[Ether]

The OP didn't mention the shooter speed(s) or the execution rate.

So I'll pick some numbers.

At 4500 rpm true wheel speed, sampling the counts of a one-per-rev sensor at a 20ms rate will give you 1 count 50% of the time and 2 counts 50% of the time. So if you compute speed by dividing counts by elapsed time, your measured rpm will be 3000 half the time and 6000 the other half, and that's assuming perfect 20ms scheduling of your control algorithm.

That's a lot of noise.

I wonder if the shooter inertia would be large enough to allow bang-bang to work with a signal like that. Maybe some inquisitive team will try it and report their results.


If it doesn't work, let the FPGA do the sampling (at 153KHz) and timing (with its 1MHz timer). You'll see +/-3 rpm, and your 20ms scheduling doesn't need to be tightly held.