I’m not sure about the first one but the other two I’m using in a song each. I tend to create an IR with a specific project in mind and then archive it separately.
I tend to work with white noise, my sample collections of sounds I’ve recorded myself and whichever type of synthesis I can use that can produce sounds which do not reinforce too much of a specific frequency. So far it’s been pretty trial and error. I feel like I haven’t fully developed a technique for it but here’s what I remember of my thought process for creating the IR’s I can recall using:
The second one was made for vocals, I wanted to create an effect similar to Phil Collin’s reverse reverb vocal effect on In The Air Tonight, but still slightly weirder and more modern. The exponential attack of the noise creates the reverse effect and that click makes for a delay with a specific EQ profile. They’re then panned to widen the sound.
The third one is used as this huge ambient reverb swell that fades in at the end of the song it’s used in. The sweeping pitch of the sample I used creates a sweeping resonance effect in the reverb tail where each frequency comes in at a different time. The samples going on in stereo were used to create more definition and width to the sound.
@zoundsabar’s post made me think about the core of what convolution does to sound and I got inspired to do use a sine sweep as an IR. I think it’s a pretty cool sound that sort of exposes what is happening to the sound when you convolve.
This is the IR, as mention a sweeping sinewave:
Here is the dry sound, a pedal steel guitar:
Here it is going through the sine sweep, completely wet:
You can tell that each transient gets replaced with what sounds like a resonant filter sweep, while the notes with a softer attack seem to have slightly less resonance.
If you look at this sound in a spectrogram, you will see each overtone of the guitar be shifted in time to the same point in time that frequency occurs in the IR (it looks a bit like as if you tilted the sound, actually).
Each frequency only occurs once in this case so it doesn’t build up much resonance but if you were to prolong any frequency it would be like adding the same overlapping sound a few milliseconds after the original. A way I sometimes think of convolution is giving each millisecond snippet of each frequency the envelope of the corresponding frequency of the IR, or perhaps adding that envelope on on top of the existing dynamics of it.
My technical understanding of it is limited though, I can only really explain how I experience working with it.
I’m really curious now about creating drum sounds with convolution though. Sending noise bursts, pops, scratches into a drum hit as an IR is an interesting proposition, perhaps combined with frequency shifting for further dynamics. I’ll try to experiment with this and report back!
