Thomas Neuhaus wrote:

> 
> The effect you experience is normal due to the nature of hetrodyning
> analysis and additive resynthesis. First the attack portion contains a
> lot of noise which cannot be reproduced by only 50+ partials, second the
> partials of a piano sound are not exactly harmonic overtones. A high
> quality resynthesis should usa a dual approach such as additive synthesis
> for the steady part and subtractive (filtered noise) for the residual
> part.

The original stated problem mentioned high frequency distortion - which
does not sound like the 'normal' output of adsyn. Adsyn can actually get
remarkably close to the original - I have analysed a male voice speech
sound, and the only significant difference is indeed that the sibilant
consonants are less 'noisy' - but then I only used 80 partials.

'High quality resynthesis' may not always be the goal (though it is an
important test of an algorithm); for example, I would like to have some
means in Csound to apply non-linear warping of the partial frequencies
of an adsyn file.  

Also, as far as I am concerned, HQR is only a valid test if it is easily
achievable with the technique. I have Xavier Serrra's Windows
implementation of SMS, and it has easily the most complex, non-intuitive
and arcane parameter interface I have ever seen, or wish to see. The one
thing I have never managed to achieve with it is a clean resynthesis! If
someone can tell me how to do that, I may look at it again.

Richard Dobson

-- 
Test your DAW with my Soundcard Attrition Page!
http://wkweb5.cableinet.co.uk/rwd