[LAU] How I built my own IR reverb in ambisonics

Michael Jarosch riotsound at riotmusic.de
Wed Sep 21 21:10:10 CEST 2022

Am 21.09.22 um 15:35 schrieb Fons Adriaensen:
> If the room is big enough, you could use a larger distance.
> This will reduce the amplitude of the the impulse but not
> the room response, so you could have a better S/N ratio.
Good hint! I always thought it was necessary to place the impulse source 
as close as possible to the microphone.
> There is NT1 preset available for tetraproc (real-time A/B processor, 
> Jack)
> and tetrafile (file processor doing the same). I can send it if you want.
Yes! I really missed using the commandline for A-/B-convertion. I'd 
really appreciate that!
> The method of choice today is not to use balloons or pistols, but
> a logarithmic sweep signal and deconvolution with the inverse sweep.
> This will give you
> 1. A more correct result, since the spectrum of the test signal is
> exactly known (apart from the speaker used to play it),
> 2. A much better S/N ratio since the test signal can be several
> seconds long so it has much more energy at the same amplitude as a
> popping balloon.

I know this procedure from speaker development. But I wonder which tool 
is used to do the deconvolution in this scenario. What do you use?

The downside is that a good speaker with a wide frequency range is 
certainly much heavier and bigger than my package of paper bags. :D (Let 
alone the fact that amplification and signal generation could afford two 
additional devices - and an additional power source.) I already thought 
in the same direction, but I dropped it in favor of mobility.

Although, I already realised, that my paper bag impulse is not the best 
option for audio quality and partly needs heavy postprocessing…
But I managed to accelerate finding the correct equalization setup, at 

Using the right head-/tailcut parameter, it's possible to isolate just 
the impulse in the "LSP IR Mono" plugin. Stimulated by a pink noise 
signal one can spectrum analyse the output and find out which 
frequencies are underrepresented in the impulse. With IEMs "MultiEQ" one 
can fill the gaps. Still - as long as you're trying to combine an 
ambisonics recording with an afterwords integrated mono signal, f.e. - 
it's necessary to use hearing and comparing with original recordings. 
Altogether, a not-so easy and quick task. Besides the fact, that heavy 
equalizing won't improve S/N ratio, at all.


More information about the Linux-audio-user mailing list