The abstracted requirements for T19 flywheel motor drives
- 3-phase PMSM controller
- Sensorless control (no encoders, resolvers, Hall sensors or the like exist for rotor angle feedback)
- Must support minimum 180,000 electrical RPM (3000Hz) - for this relatively slow application. Make that about 300,000 if we are to extend that to generalized nerf flywheel drive applications.
- Must have a closed-loop speed control mode (The Hy-Con flywheel system does not have acceptable dynamics with voltage command or "BLDC mode". There is too much inertia.)
- Inverter rated for: ~24V DC bus, ~20A continuous phase current, ~70A transient phase current (roughly)
- Must have an agile startup methodology from 0 rpm and undriven state - say, <10ms from receiving a speed command step to development of motor torque
- Must support 400Hz "RC standard" 1000-2000us PWM throttle signal input for stock blaster controller (though the controller/firmware can be adapted to any protocol)
- Must be capable of protecting the stator winding and its own power stage from excessive current during sudden shock/stoppage, prolonged locked-rotor condition, or sync loss (sensorless failure)
- Must be capable of reliably tracking and establishing control of turning, undriven motors
- Must demonstrate proper operation with the Turnigy Multistar V-Spec 2205 2350kv, or similar 12 slot, 14 pole motors with fractional slot concentrated winding
- Must be capable of applying zero torque/current to a turning motor (coasting): i.e., not some problematic types of "hobbytroller" (those that run BLHeli flavors, mostly) that always do complementary PWM and yet only do open-loop voltage command, or those that misbehave other ways (BLHeli_32, failure to restart from running) when damping modes are disabled.
- Must be tunable enough to debug motor performance issues, for example timing for scalar control or FOC parameters for vector control, current control/voltage schedule, ...
- Reasonably packaged and priced: about like a 20-30A rated hobbytroller.
Any controller that you can lay hands on which satisfies the above, and most critically satisfies them all at once, should do, if you wish to venture outside the sorts of hardware I use. I welcome anyone claiming to find or make this magic drive. Believe me, I have dug through lots of knowledge, datasheets and specs on existing gear so far and everything existing fails at least one way. Ideally it would be 32-bit, have a codebase written mostly in C, do sinusoidal FOC, and have proper current sensors.