Convolution / IR for dummies

[che_guitarra]

After owning the Renn Maxx bundle for quite a while, I just tried firing up my Waves IR-L for the first time to see what it actually does. Although only three presets - I can see this becoming a key plugin after only 5 minutes of tweaking, and am very interested in trying more.

But what is convolution / IR exactly? I don't mean in terms of the audible effect it has (it's quite easy to hear that) - I don't understand what's going on in terms of people posting .wav files and uploading them into into the plugin. I can't seem to get the Waves IR-L to import anything - i'm guessing Waves may have it's own format as a protection deterent (as opposed to .wav's I see everywhere). Is that true or am I doing it wrong?


So what's the go? What are other IR platforms/interfaces I can use within PT8/RTAS/Mac? What are the better one's, what are the free one's, and what's the difference between them?

A 101 in the topic would be fantastic - it's still a bit hazy to me

[daveb63]

here's nice article.

Convolution reverb | How convolution works, how to make recordings suitable for use with convolution reverbs

[FastlanePro]

Convolution is fancy math that allows us to determine the time-varient response of a system given a time-varying input.

I know. Blah blah...

Think of it this way:
1) At any given snapshot in time - we can figure out the instantaneous spectral content of our audio signal - whether a sine wave or complex waveform.
2) We then can figure out what effect that instantaneous snapshot (basically 0sec long) will have on the room. That response will be the typical complex ADSR type reverberant decay.
3) Now - we have to do that over and over for each "snapshot" of our audio, and zipper it all back together.
4) Keeping in mind that we also have to track the interaction of the previous reverberant decay envelopes.

CPUs do this all by approximation. We take one or more representative impulse responses of a particular room and pretend that they represent the total response of that space. Then, we do fast fourier transform algorithms at a predetermined resolution (i.e. shortness of time) to determine the frequency response "snapshots" and pretend that completely represents our audio signal.

We do all the math, then zipper everything back up. The math itself isn't as hard as you would think. The process of time-domain convolution actually becomes simple multiplication in the frequency domain. You just have to do it over an infinite continuum. The approximation often sounds pretty kickass.

I apologize to those out there for whom this is overly simplistic and full of holes. Based on the OPs questions - I figured this would be enough of an explanation rather than trying to explain Laplace Transforms, etc.

Hope it helps rather than harms,
Cheers.

Lane

[rjacobsen]

Great article daveb63, very informative!

Thanks, rjacobsen