microcontroller for e-bikes article (attn Joe Perez et al)
 

Interesting.

The current crop of motor controllers are pretty "dumb." Even though they are programmable for things like speed and phase-current limits, they mostly just translate the throttle input to a motor duty cycle. This works on on low-power units, but is very twitchy at higher voltages. (eg: 10% throttle makes the bike want to wheelie from a stop.)

I'm using a piggyback solution at present, with a device called a CycleAnalyst. This computer sits between the throttle and the controller, and takes in data from motor temp and battery current. In addition to various logging and display functions, it also allows you to implement a closed-loop throttle rescaling mode, which turns the throttle into a "current request" function, similar to how a Subaru DBW system works. It generates a synthetic throttle output (to the controller) which aims to achieve a specific amount of current for a given throttle input. Works quite well, and is something which ought to be a part of all current-gen controller ICs.

No reason this couldn't all be integrated into a single box, I'm just not aware of anyone who has done it yet. This looks like a step in the right direction.

"Current request" is approximately equal to "torque request".

I considered it really dumb to do a fixed duty cycle which translates to motor speed - ergo at low speed and you open the "throttle" slightly you get full torque until the motor is at the target speed.