Thanks for the excellent input and data, Bob!

I intend to count the time between successive revolutions to a resolution of at least 1us, which would be equivalent to sampling at 1MHz; anything less than that will not give me the accuracy I desire.

For the moment, I won't be doing any current measurement, so I won't be able to determine efficiency. But my main interest is determining power output vs timing.

I agree that gearing can keep the flywheel to sane speeds, and also reduce aerodynamic power losses (which increase as the cube of the speed). But I still need to establish that the data won't be compromised by the gear lash.

On another note, I have acquired a nice switching power supply (Meanwell RSP3000) which is adjustable over the full range of voltages equivalent to 1s through 3s LiPo. This should reduce voltage droop to negligible values during a run. But it does require 220V, so I need to unplug the clothes dryer if I want to run the dyno!