Archive for July, 2013


Melonscooter 2 Rage-Finish

Jul 25, 2013 in Melon-scooter 2, mikuvan

Last week, I actually rage-finished not one but two projects in progress. First was Colsonbot, whose test came this past weekend at Bot Blast; and second was Melonscooter 2, which has been tested every single day since then already. I’m getting to this update now because, well, Bot Blast. Melonscooter 2 is now done, save for remounting my signature orange basket.

Finishing the construction of Melonscooter was delayed by replacement timing belts and pulleys. After assembling the Epic Pulley, I measured the distance between wheel axle and motor and came up with the nearest timing belt size to order – this was done with the tensioners, which are of the axle-pulling type, in the most relaxed position so I could order the next size belt up. Turned out to be a 700mm-size (140 tooth at 5mm per tooth) belt with the 20 tooth pulley I wanted to use. Both of these were ordered from my next favorite belt and pulleymonger, BB Manufacturing.

Waiting on those parts was what delayed the build for a little while. But when they got here, it was time to take everything to completion in one night.

Here’s everything test fitted out and satisfactory. The overall ratio of this drive is 5.4:1, which is higher than my typical ratios for this size wheel (more like 4:1). For instance, previous Melonscooter had 8″ wheels, only 1″ shorter, but only a 4:1 ratio.

The reason I’m doing this is to extract a little more power out of the system by allowing the motor to spin fast. In my opinion, that’s a more tenable solution to R/C motor powered drivetrains, since R/C motors love to spin *fast*. By allowing the motor to spin fast, you can gear higher and hence end up getting more torque per amp out of the system (faster acceleration, punchier throttle, the classic stuff people want R/C systems to accomplish). This is especially crucial if you can’t have a Controller of Infinite Amps.

In my case, I’m limited to a 50/120A small Kelly Controller – 120 amps is all I get, so why not increase the peak power of the system by allowing the motor to spin closer to the Kelly’s top speed limit? At 38 nominal volts, the 80/85 “short melon” with its 170 RPM/V shoots a bit above what the Kelly can hypothetically keep up with, so it should be no problem when loaded down.

That’s also why you’d be hard pressed to ever get more than about 1000W from even a heavily modified Jasontroller – the top speed limit of the things is so low (about 4700 rpm on the average R/C motor) that the only way you have to get more out of the system is increase the current, which is on shaky component ground after 30-40A.

As BurnoutChibi shows, if you can get away with using a fast motor with a high gear ratio, it’s the way to go for sheer performance. If only drivetrain components were lossless – the high the ratios, the less coasting ability you’ll have since the friction inherent in the motor and power transmission is amplified by the gear ratio. I’ll stick with my almost-lossless timing belts, thank you very much, for daily use where I need other redeeming characteristics besides brute force (oh dear, I’m losing it…)

I broke out a charge point for the battery using a somewhat-shielded XT connector. My favorite for this is historically Deans, but I definitely find it hard to aim those things some times. Since this location is immediately inside the kickstand, getting a good line of sight would be harder, so a connector with a shroud like the XT is better.

Now, at least I’m not using the XT for actual power. I hate those things with a passion.

The utility wiring on this build is pretty sparse – just a jumper from the charge port to the battery side of the switch, and the switch outputs directly to controller. I didn’t even bother with a precharge circuit (inrush current limiter) this time. That’s how much I didn’t care.

(The tradeoff being earlier retirement of this Hella switch from arc erosion, and possibly risking detonating the Kelly from inrush current on ever power cycle. Don’t be like me, kids.)

And no, the batteries aren’t isolated. What this means is the pack is not made of two packs joined in series externally, so you could charge each half on one power supply if connected in parallel. This is to remind myself of the fact if I ever take this pack apart and try to charge it from two power supplies at once.

Flipping everything over now for the main installtion. The KBS48121 is bolted in, as is the battery box through its four side screws. At this point, it’s cleaning up and tying up wiring to keep out of the way of rotating assemblies.

This 80mm motor features a Hall Sensor board and mounting rig by Yours Truly.

A bit hard to see here, but now I’m wiring up throttle cables and the like. I didn’t go for super clean through-the-handlebar installation like this thing came with – it would have taken too much time to thread the noodly cable through the wire channel inside. Instead, I used some spirally-loop-wire-loom-twizzler-things that I stripped from the original wiring. This at least makes it look not like shit unlike Melonscooter 1.

I could ride it like this, I guess. This is the state right before the first all-systems test. The throttle wire runs above the battery pack but below the main frame, in the roughly .2″ gap between the two.

To tune the Hall sensors, I followed my own page about Hall sensor tuning to the letter. It only took 2 wire swaps to get the motor to spin, then I alternated measuring current draw and bumping the sensor board a little each time (since you were not going to convince me to reach my hand into a spinning belt drive) until the current fell to something like 12 amps no-load. Still a bit high for this system, but by then the sensor board was maxed out adjusting in one direction and I did not want to play the wire game any more.

Bundled up and bolted together!

A few test spins around the IDC hallway told me that I should not be riding this thing in the IDC hallway.  First, because it launches fairly energetically, and second, because I came close to tearing off that giant Hella switch on the starboard side several times.

Outside MITERS after some around-the-block street runs. This thing definitely tastes like suicide again. I’ll say that Melonscooter hasn’t been this jumpy and powerful since its original R/C controller based incarnation. I have yet to have someone pace me or radar gun me to get a speed reading – maybe it’s time to break out a GPS speed application, but it’s definitely well north of the 22-23mph that the late models of Melonscooter 1 could hit.

It also rides like a old Cadillac and handles like a battleship. The giant 9″ balloon tires are even better at road damping than the 8″ tires of Melonscooter 1, and the much more solid frame design means it just sort of rolls over everything. One thing I don’t like as much about this frame is its very wide turning radius. Those big shiny front forks hit the folding joint at maybe 35 degrees of steering tilt. You’d never go this far when riding at speed, but it can barely snake its way up the wheelchair ramp entrance because the turn circle is so huge. I would place this build on the very high end of what I would accept as portable – it’s definitely big for my usual tastes, and if it were any bigger I would have trouble with bringing it inside.

The possibly suboptimal motor timing means that people heavier than me who try to ride this thing full throttle some times experiences the “Kelly Cutoff”, where the KBS controllers can’t handle the fast-RL time constant of R/C motors generating high current transients and shut down to protect themselves. I’ve also experienced this once or twice so far, but not under conditions which I can reliably duplicate.

Some times the “Kelly Cutoff” is resolvable just by turning down the maximum current command to less than its max rating; for this case, maybe 100A. If it becomes a consistent problem, I’ll try that first since the acceleration on 120A is pretty satisfactory. To put it mildly. The next level of fix would be actually spending time to nudge the sensors.

For now, I’m hopefully going to get to range test this thing once the Mysterious A123 Cells of Yore get a few cycles on them – I’ll check the balance state of the pack then. Hopefully if any of the cells are substantially weak, I can replace them before moving on.

daily van bro

A few days ago, while riding Melonscooter 2 down a main street in Cambridge, I passed a white van, one like many thousands of others plying the streets here, representing about every possible contracting firm, flower business, or sandwich house.

But something just didn’t look right as I passed it. I wasn’t sure what, but I got this feeling of “this van ended too soon”

I was right. It did end too soon. Someone had brought an honest-to-goodness modern Toyota Hiace all the way up to Cambridge, which means they just saved me an entire trip to Japan just to take pictures of weird cars.

The previous generation Hiace is the type subject to all of those glowworm-porcupine-starfish-Transformer mods that I originally began being into these things for. I’m not very in resonance with the current generation styling, but they don’t look too bad with the right color scheme, kind of like driving your wireless router down the street, or at the least some sort of bagel oven. Either way – holy hell, someone brought one to what was basically my front door. Life is taunting me – just fucking taunting me.

Sadly, it looks like a rental vehicle, and from Mexico at that (which is my best shot at getting one of these that isn’t older than I am). I assume it was some visiting tour group from Mexico. Unfortunately, by the time I took this picture, their parking meter had run out, but I didn’t love it that much as to give them freebies, however. Enjoy yourselves some American parking tickets?

So what do I do when life dangles a carrot in front of me? Go get some celery and fight back. Whatever the hell that means.


I immediately whipped Melonscooter around and fetched Mikuvan just to take this picture. Why yes, my name is Mr. Va(i)n.

Speaking of Vans Next to Things (…, here’s another spontaneous picture which may appeal to a wider audience than just me.

Returning from a Home Depot trip for the summer go-karts session (more on that later), I wondered who the hell was parked like an ass on the busiest throughfare in Cambridge.

Check out this three orders of magnitude span of vehicle value (Mikuvan: $800. Lamborghini Gallardo: $200,000). This is a pretty damn near Jalopnik level of “highest vehicle value contrast” here. You can’t get much lower than 3 figures in price, nor higher than 6, typically. The only way for me to beat this is to find a Bugatti Veyron.


Now, about that orange milk crate.


Colsonbot Rage-Finish and Bot Blast Recap

Jul 23, 2013 in colsonbot, Events

Well, here it is! Colsonbot was finished in the intervening 3 days between the previous post announcing its arrival and when we all shipped off to Bot Blast. The event was a blast indeed, and I’m furthermore glad we even got there in the first place! Here’s the whole story.

This is the friction drive module fully assembled. A little printed rim attaches around the motor, around which I sling some stretched-out O-rings. I recommend everyone who is considering O-ring friction drive make sure that the rings are not bought “on-size”, since if they are, at high operating speeds they will stretch out and be more likely to pop out

Tension in the friction wheel mechanism is maintained with a stack of wave washers. I was hoping for enough space to put a very short coil spring, but it ended up that I really only had 4-5mm to play with. So, wave washers it is.

Since this bot is way more unibody than anything else I’ve built, I needed to construct most of the wiring harness first before installing the motors. Here are some of the Pololu 30:1 HP “Sanyo-flavored” gearmotors being prepped with wires. I call these motors “Sanyo flavored” since originally the first on the market in the robot world was the 75:1 Sanyo micro gearmotor.

Other electronics being prepped include two Vextrollers, one of the left over 18A ESCs from tinycopter version 0, and a gutted Hobbyking receiver. Everything had to come out of their packages for this bot since the space was extremely tight.

The wheels are also 3D printed affairs. They’re little duallies made with O-rings. The center bore is pre-flatted to mate with the Sanyo motors with a set screw hole on the flat to keep them secure.

The installation sequence is wheels on motors, then motors into their mounts.

The center shaft of Colsonbot is a single hollowed-out 5/8″-18 fine threaded bolt. The hex head was machined down to 1/8″ thick, and it rests flush with the top of the battery cavity in a hexagonal cavity which prevents it from turning as I tighten the clamping nut.

Wiring on the bot can be described as “ad hoc”, to keep it politically correct. I used no connectors at all – there’s not enough space for the ones I had. To reverse motor directions, etc. meant desoldering.

Most of the wiring is done with 30 gauge wire-wrap wire, with heavier currents such as the brushless controller and two Vextrollers done up in 24 gauge.

First power test! This was just to verify that everything was still working and hooked up the right way.

Starting to add the other snappy-lids to keep things constrained. The electronics are mostly just piled inside the cavities. I had thought of making little exactly Vextroller shaped cavities, but realized I would never have put up with that level of cleanliness. It’s just a little too Apple.

Lids attached. The external wire run is unceremoniously held down by a glob of used Kapton tape harvested from the Hobbyking battery.

Time to put everything together and find out exactly how not a 3lb bot this is. Turns out the answer was not even 2 pounds. Great…

I knew Colsonbot would have a not-enough-weight problem since it could literally be made of lead and be fine. There’s just not enough volume of bot inside the wheel. Colsonbot’s base by itself, at 10 ounces, would actually make an excellent 1lb shell spinner with a 6oz real shell. This is something I’m considering for Dragon*Con, now a little over a month away.

With a second 5/8″-11 jam nut and a bit of thin PETG plastic, I made this direction indicator since otherwise I had no idea which way the bot was facing underneath!

And here is the soft underbelly!

Check out some of the test footage:

Colsonbot being  a roughly square 4WD setup, handles extremely well and it’s also quite peppy. It really does not need this many motors. As can be seen in the video, it “gyros” extremely easily due to the large spinning mass. I originally wanted to put a “mast” up high over the wheel so it could flip back over, but I decided to keep the wheel illusion. If Colsonbot gets flipped, it will tend to stay upside down and wobble like a coin.

This work was done on Thursday evening. I left Colsonbot alone after that to help out with the two new stragglers we inhaled: Jaguar (by the way, check out his excellent Instructable on his even more excellent Orbit Wheel pod things) and Julian. Of these two, Jaguar decided to start a new robot some time around 1AM Friday.

bot brast

Great, now we’ve increased from four to six people. I was really getting worried about Mikuvan’s mechanical integrity at this point – six guys and another hundred pounds or two of robot gear times 800 miles round-trip plus mountains. If I was going to grenade something, it was going to be on this trip!

On Friday afternoon, I gave the engine cavity a more thorough check over. My primary focus was on the timing belt, since I had observed some leaking oil on the front underside of the engine, leading me to think leaking camshaft/crankshaft oil seals, which would throw oil all over the belt and cause quicker deterioration. Otherwise, I had already done a mechanical once-over after the Adafruit Adventure. To my surprise, the timing belt cavity was dry and there were no signs of abnormal wear on the belt, and the tension was still good. So where the hell was all my oil going, then?

That’s the back side of the engine where it mates to the transmission. I see a new sludgebunny colony is starting to form. This is the area I cleaned a few months ago to see where the oil leak was, and… well, this just tells me it’s coming from everywhere. The side of the block is still well-coated in grunge.  I definitely have no idea where this could be originating from. Valve cover gasket? Rear crankshaft seals?  I’ll write it off to built-in underbody corrosion prevention for now.

Since after the New York trip the oil level was low, I topped off this time with some heavier weight oil – some Mobil synthetic 10W-40 (the manual calls for 10W-30). Time to see if those “for high mileage engines” claims have any merit. The slightly more viscous oil might reduce the rate of runoff if nothing else.

Whatever the case, it was time to ship out. Space Battleship Mikuvan was temporarily commandeered to become the MITERchan Party Van. Having only like 105 original HP and surely less available nowdays to push 6 people and robot gear, this was pretty much my most conservative road trip ever – essentially spent all at less than 70mph and almost all in the right lane. If you know me well, then you know this is basically the opposite of my usual style. The van is teaching me some humility.

Somehow, 6 hours later, Waffle House.

Every time I wander out of New England, I must stop at a Waffle House, so this time we met up with aaronbot a hour outside of the event venue at a Waffle House in Scranton, PA.

On site at the event… and there’s more than one Colson! On the left is Agent Colson, built from a 6×1.25 wheel – it’s flatter, but only 2WD, and also direct drive from a small outrunner. Colsonbot was originally going to be a DD from a large diameter (e.g. multirotor motor) outrunner, and that idea is still being entertained.

Check out the lineup of newbie bots. From left to right is Glorified Doorstop, DEL-RAN Bumble (by Dane), and Speed Bump (by Jaguar). Don’t let those eyes and zip ties fool you – the front of the bot is to the right. I’m glad that somehow in the last week we managed to generate an entire Battlebots team.

Motorama should take walk-up registrations like Bot Blast, because hell if you could get MIT students to commit to anything more than 1 week ahead of time. Trust me on this one.

The event was held in a regional mall, nestled between JCPenney and Payless Shoesource. Quite a contrast of venues from the typical grungy warehouse or barn you find these ‘bot events in. Because of the proximity to curious onlookers, the audience was pretty steady and numerous for most of the day.

I don’t have as many pictures of everything this time, since I was either sleeping under a table or filming matches (And I’ve posted random robot pictures dozens of times). But here’s Agent Colson after one of its first matches, where it may or may not have hit the ceiling. Sad…

The box is a nicely build 12 footer which fit beautifully into the mall’s… band pavilion? Free Speech Corner? Not sure, but it sure worked well.

Colsonbot in line for its first fight…

Colsonbot performed flawlessly during this match (of course it didn’t win anything), but afterwards I noticed the weapon motor was shifting in its slidy-mount. The motor was chosen as a compromise – the one I purchased from Hobbyking that was intended for the bot just didn’t fit. I used instead a Hacker A20-50S, which was admittedly undersized.

It didn’t get hot enough to burn out, but it did get hot enough to warp the ABS housing, causing some tension loss in the friction drive. I literally bent it back while the motor was still hot.

Both colsons being serviced in some way…

In my second match, I sort of went for broke at the end and tried to keep spinning. This resulted in the motor melting the ABS to the point where the friction wheel just sort of welded into its mount, stopping Colsonbot. I still had drive ability, though, so spent the next half minute just getting punted around.

Colsonbot suffered no particularly permanent damage at Bot Blast, and was a great first version run. Now that I know how this version performs, I can better build the Next Generation Colsonbot.

Boy, some “joke” this has turned into.

For Dragon*Con, Colsonbot will return. I’m currently thinking this as the list of changes:

  • 2 motors, since 4 was a bit excessive, and O-ring belt drive to mimic the stable 4WD drivetrain
  • Fitting the larger and more robust motor in the bot as a result
  • Possibly switching to tinyESCs or something to kill the electronics volume even more.

Finally, here are Colsonbot’s two matches from the day!

Returning from the event was entirely uneventful, save for a short section of I-84 in southern New York which seemed to climb straight up a mountain for 5 miles. I realized it was hopeless after fully opening the throttle, falling back into 2nd gear, and was still losing speed. That stretch was done at 40mph with the heater fan on full as I watched the engine temperature climb slowly towards hot.

Clearly, I am never going on a cross-country roadtrip with this contraption as is – I will never even make it over the Rockies, and if I take the southern route through Arizona, would just light on fire regardless. By the way, Newburgh’s Stewart Airport which we passed next to is the site of the original Three Legendary Derpy Van showdown.

So now that Bot Blast is over, it’s time to start prepping for Dragon*Con. More on that later. The final thing to take care of is…

build your own colsonbot

Here’s the entire Solidworks 2013 CAD folder for Colsonbot. Also included are the ready-to-print STL files for all the major components. You’ll need:

  • A Hacker A20-50s or similar motor
  • An appropriately small ESC for the motor – I used an 18A Turnigy Plush
  • Four Pololu 30:1 micro HP motors
  • Two Vextrollers
  • A 3 cell lipoly battery, this one in particular is known to fit (but only if skinned)

Loose Ends Roundup for the Week of the 14th: Adafruit Trip Summary, DERPDrive Painting, Melonscooter’s Battery, and What does a Colsonbot Do?

Jul 16, 2013 in Bots, colsonbot, derpdrive, Melon-scooter 2, mikuvan, Project Build Reports

Here’s another one of those posts where I report up on like 17 things at once! Running (this time wholly my own – no more protection afforded by the likes of 2.007!) the summer go-kart class for the MIT-SUTD collaboration has been one hell of a time sink, so I can only get small incremental things done at any one time.

We begin first by recapping what all went down to get me on the Adafruit Ask an Engineer show this past weekend. The trip to NYC all started as a group desire to just hang out in the city for a few days; so I contacted Makerbot and Adafruit Heavy Industries Co. Ltd. to see if I can swing in anywhere and check them out. Sadly, Makerbot is too pro these days to afford a random visit to their production facility, but Adafruit gladly obliged with an invitation to their web show.

This trip was actually slated to be the very first major long distance haul for Mikuvan. None of us really expected to end up in the city – more like broken down in Rhode Island somewhere. I made sure to pack all the tools needed to service anything short of catastrophic driveline failure, and picked up a new compact spare tire (the stock full-size spare having rusted out seemingly years before, which I took in to get scrapped) beforehand from Nissenbaum’s up the street here.

I’m proud to say that it went down completely without incident. Now I have even less of a reason to dismantle the powertrain, right?

I even looped a new A/C compressor drive belt beforehand (came without one) to test the state of the air conditioning coolant circuit – and to my utter surprise, it blew totally cold. So there we go – all the amenities of a modern car with 9000% more “What the hell is that thing?”. By the way, the A/C still runs R12.

Above is a picture of the van right after arrival in Flushing, Queens.  The only downside, of course, is that it has juuuuust enough horsepower to climb the Whitestone Bridge at about 50mph constant velocity with the gas pedal floored. Horsepower is not something hastily-modified JDM cargo vans are known for, but the electric version ought to fix that. I’m aware the speed limit on the Whitestone seems to be 30mph, but the crowd of delivery trucks and NY-plated private cars huddled around me seemed to beg to differ. I’m sorry, everyone, for having no power whatsoever.

Anyways, Nancy sums up our discoveries about Adafruit well. I no longer think they are made of magic and open-source genome unicorns, but infinity work and dedication.

On this trip, I confirmed the engine oil consumption as about 1 quart per 700-800 miles highway driving, and more like 500ish-miles local (with more cold-starts and short driving trips).  This is a staggeringly high amount, but I don’t think most of it is burning up. During my pre-trip inspection, where I recorded all fluid levels and made sure things weren’t jiggly and double checked my brake rotor-pad-shoe-drum-line-fluid conditions (since it should at least be able to stop, nevermind go) I discovered some fresh oil slicks near the bottom of the timing belt cover and that area of the engine block. This tells me that I probably have a leaking crankshaft front oil seal, and could explain the terrible condition of the timing belt discovered prior to Operation: BAD TIMING. It also tells me the current timing belt might not live that long anyway. The exhaust does emit a brief burst of smoke when cold-starting after a few hours of sitting, so it could indicate a number of other things worn, like the valve guide seals which were suggested by more automotively inclined buddies. I’m willing to write it off to 20+ year old poorly maintained engine. The oil itself does not show excessive signs of burning – the shade isn’t particularly dark, nor does it smell like burned fuel significantly, so I’ll say that most of it is just physically leaking out.
The fact that I hauled ass a total of 450 miles without any hiccups is amazing in and of itself, I think…


Hey, if I’m not going full-on electric right away, let’s at least check in on the thru-the-road hybrid shop-pusher module. DERPDrive hasn’t moved an inch in the past few weeks save for painting (in the same round as Melonscooter2), and that process looks kind of the same:

I picked up a handheld sandblaster from Harbor Freight (this one) to pluck all the rust and scale off the welded steel tubing quickly. Along with a jug of 80 grit aluminum oxide, it took maybe an hour or so to reduce the major frame parts to fresh steel. Here’s a picture of the blasting in progress. By the end, I’d created a small ejecta ring of sand, and I was basically covered in sand in every place imaginable. To supply the blaster, I borrowed a 25 gallon compressor from the IDC shop.

I hung up the parts using picture hanging wire and gave them three coats of the same etching primer used on Melonscooter space a half hour apart. With some of the lessons learned from Melonscooter’s frame, and a bit more advice from more legitimate painters, these parts came out far more even in the end than the scooter frame.

Next up were three coats of black (the same black, again, as used on Melonscooter since I bought like 5 cans of the stuff). Notice how I started during the daytime and it’s now the dead of night. There’s still some “orange peel” areas, but overall, everything dried totally smooth. I ran out of clearcoat, so DERPDrive won’t get the same crisp and shiny finish (But you’re never supposed to see it anyway…)

The finished parts after sitting in cooler, drier air for a day or two.

After the paint fully cured, I began adhering rubber strips to the front and rear of the structure, the parts which will be jacking on the van frame. These are some moderately hard (70A) and thin (1/16″) BUNA rubber strips I bought, being attached with contact cement. A thin layer of compliant material will aid in the attachment in a way two metal on metal contacts cannot – especially given that I won’t be able to torque down the jackscrews fully given that the van frame is still some pretty wimpy stamped steel rails. Again, if this doesn’t work out (like I start popping spot welds), I’m just drilling through everything and attaching them with rivet nuts.The C-clamps are to keep the adhesive fully engaged with the welded steel parts.I hope to assemble DERPDrive soon – I got into another one of those cycles of opening up multiple project threads, unfortunately…


The only work I’ve been able to get in on Melonscooter2 recently has been constructing and balance-changing the battery pack. I also prepared the motor controller, a KBS48121, and most other chunks of wiring for immediate installation. What I have been missing is the timing belt and pulleys – I ordered them last week, but of course waiting for shipping is the killer here. After I receive these parts, everything ought to fall into place quickly.

This is the battery pack in the middle of assembly. I waterjet-cut some 1/32″ copper bus bars for the task. One of them, to the left, has a chunk cut out of it to act as a last-ditch +250 Fuse of Oh Shit Amps. Unfortunately, I had used the wrong design equation values to make the cross section – I think this is actually good for something like 800 amps. Oh well…

Check the fully assembled pack. I added two 6S independent balance leads just to check cell voltages with for now – I hope this pack will be maintained infrequently enough that just cracking open the battery box and alligator clipping to it every few months is enough. Worst case, now I have one of these guys that I’ll make a balance lead jack for. These cells were in wildly varying charge conditions, so I had to spend a day or two just pushing buttons on balancing chargers, but now they’re all within 20-30 millivolts of each other.


Colsonbot… Colsonbot..

Does whatever a colsonbot does

Can he spin? Can he win?

No he can’t! He’s a wheel.

The Battlebots crew up here has reached critical mass. Full disclosure: The real reason for testing Mikuvan to New York City and back was so I can take it to Pennsylvania and back this weekend! The event in question is the PA Bot Blast, and the MIT crew will comprise myself, Dane, Jamison (whom I welcome to the MITrap), and freshly dragged into the craze, Ben.

If I thought trying to wing it up a bridge with only 4 people was bad, then climbing the Allegheny Mountains with four people and robots is going to be really adventurous!

Colsonbot has been in planning since a joyous all-hands dinner at Motorama 2013. Basically, the idea is to build an entire fleet of 3-pound “beetleweight” class robots and sprinkle them about the arena  as a “multibot”, or multi-part entry, to cause trouble and mayhem. Oh, and they’d all be shaped like wheels.  They would be otherwise functional “shell spinner” type bots, but the shell itself would be made of a popular robot drive wheel, the Colson Performa.  I was basically tasked with whipping up a “mass produceable” prototype which we can make a box full and show up to any event with.

I’m proud to say that’s now well under way. To extend this post even further, here’s the work that I’ve done on the Colsonbot front in the past few months. Bear in mind that this sucker has to be ready to run in like 4 days. Luckily, all the parts are on-hand and ready, so I’m only doing some mechanical assembly work.

The way I planned Colsonbot is as a design which could be a successful shell spinner on its own, if only I didn’t put such a silly bouncy rubber shell over it. The drive should be 4WD for stability and traction, and the weapon drive should be as reliable as possible, though not necessarily the most powerful. Under all reasonable circumstances, it should keep rolling! Basically its strategy is to get smacked repeatedly and just roll away.

This is the basis of Colsonbot, a 6×2″ Colson Performa wheel. Typically you’d find these on 30 and 60lb (if not larger) bots. They were a staple of the early 2000s 60lb and 120lb pusher wedge – they paired well with the popular EV Warrior motor and some power wheelchair motors, so they were used widely by new builders. Now that the new builder typically starts in a smaller (e.g. 1 through 30lbs) class, they are less commonly seen than their smaller brethren in the 2 to 4 inch range.

One of the first things I did was to core out the Colson to as far as I thought was reasonable. This process should be repeatable for everyone in on this build, so I didn’t try making any fancy contours. The main body of the bot was consequently limited to about 4″ diameter x 1″ height, with an extra nub on top where the hub of the wheel is normally molded.

Check out those molding voids – someone just did not care at all. Typically, injection molded parts are rejected if they contain voids inside – a result of gas bubbles evolving in the material from impurities or just shitty sealing. However, an industrial caster is hardly a precision application, so I guess this is fine.

The nub in question. I found that the bore of the wheel was basically ready for two FR10 bearing (flanged R10 bearing with 5/8″ bore and 1 3/8″ OD) back to back, so the shaft support was potentially great. I hollowed out the bore as far as I was comfortable with given the Colson’s pseudo-spoked core.

Cored vs. stock, with FR10 bearing. If you actually want to buy these, be aware they are rarely sold as “FR10″ (in the vein of FR8 1/2″ bore bearings, which are very common). Try searching G10 or FR2214 bearing instead. By the way, these are exact swap-ins for the horseshit bearings in common Harbor Freight wheels, like these or these (my favorite!)

This is where the fun part starts. Time to try stuffing an entire robot drivetrain into the hollow cavity of the Colson! The only motors short enough for the job were the Sanyo-type “micro” gearmotors sold by a number of places, including Pololu. Literally no other common robot motor (i.e. which we could all buy a bundle of) could fit, even in an “offset” 2WD application, while leaving enough space for the weapon motor and batteries, at least to my sophisticated (…apparently..) specification. I have my own doubts about how robust these very tiny motors will be given the high-impact application they will be in, but we shall see. I purchased a handful of 30:1 units for testing.

After some component shuffling, this is what I came up with. It’s actually shaping up to be a great bot. The four motors are placed in a nearly square wheelbase for best handling, and the weapon motor is off to one side. I decided on a spring loaded slide assembly to keep constant pressure on the shell, which has not been modeled yet.

The hardest part about this thing is the battery. I wanted to fit at least a 1Ah, 3S lithium battery into it, but sadly there were just no options available which could fit in the space required. I had to settle for a 800mah pack from Hobbyking, and even that (as you’ll see in a bit) was pushing it.

Wow, now we’re getting somewhere. I’ve designed this frame to be very quickly blasted off on a 3D printer. As a result, it’s actually the most product-like thing I will have built, yet. The body is all plastic with lids and snaps covering the important bits.

Now with more colson and other parts. The left part of the frame is where the motor will mount – it will be on a little dovetail slide assembly.

This is the mechanism modeled in more detail. I typically just model big blocks and geometric representations of parts until I get to them in earnest. The motor will have a “tire” made of rubber O-rings mounted around the outside. The motor in question is a Hacker A20-50S, first generation (i.e. without the obnoxious tailcone) that I have a few of thanks to my weird airplane friend Ryan. It was the only motor I could get in short order that was short enough yet had enough power. In the”mass production” Colsonbot, this will be replaced with an equivalent Hobbyking shady outrunner.

After the big mechanisms were settled, I began hollowing out cavities for other components and making wire guides.

Here’s a picture of most of the guts installed. The master parts list rundown is:

  • Leftover Turnigy Plush 18 for the weapon controller
  • Hacker A20-50S 1Gen for the weapon drive
  • Vextrollers for main drive
  • Hobbyking T6A receiver guts for the receiver
  • Z800 3S 20C pack for the battery

The center axle is a 5/8″ fine thread bolt with the head machined down for fitness and hollowed out for weight. I don’t think there will be any problems with robustness for the joint between bolt and plastic frame.

I’ve moved onto designing covers and plates here. The motors mount only using the frame members to clamp them in place. They’re square and of a known length gearbox-wise, so this was actually quite easy. It is the same system in use on Pop Quiz 2 to clamp its own 4 Sanyo-style micro motors.

With the battery cover done, it was fine to export everything as STLs and 3D-print all the parts in ABS plastic.

I popped these into a Dimension 1200SST and ran out the last bits of a cartridge with it. I would have tried this on our shop Replicator 1, but just had this sense of hopelessness from the amount of weirdly sticking-out parts.

Test fitting parts now. The motors snap right in – I could almost just run these as-is without the bottom cover!

One issue I found was with the 3/4″ Dubro airplane wheels I bought. I’d never drilled them out before – Pop Quiz 1 used the same wheels back in 2005, but with their stock 2mm bores. It turns out their hubs are no more than about 3.5mm diameter in the center, so when I drilled them to 3mm to fit the Sanyo-style micro motors, there was nothing left to drill and tap into.

Well damn. I quickly whipped up a set of 3/4″ o-ring wheels to be 3DP’d to get around this issue.

Remember the battery? Hobbyking’s dimensions should be considered to be +1mm in all directions in the worst case. I designed this battery compartment using their given dimensions, but when I actually got the battery, it didn’t fit!

Just barely, however. The heavy plastic wrapping they use to shield the pack against punctures sort of got in the way. So what do you do in this case? Cut the damn thing up and just use the 3 cells totally naked. Hey, they’ll have some thicker plastic armor once in the bot anyway. I intend to do this to the 3 packs I got for this thing as spares.

Colsonbot should be all together in the next 2 or 3 days, so definitely stay tuned for this one!



Ask an Engineer References

Jul 14, 2013 in Reference Posts, Stuff

In case you didn’t know (and you probably didn’t, given how un-announced it was here), I was invited to be on the Adafruit Ask an Engineer webcast as a special guest.

Essentially it was aiming to be a quick highlight of what  someone could do in the realm of mechanical projects, since the usual target of the show is very much EEs and software folks. I gave a rundown of some of the projects (the entire build histories of which are to the left in the sidebar) and answered a few questions about EV technology. One thing that cropped up was where to start reading or joining if you want to build mechanical things, and I really couldn’t provide a good answer on the show.

I still can’t think of one – unlike the Ardusphere, there’s not really any well known big names that ‘unify’ the mechanical hobbies, such as what Sparkfun and Adafruit might do for the modern EE hobbyist, since the scope of these is so much larger than open-source electronics. You’d generally have to break it down further, like ‘modified electric bikes’, or ‘battlebots’ or something, which is why I suggested with only a bit of irony for people to google search the thing they want to build. For instance, there’s pretty much a community for every make and model of car in existence.

But here’s a classic-me onslaught of links which should help at least a few people get hooked and started:

General Knowledge

  • Instructables: search for your desired project. Here’s the ones I wrote.
  • Making Things Move is probably one of the best primers for basic mechanical concepts and parts I have seen. Highly recommended (a copy of this is in our shop library, by which I mean a pile on my desk)
  • Gizmology and RoyMech: Two of my favorites for some more in depth analysis of basic mechanical components
  • Fundamentals of Design: Still one of my all-time favorites despite the diabetes-inducing writing style, because it drills down deep into theory and concepts for many different machine components. Unfortunately, it seems the book is now behind a MIT certificate wall, so I’m hosting it locally here.



  • Big Blue Saw (custom waterjet and laser machining)
  • eMachineShop and FirstCut (custom 3D machining – $$$)
  • Shapeways (custom 3D printing) and its upmarket competitor Materialise
  • I don’t know enough about Ponoko to say anything about it besides Ponoko.
  • is for the more pro designers who want to hire out manufacturing of just about anything.


The part that we touched on the least, one of the issues limiting people from engaging in building mechanical things is the lack of a well known toolchain to do so – for example, Eagle or KICAD. Most CAD systems are proprietary and extremely expensive; $10,000 is considered quite cheap and there are a few recent systems that start at ‘only’ $1000 or so. The free ones tend to be gimmicky and limited in functionality. Given enough time and practice, of course, these limitations could be overcome.

  • Autodesk gives 1-3 year student licenses to those with a .edu or verifiable educational relation. This is how I keep running Inventor forever.
  • Autodesk 123D design seems to be the highest design software in terms of accessibility * usability product. Autodesk has been pushing hard into the hobbyist market recently, so this might explain that.
  • Alibre (now Geomagic) was a low cost CAD system which was bought by 3D Systems recently.
  • Sketchup (formerly Google) is popular with some folks, though I found the 2013 free version frustratingly limited.
  • Open Source CAD platforms: I haven’t kept up with these so much recently since they inevitably require you to be a Linux hacker to even run, but in case you are one already:
    • FreeCAD seems to be the most together in terms of a functional parametric CAD software (i.e. not just mashing cubes into spheres)
    • OpenSCAD: I am put off greatly by the thought of “programming” your shapes, but to each their own! Many of the models on Thingiverse seem to be made with this.
    • Some people do use Blender to design mechanical things.

Parts for Designing

  • Thingiverse, a repository of mostly 3D printable models
  • 3DContentCentral: Affiliated with Solidworks, many industrial suppliers
  • GrabCAD: Full of pretty design models, not so much useful individual parts.
  • One of the great things about McMaster is that they have basically every nut and bolt and sprocket and gear and everything else as downloadable solid models. 50% of any of my designs these days are just downloaded from McMaster, pursuant to the art of design-around-what-you-have-fu.

There’s also my own references page which I’ve gathered all my rather technical or design-heavy posts, or posts that use or demo one specific part. I also have some downloadable CAD models uploaded.



Melonscooter 2 Epic Post

Jul 07, 2013 in Melon-scooter 2

Much work has been done on Melonscooter 2 since the last week’s update. In fact, it’s so much that I’m wondering if I should split it up into two posts or not. Since that would be against the tradition of my style of build reports, here; have like 40 pictures. Melonscooter2 is in a state where I essentially just have to put together the electrical system – assembling the battery pack, mounting the controllers, and then wiring everything up.

Last time I left off with the new rear “stern deck” just having been cut from 12 gauge steel and origami’d into shape. It was time to hit up the welders once more.

First step is to atone for my sheet metal sins via clamps. I’ll admit my sheet metal technique isn’t perfect, and that the equipment (an old enormous 10 gauge capacity box and pan type) is pretty clapped out. So, to make the tabs fit into the slots, out come the bar clamps.

After tacking the corners, I removed the clamps and proceeded to scientifically draw metal slugs using a MIG welder.

Next was rigging up the deck on the scooter frame. I had to make absolutely sure it was welded on straight – there’s of course no turning back after a certain point. I again used almost all the clamps in the shop and adjusted things little by little, albeit still visually, for alignment. The rear fork plates are actually not parallel (which could arise from manufacturing or the thing having been crashed at least once in its past life), so I was mostly relying on the existing tube frame.

After initial adjustments and tacking, it was time for some more steel loogies. I welded entirely around the outside as well as on the two sides on the interior.

Next, I turned the frame over and removed the existing starter battery box. It isn’t anywhere big enough to house an actual traction battery, so I would have to devise a custom solution.

The finished stern deck with wheel mounted once again. I had intended on “filling” the fold slots with weld material, but determined that it was pretty unnecessary.

I have a roughly 2.5″ tall space to put a battery box in before the ground clearance becomes too low to be worthwhile. Melonscooter’s lowest point was about 1.25″ fully loaded, which I used as a benchmark. I spent a little while thinking of the approach I wanted to take – custom folded steel box, use an existing steel chassis or some sort, or Landbearshark style waterjet-cut polycarbonate box. Speaking of batteries, I haven’t introduced them yet, have I?

This is a veritable cluterfuck of A123 B456 what letters are they on now?! 32157 type automotive-grade cells. They’re 9Ah apiece and will be arranged 12S, for basically the same capacity as Melonscooter’s former 12S4P A123 pack made of the 26650 type cells. I can’t reuse the Melonscooter pack, which is just fine and functional, because of the height of the cells impacting ground clearance. The cells will be arranged 12 in a row, necessitating quite a battery box that will almost be in line with where the deck ends with the front. I’m planning on equipping this row with some 1010b-compatible cell taps, but I’m also awaiting a shady BMS shipment from my new favorite sketchy e-bike parts store, eLifeBike. We’ll see how that works out first – if I can embed that board inside, then I can even obviate old Melonscooter’s once-per-semester cell checkup.

I decided to build a waterjeted polycarbonate box over folding up a new steel box. Initially, I had even looked around in area military surplus stores to see if any old ammo cases would fit my application. Unfortunately, the only boxes I could find were for 7.62 NATO rounds, which were too long (i.e. tall in the orientation I would need to use the box in). Whipping up a sheet steel box was hampered by the lack of reasonable steel sheets on-hand – I could go out and get several 22 and 24 gauge sheets from the likes of Home Depot, but that’s pretty thin. I’d also need to make a lip with a removable lid. Making up a custom RP’d box was in fact the path of least resistance since I had a spare plate of 1/4″ tinted polycarbonate left over from some robot project that never happened (or perhaps did not happen hard enough). I elected to skip this session of sheet metal lab for now.

In my How To Kinda-ish Maybe Build Everything Instructable, I have a page on “making boxes” – which is something done often with laser cutters and their ilk. So here’s how I went about making this box. Regarding “edge precedence” in the chapter/page, it’s which side I wanted to be able to remove/install the quickest. You could imagine (and be correct in doing so) that I would want to remove the bottom for servicing the battery. However, I decided that having a material-on-material interface for the top and bottom would be the best, since this thing will most likely be shaken violently up-and-down when it attacks the “meh, it’s not a sinkhole yet” roads of the local area, and that the only thing holding in the batteries being a few #4 scres would not be the best scenario.

So I made the two endcaps the “highest precedence” components – to remove the battery itself, I’d have to at least take off one endcap. It would make it less serviceable, but if I am removing the battery that often, something’s very wrong.

What’s that notch in the corner? It’s to clear the kickstand. I could have made a totally rectangular box for easiness, but I knew that other wiring components – distribution, charging ports, switches, etc. will have to go in too, lest they be haphazardly arranged external to this. So one side is about 2″ longer to hug the kickstand and house these parts.

Getting more together now. When I make these kinds of boxes, I make the solid shapes first, then decide which things to tab into each other. The cell models are now in, and I’ve also made a cutout for the Hella battery switch.

Continuing the box design, basically all the corners are accounted for.

And the last two pieces are in. The round hole will be where I mount an actual charging jack, probably made of an embedded Deans or XT-60 connector like on the other vehicles, in a small printed carrier. On old Melonscooter, I had to disconnect the battery itself to charge it. This is just one little layer of refinement.

I made a prototype out of laser-cut plywood just to test for dimensional sanity. Result: Satisfaction. The battery box stops at where the deck stops up front, and while it isn’t nice and curvy, at least it doesn’t stick way out. I moved a few things such as the charger jack around so it wasn’t obstructed by the kickstand as much. Four little ears pop up from the top of the side rails and keep the battery box sitting flat with respect to the frame. I made it this way so it’s easier to fixture the future welded mounts.

I tried to take one shortcut in making the battery box – trying to laser cut it from PETG plastic.

This resulted in what must have been  the most dismal failure I’ve ever generated on a laser cutter. PETG is often advertised as “halfway between acrylic and polycarbonate” – unlike polycarb, it can be laser cut, but not as cleanly as acrylic. And it’s not as shattery as acrylic, but not as strong as polycarb. Well, it also melts, smells like death, and turns yellow halfway as shitacularly as polycarbonate, and takes far more energy to melt than acrylic. It’s not that it wouldn’t laser cut – it just laser-cuts like total unshaven ass. And I suppose instead of smelling like death, it smells like terminal cancer or diabetes.

It doesn’t help that the Epilog 36EXT has an almost-useless gas assist system – instead of, say, a cone over the lens that focuses pressured air into a single stream, it just has a derpy little bent steel tube that kind of puffs on the cut. So, it couldn’t really clear the melted PETG material from its own cut. If I went slow enough that it cut through the first pass, then the melted kerf becomes enormous.

I ended up having to hammer a few pieces out anyway, before just totally writing off this sheet. Luckily, it was a leftover of a previous class run in the IDC space that I’ve been hiding, so I didn’t actually have to spend money on this wreck.

PETG. Not even once. (At least, not without a laser that has a high pressure gas nozzle….)

The battery box waterjet-cut from tinted 1/4″ polycarbonate. That’s much better!

To mount the battery box, I cut up some random steel strap things which were made of 1/8″ thick, roughly 1.25″ wide steel. I literally do not know what these were – they were found in a scraps bin at MITERS.

This was the pilot application of the all-new cold saw I commissioned for the IDC fabrication space.

The steel mounts will each have a hole drilled into them to mate with the battery, and the rest will be welded to the frame. The battery box will help jig up the mounts so all I need to do is tighten them vertically and clamp the whole assembly to the frame.

Here’s the battery box mount prepared after drilling and finishing.

And it’ll go like so!

Back to the welding room for some very quick beads. I clamped the box such that the plates were in position in order to tack them once. Then, to prevent melting the battery box, I’d remove it and finish the welds.

Tacked in place and battery box removed..

…and a fat MIG slug deposited onto each side. That does it for the mounts – this is all they are.

On the same waterjet run that yielded the battery box, I also took the chance to cut out new a drive pulley for the rear wheel. The X-Treme scooter came stock with a weird 8mm pitch chain that nobody uses anywhere except on derpy scooters. Favoring HTD timing belt drive, it was clear that I was going to have to replace this.

The bigger wheel is a double-ended drill bit, a modern cousin to the double edged sword. On the one hand, it’s easier to achieve a higher gear ratio for the same motor speed since the output stage pulley/sprocket/gear can be larger and not hit the ground while turning, but the fact that the wheel itself is larger mostly negates this – your ground speed is theoretically unaffected. What the bigger output stage pulley allows me to do if I wanted to keep the same gear ratio is to use a larger motor-side drive element. This has the upside of lessening the tension in the belt and bending it far less (the curve of a larger pulley is more gradual), and lessening the load per tooth since there are more teeth in contact in the angle of wrap.

Small pulleys and sprockets wear out their belts and chains much quicker because of the increased material flex and decreased tooth contact. Melonscooter was known for going through a drive belt every few months just from them becoming tattered and separating from their rubber backings and breaking the tension elements – these weren’t cheap unbranded belts we were talking about either, it was Gates belts straight from McMaster. I was using a 15 tooth pulley on the motor to transmit north of 1500W most of the time.

With this wheel, I should be able to retain my top speed while using a much larger 22 tooth motor pulley which will have nearly double the number of teeth in contact. I hope to get more belt life this time.  The motor side pulley will be a stock one I have sitting around from playing with gear ratios in the past.

The wheel side pulley, though, is something a little weirder. Notice that it’s made of chunks of pulleys. To save material, I split the outer profile of the pulley into 120 degree arcs, fastened to each other by a ridiculous number of bolts. This was a technique I tried out first last year on a Silly Media Lab Vehicle, and it worked very well. The nice thing about this method is you can quickly build up thick dished pulleys and other elements with rings, without going through 8 plates of metal and generating lots of thermally conductive round pot and dish coasters at the end (hence ruining the point of a coaster).

The critical part of doing this right is overlapping the segments on each successive layer such that there’s not a “parting line”, which would occur if the segments were just stacked one on top of another. That’s why there’s a million bolts around the edge – so I can shift each layer like 60 degrees. In the end, when everything is tight, all the material overlap will approximate a solid pulley to a degree more than what I set out to accomplish.

Here’s the pulley installed on the former sprocket perch.

I actually generated this pulley profile with a custom template part that I made in Autodesk Inventor because of two major reasons: first, nobody seems to make a CAD program that comes with a HTD belt generator. I can’ tell if it’s because HTD is a private brand or what (I, at least, use it to refer to every pulley that has rounded tooth profiles). Inventor has a “T” metric belt line which seems to be an ancient metric trapezoidal tooth profile. I even tried Solidworks (Oh boy, using the CAD program I’m supposed to teach to people…) and it, too, has tons of English belts but only metric T belts.

And second, nobody seems to commercially make a HTD pulley this big. The largest downloadable ones I found were about 70 teeth. My pulley is 108 teeth…

So I grabbed an image of the HTD belt cog profile and made a parametric part in Inventor. With a bit of nozzle offset magic, the belt wraps with no problems all the way around.

What’s left after welding everything that needs to be welded and making the wheel driveable again? Painting!

I’ve never been a big painter or finisher, but if I left the bare metal and welds untreated, sooner or later I’m just going to be riding a small hill of oxidation again. To paint over everything, I cleaned up all the surfaces and used a self-etching primer first on the bare spots. Unaffected paint spots near the welds were hit with a fine grit sandpaper befoerhand to encourage sticking.

Next up were a few coats of black acrylic enamel.

And finally a clearcoat. I have yet to master the art of spraypainting without the “orange peel effect” – a finely textured surface resulting from uneven spray thickness, droplet sizes, time-between-coats, etc. In this application, I don’t really care, but I would of course prefer not to generate it on the van.

It’s bad karma to paint indoors, so I did it the best way possible – right next to the 300CFM laser cutter ventilation fan. Outside that day was approaching 80% humidity – I felt like almost drowning just walking around outside, and my paint would have stayed wet for the next 3 years. The ventilator kept the funky smell from spreading to any other room.

After the paint fully cures, it’s time to start putting things together. First, I still need to assemble the battery pack itself, BMS or otherwise – given that I might not receive the BMS shipment for another week, I might just pitch together some JST connectors for my balance charger. Next, the slated controller, a KBS48121, needs to be mounted. Yes, this does entail putting sensors on the old Melonscooter C8085 melon motor (which I have since re-bearinged, so it should stop sounding like a sandblaster while running!)