You Say you Want a Convolution…

Wednesday, April 30, 2008| by Will Chen


Math and physics constantly amaze me. I remember being in school daydreaming as the teacher would lecture topics which could have been in another language as far as I cared. My friends and I rolled our eyes repeating the mantra “we’ll never use this stuff in the real world!” Flash forward 20 years and the kids that were paying close attention have created an entire industry surrounding mathematically modeling acoustical processing. One of the most interesting applications in my opinion is using an Impulse Response to mimic an environment, otherwise known as Convolution Processing.

How It Works

In the most basic layman’s terms, a sample is taken of a sound in an environment. The resulting waveform is manipulated to remove the source sound leaving only the ambience. This ambience only waveform can then be “convolved” to other sound sources to make them sound as if they existed in the initial environment. Think of it as an audio version of the green screen technique used in the movies to place actors, filmed in a studio, appear as if they’re zooming down the highway. Got it? Cool. Now, to blow your mind a little, think of “environment” in the most liberal definition as basically anything that can have an effect on a waveform. See where I’m going? Impulses can be made of effects processors, speaker cabinets, preamps, etc. Convolution processing is extremely powerful and can result in some incredibly realistic sounds which very closely emulate acoustical environments. Convolution processing is notoriously processor intensive. However, as processor power continues increasing the average home computer can handle multiple instances without excessive strains.

Free Lunch!


Super Impulse Response (S.I.R.) was developed by Christian Knufinke and was an immediate hit due to its professional interface and great price…free. As with most freeware, there’s not much of a manual. Fortunately, the controls are fairly self explanatory and the typical VST installation (copy and paste) was smooth. There has been some criticism regarding the fixed 8960 sample latency, however I use SONAR which automatically compensates while using it to post process. It should be noted though that due to the latency, don’t expect to use SIR1 in real time on an input. For more information or to download SIR1, visit their home page:


keFIR was developed by Piotr “Habib” Pyrzanowski and features and extremely simplified interface which will allow even the most novice user the ability to experiment with convolution processing. While sacrificing some advanced features, keFIR advertises 0 latency processing. Unfortunately, I couldn’t get keFIR to work in my otherwise 100% stable environment and experienced dropouts on playback even with my soundcard set to it’s max latency. Bummer! For more information or to download kefir, visit their homepage:


A convolution processor is useless without impulses and NoiseVault is happy to deal our fix. Samples are categorized by description and range from Real Spaces and Springs to Roland and Lexi, with the later being samples created by recording impulses through some “industry standard” processors. On average, the quality is great and there are a few real gems as well as some cool quirky stuff. There’s also a decent forum and some articles for those who want to create their own impulses. For more information or to download a plethora of great impusles, visit the NoiseVault homepage:

Filed Under: Articles