If you are starting out, keep it simple...Record the exact same thing on two tracks. Reverse phase on one of them. Voila. Magic Disappearance.

Play with that. Undertand it. THEN you can being to apply (or not) phase cancellation to your recordings.

While the process & the explanation may seem complex? It's really just a exercise in phase cancellation.

In our studio, we had a pair of studio speakers. We could selectively turn them on or off. Generally, they were kept muted so as the announcers & singers could utilize headphones comfortably. But when cutting music tracks, it was convenient to play back the rhythm tracks to the studio speakers. This was subsequently picked up by the live microphones recording the brass, strings, woodwinds. We would want that track to be fattened up & slightly enlarged. So we would do at least a second pass with the same musicians. To make the "section"sound more normal, I would reposition the musicians. But I would not touch nor reposition the microphones. This now provides me with two passes of brass or strings, woodwinds, etc. But the fold back from the studio speakers was extreme and infiltrated both tracks. When these two tracks were evenly combined in their phase inverted, the fold back bleed would magically disappear. This is only applicable if no microphones are moved. No levels are changed. No equalization tweaked. Any variables will keep the cancellation from properly canceling. So we actually ended up with a single track that was already 2 generations down even before the mix began.

If you want to look at this from a more mathematical standpoint, time delay & phase response is affected by moving musicians around. However, it is actually the room we are trying to cancel out. This is not a new technique. This goes way back beyond 30 years. Of course one must experiment with how loud the studio monitors can be cranked? But it is the summation of two channels that have been phase inverted and levels are adjusted between the two for the best null of background interference. Should someone accidentally bump into changing a position of a microphone but even 1 inch will prevent the phase cancellation from canceling anything. So things must be kept very highly consistent. You also don't necessarily want to daisychain anything. You want things to be paralleled. Everything must be duplicated exactly for the phase cancellation to work to an adequate extent.

If you have further questions, don't hesitate to drop me a private line or give me a call at 703-532-REEL (7335) with voicemail and I'll be happy to talk you through the procedure.

Queen of the Cancelers
Ms. Remy Ann David

The concept of complete phase cancellation is fairly simple. If you have one sound wave with peaks and valleys and another with exactly opposite peaks and valleys the two will cancel each other out.
http://www.opensourcejason.info/Blog/Phase
This can be done by establishing two seperate tracks of the same recording and reversing the phase on one of them. This is done industrially to reduce noise created by for example a piece of machinery. It is my understanding that this is an imperfect solution however in real space as sound waves in an ambient enviroment are reflective and as they bounce off different walls and or objects tend to create slightly distorted or delayed waves. Also that the placement of the speaker being used to create this effect can never be perfect as the sounds that a piece of machinery makes originates in more than one precise location.
If Remy says that this can be done in a studio setting then great, I must admit that I am not certain how she is achieving this effect.
Since partial phase cancellation occurs when multiple mics are used at different spacings from the same source (partial cancellation being that the waves created by two different mics while similar are not identical especially in regard to time and this creates only a slight shift in the peaks and valleys) I do not understand exactly how Remy's concept works. Partial phase cancellation creates a pumping or wishy washy sound and is the basis for phase pedals used by guitarists.
I am certain that this very simple explanation of a complex mathematical and engineering phenomenom will garner some flak because it is not a scientific explanation. While I have encountered phase cancellation issues in recording I must admit that I do not fully understand this vastly complex problem.

While I was writing my original post Remy posted hers and mine seems a little late.

jg49,

I think you are just making it more complicated that it is. The partial phasing that you describe changes the tone of the sound, but what is important is that the sound is repeatable. Think about it without recording anything but the bleed from the speakers to the microphones. Without phase flipping the two passes will be nearly identical. It's simply a repeatable event. When you flip the phase on the speakers, you now have nearly identical but reversed phase sound. These cancel each other out nearly entirely. They aren't exactly the same because of noise and non-linearity of the system.

I understood all this, but I'd never thought to put it to use in noise reduction (though it makes perfect sense).

So, if I record an acoustic guitar in a room with an air conditioner, two minutes long, then, leaving the mics there, I record the same room for two minutes on a new track, reversing the phase, some of the noise will be reduced, is that right? And I assume that's how noise reduction tools work?

You need to nudge the tracks around to adjust for starting during different parts of the cycle. That's if the aircon unit hasn't changed sound at all...
Moral of the story, don't record music around machinery.

StephenMC wrote: So, if I record an acoustic guitar in a room with an air conditioner, two minutes long, then, leaving the mics there, I record the same room for two minutes on a new track, reversing the phase, some of the noise will be reduced, is that right? And I assume that's how noise reduction tools work?

Not quite. The noise is random and not repeatable, so it will not be canceled out. It works with the recorded track because everything is exactly the same the second time including the phase of every frequency.
Most noise reduction plug ins work by converting the information into the frequency domain, then subtracting the noise profile and removing negative numbers, then converting back to the time domain. ReaFIR from Cockos the makes of Reaper is a VST that gives you a nice real time view of this technique.

thank you ,
I'm lost again,
Well, for now I would like to learn how to do it...

This magic things happen all the time, I still remember years ago, during a duo jam with my friend (guitar-vocal, drums), suddenly a bass line perfectly in tune & time was playing with us out of the big pa speakers.!!! ... There was no bass player!!! Today we still can't believe about what happened that day!
?????????

OK back to the Magic disappearances through phase manipulation technique...
Can be explained step by step for me please? sorry, I'm kind like slow...

1) I record the band live inside the tracking room with with 8 mics , bass , drums, guitar, vocal, piano etc...
without headphones! All live .

2) At the end of the song I open the door of the tracking room let all the musicians go for a coffee, then without changing anything I record from all the 8 mics again without musician.

monitoring the silence trough my control room monitors with the door of the tracking room open?

3) ?
4) ?

sorry...step by step for me? possible?
please?

To the OP...Are you simply wanting to learn what phase cancellation is or do you have a specific need for phase cancellation? Are you sure that this is something that you need to do?

Please note that this technique is really only needed when the band is recording to already recorded tracks that are being played back through speakers in the studio rather than headphones. If they are recording but you are not playing back recorded tracks, then this doesn't really apply. Think of it like this...you record drums and bass. Now you want to record horns but instead of having each musician monitor the pre-recorded drums and bass through headphones, you play them back through speakers in the studio. The drum and bass that is being played back through the speakers will be picked up through the horn's microphones and you need to eliminate it...here's how.

Remy's post gives you the steps but if you need numbers...

1) Record some sort of backing track

2) Record a new instrument track where the musician(s) monitors the backing track through speakers rather than headphones - Note this new track will have the bleed from the backing track and that's what you want to get rid of

2) Don't move any microphones...don't change any EQ, don't do anything other than ask the musician to leave

3) Record a new track with the same backing track being played through the speaker(s) in the studio. NOTE: no new parts would be recorded here. You are really just recording the "bleeding" backing track.

4) Reverse the phase of the track recorded in step 3

5) Play back the instrument track with the phase reversed track and then you should no longer hear the bleed from the backing tracks.

Why?

Also, about the aircon as it's the only thing I can grasp at this point in time, from this thread. An aircon doesn't just make random noise. Airflow is white noise I guess, but a compressor is an electric motor and that hums pretty nicely. So long as you don't alter the temperature.
Which makes me wonder if the aircon is working effectivly, because this would alter the temperature making it quite dynamic in sonic signature. However, given that nearly everyone that hangs around here is a digital editing wizard. Would it be difficult to figure out the sonic footprint of an air-conditioning unit and run it through some sort of digital manipulation that could make a plug in that you could use to filter out these unwanted recording studio hassles?
I'm thinking that with enough sampling under different conditions, a mathematical model could be derived that accounts for things like different initial operating temperatures and conditions, which would affect the sonic footprint over a given session with the musicians. This product would work much like auto-tune does, you feed it the known variables, it takes the given input, compares to a mathematical equivalent, adjusts to meet the known variables and outputs the sonic gold. Or whatever.

Man, I love this forum. Absolute genius! Awesome ideas that I would have NEVER thought of! Thanks!

about the aircon,

Right, some of the noise is broadband hiss and whir, and other is narrow band hum. But it is not really a repeatable waveform. So the phase trick doesn't work.

Recording and reversing the phase will not eliminate the aircon noise. It will probably make it worse. If the pitch varies every so slightly, which it will, even careful alignment of hills to valleys at the start of the track will not be lined up at the end.

Of course the best way to remove this sort of noise is not to have recorded it in the first place. Although comfortable musicians are happy clients, so it is sometimes unavoidable.

Next best is with a subtractive EQ like the one I mentioned earlier, or Elevayta Klean Boy. These however do dampen your high end to some degree. Greener, this "noise sample" is probably similar to your mathematical model idea.

Real life situation where it worked and didn't work. Technically, it should always work, but I don't have the technique.

My piano accompanist couldn't make the recording session. So...

1. I record the piano in advance
2. I have a group of singers sing along to this "karoake" track and record them together with the piano's sound in the room.
3. I play back the piano track in the room and record it with the singers standing silently
4. I combine the 2nd and 3rd take, flip the phase and...voila I have a track with a "vocal only" track - the piano has disappeared. If its perfect this could be an acapella track that has perfect time or...
5. I can edit/ eq this vocal track to get its best sound and then mix it with the original piano track I made at the beginning.

A setting where it didnt' work was trying to use an XY and 2 omnis recording a choir and then trying to get the accompaniment to disappear.

It just worked better to put the monitors in the XY's null and go for it and control the amount of volume of the playback.

Does that make more sense?

Phil

And in a similar application.... Where I asked the live sound guy to give me vocals on Channel 1. Guitars, Channel 2. Keyboards, Channel 3. What I got was hot vocals on Channel 1. Hot vocals with low guitars on Channel 2. Keyboards on Channel 3. Vocals needed to be removed from the guitar track. So the guitar track was combined with the vocal track which was added in out of phase. With careful balancing, the vocal disappears from the guitar track . Then the guitar track can be manipulated any way you wish without affecting the vocal sound, now that it is isolated from the vocal. This can be done in real-time as you mix. Or, you can create a completely new, corrected & processed track to be used within the mix.

Phase manipulation can do all sorts of fascinating things. Try spectral isolation with phase manipulation. Then recombine. Some of this is required in audio forensics. You really have to know how to filter in the broadest sense of the term.

I can't hear you because my fingers are in my ears. LA LA LA LA (click)...............................................................................................

I've used the phase trick on a choir. Three men in the middle surrounded by , 12 sopranos and 8 altos. It was 1 bass, 2 tenors, and 7 times that many women! It sounded a little unbalanced live. :D Put an XY up with one pointing at the men, and the other pointing at the loudest and closest sopranos (right in front of the men). Inverted phase on the sopranos, and mixed until they were as quite as possible. There was also a left right pair involved. The choir director was happy, in the end you could hear the men.

Can the trick be done digitally during mixdown instead? logic, cubase ,protools etc...

I really need an analog console with phase switch and routing and do the trick during tracking?

You can flip the phase digitally or on the analog console or by switching the wires on the back of the speaker.