Re: [fluid-dev] Testing synth.overflow parameters - more questions

[Aere Greenway]

On 01/27/2013 08:15 PM, David Henningsson wrote:
> On 01/27/2013 09:35 PM, Aere Greenway wrote:
> > David and all:
> >
> > One more thing I need to know, to try different values for the
> > "synth.overflow.volume" parameter, is what range of numbers are produced
> > from the "volume_score" and "attenuation" calculations.  Is tho
> > "volume_score" just the value from the MIDI volume, velocity, and
> > expression values?  Added together? Is "attenuation" that value,
> > multiplied by a number between 0 and 1.0?
>
> Hi, and thanks for experimenting with it.
>
> The attenuation is based on the attenuation modulator it seems like. 
> It ranges from 0.1 to 1440, where 1440 is the softest ( -96 dB) and 
> the calculation is (volume_score / attenuation).
>
> There are default modulators in a soundfont for making the MIDI note 
> velocity as well as CC 7 (main volume) and CC 11 (expression) affect 
> the attenuation, but this can be overridden by the soundfont.
>
> // David
David:

Thanks for the additional information.

I tested this primarily on SUSe Linux, where I can't install your PPA 
(that I re-packaged).  My hope was that I would be able to work around 
the problem with libfluidsynth 1.1.5-3.1.2 by configuration changes.

For all this testing, I used a polyphony parameter of 64.

I was not able to configure my way out of the problem with 1.1.5. There 
is a real problem in that level (which most users now have), which is 
fixed in 1.1.6.

I did a lot of testing with my most demanding pieces, and my 
configuration changes did not change the sound in any obvious way. Here 
is the fluidsynth command I last tested with:

fluidsynth -a pulseaudio --chorus=no -o synth.polyphony=64 -o 
synth.overflow.age=8000.0 -o synth.overflow.released=-6000.0 -o 
synth.overflow.sustained=-4000.0 /dos/Shared/SoundFonts/FluidR3_GM122501.SF2

My thinking was to to make released (and not sustained) notes the first 
candidates for scavenging, followed by sustained notes, trying to have 
recent notes be less-likely to be taken-out.

I connected to it (for output) with my new MIDI device, using both the 
integrated MIDI player and the keyboard, simultaneously.

I tried (with the sustain pedal held down) multiple ascending and 
descending full-keyboard glissandos, and the sound was as expected, with 
the oldest notes being taken-out, and the newest notes sounding the 
proper direction of the glissando.  My device has 84 keys, but I used 
only the white keys, so each glissando was 48 notes.

So the results of my testing, other than finding I could not 
configure-away the bug that affects me in 1.1.5, were not conclusive.  
It sounded the same as with the defaults, as far as I could tell.

I do encounter the bug using just my keyboard, though I have to work 
hard to encounter it.

--
Sincerely,
Aere