Applying the inverse of an impulse response?

[MaxC]

Thanks guys for taking the time to respond so thoroughly. As I suspected, this process appears to be well beyond my comfort zone. Considering it's more of a luxury than a necessity for my purposes, I'll leave the matter for any others who may be more inclined to pursue it.

Cheers .

[xoxos]

after i was finished being clever yesterday, i realised that i was wrong. i was right too.. the phase data is still good. i remembered later, only the magnitude is used for convolution, the phase data is not used at all

as you wouldn't want to use this unless you were undoing something you had done earlier, it could be a while before it's available. if you need it for one specific file task, i expect you can be accomodated.

[Rock]

I stumbled upon this looking to find a way to bring uniformity to various types of pres, so that my outboard pre's could be made to sound like my mixing board's pres and vice versa. I figured it might not be perfect, but might result in something close.

It seemed only natural that the coloration of the pre would need to be removed first. If anyone can explain a (free) set of steps to take, I would appreciate it.

[dewgong]

Does this process have any creative uses or is it just a creative way to idle away a few hours? If you have the original dry sound and a convolved version what utility is there in deconvolving it unless you've suffered data loss?

I can see that you just need to undo the multiplication of the IR with the audio signal and being able to manipulate the raw data would prove to be extremely interesting but aside from that I dont really understand why there's a $400 plugin for this

[nineofkings]

That's the problem: you don't have the original dry sound. You have the convolved sound and the impulse. This is good for things like impulses of preamps where there's no way to obtain the dry sound, as you literally have to use the preamp to record sound in the first place.

[Gribs]

The field of "inverse problems" in mathematical sciences and engineering is rich and deep.

http://en.m.wikipedia.org/wiki/Inverse_problem

It encompasses the various forms of medical and radar imaging, sonar, ultrasound imaging, geophysical imaging with acoustics or electromagnetic waves, and many other topics of deconvolution and "unfolding", not necessarily of linear operators though the most common topics are linear operators such as Fredholm integral operators (e.g. convolution) where you have an operator over a function space

Lu=v

and you know L and v and want to compute u.

One of the common properties of the operator is that it is a bounded operator, which means that it throws out information contained in u. For example in linear acoustics the response of a speaker cavity might squash out frequencies above some threshold to the point where they are indistinguishable from noise in the measurement. A simple inversion by FFT or other more complex harmonic method (decomposition in terms of basis functions other than sines and cosines) would yield noise. Numerically a matrix inversion may be required and if you try to reconstruct u with too many basis functions the matrix might end up being poorly conditioned and a small perturbation in the measured v (output sound) might yield large variations in the estimated values of u.

I did some work in inverse problems and linear operators on Hilbert spaces when I was a summer intern at Kodak (late 1989's before Kodak took its major dump) during grad school. There were problems in imaging and also in phototheory.

[Rock]

I have the gear I want to deconvolve.
I have the gear I want to convolve.
I have no source material to worry about, I was anticipating using something like a sweeping sine wave or pink noise, whatever is used to create impulse responses.

I guess I want to create my own version of something like the Antares mic modeller but for pres, well for mics too for that matter.