Subject: Re: ANALYSIS USING SPECTOGRAM & TIMELINE (fwd)
From: Chris Rolfe (rolfe@sfu.ca)
Date: Wed Nov 24 1999 - 18:15:31 EST
Hi,
There's a small page with spectrograms of whale-song, singing voice and a
creaky door at:
http://www.monumental.com/rshorne/examples.html
Comparing spectrograms is a good way to learn to read them.
Spectrograms display the spectrum or (roughly) timbre of a sound over time,
making it easy to discern a harmonic relation among overtones, for example,
or the bandwidth of noise in a sound. Changes in amplitude, however,
represented by density or colour are usually less precisely displayed.
Waveforms display amplitude over time, which is a more traditional way of
viewing sound (derived from oscilliscopes, I suppose), but possibly less
intuitive. For instance, it's difficult to distinguish the waveforms of the
vowel "a" and the vowel "e", while the spectrograms are clearly different
(phoneticians have used spectrograms for decades). Also, a slight change in
phase in an overtone can result in a very different looking waveform,
although the sound is almost identical.
For musical analysis spectrograms would be far more informative than
waveforms.
Don't look just at enduring frequencies, but also at in/harmonic relations,
missing overtones, start time, end time and solidity of horizontal lines,
sloping lines (glissandi), density and use of bandwidth, rhythmic
distributions (vertical groupings), etc.
Waveform display is probably most useful while editing. Spectrograms
average a sound over a windowed period of time, which is a disadvantage
when precision in the time domain is needed. It's much easier, for
instance, to find and to fix a digital click, or to find the precise onset
of a sound while viewing the waveform. Very expert editors learn to read
waveforms, but it takes years to acquire the knack.
And, here's an interesting wrinkle for you:
Try to find the score for one of Xenakis' UPIC works; they're quite similar
to spectrograms, but used in this case to generate the sound.
Also look at the recent Computer Music Journal for compositional analysis
using spectrograms:
Barrett, Natasha. "Little Animals: Compositional Structuring Processes"
Cheers,
Chris
>First, I would appreciate any input offered about spectograms in general,
>Secondly, I would appreciate your comments about how you feel the waveform
>is/is not more useful (if not at least quite different) than the
>Thirdly, do you find that the spectogram is useful in terms of analysing
>the frequencies of the sound(s) represented through a piece and/or in
>layers of a particular section of the piece? I ask this because, I'm at
=======================================================================
Third Monk Software email: rolfe@sfu.ca
262 6th St. West email: thirdmonk@bc.sympatico.ca
North Vancouver, BC Web: http://www.thirdmonk.com
CANADA V7M 1K6 Tel: 604-985-7513
=======================================================================
This archive was generated by hypermail 2b27 : Wed Jun 11 2003 - 13:09:14 EDT