< Back to IRCAM Forum

convolution reverb

Spat
#1

Hi all,

I’ve a feature request and a question.

I would like to get rid of delay line and FDN based reverbs, and would like to use a convolution reverb.

The feature request is:
it would be useful if the user could choose between the default FDN reverb and a reverberation based on convolution directly in the spat.oper.

The question is:
Which is the best way to realize a convolution reverb in the spat environment?
I’ve seen that in the spat.converb~ helpfile there are two examples: one spat.oper, using also spat.oper_ to compute low level parameters, the other with spat.rat and the filter compensation.

thanks

#2

Hi again Lorenzo,

“Get rid of FDN reverb” ! What a crazy idea.

More seriously:
Replacing spat~ reverberation engine with a convolution-based reverberator makes sense.
However this is not trivial at all. This is actually part of ongoing research.

The task is challenging because your spat~ is not “just a reverb”, it is a spatialized reverb.
Dealing with the timbral aspects is not too complicated, but the spatial aspects (i.e. how the direct sound and reflections are spatialized) is a completely different story.

With the current state of the art, you could replace spat~ with a convolution engine only if you target mono or stereo output.
If you plan to use other speakers arrangement, it is not yet possible to fully replace spat~ with a convolution.

What is currently available:

  • spat.conv~ : this is just a realtime convolution engine, without control.
  • spat.converb~ : this is also a realtime convolution engine. It further splits the impulse responses in 4 sections, and filter them, similarly to what the spat FDN does. So you can use spat.oper and the perceptual approach to modify the convolution reverb. But you can only modify the timbral aspects! The spatial aspects are not handled.
  • spat.rat : this object analyzes a given room impulse response and tries to compute a spat.oper preset that is equivalent (or similar). So you can use this object to tune your FDN so that it mimics a given convolution.

So… there is no “best way” to realize a convolution reverb. It really depends on what you want to do.

T.

1 Like