Home |
Search |
Today's Posts |
#1
|
|||
|
|||
can you use phase to create an a capella track?
hello,
i've been listening to my old Ella Fitzgerald cd's on headphones. I notice that the instrument groups are often very separated into the left and right channels, with the vocals dead center. i think this has something to do with "3 track" recordings that were done for a while. could you apply some sort of phase processing in a daw to cut out the background tracks and keep the vocals. there are software plugins that remove the vocals by eliminating info that is common to both channels. i'm looking to do the opposite: keep the stuff that is the same in both channels, and eliminate the different stuff. |
#2
|
|||
|
|||
No this does not work. I suggested to someone who asked this question
in the past to do just what you describe and see if it works. I was ridiculed by some dude who had way to much free time on his hands and did a math proof to show how brilliant he was and thump his chest. However, he was correct and it made sense after looking at his proof as to why it does not work. Do a scan of the newgroup and you may find it. |
#4
|
|||
|
|||
thanks, that clarifies things.
|
#6
|
|||
|
|||
|
#7
|
|||
|
|||
In article ,
says... You need to run the stereo mix into a Mid/Side decoder. This will actually encode it into a Mid (center info) and the Side (both left and right only info). From there you can just kill the Side and you're left with what ever is left in the center. If the Ella material is vocal right down the center and the instruments on the hard left and right as you describe, then you should have a very clean "vocal only" track. If it's more like a mix of today, where instruments are panned not only hard left or right but also mid left/right and center, than you'll still hear those instruments. This doesn't make sense. A "mid-side decoder" does nothing but output L+R (mid) and L-R (side). Subtract side from mid, and you get: L + R - (L - R) L + R - L + R 2R Don't trust the math; make yourself a simple "left - center - right" track and hear it for yourself. If you're hearing a "perfectly isolated vocal" in mastering, it was mostly on one channel to begin with. The Dolby Pro-Logic decoder is a bit more sophisticated. Below are results I obtained from it, as posted back in August 2004: ----- Experiment 1: A CD track of me saying "left, right, center" panned appropriately. At first blush, only "center" is heard from the center; upon closer investigation, "left" and "right" are still present, but at a much lower volume and with nothing present over (guessing) 250Hz. Interestingly, some of the background noise and breath right before each word appear to be present, full spectrum, in the center. Experiment 2: If I take the above and mix in continuous sounds panned left, right, and center, the "main vocal" *still* is panned mostly to the appropriate speaker. However, the "background vocals" are still present in the center speaker. My surround setup is horrible - the center speaker's actually a different type of speaker altogether - so I can't be sure if it's at the same level as the side speakers or not. It seems to be either slightly softer or slightly rolled off. ----- Adobe Audition does have a fairly fancy FFT-type center-extraction function, and there's also some plugin called "CenterCut" that does it. -- Jay Levitt | Wellesley, MA | I feel calm. I feel ready. I can only Faster: jay at jay dot fm | conclude that's because I don't have a http://www.jay.fm | full grasp of the situation. - Mark Adler |
#8
|
|||
|
|||
wrote in message
ups.com hello, i've been listening to my old Ella Fitzgerald cd's on headphones. I notice that the instrument groups are often very separated into the left and right channels, with the vocals dead center. i think this has something to do with "3 track" recordings that were done for a while. could you apply some sort of phase processing in a daw to cut out the background tracks and keep the vocals. You probably can't eliminate the instrumentals but you might increase the prominance of the center-channel vocals by pushing the stereo signal through a Dolby Pro Logic decoder and capturing the center channel. |
#9
|
|||
|
|||
|
#10
|
|||
|
|||
On 3 Aug 2005 14:12:44 -0700, "Mark" wrote:
there are software plugins that remove the vocals by eliminating info that is common to both channels. i'm looking to do the opposite: keep the stuff that is the same in both channels, and eliminate the different stuff. Interesting question.... please ask it over at comp.dsp. Adobe Audition 1.5 has a plugin that makes a fair go at extracting as well as suppressing centre channel. Simple phase tricks can only suppress it. |
#11
|
|||
|
|||
On 8/3/05 3:41 PM, in article , "Allen
Corneau" wrote: On 8/3/05 12:30 PM, in article , " wrote: i've been listening to my old Ella Fitzgerald cd's on headphones. I notice that the instrument groups are often very separated into the left and right channels, with the vocals dead center. i think this has something to do with "3 track" recordings that were done for a while. could you apply some sort of phase processing in a daw to cut out the background tracks and keep the vocals. there are software plugins that remove the vocals by eliminating info that is common to both channels. i'm looking to do the opposite: keep the stuff that is the same in both channels, and eliminate the different stuff. Yes, this is possible, quite easily. We sometimes do this in mastering. The results will vary from "almost perfectly isolated" to "better than the full stereo, but not by much". You need to run the stereo mix into a Mid/Side decoder. This will actually encode it into a Mid (center info) and the Side (both left and right only info). From there you can just kill the Side and you're left with what ever is left in the center. Ummmm... IIRR the MID of MS is L+R ... Not sure how you derive both C-L -and- C-R without first deriving C-only... Which is the thing you're trying to get at in the first place. How does the math work to get from L+R And L-R To Anything ressembling Center-Only? |
#12
|
|||
|
|||
no, that trick doesn't work.
This and more is explained in detail here http://www.csp-audio.com/vocalremoval_article.htm schreef in bericht ups.com... hello, i've been listening to my old Ella Fitzgerald cd's on headphones. I notice that the instrument groups are often very separated into the left and right channels, with the vocals dead center. i think this has something to do with "3 track" recordings that were done for a while. could you apply some sort of phase processing in a daw to cut out the background tracks and keep the vocals. there are software plugins that remove the vocals by eliminating info that is common to both channels. i'm looking to do the opposite: keep the stuff that is the same in both channels, and eliminate the different stuff. |
#13
|
|||
|
|||
On 8/3/05 6:39 PM, in article ,
"SSJVCmag" wrote: Ummmm... IIRR the MID of MS is L+R ... Not sure how you derive both C-L -and- C-R without first deriving C-only... Which is the thing you're trying to get at in the first place. How does the math work to get from L+R And L-R To Anything ressembling Center-Only? Actually, a M/S decoder works like this: M=L+R S= L+(-R) If you put a stereo signal into this device then you get: M= The sum of the left and right (mono) S= What's left over after the left channel is added to a polarity-reversed right channel. Anything common between the two will cancel out (to varying degrees), leaving you you left-only and right-only information. If you turn off the Side signal, you get what's left as "center" information. Don't get me wrong, I'm NOT saying that all you'll be left with is only the information the the EXACT center, just that center info will be strong and things are panned away from center will be reduced in volume to where anything hard left or hard right will be gone. Does that help explain it? Perhaps I should send you some small clips of audio to prove it to you. Better yet, try it out for yourself. Allen -- Allen Corneau Mastering Engineer Essential Sound Mastering Houston, TX |
#14
|
|||
|
|||
Allen Corneau wrote:
Actually, a M/S decoder works like this: M=L+R S= L+(-R) If you put a stereo signal into this device then you get: M= The sum of the left and right (mono) S= What's left over after the left channel is added to a polarity-reversed right channel. Anything common between the two will cancel out (to varying degrees), leaving you you left-only and right-only information. If you turn off the Side signal, you get what's left as "center" information. Don't get me wrong, I'm NOT saying that all you'll be left with is only the information the the EXACT center, just that center info will be strong and things are panned away from center will be reduced in volume to where anything hard left or hard right will be gone. No it won't. A signal panned hard left or right will at most be attenuated 6dB relative to a signal at the same level in both L and R. -6dB is hardly "gone" ! In fact it will be exactly what you get when you listen to the mono mix on (for example) an AM radio. To get it to vanish completely in the mono mix a signal would have to be in equal level and opposite polarity in L and R, and that's not what panned hard left or hard right means. -- Anahata -+- http://www.treewind.co.uk Home: 01638 720444 Mob: 07976 263827 |
#15
|
|||
|
|||
On 8/4/05 5:08 PM, in article , "Allen
Corneau" wrote: On 8/3/05 6:39 PM, in article , "SSJVCmag" wrote: Ummmm... IIRR the MID of MS is L+R ... Not sure how you derive both C-L -and- C-R without first deriving C-only... Which is the thing you're trying to get at in the first place. How does the math work to get from L+R And L-R To Anything ressembling Center-Only? Actually, a M/S decoder works like this: M=L+R S= L+(-R) If you put a stereo signal into this device then you get: M= The sum of the left and right (mono) S= What's left over after the left channel is added to a polarity-reversed right channel. Anything common between the two will cancel out (to varying degrees), leaving you you left-only and right-only information. If you turn off the Side signal, you get what's left as "center" information. Don't get me wrong, I'm NOT saying that all you'll be left with is only the information the the EXACT center, just that center info will be strong and things are panned away from center will be reduced in volume to where anything hard left or hard right will be gone. Well, in a word... No. M = L+R Simple ALL-left and ALL-right is there, no 'reduced levels' are there. Levels get reduced L or R when you do the recombining. Again, M is NOT the same as C. Does that help explain it? Perhaps I should send you some small clips of audio to prove it to you. You can if you want, I could also send you 20 years of various MS work and we could compare/contrast... But that won't change the math. |
#16
|
|||
|
|||
On 8/4/05 7:02 PM, in article ,
"SSJVCmag" wrote: Don't get me wrong, I'm NOT saying that all you'll be left with is only the information the the EXACT center, just that center info will be strong and things are panned away from center will be reduced in volume to where anything hard left or hard right will be gone. Just a note, what you describe here is pretty much the trick that Dolby Pro-Logic pulls off to derive a fairly believable CENTER channel from a stereo pair. It doesn't quite happen the simple way you described though. |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
Speaker Phase? | Car Audio | |||
Cool Edit Pro 2.1 Questions | Pro Audio | |||
Countdown on track one | Pro Audio | |||
Blindtest question | High End Audio | |||
Negative/Positive Phase Shift in a Transformer | Pro Audio |