Re: [fluid-dev] FluidSynth problem

[Josh Green]

On Tue, 2003-12-02 at 23:51, Joerg Anders wrote:
> On Tue, 2 Dec 2003, Josh Green wrote:
> 
> > 
> > It would be good to give FluidSynth a good comparison workout between
> > other synths. I did a little bit of testing of that nature myself, but
> > hardly scientific :)
> 
> First of all :Thank you for your answer!

You're welcome, sorry I couldn't help you out more. I'll add your
problem to the TODO list. Unfortunately FluidSynth development has been
kind of slow lately, due to the original author having much less time to
work on it, and I as the maintainer haven't really been able to devote
enough time to it either due to other projects (Swami, libInstPatch,
FlacPak) that I have been working on.

> > 
> > > 
> > > Unfortunately, the string rendering is very important.
> > > Please have a look at:
> > > 
> > I tested out 48 (strings), I don't think I have the best ear for this
> > type of stuff, but it didn't sound like it was as dramatic an attack as
> > the FluidSynth sample you put on that web page.
> 
> It is dramatic. It is very dramatic! Please have a look at
> 

I wasn't saying that it was not dramatic. I could definitely tell a
difference with your samples that you had on the web site. I was saying
that when I played that instrument myself, it didn't sound like quite as
dramatic an attack as the one you have on your web site. Maybe that was
due to other conditions that were not the same between our two tests?

>  http://rnvs.informatik.tu-chemnitz.de/SYNTH/synth.html
> 
> and compare the hardware synthesis with FludiSynth synthesis.
> You need practically "no ears" to hear that there are no
> strings, at all. That disfigures the whole bigband sound.
> 
> But I think I can stop the discussion here: The newest CVS version
> of TiMidity++:
> 
>   http://rnvs.informatik.tu-chemnitz.de/SYNTH/synth.html
> 
> does the job very well. You cannot hear any difference
> between TiMidity++ synthesis and hardware synthesis.
> And if: TiMidity++ is better.

"And if: TiMidity++ is better?" Perhaps, I've never really done a
comparison feature wise between FluidSynth and TiMidity++. I know that
FluidSynth has fairly complete support for the SoundFont2 specification,
modulators, etc. Although it could use some attention in regards to
compatibility with other synths (as you have proved yourself).

Can you explain in words what the problems are you are experiencing with
the strings? From the first samples I listened to it sounded like you
were referring to the attack envelope being too dramatic and causing an
excessive surge of volume at the beginning of the instrument. From your
big band tests it sounds more like you are referring to the strings
being too quiet? Are those the same instrument?

I'm not sure what would cause the attack envelope to be too sharp, that
is new to me. The too quiet strings on the other hand I have a good idea
of what could be causing that. FluidSynth's volume envelope attenuation
currently follows the SoundFont specification. This breaks with a lot of
EMU8k patches because they don't follow it. The solution would be to
have FluidSynth detect that an EMU8k is the target for the SoundFont and
configure itself to the broken (in regard to specs) way in which it
deals with the attenuation parameter. You can verify that this is the
cause by changing the attenuation parameter for that instrument with a
SoundFont editor (such as Swami). Anyways, if TiMidity++ works for you,
then by all means use it. Your feedback is appreciated though, and in
the the long run will help FluidSynth improve. Cheers.
        Josh Green