[linux-audio-user] Decent reverb, and DSP in hardware?

davidrclark at earthlink.net davidrclark at earthlink.net
Wed Jan 26 13:59:51 EST 2005


Greetings All,

I have recently posted to LAA an announcement of Mixster and other programs.
These are aimed specifically at convolution reverb.  They run under Linux,
and with the appropriate libs, under XP which is where I will probably use
them.

The program called "Boxster" is for generating 3-D impulse response functions.
The program called "Mixster" is for convolution, albeit nonrealtime, and
for combining the tracks.  It has other utilities which are useful for 
performing tasks associated with convolution reverb.  To novice users,
I'm sure it appears to be a miscellaneous grab-bag; it's not.

-----------------------------------

Boxster uses Green's functions rather than ray-tracing for generating the
IR's.  This is a complete solution --- as complete as the finite number of
samples allows --- for the ~3 billion modes in a typical concert hall.  The
walls are hard, but I have myself developed techniques for engineering the
IR's so that they sound very realistic.  I have a set of IR's that I have
used for about two years now that I would be willing to share.

This technique requires monophonic, dry recordings, not what people are
used to.  The programs are not especially easy to understand, but I'm 
willing to work with serious users and eventually hope to create more
docs, help, and probably yet another interface.  I believe that the 
current interface is useable by anyone who has ever solved an ODE or
PDE.

Steve Harris mentioned that the code for convolution is pretty easy.  That's
true.  The hard part is engineering IR's.  That is the purpose of Boxster
as well as the tools that are included in Mixster.  This is anything but
easy, but even simple-minded approaches produce better reverbs than most
DSP techniques.

Several posters have suggested that a few good reverbs would be very
helpful.  The more I work with this, the more I think that this really is
not true.  Just as there are many different types of drum sets, there will
probably be many different types of reverbs.  As people get used to 
good reverbs, I think they'll become even more discriminating and want
more variation.  The variations will become harder and harder to distinguish
for novice listeners, but more and more obvious to experienced ones.  The
drivers for the technology will be the discriminating engineers and listeners,
not the novices.  I'm amazed at just how different various IR's can be when
applied to the same recording, even though the rooms are similar.  There
is a completely different "feel" for each one.

For those who are really serious about reverb: Please don't hesitate to
contact me.  I'd be happy to improve the programs; they definitely need some
more docs and probably yet-another-interface.  All I can say is that the
command-line versions were a lot worse.

Thanks for your attention.  I would not have posted if someone had mentioned
these programs, even if only to complain about them!  But no one did (that
I saw).

Regards to all,
Dave.

http://home.earthlink.net/~davidrclark






More information about the Linux-audio-user mailing list