## What to use for deconvolution (library, software, ...)?

13 posts
• Page

**1**of**1**- KVRian
- 1289 posts since 26 Apr, 2004, from UK

Deconvolution in what sense, as it seems that you are confusing convolution with what people market as non linear convolution (which means it's not even close to convolution)?

- KVRian
- 574 posts since 28 May, 2010, from Finland

Then again, I also think that the deconvolution operation may be simple enough to actually write it oneself, if one has a high-quality FFT implementation already in hand.

https://en.wikipedia.org/wiki/Deconvolution

Unless the division F=H/G can use some meaningful optimization?

https://en.wikipedia.org/wiki/Deconvolution

Unless the division F=H/G can use some meaningful optimization?

- KVRian
- 1289 posts since 26 Apr, 2004, from UK

You could do that (which is version 0 of deconvolution), but you end up with division by zeros lots of the time.

You need to accommodate for the noise, the zeros...

You need to accommodate for the noise, the zeros...

- KVRian
- 574 posts since 28 May, 2010, from Finland

Miles1981 wrote:You could do that (which is version 0 of deconvolution), but you end up with division by zeros lots of the time.

You need to accommodate for the noise, the zeros...

How ought the zeros be treated? Can I merely replace them with zeros?

Since convolution of Fourier transforms is zero <=> convolution of their derivatives is zero, i.e. at least one of the original contains zero at that point and thus the output would also be a zero at that point?

- KVRian
- 574 posts since 28 May, 2010, from Finland

- KVRist
- 244 posts since 7 Feb, 2017

The IR magnitude spectra shouldn't be flat zero but rather the continuous magnitude spectra is rippling close to zero and on occasion touches zero if the FFT bins happen to align with those frequencies. Maybe raise the noise floor of the IR to compensate.

- KVRian
- 885 posts since 11 Nov, 2009, from Northern CA

So let me ask what may be a very naïve question.

Can you not convolve an impulse signal with a unit impulse (1 followed by as many zeroes as needed to complete the representation) and get back the impulse?

So, for example, if you have a convolver with encrypted impulse signals, you could send the unit impulse into that plug-in and out would come the unencrypted impulse signal. Is that not correct? Seems to me that it should work (at least "in theory" ).

Can you not convolve an impulse signal with a unit impulse (1 followed by as many zeroes as needed to complete the representation) and get back the impulse?

So, for example, if you have a convolver with encrypted impulse signals, you could send the unit impulse into that plug-in and out would come the unencrypted impulse signal. Is that not correct? Seems to me that it should work (at least "in theory" ).

- KVRian
- 1289 posts since 26 Apr, 2004, from UK

dmbaer wrote:So let me ask what may be a very naïve question.

Can you not convolve an impulse signal with a unit impulse (1 followed by as many zeroes as needed to complete the representation) and get back the impulse?

So, for example, if you have a convolver with encrypted impulse signals, you could send the unit impulse into that plug-in and out would come the unencrypted impulse signal. Is that not correct? Seems to me that it should work (at least "in theory" ).

You can, but you don't get the "best" result. One good way of doing such a thing is to use a Hadamard signal.