[iiwusynth-devel] Some FluidSynth notes

[Josh Green]

I thought I would update the list on current bugs and un-implemented
features with FluidSynth. I haven't done thorough testing on recent
versions of FluidSynth, since the Swami plugin currently does not work
with it (is there an API for configuring settings at the moment?)

1. Sample start and end offsets not implemented at the instrument
(generator) layer.

2. It appears that Filter Q has no effect when Filter FC is maxed
(16khz+). This was a problem a while ago, but was fixed. Probably got
re-introduced with new Filter optimizations?

3. Filter FC optimization is a bit rough. Modulating the Filter FC
around 16khz causes audible change as Filter is turned off/on (turn off
value should probably be higher).

4. Default reverb is pretty weak. I remember the suggestion before of
having some presets for Reverb and Chorus. I've seen lists of
Reverb/Chorus effects for MIDI devices before. Perhaps something like:

Reverb
--------
Small Room
Medium Room
Large Room
Short Hall
Long Hall
Plate Reverb
Delay Reverb

Chorus
--------
Chorus 1
Chorus 2
Chorus 3
etc.

Having these presets built in the API would allow programs like Swami to
query and select them (leaving the actual values under the control of
FluidSynth).

5. Sample looping doesn't work when the loop length is < 32 samples.
While I know this is defined in the spec as being the minimum loop
length. I believe this is talking about a restriction when writing a
SoundFont to ensure that it can be synthesized on any SoundFont capable
hardware. Therefore a SoundFont synthesizer is free to exceed these
requirements (provide looping at less than 32 samples). The AWE/Live!
cards allow looping at less than 32 samples and many SoundFont files
have these smaller loops. Its important to distinguish between parts of
the spec that are minimal requirements of a synthesizer (those places in
which we can exceed the requirements) and those that are set in stone
(not complying would somehow break format/compatibility).


It seems important to me to get the synthesis of FluidSynth right before
releasing an official version. FluidSynth is setting a standard by which
SoundFont files will be created, so it would also be good to compare its
synthesis to other devices. It would not be very nice to have to
radically change some part of the synthesis later that would cause
created banks to sound different.
Anyways. Keep up the great work. I'm anxious to get Swami back in the
game and start adding some real functionality to the API that I have
been working on for so long. A lot of this API work will make the
addition of other patch formats feasable (among a lot of other cool
things :) Cheers!
        Josh Green