The Derpbike Conclusion: more volts more problems

This backlogged post concludes the Brems-Saga for the time being.  It was an excellent journey into motor control and hybrid-car-parts hacking, but was ultimately doomed by my inability to automatically balance the terrifying 400V battery.  dang!  cell balancing is kind of the worst.

OK! It's been a while.  I've been working on this all the time now that school is done and i've turned in a thesis (be warned, 37 pages of mostly pictures and blogpost copypasta), there's nearly a month of inter-lifechange-space where I can bum around MITERS before I start a job.

If this thing is gonna have any chance at being street legal, it needs lights, a horn, and some DOT approved tires.  The original wiring harness was pretty ratty, so I tore it open and installed new connectors where necessary.

IMG_0792 IMG_0793 IMG_0794 IMG_0795 IMG_0796

Thankfully, all of the old engine related stuff could be left out, but it still took a good amount of staring at the XT-350 wiring diagram to get things working right.

XT350 wiring diagram

I hacked together a bar clamp for the left mirror.  The right mirror screws into the aluminum casting that contains the front brake master cylinder.  In most US states, motorcycles need only one mirror on the left to be street-legal.

IMG_0887 IMG_0886


next up: tires.  A new set of DOT approved tires + tubes cost $130.  I stupidly chewed up a couple tubes in a frustrating afternoon using some tire spoons borrowed from a friend.  The spoons were too big.  A smaller set I bought from harbor freight ($6 each) did the trick just fine.

IMG_0803 IMG_0804

There were many early/late nights spent at MITERS in a rush to work on the brems-thing before the end of school.  I showed up at 8AM once to find peter and mike there.  They had just finished coiling for the night.

IMG_0823 IMG_0824

If we're gonna have any decent range, we're gonna have to burn gas. The general series-hybrid idea looks like this:

series hybrid schematic

I took the T out to Wellington where there's a harbor freight.  IMG_0825

I bought a 212cc 6.5 hp gas engine for $100.  Thank you, UROP pay.


There is a strong online community around use and modification of this relatively inexpensive engine for minibikes and go karts.  I played tetris and chopped off some nonessential bits to try to get it to fit inside the frame.


IMG_0843 IMG_0844 IMG_0845 IMG_0847

this can't possibly end well


The synqor brick was uneventfully plugged in.


it eats 400V and puts out about 12.


Next I soldered up a rather absurd quantity of A123 26650 cells.  These are 3.3V and 2.2Ah each.  I made five packs, each consisting of two ~40V 6.6Ah 12S3P subpacks.  This many-hour ritual was fuelled by caffeine and the desire to give the prius inverter the many volts it deserves.

IMG_0849 IMG_0850

This brick of 72 cells is split into two packs of 12s3p.

IMG_0851 IMG_0852 IMG_0853 IMG_0854 IMG_0855 IMG_0856 IMG_0857 IMG_0858 IMG_0859 IMG_0860 IMG_0861

this is the hardest i have ever batteried ever.


The packs fit nicely in a “30mm” ammuntion container, available from various internet sources for about $40.  I hastily welded up a pannier rack thing to attach the box of battery to the bike.


I used strips of rubber to pad the packs


the bremsvehicle received a blessing from ken lager's parrot.

IMG_0870 IMG_0871

bayley and me did a quick load test on the synqor brick to see how efficient it is (how much heatsink it will need)


yep it's a battery alright.  many volts deserve maximum caution and respect.


with the pack made, we did some legit throttle step response testing and debugging.  Dumping 40kW into the rear wheel spins it up fairly quickly


it soon became apparent that charging the mega 400V terrifying bremspak was not an easy affair.  The 10x12s subpacks could be reconfigured in parallel for bulk charging.  That worked OK.  The real issue was balancing.


hobbykinging it is not terribly effective.  It takes an hour or two to balance 6 series cells using an IMAX-B6


In the final weeks before the end of everything (leaving school and starting an internship), the project lagged.  Testing was a PITA because of the weight of the bike, the cramped space in MITERS, and the difficult pack charging and balancing situation.  Even so, we managed to make some good findings:

  1. phase lag in the rotor position detection system (2x analog hall sensors) screws up commutation at high speeds.

This could have been fixed by specifying analog halls with a frequency rating.  Or by characterizing the phase offset vs motor speed and compensating for it in code.  We never got around to characterizing the offset, but instead adjusted the (constant) phase offset in the controller to align at higher speeds.  The result was a very peaky feeling throttle curve, where torque was limited at low speed but picked up greatly at higher speeds.

By doing this, we were able to spin the motor up to terrifyingly high speeds of 10k RPM or so.  In one such test, we experienced the centrifugal force induced detonation of a 3D printed frame of a long forgotten external magnet array attached to the rotor.  (pictured below in freshly printed form)

IMG_0453 IMG_0350

that sensing scheme never worked due to the long story of rotor position detection sadness and learning, and the magnet array was never removed from the rotor.

During the climax of high speed testing, an impressive crunch clack smash noise came from the motor, accompanied by a shower of white sparks.  This was very exciting, as it was reminiscent of the bremsfire incident just a few months before.  Reenacting the bremsfire toasted rotor scene, I removed the rotor to find a ball of magnet chunks and a few nicked windings.


Sintered neodymium magnets make white sparks when you smash them together.  The damage was surprisingly small considering the hard chunks of magnet being thrashed around.  One phase increased resistance by about 5%.


I coated the damaged spot with epoxy and put the motor back together for more.


Everything was naturally pretty hasty and poorly documented toward the end.  You'll have to take my word for the 60mph speed run at 3AM.  That night, we figured out that it was phase lag that was causing the drive to commutate incorrectly.  The symptoms were a loss of torque at around 30mph and excessive current draw.  By advancing the phase offset so that it lagged into the proper alignment at speed, we were able to extract much higher power at speed.  read: wheelie torque at 20mph

This is a little demo I did for my folks the day we packed it all up and headed back home to new jersey.


that concludes the brems-saga for the foreseeable future.  I moved to the San Francisco Bay Area (as a large fraction of MIT grads do) and took an internship at Alta Motors, an electric dirtbike/supermoto company.