Noise Cancelling Technique

[alwayslearning]

I remember quite a few years ago, taping an event where there was a rope in the wind against a flag pole in the background. Wondering what to do about it when that or something similar comes up in the future, I was doing some research and came across an idea.

Flip pins 2 and 3 on an XLR "adapter" and use the channel it's plugged into on a mixer to flip mic phase 180 degrees, to record, on purpose, the unwanted sound(s). Then on my mixer, I could adjust the levels to a point where the unwanted noise, being 180 out of phase with the same unwanted noise on the talent mic, would be cancelled, or at least minimized.

After reading something to this affect, I tried it. I remember it doing something like I wanted, but I can't remember the details and I cannot find my flip adapter that I made up. Before doing this again, has anyone done anything like this and have any tips on doing it? Is there a thread here about that or anything similar? I tried a search on "cancelling" and found some stuff, but not really about this idea.

Thanks,
Larry

 

[khmuse]

The idea of inverted cancellation is neither new or unique. This is how noise canceling systems typically work (noise canceling headphones are a common example). The problem is that unless the noise is identical in amplitude, spectral shape and phase relationship, the amount of cancellation will be limited. You typically don't have to use a wired inverter just flip the "phase" switch on the board or do the inversion in post.

If you have the extra channel available, it can't hurt to record the offending noise source to see if you can do something usefull with it later.

Don't count on massive improvements, but it can be better than doing nothing at all.

PS The more correct term for "flipping the phase" is "invert the absolute polarity" as phase is only a relative term.

 

[alwayslearning]

Thank you for your reply khmuse.

"The problem is that unless the noise is identical in amplitude, ..."

I'm thinking that aspect can be controlled with the fader on the mixer?

"... spectral shape ..."

Could you please explain for me what that means exactly? I'm thinking that since it is the same noise source (i.e. rope against flag pole), that everything about it would be the same?

"... and phase relationship, ..."

Tell me if I am understanding this ... Let's say the noise source was 70 feet away. If I put a mic near that, the sound will be recorded BEFORE the noise is picked up by the talents mic. Therefore, it is "out of phase"? When flipped, whether at the board input or in post (I'm learning/using Adobe Audition - Formerly and still, Sound Forge), the peaks and valleys won't line up. I suppose I could delay the distant mic signal by about 62 mSec (using 70/1132?) with some sort of a delay unit or shift it that much in post (if recorded on a separate track).

"You typically don't have to use a wired inverter just flip the "phase" switch on the board ..."

I'm only using a Mackie 1402-VLZ and it doesn't have a phase switch.

"If you have the extra channel available, it can't hurt to record the offending noise source to see if you can do something usefull with it later."

Ya ... When I originally experimented with this, I only had all analog equipment.

"Don't count on massive improvements, but it can be better than doing nothing at all."

I'm hoping. Sometimes I have to shoot stuff with no choice about air conditioning unitys, hammering ....... lots of unwanted stuff.

"PS The more correct term for "flipping the phase" is "invert the absolute polarity" as phase is only a relative term."

Thanks. I can figure out some stuff ... sort of .... on a technical level but I don't always no the "lingo". I googled "absolute polarity" and found some truly amazing/interesting info. Thanks again.

Larry

 

[alwayslearning]

Still thinking ....

ME: "If I put a mic near that, the sound will be recorded BEFORE the noise is picked up by the talents mic."

I suppose I could shotgun the offending noise from a distance that is more equal to the talents mic?

I guess I'm thinking that the more I can get right in the field and the less I mess in post, the better?

Larry

 

[khmuse]

You are grasping the issues nicely.

Yes, amplitude can be controlled with any number of gain setting devices, like the "fader" example that you mention.

If the one microphone is nearer not only will its signal arrive first, it will have a different spectral shape (think of this as the amount of energy that is present at each and every frequency).

In order to work best, the cancellation microphone needs to be as close to the microphone that is picking up the intentional sound source, but oriented (pointed) directly at the noise source. Both microphones need to be as directional as possible and should be identical in model and brand to help ensure their similar spectral response.

Delaying the one microphone will help, but only ever so slightly. The accuracy that you would need to obtain is frequency dependent. Say that you offending noise is 1 kHz in foreigner. That means is has a period of 1 ms (milli seconds). In order to cancel its inverse energy needs to be added that inverse energy would be offset by 1/2 of the period or in this case 500 us (micro seconds). To get a reduction of 40 dB (100:1) you would need to hit the time within 5 us (5 / 1,000,000 of a second). This is highly unlikely and even if you could do it, you would still be contending with the spectral issues. Now if the offending signal has harmonics that extend up to say 10 kHz, then the difficulty is increased 10 fold.

Experimentation is the only way to learn how these requirements actually manifest in a real world condition. If time allows, set up a simple test and try different microphone placements and different orientations, you will learn what works best for a variety of conditions.

The less you have to fix in post, the better in my opinion. Hope that this helps.

 

[pmark23]

In the studio we often used to use that trick to get rid of headphone leakage. It can work if it's buried in a mix, but standing alone (as most audio-for-video is) you'll get weird phasing that'll only please those tripping on acid.

Nowadays FFT cancellation is used, and it works great. Today I cleaned up a track that had terrible intermittent wind noise. This track would have been written off as useless a few years ago.

Very basically, FFT cancellation works like this: You know that sound is just a varying sine wave, right? You'd highlight the bit of sound you want to get rid off, then the software breaks that down into a mathematical representation of the sine wave. Then the software breaks down the entire sound file you're processing into a bunch of mathematical representations, and removes any matches. It's actually a lot more complicated than this, but it works very well when used in moderation.

Incidently, Autechre and Aphex Twin use FFT to get a lot of their phase effects, so don't go too overboard when cleaning your audio or you'll trigger heavy flashbacks .