The following figure shows the result of a little engine noise analysis based on the on board audio from Takuma Sato pitstop in Melbourne.
The analysis starts with the car already stationary and stops when the guy in the radio talks. The black crosses are the various samples while the red line just connects them for a more obvious visualization. Notice that the instantaneous peaks caused by a single data, as for example the peak at about 4 s (or the rapid up-down between 6500 and 7500 @ about 1s), are probably results of some odd noise hence should be discarded. What happens at 2 and 3 s on the contrary is probably an rpm variation.