Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Deconvolution using sine sweeps (Read 4612 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Deconvolution using sine sweeps

Hello.
Somewhere on the webpages of professor Angelo Farina (pcfarina.eng.unipr.it), he said that an "inverse filter" for a sine sweep could be made by simply reversing the sweep in time. I wanted to try it so I made an exponential sine sweep of 10 seconds in length which went from 20 to 22000 Hz. To compensate for the exponential frequency curve, I was also gradually changing the amplitude in a square root relationship to the changes in frequency -- i.e. while the ratio of the final frequency to the initial one was 22000/20, the ratio of the final and initial amplitudes was the square root of that. Finally, I made a 50 ms fade-in and fade-out to prevent clicks. When I convolved another sample with the sine sweep and then with its "time reversed" copy, the amplitudes of all the frequency bands seemed okay to me but their phases were not preserved. My question is: Is there a way to do something similar which could retain the original phases?
Thanks in advance.
Petr

Deconvolution using sine sweeps

Reply #1 –
Quote from: Petr P. on
My question is: Is there a way to do something similar which could retain the original phases?

My apologies to all of you ... The problem probably isn't caused by the sweep itself but by the convolution algorithm. I've found an old copy of CoolEdit which does the convolution explicitly in the time domain, I tried it with that, and the result was almost identical to the original. What I was using beforehand was Sonic Foundry Acoustic Mirror which probably does the convolution via FFT -- for one thing, it's terribly fast (actually so fast that it makes me doubt about the quality); for another thing, when I subtracted the result from the original, instead of hearing silence, I could hear some very soft strange noises. In case Acoustic Mirror is to blame here, then I'll have to look for another piece of software which does the convolution in the time domain -- the disadvantage of CoolEdit is that you have to load ".imp" files there and can't just give a ".wav" file or something more common.
Once again, sorry for disturbing. Hope I'll eventually work it out.
Greetings.
Petr