Hi Raimund
>
> as Scott Fraser already pointed out, looking at the waveform alone might
> be misleading here. Only slightly more than two samples per cycle can
> really look terrible ;-)
>
> You should instead look at the spectrogram, which inherently integrates
> the individual samples over a longer period of time. You will then find out
> that even very high-pitched calls that are close to the Nyquist frequency
> exhibit a perfect shape.
>
> Here is an example of a Blue Tit song sampled at 22.05 kHz, whose
> maximum signal frequency is 8.7 kHz at the beginning of the first syllable:
Forgive me for my (probably) stupid question.
I don't understand the correlation between what you're saying and the
spectogram you're referring to.
Previous posters on the subject mentioned a 10% security margin between
the Nyquist frequency and the highest frequencies that might be of interest.
In your example the ratio between the Nyquist frequency of the recording
(11.025 kHz) and the highest frequncy of the Blue Tit song (8.7 kHz) is
Message: 0.
Subject: 789, i.e. 21%.
In any case, I, of course, agree with you that it's better to look at
spectograms
rather than guestimate with the "10% formule" (TM) !
Best regards.
--Serge
|