Noise Cancellation?

[Thane of Cawdor]

Hi everyone,

I was wondering why you can't make a noise cancellation circuit with just a microphone and an inverting op-amp to make the noise signal 180 degrees out of phase with the original outside noise signal . Can this nullify the noise?

Thanks

[Richard Crowley]

No. If it were that simple it would have been done decades ago.
The closest you can get is those "noise cancelling" headphones/earbuds.  Even they don't use simple inversion. They sample the noise waveform and synthesize an opposite waveform to "cancel" the noise.  That works OK for many people through the mid-range, but you don't have enough power to counteract low frequencies,  and you can't deal with very high frequencies because of the wavelength of the sound waves vs. your operating space.

The problem is that sound waves in open space is an extraordinarily complex phenomenon and we don't have anything nearly complex enough to significantly control it.

[tszaboo]

The simplest (not headphone) system I saw was using multiple microphones, a DSP to FFT the input and multiple speakers. And that was only working with low frequency repetitive signals, long development times.

[Rudane]

That is a great question, and it shows you are thinking! The answer is: Not yet, but you are welcome to have a go! The problem with noise is it's random. The only tool we have to analyze random signals is statistics, and statistics can only tell us about the average behaviour of random signals. Any single event is unpredictable, and canceling something that isn't there yet would actually induce noise.   

Edit: To clarify I was talking about noise in general, not the microphone case. The microphone case is calling unwanted signals noise. Unwanted signals can be removed.

[ejeffrey]

With the exception that it is 180 degrees, you can and this is basically how noise canceling headphones work.  The tricky part is calibration, you need to make sure the amplitude and phase are just right at every frequency.  It also only cancels the sound in a specific direction or directions.  In other directions the signals will be in phase and the sound will be reinforced.  That is why noise canceling headphones work well but you can't just silence a room.

[IanB]

I was wondering why you can't make a noise cancellation circuit with just a microphone and an inverting op-amp to make the noise signal 90 degrees out of phase with the original outside noise signal . Can this nullify the noise?

Sure you can. I have a $15 pair of unbranded noise cancelling headphones that I got from somewhere like Walgreens, and they work. Not brilliantly, but they do reduce background noise. Given the price, I am sure they do not have really sophisticated electronics inside them. Most likely just what you describe, a microphone and a phase inverter.

By the way, you need to invert the signal, not make it 90 degrees out of phase. 90 degrees is a quarter turn, not a mirror image.

[Alexei.Polkhanov]

I tried this headphones from "Bose": http://www.bose.ca/controller?url=/shop_online/headphones/noise_cancelling_headphones/quietcomfort_15/index.jsp&color=SI. Co-worker got them on sale, but I am not concerned about noise enough to spend $300 on noise cancelling. These model probably is one of top range noise cancelling headphones. I'd say it works pretty well against most noise sources - I was semi-impressed. Would love to see tear-down, but he would not let me take them apart 

Here is some photos of another model inside I found on internet: http://hackaday.com/2013/12/07/repairing-bose-active-noise-cancelling-headphones/

[Richard Crowley]

The simplest (not headphone) system I saw was using multiple microphones, a DSP to FFT the input and multiple speakers. And that was only working with low frequency repetitive signals, long development times. 

To be sure, there are some examples of ACTIVE noise-control systems. In some cases heavy equipment that puts out predictable noise can have larger-scale cancellation systems put in place. But, as NANDBlog suggests, it is a customized, specialized solution for a very narrow situation.

[Alexei.Polkhanov]

I have read about a sound pick-up system implemented with 34 and 94 GHz Doppler radar sensors. (I have attached paper to this message).
Perhaps combining Doppler radar and noise cancellation together will make it possible to create noise cancellation headphones that can cancel any noise including random? Will someone buy $5000-10,000 noise cancelling headsets?

[Bassman59]

I was wondering why you can't make a noise cancellation circuit with just a microphone and an inverting op-amp to make the noise signal 90 degrees out of phase with the original outside noise signal . Can this nullify the noise?

An inverting op-amp won't move the noise signal 90 degrees out of phase, nor will doing so cancel the noise.

Active noise cancellation systems use an adaptive filter whose taps are updated using the least-mean-squares (LMS) algorithm. The algorithm input is generally a signal corrupted by uncorrelated noise, and that noise signal, and the loop trains on the noise to create a matched filter capable of canceling it. The length of the filter is one determinant in how effectively the noise is canceled; a longer filter means more latency and longer training time.

Such systems can be very effective for steady-state noise (such as the low-frequency drone of an airplane), much less so for impulsive noise (someone clapping their hands).

See the canonical text on the subject, Adaptive Signal Processing by Widrow and Stearns.

[Richard Crowley]

90 degrees will NOT cancel out anything.  It takes 180 (i.e. opposite polarity).
Why do people think 90 degrees has anything to do with signal cancellation?

And, of course, even using an inverting op-amp (to produce the proper 180 degree signal) won't do the job for half a dozen good reasons, and likely scores of more arcane violations of the laws of acoustic physics.

[Thane of Cawdor]

I have edited the original question from 90 to 180 degrees   .

[IanB]

I have edited the original question from 90 to 180 degrees   .

That's better, but you are still not quite there. A rotation through 180 degrees is not always equal to a mirror image (a reflection). Consider for instance a duck: if a duck has its head facing right and you rotate it through 180 degrees its head will be facing left. But if you reflect it in the surface of the pond so it's upside down its head will still be facing to the right. The rotated duck will not cancel out the original, but the reflected duck will.

[Thane of Cawdor]

Ok, so an inverting amplifier cannot create a symmetrical waveform which would effectively 'cancel out' the noise?

Thanks

[IanB]

Ok, so an inverting amplifier cannot create a symmetrical waveform which would effectively 'cancel out' the noise?

That's not what I'm saying.

I'm trying to explain that "inversion" is not the same as "rotation through 180 degrees".

If you want to cancel out the noise, you need to take the input signal from the microphone and invert it, or make a mirror image of it. This is what an inverting amplifier does, it turns plus into minus and minus into plus.

If you shift a signal through 180 degrees, you are moving it forwards or backwards in time by half a wavelength. For a regular signal like a sine wave this has the same appearance as inversion, but for an irregular signal like noise it is not the same thing at all.

Here's an illustration to help:



Do you see the difference?

[Rigby]

Assuming an ideal opamp, an ideal audio driver in the headphone, an ideal mic that is the same distance from your ear canal as the driver, and you invert the audio from the mic, then sum it with the incoming audio with another ideal opamp, and put that audio in the ideal driver, you will have a perfect noise canceling device, if and only if the headphone is perfectly aurally transparent and does not passively attenuate external sounds before they reach your ear.

The truth is that every step I mentioned, every component in the circuit, electrical and aural, has a different frequency response from everything else, and there is no such thing as an ideal component, nor a perfectly aurally transparent headset.

So, in theory, if you invert the ambient noise, sum that inverted ambient noise with the incoming audio, the audio given to the driver will cancel out the ambient noise.  In theory only.

[Richard Crowley]

Sure, you can create the inversion of a signal using an op amp (or similar circuit).
And, in fact, we do that to cancel electrical signals all the time.

The problem comes when you try to do it ACOUSTICALLY.
As I suggested earlier, the nature of sound waves loose in the air is ENORMOUSLY COMPLEX.
You don't even perceive how complex the issue is, or you wouldn't have asked the question in the first place.

[Rigby]

Can you explain a bit? I don't understand the complexity either.  When I look at a waveform and think of it as a one-dimensional movement through time it is hard for me to see possibilities for a lot of complexity.

[Richard Crowley]

The exact pressure waveform is DIFFERENT at every place in the system.  The reason the "noise-reducing" earbuds, etc. are only partially effective is because you aren't going to allow them to put a microphone RIGHT ON your eardrum.  So, they have to sample the "audio" several millimeters away, and that limits how accurately they can measure what soundwave to cancel.  And THAT is inside a tiny space within your ear canal.

When you get out into the real world, the difference between where and when you can measure, and where and when you can "inject" some correcting pressure wave is a formula so complex, whole teams of PhD mathematicians and and physicists and acousticians have never been able to solve the problem. Who knows, maybe someday we will have better ways of sensing and generating sound waves, and better understanding of the complex mathematics and physics phenomenon.

[Rigby]

Ah I understood that stuff.  I guess I just assumed that the math was all worked out and well understood by now.  That's where my gap was.

[Richard Crowley]

The FIRST-ORDER math is indeed trivial.  x - x = 0
But the phenomenon of sound in 4 dimensions (x.y.z,time) is the kind of thing they put super-computers to work on for weeks just to solve something that takes a few seconds.

[miguelvp]

Since the duck is symmetrical it doesn't convey the point, because one could say that is 180 degrees rotated around the X axis instead of the Y.

A mirror of an asymmetric shape will make the point better.

[Thane of Cawdor]

I think I understand now. So unless the noise is completely symmetrical, the inverted signal will not cancel out the original signal due to the different orientation. Is that correct?

Many thanks

[IanB]

I think I understand now. So unless the noise is completely symmetrical, the inverted signal will not cancel out the original signal due to the different orientation. Is that correct?

No. An inverted signal is what you want.

However, do you see that inversion is not the same operation as shifting through 180 degrees?

[Richard Crowley]

However, do you see that inversion is not the same operation as shifting through 180 degrees?

At a single frequency, inversion and 180 degree phase shift is, in fact, exactly the same thing.

The problem is that noise typically is not one single pure sine-wave  frequency.  So it is practically impossible to fully CANCEL using this scheme. Especially in a 4-dimensional space.

[Jarrod Roberson]

The way the Jawbone works is the best noise canceling technique for speech I have ever used.

When they say "military grade" they mean it!

I can talk on my Jawbone in my convertible with the top down at 85MPH on the interstate highway and people never know that I am driving down the road with the top down!

With most other noise-blanking headsets, there’s a distinct difference in the quality of the speaker’s voice when the feature is turned on, there isn't any difference in the quality of your voice with the Jawbone.

The Jawbone does this well because it has a sensor on the underside of the device that touches your face. When you speak, this Voice Activity Sensor can tell, using the resulting vibration to separate your speech from noise around you. According to Aliph’s marketing materials, this technology was originally developed for use in the military.

My Jawbone 2 worked great, and my ERA seems to work even better.

[TMM]

The closest you can get is those "noise cancelling" headphones/earbuds.  Even they don't use simple inversion.

Oh? i thought most of them did work by simply inverting the signal from a mic that is exposed to the outside and then playing it through the speaker/driver. The operating distance is small enough (a couple cm) that they have reasonable attenuation from low frequencies up to a couple KHz. Above this there would be a lowpass filter in order to prevent adding positive feedback to the noise as the wavelengths approach the path length difference of the mic, speaker, noise sources to the listener/ear. High frequencies have good passive attenuation from the closed ear cups so it's not really too much of a limitation.

There would obviously need to be some equalisation to the signal to compensate for the frequency response of the mic and the driver and achieve the best attenuation. Most mic capsules and small headphone drivers are well behaved up to a few KHz anyway.

[Richard Crowley]

The Jawbone does this well because it has a sensor on the underside of the device that touches your face. When you speak, this Voice Activity Sensor can tell, using the resulting vibration to separate your speech from noise around you. According to Aliph’s marketing materials, this technology was originally developed for use in the military.

In other words, it uses a figure-8 microphone. Sound entering from the BACK of the microphone capsule is "subtracted" from the sound entering from the FRONT. See Dave's microphone 101 episode on patterns.  This is how all noise-cancelling microphones operate.

[IanB]

At a single frequency, inversion and 180 degree phase shift is, in fact, exactly the same thing.

That's a special case that's not true in general.

In general, phase shift and phase inversion are different operations. It's a good idea not to use the term for one when you mean the other.

[Jarrod Roberson]

The Jawbone does this well because it has a sensor on the underside of the device that touches your face. When you speak, this Voice Activity Sensor can tell, using the resulting vibration to separate your speech from noise around you. According to Aliph’s marketing materials, this technology was originally developed for use in the military.

In other words, it uses a figure-8 microphone. Sound entering from the BACK of the microphone capsule is "subtracted" from the sound entering from the FRONT. See Dave's microphone 101 episode on patterns.  This is how all noise-cancelling microphones operate.

No it doesn't, it only has a single microphone. Which is not "how all noise-cancelling microphones work".

It has a vibration sensor ( a little silicon nub ) that reads the vibration on your skin from your voice to include only what matches the vibration.

Completely different approach to dual microphone solutions, which don't work worth a flip on battlefield level noises. Or convertibles at highway speeds.

[Richard Crowley]

A "figure-8" microphone IS a "single microphone".  Perhaps you should view the recent videos on Microphone 101.
They even mention SPECIFICALLY this application (microphone on a telephone headset, and went into some detail about how it works. Recommended.

You seem to be talking about a "throat mic" where they strap a transducer around your neck to pick up vibrations from your vocal cords.

[Jarrod Roberson]

Aliph uses a patented transducer based system. It is not this figure 8 thing that everyone else does. The results are in a class of their own. I have had both non-noise canceling and "typical" noise canceling headsets, the Aliph Jawbone sets are completely different in results. Nothing compares and most people don't even "get it" until they use one.

http://www.google.com/patents/US8340309 - Recommended

an acoustic vibration sensor in the housing, the acoustic vibration sensor comprising, a protrusion that extends from the housing to contact a skin surface of the user, wherein the acoustic vibration sensor detects human tissue vibration associated with near-end speech of the user, wherein the acoustic vibration sensor comprises a diaphragm positioned adjacent a first port and a second port of the housing;
a noise suppression system executing on a processor in the housing, the processor coupled to and using signals from the microphone array and the acoustic vibration sensor to separately identify voiced speech and unvoiced speech of the acoustic signals and denoise the acoustic signals; and