/dt). For the time being, it’s “poke the battery, come back 5 minutes later and poke it again”, and if the answer to “how much hotter is it?” is between “quite a bit” and “alot”, I disconnect the power. This would be easy to systemize with a few thermistors jammed between the cells.

Currently (LOL PUN) though, I have to babysit it through the charge, lest I want to fill MITERS with a cloud of heavy metals and sodium hydroxide.

The bottom line is, the NiCds needed some ass-kicking to wake up. Work on LOLrioKart can now continue – I was losing interest for a bit because of the battery issue, since without batteries, having an Etek is sort of pointless. The goal is to get it at least running by Campus Preview Weekend, because…

What better preview of MIT is there than this?!


While you were away…

Mar 04, 2009 in Project Build Reports, Project RazEr

I finished the rebuild of Project RazEr. Because the site has been down, I never wrote up any day-by-day build reports, but still took pictures accordingly.

So, here’s an epicly long build report all the way from start to finish. The whole process wasn’t long and drawn out this time, mostly because the fabricated components were conceptually simple. The majority of work was done in a few evenings at MITERS (as usual.)