[Audacity-quality] Risset Drum

[Steve t. F. <ste...@gm...>]

Continuing with updating the bundled Nyquist effects to make the GUIs
more consistent,
I notice that the Risset Drum
http://manual.audacityteam.org/man/Generate_Menu#risset has 3 sliders
for setting the "Noise" component, yet when used, even with the Noise
Mix set to 100%, the noise component is hardly noticeable.

The Risset Drum sound is based on work by the composer of electronic
music, Jean Claude Risset
Other implementations exist, which seem to be mostly based on the
algorithm described on page 104 of "Computer Music" by Dodge and
Reese.

Comparing the sound of the Audacity Risset Drum with audio samples
found elsewhere, for example this sample from csounds.com:
http://www.csounds.com/catalogfrom/audio/drum2.mp3
the noise component is noticeable in other implementations, but not in
the Audacity version.

Looking in the Nyquist code, the "Noise Mix" parameter has been
converted from % (a scale of 0 to 100) to linear (a scale of 0 to 1)
twice, thus causing the noise mix level to be in a range of 0 to 0.01
(0 to 1%). I presume that this is a bug, so I've added it to bugzilla
as bug 556.

The question is, do we want to fix it? The Risset Drum generator has
been as it is in Audacity for years and AFAIK without complaint.
If we do fix it, I could change the default setting to a lower "Noise
Mix" so that the default sound remains similar to the familiar
setting.

Also, I notice that the output level is fixed at about 0.8.
I suspect that this is because the level is calculated by generating
the sound once, then normalizing a second instance of the sound.
Because the sound contains a random (noise) component, the sound will
be slightly different each time it is generated, thus it would not be
possible to accurately calculate the exact normalization amount. There
is a comment in the code about this:

;; Generate signal and normalize.
;; ISSUE: Is there any way to normalize signal without
;; generating it twice?

Yes it is possible to normalize without generating the sound twice,
thus normalization can be done accurately.
This means that the sound could be generated precisely to any level up
to 0 dB without clipping, so do we want an "Amplitude" slider (like
other Generate effects) with a default level of 0.8 ? If the
"documentation" is removed from the interface there is plenty of room
for an additional slider.

Steve