Hub Motors on Everything: Johnscooter, the Testbed for Kitmotter 0002

This scooter might rank as both my fastest-built and shortest-lived vehicle. It was called into existence solely to test Kitmotter 0002 on the evening of the 14th of July, and lasted about an hour and a half. That should tell you that Kitmotter did not fare well under load testing, but it might not be for the reason you think.

The base of Johnscooter is…. well, a scooter frame left behind by someone named John in MITERS almost a year ago. Occasionally, people drop things off and swear they will come and fix it or upgrade it, but it ends up not happening because MIT eats them shortly thereafter, or they ran away to Nicaragua, or some other life event detrimental to finishing scooters occurs. John, if you want your scooter back, I’m done messing with it.

This frame is quite interesting. A little research led me to find that it was called the FX-1 EZRider, which is probably one of the more hardcore names I’ve ever heard given to something like this. They seemed to have been produced a decade ago and used nickel metal hydride batteries, and oddly enough, sold at Walmart. It’s very compact, with 6″ pneumatic wheels native (though I’ve been unable to find where John left his rear wheel), and weighs a bit south of 20 pounds.

I started tearing down the frame to extract the components inside, and I must say I am a big fan of the design. It’s not cheap and thin like a Razor scooter – there’s 5mm and 4mm aluminum all over the place. The frame comprises a single aluminum extruded tube about 3.5″ wide and 1.5″ tall, with the forks (made of 5mm bent aluminum plate) bolted through the sides. The same bolts retain all the internal electrical components. Up front, there is a solid endcap which seals the aluminum tube from the weather.

I’m also a fan of the combination fender brake. You can actuate it both with the brake lever on the handlebar or by stepping on it.

The batteries are in stick form and they slide right into the aluminum frame tube, surrounded by rubber. I was expecting these to be lithium, and therefore toast if they’ve sat for too long, but after dismantling a small portion of the heatshrink, I found that they’re actually 4.0Ah nickel metal hydride cells, of the 4/3 A size. That’s alot of battery. Each pack (there are 2) is made of 24 cells in two sticks, each made of a 3-long string of 4 cells in parallel. Then there’s two of them. That’s 230 watt hours of battery, a number that is only rivaled in this volume by using prismatic lithium polymer or LiFe packs. Seriously, I could not make an A123 pack this good.

As far as I can tell, each pack is only 14.4v nominal, and they are used in series to run at 28.8v. To my surprise, the whole pack charged right back up and peaked at 33 volts after sitting for Robot Jeebus Knows How Long.

Some times you just can’t beat a little bit of old school technology. Unfortunately, for me to actually but 230Wh of those cells is about $140 (nearest match) if I don’t mind something generic, and a cool $570 if I want Sanyo cells. Nickel is still classy, and now you understand why I consider the advent of Hobbyking batteries to be a pivotal moment for the creation of actually useful small vehicles by hobbyists.

Now, this is actually an aftermarket pack – John mentioned when he visited that he purchased this scooter from a builder in California, nicknamed Deafscooter. I’ve actually snooped around Deafscooter’s work on the EVAlbum back before I built my first EV in 2007. In fact, I am fairly certain it is this exact scooter, because the date of sale matches roughly with John’s description. Something about going full circle…

To mount Kitmotter, I decide to cheat a little and just use the axle hole that was already in the rear forks. This involved end-drilling and end-tapping the shaft. Another solution using only ‘garage tools’ is to drill out the forks’ axle holes to 5/8″, the shaft diameter, then drill and tap set screws perpendicularly. I decided to just go with easy since testing the motor was a higher priority.

I cleaned up the battery pack wiring a little (putting them permanently in series) and then stuffed the now-charged batteries right back into the tube from whence they came.

And on the Kitmotter goes! Very little engineering occurred here: Kitmotter had a 3″ width between shaft end faces, and the scooter fork natively spanned 2.6″. Solution: Mash the bent plate forks in a vise until they became unbent.

A view from the other side, showing the quite sophisticated brake mechanism – the little off center thing at the bottom actually rotates with the brake cable, giving the brake lever variable leverage as it is depressed. Did I mention I love this frame design?

For sheer simplicity and speed of setup, I plucked a spare Jasontroller off the table. Jasontrollers handle hub motors very well (almost like they were designed to do it or something…) and Kitmotter2 is indeed a hub motor – many poles, low speed, high resistance and inductance.The throttle that came with the scooter was a generic thumb lever type that was spliced right into the controller.

And here it is!

I replaced the front 6″ pneumatic wheel with a spare 125mm wheel because it otherwise caused the thing to look very strange, and ride ‘nose up’.

Ride testing of Johnscooter indoors revealed that Kitmotter’s hardboard sides seemed to work just fine. I took it through a few bunnyhops to make sure. The #2-56 studs and nuts seem to take a little time to ‘wear in’ the hardboard surface – more washers and Loctite may be necessary in the future. Even though the wheel was bored manually, it still came out reasonably on center because of the hole saw pilot bearing hack, and the wobble wasn’t noticeable in riding.

It was, however, a little underpowered because I never actually wound it to produce torque, being a demo motor and all. I could have definitely put another 2 strands of 28 gauge in parallel and lowered the resistance, or just put more turns on to attain a higher torque per phase amp. Or both. The stock Jasontroller puts out only 25 amps, limited by hardware, so it wasn’t too much.

Here’s some indoor ride testing of Johnscooter:

Next, we took Johnscooter to Ye Olde Silley Vehicule Proving Grounds, the garage (riding the whole way just to abuse the hardboard endcaps some more, just in case). Since it was comically slow, there’s no test video, but it finished our defined test interval in 120 seconds while consuming 10.9 Wh of battery, for an “action score” of 1308.

This is actually not bad at all given its non-optimized motor. For comparison, RazEr Rev finished the same test in 82 seconds while pulling 11.1Wh, in part because it generates so much more torque per Jasontroller-limited amp. Pneu Scooter offers a more fair comparison, having performed a 99 second @ 13Wh run previously (1287).

Unfortunately, Johnscooter only lasted for one run. After the first, the motor was very unhappy – the heat could clearly be felt through 1/4″ of wood, which is a pretty damn good insulator. After letting it cool for about 20 minutes, I tried a run, but only made it up 3 levels before it totally cut out. After that, the response seemed to be intermittent, which I initially blamed on using the aged battery too roughly. As some of the other guys were riding it around, the motor cut out again and this time it felt like it failed short.

Post mortem analysis once Johnscooter was back in the shop revealed…

…that Shapeways’ “White, Strong, Flexible” is definitely not “White, Strong, Flexible, and Heat Resistant Too”. Basically, it seems the nylon stator hub heated to the point of losing structural integrity, stripping its D-flat against the steel shaft keyway and causing the whole thing to shift. This pinched and shorted the wires against the shaft.

Additionally, melt lines can clearly be seen at the stator interface. This thing definitely got hot – though not hot enough to damage the magnet wire, as it was neither discolored nor smelled funny.

But it’s clear that I can no longer suggest 3d printing a stator hub as a viable solution for anything but the lightest duty motors. I’m thinking now of doing a wood (yes, more wood) laser-cut stacked hub that uses the 3-eared shape of the stator bore fully, and also grips the entire depth of the keyway. Yeah, sure, make fun of it for being more wood, but it wo…..uld have a higher temperature resistance, especially composite wood-like substances like hardboard.

I will put Johnscooter back together – it worked very nicely for what it is, and Kitmotter needs far more abuse heaping before I can confidently tell other people to copy and paste.

Kitmotter 0002: It’s made of wood!

Two weeks ago I said that I would build the redesign of Kitmotter in the next 2 weeks. Well, like any good college student, I did it the day (and night) before it was due. So in another installment of the summer of short, one-day builds, I present the NEW AND IMPROVED Kitmotter, now with 99% more wheel.

Most of the design intent of Kitmotter is centered on making it accessible to people who do not have a shop full of machinery at their disposal, and this design was definitely a leap in the correct direction. It was finished without the use of a mill or lathe…or even a drill press for that matter, since all the holes were laser-cut and did not need finish drilling.  The stator was harvested from a HP Laserjet 8000 series main drive motor (copier motors spreadsheet here), and the wheel was cored manually with a hole saw that had a 5/16″ pilot drill swapped in place of the 1/4″ one. The shaft is a pre-cut chunk of 5/8″ keyed steel drive shaft that McMaster sells to you for one hell of a value-added markup, but at least it’s better than buying 6 feet of drive shaft to use 3 inches of it. The hub that mounts the stator onto the shaft was hired out to Shapeways to be produced in laser-sintered nylon plastic. And the rotor, the hardest part to make conventionally, was hired out to Big Blue Saw (for real this time) as a stack of steel plates based on the original “kitmotter principle” seen in Pneu Scooter and Kitmotter 1. Other than that, there’s no special hardware.

As I mentioned before, Kitmotter 2 is a prime candidate for my next DIY vehicle centered Instructable, but there are some unresolved issues with this version that I might try and take care of with another one. But in the mean time, here’s how it was built.

Over the past few days, my order to Big Blue Saw which consisted of both waterjet cutting in 1/4″ steel and laser cutting in…. wait, is that particleboard?! Yes, in the interest of not making it cost a billion dollars to try the idea out (the minimum charge for waterjetting is about $80, plus or minus), I elected to make the side plates out of wood. None of the plastics were really appealing in terms of mechanical strength, and wood is some times underappreciated as a material. I figured I would make it out of the most plastic-like wood, Masonite/hardboard, or high-density fiberboard. This stuff is pretty fantastic in compression because it’s basically a solid brick of cellulose.

Only downside, I suppose, is that this Kitmotter should never be operated in the rain…

The rotor rings were split in half and arranged together with “sprues” so they were one closed profile and could be cut without falling into the tank or wasting huge swaths of material in the center (one of the downsides of this kind of design). Because there are so many perimeter screws, alignment and concentricity shouldn’t be an issue.

The motor’s design constraints (namely the need to only increment the axial thickness in 1/8″ and 1/4″ steps) meant it had to be about 2″ wide, which is wider than they make #2-56 screws long. I had to make meta-bolts using #2-56 threaded rod chunks and locknuts. Which, by the way, McMaster will also sell to you.

I ‘laminated’ the wood endcap together with slow-setting CA glue in between the layers, using the bearing as a centering jig. Slow-settingness was critical to this build because of the extra time you have to push the parts together, and if needed, align them radially. Slow-setting adhesives also tend to be stronger, and the motor is made of wood.

Putting all of the endcap bolts in forced the wheel mounting flange to be ‘geometrically averaged’ so it was the least out-of-round.

The rotor is built up similarly. The screws are positioned such that they are internally tangent to the rotor’s outer circle. They should never be seeing any cyclic ‘wheel loading’, unlike my through-the-wheel bolting scheme in the original Razermotor.BBS’s waterjetting tolerances are fairly typical of standard waterjetting – one side is on dimension and the other side is usually 0.003″ to 0.005″ bigger. I pre-compensated for this in the size of the slots, so even with standard waterjetting the screws could still pass through.

The “taper free” waterjetting is more expensive but can produce true square ends to better than 0.003″ tolerances.

Notice the irregular spacing of the ‘seam’ between the two semicircles of the rotor. This is once again an exercise of geometric averaging – by rotating where the seam is throughout the stack, I not only make sure there’s not a single ‘weak spot’ in the whole rotor, but also average any inconsistencies the waterjet may introduce. While a professional shop is going to keep their machine running pretty tight, waterjetting is still fundamentally machining something with a wet noodle (especially noticeable on hard corners using non-high-quality settings).

Here’s the stator assembly. I used a RH7-5219 core this time, but that particular hub also fits the Laserjet 8150 motor (RH7-1260). There is a flat in the hub to grip the keyway of the shaft (but leaves the keyway itself open to run wire outside). The whole assembly is pressed together and sealed with CA glue in the middle. A thinner CA was used here so it wicked into the semi-porous top surfaces of the laser-sintered nylon.

The nylon parts are easy to press fit because their outer surface is not smooth and still a bit powdery. Not only does a disturbingly cocaine-like powder fall off them as you press, but the added compression hopefully helps part retention…

(say, can you laser sinter cocaine?)

I’m going to skip over the gory details of winding for now, since the process was the same for this as the Chibikart motors (an Instructable would go a little more in depth about how painful it is). I decided to just use the hex-28 gauge rig I put together for Chibikart, though these motors can definitely stand another strand or two in parallel, which would also cut my resistance. There are 30 turns per tooth (120 per phase).

I did a fairly standard hex-28 termination and ran the wires, heat-shrunk for insulation, straight out of the keyed slot. That’s why I chose this drastically oversized shaft – the stock keyway is enough to pass some real wires through. If I increase the wire size, though, this may no longer be true, so I’ll have to test it out anyway.

Now, how do I install this thing? 5/8″ is too big to grip inside a drill press chuck to safely lower it in. I actually have no good answer for this at the moment, but I know it is definitely not “grab it and try to hold on”.

The best I could do right now was to ‘angle’ it in using  a set of giant channel-lock pliers to grip the shaft, then wiggling it until the other side centered in the bearing and slipped through. Because this motor is a fairly low aspect ratio (pancakey), I could do that given the loose airgaps. Even this was a bit of an adventure.

Now with wheel installed. The other endcap doesn’t support the wheel in any way, but is larger in diameter than its bore, so it will at least stop the wheel from falling off.

And all closed up!

I decided that the little short wire stubs were unsatisfactory, so some 16 gauge noodle wire was used to extend it. The termination is 2mm bullet connectors, my new favorite after 4mm bullet connectors – 4mm is just a little ridiculous for 16 gauge wire.

Alright, now to dyno it so I can find out its properties:

Poor scope.

Basically, RazEr REV2 was acting purely as a speed source here, with the scope on averaging mode to get a cleaner reading. I’d run the motor up to speed for a few seconds, then hit stop to capture the waveform. Hey, this thing was impossible to “lathe-o-mometer”, so I had to think of a way around it. Here’s the resultant waveform:

V peak-to-peak of 16.8 volts at speed 100Hz – that works out to a Kt in V/(rad/s) of right around 0.191. Not bad, and it makes sense given the motor geometry. For comparison, Chibikart’s motor is the same stator height but  73% of the radius (50 vs. 68mm), about 50% more airgap (for a decrease in stator surface flux of 87%) and has 90% of the turns (27 vs. 30). Just direct scaling from these factors alone from this motor gets me (0.19 * 0.73 * 0.87 * 0.9) = 0.11 Nm/A….. which is exactly what I found.

There you have it, the beauty shot. Now I need to verify its durability by shoving it onto a vehicle and riding it, but for the time being, it spins quite nicely I SERIOUSLY PAID MONEY FOR THOSE BEARINGS? Seriously, watch the video and listen for the bearings.

They’re the cheapest, “not rated” grade of bearings found on McMaster. Now, I bought them because I figured if I was going to tell someone else what to buy, the more goods on McMaster the better. I’ve had good experiences with the “Not Rated” bearings before, but it looks like cost-cutting has taken their toll on these things, because holy shit they’re bad. Built-in radial backlash and tons of axial wobble. They’re lawn mower or handtruck wheel bearings, not electric motor bearings – hell, the handtruck wheels we get to harvest the tires from have better bearings.

Unfortunately, I was mostly after these bearings for their flange, which makes installation easy. They don’t seem to make many precision bearings in this size – only for said lawn mowers and handtrucks. I might have to just deal with that.

Stay tuned for the next episode, where I ride Kitmotter in the rain and the MDF sides melt!