Active noise cancellation for lavs?

[Oren Arieli]

I've been wondering: Noise cancelling headphones are common. There are DSP's for noise reduction on headsets and cellphones. How come someone hasn't created a dual-mic lav that can perform active noise cancellation?
The question came up again when a friend was discussing removal of white noise (ocean surf, waterfalls, etc.) from a wedding ceremony. He suggested phase cancellation noise reduction. As far as I know, that depends on keeping mics at fixed distances. Having a double-mic lav, wouldn't this be quite easy to accomplish? I'm pretty sure there is a market for this, so am I missing something?

[Richard Crowley]

What you are missing is that in order to do this kind of noise-cancelling you need TWO signals which are as identical as possible, but one WITH the desired signal + noise, and the other with ONLY the noise. Then, in theory, you can take the S+N signal and subtract the N signal from it, leaving only "S". You can (more or less) do this is a tiny space inside an earcup at least up to mid frequencies. But the shorter the wavelength (i.e. the higher frequency of the noise) the more different the S+N signal is from the N signal because they aren't both sampling exactly the same place.

Noise-cancelling headphones work (at least for repetitive noise like aircraft engines) by creating a synthesized copy of the noise signal and feeding that (in opposite polarity) into your ears, thus, cancelling the noise. This doesn't really work very well for low frequencies unless you have headphones with good low-end response and lots of power to drive them. And it doesn't work for very high frequencies because the laws of physics and acoustics prevent that in the Real World.

Now there ARE noise-cancelling microphones that were developed decades ago for military use (for aircraft crew communication). And there are modern equivalents of communication-grade noise-cancelling mics made for severe use like this. But in order for them to work properly, the front side of the microphone must be literally touching the lips of the user to get adequate signal into the S+N side of the microphone. The other side picks up the N-only signal so that they can be subtracted. While this is practical for communication and other kinds of utility applications, you won't find this kind of technology used for on-stage or on-camera use because of the size of the hardware and the fact that it covers the user's mouth.

If you tried to do this with something like a lav under the chin, or even an earset mic 1/2 inch away from the corner of the mouth, you would not get enough difference between the "S+N" and the "N-only" to perform any significant noise-cancellation.

Note that the iPhone S4 appears to have a second microphone up by the headphone jack to allow their DSP processor chip to achieve some level of noise cancellation, but even that is only possible because your mouth is so close to the primary microphone down at the bottom, AND the users ear and head are blocking much of the direct sound from the mouth from reaching the noise-cancelling mic at the top.

Beware that many of those cheap little plastic plug-in microphones and headsets you see for a few bucks only claim to be "noise-cancelling" and don't actually use any of the traditional methods.

[Oren Arieli]

I figured as much, but I'm guessing that the DSP technology is pretty well advanced (based on some demos I've seen). Even 10 dB of noise attenuation would be a welcome addition. Perhaps by using two polar patterns (omni and cardioid) there would be a bigger difference between the signal and noise.

[Ty Ford]

Hi,

The Sony ECM 88b uses two opposing diaphragms. AKG made one years ago. I don't recall the model number.

Regards,

Ty Ford

[Oren Arieli]

It appears that the dual diaphragm of the Sony is to improve sensitivity. They make no mention of ambient noise reduction (which would be a bigger selling point IMO).

[Greg Miller]

Originally Posted by Oren Arieli

Perhaps by using two polar patterns (omni and cardioid) there would be a bigger difference between the signal and noise.

If you use two different patterns, then the pickup of noise will not be the same, so subtraction won't work.

Richard has explained it very well. The distance between the "desired" mic and the "noise" mic is important, in terms of level difference, and also in terms of phase difference (and to an extent frequency response).

Let's say that the "noise" mic should be roughly four times as far from the mouth as the "desired" mic. Let's say that the "desired" mic is roughly an inch from the mouth. Then the "noise" mic needs to be up around the ear. If both mics are picking up the same noise signal, you could theoretically subtract them and get rid of most of the noise.

However we can't forget the phase issue. Depending on where the noise source is located, it might be three inches farther from one mic than it is from the other. Now as an example, let's assume the noise frequency is 2,200 Hz. Sound travels roughly 1,100 feet/second. 2,200 Hz will have a wavelength of 1/2 foot = 6 inches. Three inches is exactly one half of a wavelength. So with our three-inch mic spacing, the two mics will pick up that noise frequency exactly out of phase. When we then invert the polarity of one of the mics (for the noise subtraction configuration), we will invert the acoustically out-of-phase signal, and the two signals will end up being exactly in phase. At that point, the two mics will reinforce, and we will have a loud noise peak!

The solution is to reduce the spacing between mics, so that this peak moves up out of the desired frequency range. Let's say we define the high frequency limit as around 8,000 Hz. If we make the spacing between mics 1/4 of our initial example, or 3/4 inch, we move the above described peak from 2,200 Hz up to 8,800 Hz, where it won't be a problem. But now the "desired" mic needs to be 1/4 as far from the mouth as it was in the first example: 1/4 inch. You can see how the "lip mic" concept evolved.

Suffice it to say that sound picked up right at the lips does not sound realistic, because that's not where our ears are normally located when listening to someone speak. There are other issues, too. But the bottom line is that simple subtraction is not a solution.

Originally Posted by Oren Arieli

I'm guessing that the DSP technology is pretty well advanced

I think that's what would be needed to solve the problem. Think about the noise filters in a program like Adobe Audition. You first get a noise spectrum, then use that to set the threshold on a narrow-band downward expander, and finally pass the "noisy" audio through the expander. The subtleties (of the desired signal) which are below the threshold are somewhat lost, but the noise reduction is good. Most importantly, there is no phase issue between the noise and the desired signal; it's only a matter of threshold at each narrow frequency band.

If one could design a DSP system where the "noise" mic was continuously sampled and processed to obtain the threshold settings - in real time! - and then the "desired" mic was fed through the expander, that might have pretty good results. Certainly there is some DSP being done now to achieve echo cancellation and noise gating. But doing it realtime in a very narrow-band application would be more intensive. I'll bet someone is working on it in a back room somewhere...

[Oren Arieli]

Well said, Gregg. I've witnessed amazing technical advancements that seem to defy the laws of physics. This type of mic would be another win for science (and us). It's not a cure for cancer, so it may seem frivolous. But think of all the other seemingly insignificant scientific steps that lead to real breakthroughs in other fields. As Issac Newton said "If I have seen further it is by standing on the shoulders of giants."

[Greg Bellotte]

Forget the back room somewhere, I have one of these...

iZotope ANR-B - Adaptive Realtime Noise Reduction for Broadcast Audio

I use this in a broadcast environment, on shows where I have many lavs open at once on outdoors sets. Between this and a Dugan automixer I've had some really good sounding results in far less than ideal settings.

[Oren Arieli]

Perfect. Now is there a way to shrink that 1RU box down to a ASIC that can be incorporated into a wireless receiver?

[Greg Miller]

Originally Posted by Oren Arieli

is there a way to shrink that 1RU box down to a ASIC that can be incorporated into a wireless receiver?

Someone is working on that in a back room somewhere... ;-)

[Benjamin Maas]

Sanken has a mic that allows for you to deal with that in post- a 2 channel Lav which would give you either a main and a backup or a mic with a second that you can flip phase on to use phase cancellation for background noise removal.

SANKEN MICROPHONE CO .,LTD. | Product [ COS-22 ]

--Ben

[Richard Crowley]

Originally Posted by Oren Arieli

Perfect. Now is there a way to shrink that 1RU box down to a ASIC that can be incorporated into a wireless receiver?

Chances are excellent that it IS only one ASIC DSP chip. You need all that other stuff to control it, monitor it, and get signals in and out.

I would be amazed if people like Lectrosonics, Zaxcom and other high-end wireless mfgs aren't already working on integrating this (and other DSP features) into their products. Digital makes it so much easier (and less expensive) to add these kinds of features virtually without changing the physical product at all.

[Greg Bellotte]

I see this as a feature on wireless soon. Even my HTC smartphone has some kind of DSP noise reduction for the mic, too bad it doesn't work with external input or bluetooth...

[Tom Morrow]

The Sanken cs-3e and Schoeps CMIT are shotguns with active cancellation built in.

There are cardioid patterned lavs which is a form of noise cancellation compared to the standard omni lav.

Cardioid pattern is (at least in theory) made by two microphones: front and back. The front one captures the intended signal, and the back one subtracts the unintended signal. So any cardioid mic is effectively doing noise control... this is I think what the "noise cancelling" microphones that you see on consumer headsets are (please correct me if I'm wrong): Just slightly more-directive-than-omni transducers.