Dan Dugan wrote:
>> And example I've recently come across is Little
>> Button-quail (I think). They're around 400Hz, so hard to see
>> at the bottom of the image if the scale goes up to 15kHz, and
>> the call is long and slow. If the horizontal scale is too
>> stretched out they appear on the spectrogram almost as a
>> horizontal line, but if a more compressed scale is used
>> (about a minute per screen width) then they become very
>> obvious little horse-shoe shapes.
>
> The XC example used a linear frequency scale on the vertical
> axis. This is great for high-frequency insects and birds but
> compresses lower frequencies to the point of invisibility.
> It's kind of a left-over from early spectrogram technology,
> but academic people are used to it and consider it to be the
> the standard method. A spectrogram with a dB (log) vertical
> axis shows the full range in proportion.
That's true, although I'm not sure I'd use the expression "in proportion" f=
or a log scale. I've just had another look at the "spectrogram log(f)" view=
in Audacity, and I can display 0 to 20kHz and still see the horse shoes cl=
early.
However, unless I use a very high FFT window size, the lower frequencies lo=
ok like mush. At size 4096 there's not much point displaying 0-10Hz, which =
otherwise fills 1/3 of the screen with pixelated mush. It's only cosmetic, =
but it's very distracting.
At higher window sizes the low frequency detail emerges, but the higher fre=
quencies start to blur out.
I wonder if it's better to give the viewer the option of several frequency =
ranges to look at, with appropriate window sizes for each. But the low freq=
uencies should be somehow viewable. Given that most programs start with a l=
ow default FFT window size for spectrogram generation, I'd say that many pe=
ople have never had a good look at their low frequencies, and I've suddenly=
realise how much interesting stuff there is down there.
Having said that, anything's better than nothing.
Peter Shute
|