[music-dsp] Effective Waveform Visualisation
joshscholar at yahoo.com
Fri Feb 18 13:41:15 EST 2005
----- Original Message -----
From: "Maciej Bartkowiak" <mbartkow at et.put.poznan.pl>
To: "music-dsp" <music-dsp at shoko.calarts.edu>
Sent: Friday, February 18, 2005 8:42 AM
Subject: Re: [music-dsp] Effective Waveform Visualisation
> This is really a very nice idea and I am sure a proper waveform disply
will be greatly appreciated
> by many people. Most audio editors don't properly antialias the waveforms
and there are many
> confusing and distrubing graphical artifacts.
> I think that apart from proper antialiasing it would be great to take care
of a proper presentation
> of the continuous wave shape on the basis of the samples in the buffer.
This is particularly important,
> when the signal contains significant high frequency components, especially
those close to Nyquist
> frequency. For example, if you generate a frequency swept sine up to
Nyquist, you will observe
> a strange modulation of the magnitude when the waveform is zoomed out. In
order to avoid this visual
> artifact, and to be able to intuitively analyse the continuous signal, it
is necessary to apply some interpolation.
> I have tried spline interpolation but it is still not perfect and I am
affraid sinc-based interpolation may be
I know exactly what you mean. It was a revelation to me when I first applied
sinc interpolation to a sine wave that was nearly nyquest and it went from
looking like it was amplitude modulated to looking like a perfect sine wave.
Anyway I've written resamplers before so I know how to do this. Good idea!
> The only piece of software I know that does it properly is Adobe Audition
(they improved the interpolation
> after acquiring CoolEdit Pro and now it is almost perfect) - as opposed to
Wavelab and SoundForge,
> Audition displays a high-frequency sweep like an analogue scope. Thus, you
can e.g. observe whether your
> analogue signal will be clipped, which may happen even if all samples are
> But the display in Audition it is still not antialiased :-(
I had planned on letting the user place arbitrary filters in front of the
display, but it never occured to me that they could be interpolation
filters - that they could change the sampling rate.
Unfortunately it's a bit confusing to specify interpolation filters... For
instance I've often used special cased recursive doubling sinc interpolation
that don't effect the existing samples, or I've used variations on those
that do some filtering as well. But the simplest form is just to run a
regular iir filter over zero stuffed oversampled data.
I guess the best thing would be to give the user three options:
1. Display not oversampled
2. Display oversampled with recursively applied cosine-windowed sinc
3. Display oversampled by 0 stuffing and bidirectional IIR filter (specify
oversample rate , specify filter)
> I think that with your idea, it is a great opportunity to create a really
very usefull audio tool for research
> and mastering applications.
I had been thinking that this could eventually be turned into a full fledged
multitrack editor, but that would take a lot of time to write. I imagine
that the hardest parts for me would be to figure out Microsoft's horribly
undocumented DirectX system, and (also not thoroughly document - or so I've
heard) VST plugins, as well as trying to support file standards and various
hardware configurations... All the stuff I can't make up out of my own head.
And there are other challenges...
I haven't actually been employed doing sound processing except to make sound
drivers for video games (and a resampler for a telephone application), so I
don't really know first hand what tools people need.
If people need simpler tools than the all in one swiss army knife sort of
thing, that would be good to know.
> dupswapdrop -- the music-dsp mailing list and website:
> subscription info, FAQ, source code archive, list archive, book reviews,
More information about the music-dsp