This post focuses on convolution as it applies to audio – the capturing and reapplication of the qualities of sound.
A few years ago some new-fangled audio processing engines started to leak out of the lab and into the market involving a process called convolution. The name convolution comes from the type of math involved, but applying the process to audio signals provides for some pretty interesting opportunities.
Essentially, the process of convolution really requires two things: deconvolution and convolution. It’s actually a little backwards from the conventional use of the prefix ‘de’, although that again comes from its math roots. Deconvolution is the process of capturing the audio changes imposed upon a signal when fed through an environment, whether that be a physical room or a piece of equipment. Convolution is the process of applying those changes to a different signal. So, in essence, deconvolution would be ‘sampling’ the way an environment affects the signal, whereas convolution would be making that sample into an effect, which is applied to other signals.
Hey I’m not Lost Yet!
So convolution can be used to reproduce ambience, reverb, echos, dynamics, tone, frequency response and other time and frequency qualities. This is an awful lot of information. But, what it cannot reproduce is harmonic distortion. This means that, while you can capture the frequency response and compression of a guitar amp, for instance, you probably won’t be able to capture the harmonic distortion of the amp.
What convolution can do is sample a large variety of situations for reuse later. What the majority of people who use convolution effects do is simply purchase or download convolution libraries to use in their productions. But many of the convolution effects packages actually come with deconvolution programs, allowing adventurous engineers and producers to sample their favorite gear and rooms. This is powerful. Some of these packages, likes the Waves Q-Clone, are actually designed to make his user-friendly (although that package is only EQ-oriented). The subject of how to go about sampling with deconvolution is going to the subject of a different post, as it’s a big process in and of its self.
The natural application for convolution is reverb and ambience. There are lots of excellent, professional convolution packages out there for reverb including the Waves IR bundle, the Altiverb bundle, the convolution package that comes with the new Apple Logic, and Voxengo’s Pristine Space. Also available is the free SIR1 Convolution VST plugin and the not so free SIR2 (I used SIR1 quite a bit before I switched over to OSX).
There’s even hardware running convolution such as Focusrite’s pricey but neat Liquid Channel.
There’s a big library of free convolution samples (of varying degrees of quality) available at Noisevault. I’m a fan of the Lexicon samples available there. The cool thing about products like SIR and the stuff available at Noisevault is that people are trying convolution out on all sorts of stuff, from amps to speakers to compressors to EQ. Yeah! It works on compressors too.
Convolution is one of those things that, the more you play with it, the more you’ll get out of it.
One of these days I want to go into my favorite studio and sample the room and some of the vintage gear they have with a deconvolver. When I do that, you’d better bet I’ll feature that process right here on The Stereo Bus!