Subject: Re: [linux-audio-dev] Dithering idiots
From: Nick Bailey (n.j.bailey_AT_elec.gla.ac.uk)
Date: Mon Jun 17 2002 - 19:33:47 EEST
Bill Bland wrote:
>Probably wrong, irrelevant, or both, but I was thinking: Plotting a
>waveform is like sampling it, and at certain sampling frequencies you will
>get strange effects. Say you have a waveform representing a 1Hz sine
>wave, and you sample it at 1Hz. Then you could draw the waveform on the
>screen as complete silence, or a constant DC, when in fact that's a rather
>misleading picture to draw.
>
>
Absolutely right
>Obviously if 1 pixel == 1 sample, you won't get this effect, but if you
>zoomed out enough that 1 pixel == 44100 samples (for example), then you
>could see complete silence (or an arbitrary DC value).
>
>Just a thought. Correct as necessary ;)
>Best wishes,
> Bill.
>
>
But you don't want to plot the *signal*, you want to plot the *signal
peaks*. This is nasty and nonlinear. If you did the filter you suggest,
so that eg 10 mins audio was represented by 1000 "samples" on the
screen, all the signal over 1000/(2*600) Hz would be completely lost,
and what was left (less than 1Hz bandwidth!) wouldn't tell you very
much. So you need to search for and display maximum signal excursions in
the timeslice which one pixel on the screen represents. Obviously, this
can be computationally quite expensive, just because of the number of
bytes being thrown around (esp if your canvas supports zooming in and
out) so you probably want to cache results some how, which is what all
those peak files do.
Nick/
-- Dr Nick Bailey n.j.bailey_AT_elec.gla.ac.uk Centre for Music Technology http://cmt.gla.ac.uk/ Dept of Electronics and Electrical Eng http://www.elec.gla.ac.uk/ The University of Glasgow http://www.gla.ac.ukPlease take our survey about how the arts can be delivered over the internet! http://cmt.gla.ac.uk/CMT/OD-Survey Tell your friends! Thanks! From the OpenDrama team.
This archive was generated by hypermail 2b28 : Mon Jun 17 2002 - 19:46:37 EEST