Yes, I only need the main dominant frequency, one single number. I am not familair with the autocorrelation frequency method, so will look into it to understand the concept.
EDIT: here are two interesting links on the autocorrelation subject with sample codes for me to digest. (So it's not FFT at all ! ):
This is what a typical signal will look like. For my purposes, it will always be between 1 Khz and 25 Khz and the frequency never exceeds the 1ms period length. But the entire signal is about 6-8ms in length (so the sampling envelope should be minimum that long, or continuously sampled. And from that array the FFT must grab that dominant peak area at the beginning. And sampling needs to be accurate down to 40us with about 1% accuracy.
What's ironic is that my oscope gets it correct every time. I wish I could just transplant the algo from it into a smaller footprint
