OMG! You guys screw with my head! I have no idea what your on about or how your getting your readings/outcomes but I find it so interesting reading what your findings are. Even if I don't understand them!
WTF???
It's relatively simple, though it looks complicated. A wave file makes 44,100 samples per second, same as compact disc sampling rates. Using a sound analyzing/editing program, we can see the amplitude waves generated by the engine, and b/c the sampling rate is so high, we can literally zoom in to look at the amplitude waves for the individual combustion strokes. By measuring the number of combustion cycles over a period of time (usually a few tenths of a second), engine rpm can be determined. We are "listening" (visually with the software) for the explosion per minute b/c it is directly related to the crankshaft revolutions per minute.
In this case we are using an analysis function to help us. It displays the amplitude for each frequency, and by examining the frequency rates we believe to be relevant for MotoGP engines (130-150hz for the new engines) we can determine rpm. For example, we select a piece of the engine noise just before the rider shifts. The frequency analyzer creates a file with a bunch of peaks and valleys for each frequency. We look at the 130hz to 150hz area, and if we see a distinct amplitude peak, we can determine the engine's rpm. For an onboard sound recording with no additional audio, the loudest noise in the 130hz to 150hz is going to be engine combustion. Therefore, the amplitude peak represents engine frequency which represents engine rpm. It isn't fool proof, but it works quite well.