After sustained input of the same pitch, output freq becomes noisy

[benmurrell]

If I whistle a sustained pitch, the resulting frequency from the FFT & from autocorrelation will hold steady for some time, then suddenly begin to jump all over the range of 0-3000hz. I would expect the reported frequency to hold steady while the input consists of a steady frequency.

This behavior is not seen if the input frequency is constantly changing (say, wavering in pitch, up and down) - the FFT & autocorrelate methods both act as expected. Only when a constant frequency is maintained past a certain amount of time does the output become erratic.