[sdiy] IR Reverb

rburnett at richieburnett.co.uk rburnett at richieburnett.co.uk
Fri Feb 16 12:08:11 CET 2018

On 2018-02-15 14:38, Richard Wentk wrote:
> But only from one monitoring location.

Yes.  From one location, to one location.  With no movement whatsoever.  
The musician must play their instrument without shuffling, fidgeting or 
moving anything in any way, and the listener must keep their head and 
body equally still, otherwise the impulse response changes.  An impulse 
response is for a Single Input Single Output linear time-invariant 
system.  Or okay, single input dual output system for people with two 
ears if you capture two IRs.

As others have stated, real instruments aren't point-source isotropic 
radiators either.  Anyone who's miked up a guitar or drum will know that 
it radiates noticeably different sound on-axis vs off-axis.  A single IR 
based convolution reverb doesn't capture this.

Now you can see why Convolution Reverb is not perfect, and often sounds 
very static and sterile when compared to algorithmic reverbs that 
purposely have movement (modulation) programmed into them to make them 
sound alive.

A long time ago I was asked about the feasibility of using something 
called deconvolution processing to reduce "clutter" and increase 
intelligibility on Public Address systems installed in severely 
reverberant buildings, (train stations, swimming baths, etc.)  The 
conclusion was that it can be done reasonably effectively with enough 
DSP power, but only for a single listener located in one specific 
location in the reverberant sound field.  For all of the other listeners 
not located in the sweet spot it increased the level of reverberant mush 
and decreased intelligibility further!


More information about the Synth-diy mailing list