October 29th, 2010 by peter
The major question when building the 72 volt folding electric bike was: will the geared hub motor survive 1500 watts in the long term?
The Bafang geared hub motor contains 3 nylon gears… yes, plastic. Run within it’s limits, those gears should be fine. But this motor was being run at double the normal voltage.
To hopefully avoid these problems, a metal gear was installed in our motor.
Today, after about a year of use, I decided to open up the motor and assess the conditions inside. I can report back that everything looks beautiful inside!
August 7th, 2010 by peter
We are proud to release a video about the folding bicycle prototype I’ve been riding this past year. It’s been by far the best bike yet. Not only does it have ridiculous amounts of power, but the weight is finally below that magic threshold where it get too difficult to lift– roughly 45lbs. It was ridden through San Jose’s Indian summer heat and the pouring rain of winter without a single hiccup (hint: the electronics are wrapped with rubber).
We’re looking at building the next iteration, where we’d like to tame the top speed but improve hill climbing performance, a handy feature in San Francisco. Stay tuned!
Without further ado, here is the video (you can also watch it directly at Youtube here in full 1080p HD):
We had a lot of fun filming this video and spent many hours editing. Thanks to Mike Steczo also for the sound track! Also dear readers, note the homage to Jeremy Clarkson near the beginning…
April 29th, 2010 by peter
So with all the difficulties of the original RC motor driven bicycle concept (particularly the clearances to make the transmission work), we decided to take a simpler approach for the time being: use a tiny driven wheel. A small wheel needs to spin faster for a vehicle to reach the same speeds as a larger wheel, so the high RPMs of the RC motor becomes an advantage rather than a hindrance when the wheel it’s attached to is very small. Instead of building a whole new frame from the start, the fastest solution was to reuse an existing vehicle…
The stars aligned in this case. Earlier, Eric T. graciously donated a red Chinese stand up scooter and Brian T donated some scooter motors with belts and cogs. Todd D. had also donated a Razor E150 scooter earlier, but the frame was too small for an adult and it lacked pneumatic tires. What does an EV with solid rubber tires feel like at 15mph? Once my teeth stopped rattling I had the good sense to throw the scooter in the back of the garage until I could find a use for it.
Well, that time finally came: many of the parts were interchangeable and the end result was all the parts (besides electronics and a motor mount) to put together one working scooter.
I used Ben’s TIG welder to make a motor mount out of aluminum and then modified the scooter frame to accept it.
The performance was tested using a sensorless ebike controller and it was nothing to write home about. Hmm.
Now, RC motors lack the sensors ebike controllers require for reliable low speed operation. RC airplanes don’t need low speed torque, so the sensors (which can often be problematic) aren’t built into the motors. Some larger RC motors can be modified to accept the sensors, but this wasn’t the case with our motor.
We tested a hypothesis: that external mounting might be a feasible solution. It worked! We attached the sensors using blue tape on a work bench.
So Ben used his CNC mill to build an external mounting. The positioning of the sensors is critical, so the holder has the ability to be adjusted, much like an automotive distributor, for proper placement.
Now it was time for some on road testing…
Holy torque! The scooter will simply throw the rider off without gentle throttle from a stop. The original Ecrazyman controller had some issue above 75% throttle, however, full throttle is only necessary for a higher speed. Top speed: not too much.
The problem with the Ecrazyman controller has been reported elsewhere: the power suddenly cuts out above 75% throttle until the throttle is rotated to totally off. Strange.
I ordered another completely different controller to test whether it is simply the fault of the controller design.
December 21st, 2009 by peter
While initial work on converting a bicycle to use a high speed motor went well, little details came back to haunt the prototype.
- Although a great amount of attention had been given to making sure the go kart sprocket that was bolted to the Sturmey Archer hub didn’t cause chain interference problems and the newly built up wheel fit in between standard drop outs, one detail was overlooked: adding a sprocket on the drive side caused the rear wheel to be offset to the left just enough that the axle did not protrude sufficiently from the drive side drop out to thread a nut. Gary on Endless Sphere forums who did a similar build earlier in 2009 used a steel framed bicycle. Our bike uses a lighter aluminum frame. Aluminum is less dense than steel and so the drop outs are about three times thicker. Oops.
- The magnets in the RC motor are very strong and picked up metal shavings from the shop. These shavings got into the gap between the outer rotating magnets and inner stator. To avoid damage to the motor, the motor was disassembled and cleaned using duct tape and fine tweezers. It is advisable that if the motor will be stored in the same room as metal working equipment such as grinders and mills that all openings be taped off or the motor be stored in a sealed box or bag. Wash your hands before handling it. These magnets are STRONG!
- The sensorless Shenzen Sucteam controller was damaged yet again when applying full throttle from a stop. This is the third failure so far. The usual failure mode is that a MOSFET (an electronic switch) on one of the three phases fails and gets stuck on or off. This manifests itself as a stuttering when the motor turns. The 3 phases are arranged around the hub so that each phase covers 120 degrees of the hub. The controller should turn the phases on and off to make the wheel rotate, but instead a phase is stuck on or off and either the motor shakes back and forth or has no power in part of its rotation. The Shenzen Sucteam sensored controllers have been rock solid so far, but RC motor’s don’t come with sensor. There are only two solutions: a) find a more reliable controller or b) attach sensors to the motor.
September 29th, 2009 by peter
Problem: The original rear end of the compact tandem was actually too short. The rear axle was just a couple of inches forward of the passenger’s center of gravity. This was enough to introduce an unnerving wobble at low speed. At least, that was the hypothesis.
How do you test such a thing?
I had a passenger lean forward in the rear seat to shift the center of gravity forwards. This seemed to make a difference, but to get a more noticeable result I also had the same passenger sit backwards on the handlebars. Hypothesis confirmed: the wobble was gone.
Next question: how do you extend a bicycle’s wheel base?
The solution: cut off the old swing arm and fabricate a new, longer swingarm.
Using a large triangle hanging from the passenger seat and measuring tape, it looked like an extra 8 inches would be plenty. I decided to add 10 inches for good measure.
Since some guests were dropping by the next day to ride the bike, I had to get the job done quickly. I just threw the whole thing together out of 3/4 inch “everyman’s tubing”– electro mechanical tubing (EMT)– available for about $4 per 10 feet at Home Depot. Now, a swingarm attaches to the frame via a pivot made of two tubes and a plastic bushing and rather than pull the whole rear end off and build another pivot, I just chopped the old seat and chain stays off and welded new ones on. A jig would have been very handy and produced much more accurate results, but I had one evening to finish and didn’t have time to make a jig.
One other annoying thing about the previous seat and chain stays was that they were angled outwards to make room for the bicycle’s pedals. This is a good design for an upright bike because otherwise the stays would interfere with the pedals, but since this bike is a recumbent it just made mounting the motor obnoxious. Standoff washers were needed to get the motor (which attached to the stays) in the same plane as the wheel. Getting the right motor on and off was an aggravating trial and error process.
The jig would have helped with the fit up of the various tubes. There were some large gaps to fill which is usually inadvisable using braze rod, but there was plenty of surface area and braze rod to go around so it worked out OK.
The precision of the final construction will surely inspire confidence in passengers. Anton W., visiting from Ohio, was the first test passenger and to his credit he didn’t leap off at the first opportunity. On a whim I also rewired the bike for 72 volt operation. At 72v with a Bafang hub motor the bike really hauled with much better performance than its previous 36v incarnation and without that annoying low speed wobble.
Next time, I’ll upgrade the Big Lots trash cans– er, luggage.
September 24th, 2009 by peter
Most electric bicycles are built on an existing frame or a slightly modified one (usually to provide mounting points for the battery and motor). This is the approach we took with the compact tandem.
The problems with using existing bicycle frames for electric bicycles:
- No convenient mounting for the motor or batteries
- Motor and electronics exposed to the elements
- The medium quality frames are not as overbuilt as the cheap ones and aren’t up to the stresses of high performance ebikes, relegating most home builders to using very poorly built department store bikes
The lack of convenient mounting for both motor and batteries comes down to the frame being essentially two dimensional. This problem is especially acute for the motor. When a motor is attached to a round pipe, the torque of pulling a chain tends to rotate it around the pipe, loosening the chain and causing all sorts of problems. This can be overcome by mounting the motor in a location where there is a 3rd attachment point available. Three points are needed for stability– imagine a stool for example. Bicycles have two such locations, the triangle made by the chain and seat stays and seat tube or in the main triangle between the riders legs. The rear end mounting exposes the motor to the abuse and the chain and seat stays are usually not parallel to the wheel, but rather at an angle, making mounting somewhat obnoxious. The in between the leg mounting is kind of scary because it’s quite easy to stick your leg in a spinning motor’s chain. The batteries also have nowhere good to go except behind the seat, which is why so many Chinese ebikes have the rear end stretched out. Not really the ideal position because their also tend to mount the motor on the rear, leading to a very unbalanced machine.Let’s look at the other end of the two wheel spectrum: motorcycles. Motorcycles have frames that are roughly like two bicycle frames with cross pieces holding them together. This provides a convenient place to mount the gas tank and motor, which are both much heavier than on an electric bicycle (and more powerful). The frames are also much stronger thanks to this design. But, especially when built of big thick steel tubing, they’re heavy. Since they have humongous gas motors, it doesn’t matter very much. Most motorcycles weigh over 300 lbs. Forget picking it up. At that weight, you just want something that can be heaved upright after it falls over. I don’t want a bike that weighs that much, do you?
Since most electric bikes are already too heavy to casually lift (about the same weight as a bag of cement-50+lbs), throwing a bit more metal onto the frame shouldn’t really make any difference.
Also, most bicycles are of the upright diamond frame type, where the pedals are located in the middle of the frame. The width of the frame is limited by the bottom bracket width and the spacing between the pedals (there’s actually evidence that widening the distance between the pedals is good for your knees– search Google for Q-factor). On a recumbent, however, there is much more flexibility since the pedals are located at the front of the bicycle and we have room to play with on the rest of the frame.
So how about a motorcycle style frame but with the tubing sized for bicycles? It’ll add a bit of weight, but hopefully make up for this deficiency with the benefits of secure storage and better weight distribution. And who knows, maybe if it’s built out of aluminum and the rest of the parts are half decent, it’ll actually weigh less than many cheapo ebikes. Let’s rock and roll!
September 9th, 2009 by peter
After over a year of effort t I have finally built a boost circuit to charge an EV battery pack from a lower voltage. It’s been frustrating, expensive, and slow but now we’ve got something that works– not great, but a place to start. It has been an exercise in patience, applying systematic processes and electronics gear acquisition.
A boost circuit uses an inductor (basically, a coil of copper wire) and a power MOSFET (an electronic switch) to increase the voltage of the electricity that enters the circuit. Whenever the current flowing through the inductor decreases, the voltage of the electricity coming out the other side increases temporarily. By turning the MOSFET on and off quickly, the voltage can be regulated to something higher than the input. This is important because most motors/generators need to spin very fast to generate enough voltage to charge EV battery packs. This is often impractical to do mechanically: a pedal powered generator needs to be pedaled faster, a wind generator needs to have more wind, a solar panel needs more light.
I’d hoped initially that I could just throw together something I found on the web and off I’d go, but of course it’s never that easy. I scoured the web and selected a Linear Technology chip to be the brain of the operation. Linear provides a decent modeling tool (LTSpice) and a huge datasheet. My initial enthusiasm was slowly eroded with each hurdle. First, Linear only provides the chip in TSSOP format rather than hobbyist friendly DIP. This is an itty bitty chip that was impossible to solder with my $30 Home Depot soldering iron. Of course, the next natural step was to buy Ben dinner and several margaritas later, Ben had soldered a few LTC samples to some TSSOP to DIP converters he had lying around. I fiddled with LTSpice until I built a circuit that worked, at least in the simulator. The final hurdle with the simulator was that the LTC datasheet had an error, sizing the startup delay capacitor thousands of times too large so that the chip took several minutes to startup. I never waited long enough for it to start working, so I thought it was broken. I built the circuit on my protoboard, hooked it up to a 12v battery, and POOF! the protoboard melted. I bought another protoboard from Jameco, and to prevent a repeat of this incident, a Mastech power supply with a current limit. Ben and I analyzed the circuit from every angle on a couple of Saturday afternoons with no luck– it would work for about 5 seconds, and then the LTC chips always fried themselves. One more dinner and some homebrew later, Mike professional stepped through the circuit methodically and couldn’t figure out what was wrong. Instead, he sold me his Tekronix oscilloscope and wished me luck.
Since Ben and Mike were stumped, I appealed, on advice of the AVR Freaks forum, to Linear themselves.
Amazingly, they sent out two application engineers.
Not surprisingly, the engineers were more interested in my collection of electric guitars than my circuit. After examing a few guitars, they took one look at the protoboard on the kitchen table and said “that’s never going to work.” That was the last I heard of them.
I took a few more stabs at making the circuit work. One evening while I tried to solder a Linear LTC to a converter board, the $30 soldering iron burned two of the $7 chips and $15 adapters to a crisp and I ran out into the garage in a fit of rage and smashed it with a hammer. I gave up on the LTC chip.
I took a break from the circuit until Mike came over one day and handed me a rough schematic of the boost circuit in his battle bot battery charger. It outputted 70v and was nothing more than a 555 timer, an inductor, and a few other common parts. I built the circuit in Electronics Workbench and it didn’t work. I’d just started a new job 30 miles away in downtown San Jose, so each day for two weeks I read on the train about circuit theory and how boost circuits work, and built the circuit, piece by piece, in Electronics Workbench with Mike and Ben’s help over email. It was a fun and exciting time and then came building the circuit on a protoboard. Once again, I broke the problem down into lots of smaller problems and tested each piece before integrating it into the whole. First, just the timer circuit, then, just the MOSFET switching, then, the drivers for the MOSFET gates. This process was slow but much more predictable– each portion took one or two days and integrating usually took just a few hours for each part.
After this string of good luck, something had to bring balance. I was tired of the 555 timer chip, so I used my one good piece of lab equipment– a fancy HP function generator– to generate the fast on/off signals for the MOSFET. Well, such a sensitive piece of equipment is not meant to drive a crude thing like a power switch, and while I was microwaving dinner in the kitchen I left it hooked up and running and something blew up inside it. The oscilloscope seemed to wilt at this development and the trace– the little line that runs across the screen– expanded until it was as fat as the big toe on my foot. I twisted the knobs but it too, was broken.
Mike and I spent two evenings troubleshooting both pieces of equipment and figured out what was wrong with each– but it was still frustrating.
Still, within a month, I had the whole circuit working.
Then I tried to solder it together using my Wen soldering gun (the sort of thing you might use for plumbing) but this was like trying to play a piano with my feet.
I caved into Mike’s suggestion and purchased a professional Metcal soldering iron. After this arrived, building the circuit onto a board went quickly and smoothly. I had to modify the size of the timing capacitor for the circuit to work at higher power.
Finally, last night, I tested the circuit. It outputs 28 volts for input voltages over 10v. Success! Without tuning it at all, the efficiency was between 25-60% from 10v to 24v. These efficiency numbers aren’t very good, but they’re better than the 0% efficiency of the previous setup with the generator hooked up to the batteries directly and with further tuning, my calculations lead me to believe that efficiencies of 80% could be a realistic goal.
Now I’ll go and start that tuning…
August 19th, 2009 by peter
I am working with several folks to complete a RC motor driven e-bike. Is a 2 pound, 3000 watt motor spinning at 10,000 rpm a good fit for e-bikes? We’ll see.
Using RC motors for propulsion is the latest craze in the e-bike world. I first became aware of the power of these RC motors when the guy from EVPlasmaman.com rolled up to our booth at the 2008 Maker Faire (forget his name now, sorry). His little aluminum scooter could wheelie effortlessly using an AXI 5345 motor designed for model airplanes.
Around the same time, Matt Schumaker out of Illinois posted his e-cumbent bike to the IHPVA web site. A discussion started on the Endless Sphere forum about that posting, so I emailed Matt to let him know about it. He’s now gone on to refine the e-cumbent and a number of forum members have built similar drives. Matt also sells the final version of that unit, a CNC milled two stage reduction transmission.
How do you build a motor that spins slowly? With more copper windings. Unfortunately a lot of copper is necessary to get wheel speeds. This is why the very common direct drive hub motors are so terribly heavy (20-50lbs). They spin at wheel speeds– about 300-450 rpm. Most motors in tools and other household items spin at about 3500 rpm. Motors like the Kollmorgen that I used on several of my bikes spin at ~3200 rpm too. Gears are used to reduce the rotational speed to about 350rpm at the wheel. For example, the compact tandem has an 11 tooth sprocket on the motor and a 72 tooth sprocket on the wheel. For each complete revolution of the motor, the wheel only turns 11/72 of a full revolution. So when the Kollmorgen is maxed out at 3200rpm, the wheel is spinning at about 490rpm, which is about right.
The Kollmorgen weighs a fraction of the weight of a direct drive hub motor– something like 5lbs. RC motors weigh from just a few ounces up to 2lbs. But they spin at 10,000 rpm! The gearing on the compact tandem would result in a wheel speed of 1500rpm or a road speed fast enough to keep up with traffic on the freeway. While that might be entertaining for bragging rights, the acceleration of a one speed transmission with such a tall gear would be so horrible that you could never actually reach that speed without being pulled by another vehicle.
Reducing 3500 rpm to 350 isn’t too difficult– two sprockets and a chain can do that. But reducing 10,000 rpm to 350rpm is quite a bit harder to do in a compact package. Also, the stresses of the extremely high rotational speeds and high power output mean the system has to be carefully designed. A popular but complicated approach is a 2 stage reduction using two separate sets of belts or chains. The first stage reduced 10,000 rpm to ~2,000 rpm, and the second stage to wheel speed. But the experience of Matt and others has shown that building such a reduction can take a trained machinists weeks of full time effort.
Why not build a single stage reduction? In order to get the necessary 50:1 gearing, one of the gears would need to be much bigger than the other. But this isn’t actually a totally crazy idea. Back in the first few decades of the 20th century, motorcycles were built in exactly this way, with pulleys attached to the rear wheel that were nearly as big as the wheel itself.
But is it practical? Chains only come in several standard sizes. A sprocket with the necessary number of teeth to fit a bicycle would need to use a chain with a very tiny pitch (pitch is the distance between adjacent links)– something like the No. 25 chain I used on my first e-bike. This chain is very weak and breaks easily. Stepping up to the next size (no. 35) is a problem because the increased chain pitch would require a sprocket larger than the wheel itself.
One solution is to use belts. Belts come in greater varieties than chain at the needed size. It is easier to make a pulley than a sprocket, since a pulley has no teeth that need to be cut out. For a flat belt, any round object with sidewalls would work, even a bicycle rim. I spoke with Matt and he thought non toothed belts would work just fine.
The next problem is mounting the pulley to the wheel. Motorcycles of old used different methods to attach the pulley to the wheel like bolting the pulley to the wheel rim or tying the pulley to the spokes of the wheel. A modern gasoline bike conversion kit, the StokeMonkey, uses the spoke mounting method where the pulley has slots cut in one side so attach to the spokes of a bicycle wheel.
This is the approach I will be taking. It’s almost guaranteed to work and it’s simple.
In a future posting, I hope to talk more about the electronics side of using these high speed motors… so until next time!