Default sampling rate

[Marcel D]

In the programs that I have seen that analyse speech, I sometimes saw a Nyquist frequency of 5000 or 5500 Hz, while in other cases this frequency is set to 8000 Hz. To me, it seems like there is no default standard sampling rate. What would be a reason to use the 5000 Hz frequency rather than the 8000 Hz, and vice-versa?

[Simon]

I think you’re referring to the way Praat defaults to plotting spectrograms with a frequency axis from 0 to 5000Hz, regardless of the sampling rate. In other words, Praat often does not show the full frequency range of the signal.

Why do you think Praat limits the frequency range when plotting spectrograms?

In Wavesurfer, the spectrogram is always plotted from 0 to the Nyquist frequency (half the sampling rate).

[Marcel D]

What I remember, it had something to do with better recognition of the first five formants if the frequency axis was set to 5000 or 5500 Hz. The fifth formant usually does not exceed 5500 Hz (if I remember right), so in that case there is no need for a broader frequency axis.

[Author]

Posts