Subject: Re: [linux-audio-dev] Dithering idiots
From: Richard Bown (bownie_AT_bownie.com)
Date: Mon Jun 17 2002 - 14:57:22 EEST
Paul Davis wrote:
> i don't quite get it. when you render a waveform on the screen, you're
> normally not doing anything like interpolation. you're subsampling the
> amplitude waveform to a given frames-per-pixel density.
But what if (as Bill says) a peak falls initially between the graphical
sample points and then when you change the graphical resolution and resample
it suddenly appears? Ok, this would probably only occur for some potentially
very steep transients but surely peaks could get missed? When your graphical
resolution is lower than your sample resolution (which, let's face it is
going to be true always) then how should you interpret the underlying envelope?
It's more than likely that my code is a buggy but at the moment comparing
the waveform on my previews against those on other packages produces varying
results. While the underlying shape is usually good at some (graphical)
resolutions there are occasional transient peaks which I'm guessing are high
frequency components that the other packages are "interpolating" (or whatever)
out. In these cases my naive sampling algorithm is showing them when they're
actually too brief to actually figure in the overall shape of the waveform.
I guess the simple answer to that is to LPF the audio file to remove any
high freq components (it's a kick drum so I doubt it) and see what the
preview of that looks like.
R
-- http://www.all-day-breakfast.com/rosegarden http://www.bownie.com
This archive was generated by hypermail 2b28 : Mon Jun 17 2002 - 15:04:27 EEST