[music-dsp] interpolator testing

[Muon Software Ltd - Dave]

Olli,

I often use a bandlimited saw or square wave for this sort of testing
(generated, rather than sampled from some synth source). The harmonics are
nicely spaced out and any nasties are easily detectable inbetween the peaks.
They're also easy to generate for any pitch so you could test by say for
example making two "perfect" waves at 1khz and 0.75khz, for example, then
pitching the first one down to match the second using your interpolator of
choice and doing a comparision. A sort of "perfection" test?? If you
experiment with a range of pitches in all the musical octaves I think this
gives a good idea of how an interpolator will perform.

One other thing I found was to make sure that you have no "part-cycles" at
the beginning or end of the test tone, though. As someone pointed out
recently FFT's assume periodicity - sticking to whole cycles will help to
keep a clean plot and so you can be sure you're only looking at the noise
generated by the interpolator.

Dave

----- Original Message -----
From: Olli Niemitalo <oniemita at mail.student.oulu.fi>
To: music-dsp mailing list <music-dsp at shoko.calarts.edu>
Sent: Tuesday, September 12, 2000 9:26 AM
Subject: [music-dsp] interpolator testing


> I plan to start testing interpolators and making a quality versus
> processing time chart. Now i started wondering if a flat-spectrum test
> signal is really the best...



dupswapdrop -- the music-dsp mailing list and website: subscription info,
FAQ, source code archive, list archive, book reviews, dsp links
http://shoko.calarts.edu/musicdsp/