Home |
Search |
Today's Posts |
#1
Posted to rec.audio.pro
|
|||
|
|||
Hilbert transform to "fake" X/Y from a mono source.
I recently wrote a Hilbert transformer* for .wav files, as an exercise. So I got curious - since X/Y is basically "acoustic quadrature", how would it sound to spread a mono source left, with the Hilbertized signal panned right? *does a 90 degree phase shift on the original signal - the algorithm is painfully poorly exposed on the Wikipedia page - took me three tries to find it. This seems to work remarkably well. I know it's not really stereo, but it's the best fake I've been able to find so far. One should be able to use a fairly simple kernel with a convolver to produce this. Thoughts? -- Les Cargill |
#2
Posted to rec.audio.pro
|
|||
|
|||
Hilbert transform to "fake" X/Y from a mono source.
On Thu, 22 Sep 2011 14:27:25 -0500, Les Cargill
wrote: I recently wrote a Hilbert transformer* for .wav files, as an exercise. So I got curious - since X/Y is basically "acoustic quadrature", how would it sound to spread a mono source left, with the Hilbertized signal panned right? *does a 90 degree phase shift on the original signal - the algorithm is painfully poorly exposed on the Wikipedia page - took me three tries to find it. This seems to work remarkably well. I know it's not really stereo, but it's the best fake I've been able to find so far. One should be able to use a fairly simple kernel with a convolver to produce this. Thoughts? I think I'd like to hear it. Can you post a sample? d |
#3
Posted to rec.audio.pro
|
|||
|
|||
Hilbert transform to "fake" X/Y from a mono source.
Don Pearce wrote:
On Thu, 22 Sep 2011 14:27:25 -0500, Les Cargill wrote: I recently wrote a Hilbert transformer* for .wav files, as an exercise. So I got curious - since X/Y is basically "acoustic quadrature", how would it sound to spread a mono source left, with the Hilbertized signal panned right? *does a 90 degree phase shift on the original signal - the algorithm is painfully poorly exposed on the Wikipedia page - took me three tries to find it. This seems to work remarkably well. I know it's not really stereo, but it's the best fake I've been able to find so far. One should be able to use a fairly simple kernel with a convolver to produce this. Thoughts? I think I'd like to hear it. Can you post a sample? d I don't have anything publishable saved. Lemme see. -- Les Cargill |
#4
Posted to rec.audio.pro
|
|||
|
|||
Hilbert transform to "fake" X/Y from a mono source.
Poster wrote:
"Les Cargill" wrote in message ... I recently wrote a Hilbert transformer* for .wav files, as an exercise. So I got curious - since X/Y is basically "acoustic quadrature", how would it sound to spread a mono source left, with the Hilbertized signal panned right? Would it be mono compatible? If you sum the two, you get comb filtering. To make it mono compatible, you mix a sum and difference of the original signal and the all-passed signal, pan them right and left. This means when you sum it to mono, you get the original signal. This is how the classic Orban stereo simulator works. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
#5
Posted to rec.audio.pro
|
|||
|
|||
Hilbert transform to "fake" X/Y from a mono source.
On Fri, 23 Sep 2011 09:32:49 -0400, Scott Dorsey wrote
(in article ): Poster wrote: "Les Cargill" wrote in message ... I recently wrote a Hilbert transformer* for .wav files, as an exercise. So I got curious - since X/Y is basically "acoustic quadrature", how would it sound to spread a mono source left, with the Hilbertized signal panned right? Would it be mono compatible? If you sum the two, you get comb filtering. To make it mono compatible, you mix a sum and difference of the original signal and the all-passed signal, pan them right and left. This means when you sum it to mono, you get the original signal. This is how the classic Orban stereo simulator works. --scott I love rec.audio.pro! Regards, Ty Ford --Audio Equipment Reviews Audio Production Services Acting and Voiceover Demos http://www.tyford.com Guitar player?:http://www.youtube.com/watch?v=yWaPRHMGhGA |
#6
Posted to rec.audio.pro
|
|||
|
|||
Hilbert transform to "fake" X/Y from a mono source.
Poster wrote:
"Les wrote in message ... I recently wrote a Hilbert transformer* for .wav files, as an exercise. So I got curious - since X/Y is basically "acoustic quadrature", how would it sound to spread a mono source left, with the Hilbertized signal panned right? Would it be mono compatible? Good question. I'd think so. -- Les Cargill |
#7
Posted to rec.audio.pro
|
|||
|
|||
Hilbert transform to "fake" X/Y from a mono source.
I recently wrote a Hilbert transformer* for .wav files, as an exercise. So I got curious - since X/Y is basically "acoustic quadrature", how would it sound to spread a mono source left, with the Hilbertized signal panned right? Would it be mono compatible? If you sum the two, you get comb filtering. 1) An X/Y pair is not acoustic quadrature. The mics are coincident so there is little or no phase information, it's all intensity information. 2) A HT is a 90 deg shift at all frequencies. It's not very intuitive. I don't think you will get comb filtering summing to mono because it is 90 deg at all frequencies. Comb filtering is caused by a fixed time delay that causes a variable phase shift that varies through 0 and 180 causing additions and cancellations. An HT is 90 deg at all frequencies. 3) All that being said, it probably does spread a mono source pretty well. Mark |
#8
Posted to rec.audio.pro
|
|||
|
|||
Hilbert transform to "fake" X/Y from a mono source.
2) A HT is a 90 deg shift at all frequencies. It's not very intuitive. I
don't think you will get comb filtering summing to mono because it is 90 deg at all frequencies. Comb filtering is caused by a fixed time delay that causes a variable phase shift that varies through 0 and 180 causing additions and cancellations. An HT is 90 deg at all frequencies. The only practical way to achieve it is with a pair of all-pass filters, of phi and phi+90 (or phi-90). These were uised in most matrix-quad systems. 3) All that being said, it probably does spread a mono source pretty well. One way to "spread" a mono source -- without altering it in any way! -- is to synthesize hall reverb and play the reverb through additional speakers. You'd probably be amazed at the improvement. |
#9
Posted to rec.audio.pro
|
|||
|
|||
Hilbert transform to "fake" X/Y from a mono source.
MarkK wrote:
I recently wrote a Hilbert transformer* for .wav files, as an exercise. So I got curious - since X/Y is basically "acoustic quadrature", how would it sound to spread a mono source left, with the Hilbertized signal panned right? Would it be mono compatible? If you sum the two, you get comb filtering. 1) An X/Y pair is not acoustic quadrature. The mics are coincident so there is little or no phase information, it's all intensity information. I didn't mean *literally* quadrature. I put that in scare quotes for a reason Both involve the phrase "90 degrees". I'm not being severely rigorous here, just musing. it's more an art concept than a science one. So the idea is to use the HT to "simulate" the second mic of an X/Y pair. And yes, it's a naive initial idea, not something derived from the ... solution of differential equations but apparently, Bob Orban did something similar, so it's not a completely stupid idea And they say "it's intensity", but each card element is a "phase" based cancel-ifier, so ... it might be interesting to see the convolution between the two mics in an X/Y, but I don't know where to put the impulse... 2) A HT is a 90 deg shift at all frequencies. It's not very intuitive. I don't think you will get comb filtering summing to mono because it is 90 deg at all frequencies. Comb filtering is caused by a fixed time delay that causes a variable phase shift that varies through 0 and 180 causing additions and cancellations. An HT is 90 deg at all frequencies. 3) All that being said, it probably does spread a mono source pretty well. Mark -- Les Cargill |
#10
Posted to rec.audio.pro
|
|||
|
|||
Hilbert transform to "fake" X/Y from a mono source.
William Sommerwerck wrote:
2) A HT is a 90 deg shift at all frequencies. It's not very intuitive. I don't think you will get comb filtering summing to mono because it is 90 deg at all frequencies. Comb filtering is caused by a fixed time delay that causes a variable phase shift that varies through 0 and 180 causing additions and cancellations. An HT is 90 deg at all frequencies. The only practical way to achieve it is with a pair of all-pass filters, of phi and phi+90 (or phi-90). These were uised in most matrix-quad systems. 3) All that being said, it probably does spread a mono source pretty well. One way to "spread" a mono source -- without altering it in any way! -- is to synthesize hall reverb and play the reverb through additional speakers. You'd probably be amazed at the improvement. Absolutely. Indeed, this, then a reverb sounds pretty lifelike. -- Les Cargill |
#11
Posted to rec.audio.pro
|
|||
|
|||
Hilbert transform to "fake" X/Y from a mono source.
Don Pearce wrote:
On Thu, 22 Sep 2011 14:27:25 -0500, Les Cargill wrote: I recently wrote a Hilbert transformer* for .wav files, as an exercise. So I got curious - since X/Y is basically "acoustic quadrature", how would it sound to spread a mono source left, with the Hilbertized signal panned right? *does a 90 degree phase shift on the original signal - the algorithm is painfully poorly exposed on the Wikipedia page - took me three tries to find it. This seems to work remarkably well. I know it's not really stereo, but it's the best fake I've been able to find so far. One should be able to use a fairly simple kernel with a convolver to produce this. Thoughts? I think I'd like to hear it. Can you post a sample? d Try this: http://soundcloud.com/les-cargill/phase It's: - Original panned center. - HT left - Polarity-inverted-HT right. The first half is just the "fake stereo"; the second half is fake stereo plus some room reverb. I dunno. Sounds kinda weird by itself now, kinda "twist your head off"-ey. -- Les Cargill |
#12
Posted to rec.audio.pro
|
|||
|
|||
Hilbert transform to "fake" X/Y from a mono source.
On Sat, 24 Sep 2011 15:15:12 -0500, Les Cargill
wrote: Don Pearce wrote: On Thu, 22 Sep 2011 14:27:25 -0500, Les Cargill wrote: I recently wrote a Hilbert transformer* for .wav files, as an exercise. So I got curious - since X/Y is basically "acoustic quadrature", how would it sound to spread a mono source left, with the Hilbertized signal panned right? *does a 90 degree phase shift on the original signal - the algorithm is painfully poorly exposed on the Wikipedia page - took me three tries to find it. This seems to work remarkably well. I know it's not really stereo, but it's the best fake I've been able to find so far. One should be able to use a fairly simple kernel with a convolver to produce this. Thoughts? I think I'd like to hear it. Can you post a sample? d Try this: http://soundcloud.com/les-cargill/phase It's: - Original panned center. - HT left - Polarity-inverted-HT right. The first half is just the "fake stereo"; the second half is fake stereo plus some room reverb. I dunno. Sounds kinda weird by itself now, kinda "twist your head off"-ey. That's really interesting. Through speakers it sounds as you describe. But through headphones, the first half just sounds like mono, while in the second I can hear the fake reverb. d |
#13
Posted to rec.audio.pro
|
|||
|
|||
Hilbert transform to "fake" X/Y from a mono source.
Les Cargill wrote:
Scott Dorsey wrote: To make it mono compatible, you mix a sum and difference of the original signal and the all-passed signal, pan them right and left. This means when you sum it to mono, you get the original signal. This is how the classic Orban stereo simulator works. Aha! There we go. I knew it couldn't be new. Well, in one sense it's new in that the transform gives you a much cleaner and more accurate phase shift than an analogue all-pass network gives you. So yes, the comb filtering issues are eliminated which is something I hadn't thought about. Phase shift networks for generating single sideband audio have been around since the early fifties, but doing it in the digital domain to generate single sideband audio that actually sounds decent and has low group delay is a new thing. Likewise SQ and QS matrix quad decoding, which the Hilbert transform can do far more cleanly than the all-pass networks. It is interesting, though, to note that almost all audio design techniques we use today can be found in 1930s copies of the Bell System Technical Journal in some form or another. It's amazing how little there is new under the sun. --scott -- "C'est un Nagra. C'est suisse, et tres, tres precis." |
Reply |
Thread Tools | |
Display Modes | |
|
|