well over here on windos the help patch couldn't be found until i renamed it
interesting what you found out... i was more talking about the clip 100 problem and the bump, at 4db (completely insignificant), which appears at high compander point settings

All tech aspects aside, I've been using qompander~ for about six months and now it finds its way into every patch I make.

It is a beautiful thing: simple and powerful. Thanks Katja.

I was wondering: why not apply the compander curve to the original signal instead of the complexified one? Wouldn't we get rid of the harmonic distortion?

@kyro said:

I was wondering: why not apply the compander curve to the original signal instead of the complexified one? Wouldn't we get rid of the harmonic distortion?

The filters used to create the analytic signal cause phase shifts (= frequency-dependent time shifts). From http://www.katjaas.nl/compander/compander.html:

"Since the measurement delay with the analytic signal is frequency-dependant, there is no practical way to sync the amplitude curve with the original signal. To cut things short, I decided to use the analytic signal itself as the audio, by summing real and imaginary phases and reduce the resultant amplitude with factor 1/sqrt(2). Then, the measured amplitude will by definition always exactly match the processed signal."

Oh. Turns out I was absent-minded.