[Martin Taylor 970121 17:30]
Bill Powers (970121.1230 MST)]
Another comment I should have made in my earlier message.
> In a human control system which has a
> maximum frequency response of about 2.5 Hz, the implied Nyquist sampling
> rate would then be about 0.8 samples per second. Using a safety factor of 10
> for comfort, that would still call for only 8 samples per second. Simulating
> integrations in a typical environment accurately would certainly require
> using a dt a lot smaller than 1/8 second.
2.5 Hz implies a Nyquist rate of 5 samples/sec, doesn't it? Using a safety
factor of 10 yields 50 samples per second, not far from your recommended
60 samples/sec, which I assume you use because you live in N. America
rather than in Europe, with its 50 Hz electricity supply.
So the ad-hoc "works pretty well" way of doing it jibes fairly well with
the bridge-engineer's "safety factor of 10" approach to theoretical values.
Both ways we get a few tens of samples per second as a reasonably valid
value for dt when we are dealing with human low-level control systems.
What more can one ask?