Active Topics
-
maemo.org (1)
to Community by sixwheeledbeast - 13 hrs ago -
Old stuff for Nokia 770 (0)
to Nokia 770 by teroyk - 16 hrs, 58 mins ago - more...
|
2008-02-20
, 15:39
|
|
Posts: 12 |
Thanked: 3 times |
Joined on Feb 2008
|
#12
|
Originally Posted by Saturn
So I'm thinking of the obsolete passive noise-cancelling headphones that have microphones and DSP built-in? Did you READ my post? Thanks for being a jerk right out of the gate.
Maybe you confuse this with something else. That's not how state-of-the-art active noise cancelling headphones work..
Originally Posted by polossatik
This is more or less what I said, except I concluded that it's pretty much not worth any effort to try and adjust for delay; as soon as the device (bearing the mic) moves to either side of your head, you can only [technically] cancel noise for one ear. Since you're most likely canceling a hum or other steady periodic signal, you could theoretically just go out of phase by some number of full periods, but you'd still have to guess at the offset. It really is key to active noise-cancellation that the distance between mic and output are a known quantity in order to avoid having to guess or have the user specify an offset.
From electronic point of view it's rather simple,mic to pick up the ambient noise, low pass filter (R/C network) a Opamp who inverts the signal and a amp to bring it to a decent level for the speaker. You can then add things like feedback control by using second mic next to the speaker to measure the diff between the surround and the inverted signal to control the gain of the inverted signal.
Bottom line is that with all this stuff the real pain point is the placement of the mic(s) vs the output. you need to know the delay of the sound, etc...
If you don't know that then you might make a software that inverts some input (like from the stock N8x0 headphones with mic) and the a few sliders who control delay, cut off frequency and gain. But no success is warranted. And in worse case you can actually amplify (a part of) the noise you want to cancel and have a really load humm in your ear which may be painful or damaging even.
| The Following User Says Thank You to EvilBit For This Useful Post: | ||
 |
|
2008-02-20
, 16:05
|
|
Posts: 4,930 |
Thanked: 2,272 times |
Joined on Oct 2007
|
#13
|
Originally Posted by Saturn
Please help a poor clueless noob who doesn't know anything about state-of-the-art active noise-cancelling headphones and how they differ from real, ordinary, (passive?) noise-cancelling headphones, and hence has no clue what to search for or what library book. I keep coming up with stuff like http://en.wikipedia.org/wiki/Active_noise_control and http://en.wikipedia.org/wiki/Noise-cancelling_headphone; they just seem to be describing real noise-cancellation, not some
There's plenty to read. Please use a search engine or a library.
Hint: Notice the part that those types of headphones work best for repetitive sounds, i.e. plane engines, train etc.
Hint: notice that a fixed phase relationship is required; constructive interference != noise cancellation, state-of-the-art active or otherwise.
|
|
2008-02-20
, 17:09
|
|
Posts: 12 |
Thanked: 0 times |
Joined on Dec 2007
|
#14
|
What if you first recorded just background noise. Then software would do simple FastFourierTransform then you would have spectrum and and periodic noises (engine noise, snoring, etc.). Then you would manually set graphical eq to silence random but distinct noise frequencies and set the delay for periodic noises cancellation. You should find easily bits of ready code for EQ and FFT but everything else beyond me.
 |
|
2008-02-20
, 18:38
|
|
Posts: 1,648 |
Thanked: 2,122 times |
Joined on Mar 2007
@ UNKLE's Never Never Land
|
#15
|
Originally Posted by TA-t3
With all the respect TA-t3, he's wrong (since we are in the context of how the noise cancellation works in today's headphones of course!). Those headphones treat only the low frequencies so a variation in the distance of ~0.5m between the speakers and the microphone won't make much difference.
Still, as far as I can see EvilBit got it pretty much right.
Active noise cancellation won't work if you can't guarantee that the phase of the "anti-noise" sound is exactly opposite of the noise you want to cancel when both waves reach the eardrums.
If it's not exactly opposite (i.e. if it has a small phase advancement or delay) it will still 'work' with a less perfect result. That is, it will remove the original noise but the remainder will introduce another noise, usually in the form of high-frequency hiss. Which is what exactly you'll hear more or less in all of this type of headphones currently in the market.
PS: Sorry if I sound like a smart-a55, actually I'm quite dump and it took me more than 3 years to get some of what that crazy professor was talking about.
Last edited by Saturn; 2008-02-20 at 18:40. Reason: grammatical error
|
|
2008-02-20
, 18:38
|
|
Posts: 12 |
Thanked: 3 times |
Joined on Feb 2008
|
#16
|
Originally Posted by ertszi
I'm not saying that there's no way to do active cancellation on an N8x0. I'm just saying that it would only work on periodic noise (which, granted, is pretty much the domain of active noise cancellation: low-frequency periodic noise) and most importantly, would require significant and unwieldy user input.
What if you first recorded just background noise. Then software would do simple FastFourierTransform then you would have spectrum and and periodic noises (engine noise, snoring, etc.). Then you would manually set graphical eq to silence random but distinct noise frequencies and set the delay for periodic noises cancellation. You should find easily bits of ready code for EQ and FFT but everything else beyond me.
The user's not going to put on their headphones, fire up the app[let], listen for a second, then say, "oh, this is about -0.413s out of phase in the left ear and -0.118s for the right." They'd have to sit there and waggle two sliders around until it sounded a little quieter. Then if they moved their tablet, they'd have to do it again.
All the FFTs in the world aren't going to infer where your headphones are relative to the mic unless they're playing REALLY LOUD and your N8x0's mic can pick them up and do some acoustic rangefinding. In which case we're talking about a completely insane scenario.
|
|
2008-02-20
, 18:49
|
|
Posts: 1,648 |
Thanked: 2,122 times |
Joined on Mar 2007
@ UNKLE's Never Never Land
|
#17
|
Originally Posted by Benson
If you got offended because I said to read a book instead of asking me, then you have a problem.
Please help a poor clueless noob who doesn't know anything about state-of-the-art active noise-cancelling headphones and how they differ from real, ordinary, (passive?) noise-cancelling headphones, and hence has no clue what to search for or what library book. I keep coming up with stuff like http://en.wikipedia.org/wiki/Active_noise_control and http://en.wikipedia.org/wiki/Noise-cancelling_headphone; they just seem to be describing real noise-cancellation, not somefoo-foo magicstate-of-the-art active stuff that works no matter where the mic is.
Hint: notice that a fixed phase relationship is required; constructive interference != noise cancellation, state-of-the-art active or otherwise.
Actually, I was going to tell you open a dictionary but I'm afraid so I looked it up for you:
state-of-the-art
the highest level of development at a particular time (especially the present time); "state-of-the-art technology"
|
|
2008-02-20
, 21:31
|
|
Posts: 12 |
Thanked: 3 times |
Joined on Feb 2008
|
#18
|
Originally Posted by Saturn
Except aircraft cabin noise, for example (which is honestly going to be one of the lowest-frequency noises one might wish to reduce) carries a range of frequencies between 50 and 2000 Hz. Some of the more significant components of the vibrations fall between 40-100 and 100-200 Hz. At 100 Hz, the wavelength of the sound is 3.43m, of which 0.5m distance would be 15% out of phase. At 200 Hz, the wavelength is 1.715m, of which 0.5m is 29% out of phase. I'm pretty sure those variations aren't exactly insignificant.
With all the respect TA-t3, he's wrong (since we are in the context of how the noise cancellation works in today's headphones of course!). Those headphones treat only the low frequencies so a variation in the distance of ~0.5m between the speakers and the microphone won't make much difference.
If you're trying to cancel out 20 Hz vibrations, then you MIGHT actually be getting into an acceptable range of error, but 29% isn't exactly "close enough" for active cancellation. The high-frequency hiss you hear in "more or less in all of this type of headphones currently in the market" comes from an error of millimeters.
|
|
2008-02-20
, 23:45
|
|
Posts: 4,930 |
Thanked: 2,272 times |
Joined on Oct 2007
|
#19
|
Originally Posted by Saturn
I guess my sarcasm was too opaque... I don't want you to explain how they work; I know how they work. You, on the other hand, don't seem too.
If you got offended because I said to read a book instead of asking me, then you have a problem.
Actually, I was going to tell you open a dictionary but I'm afraid so I looked it up for you:
I'd still be open to an explanation of how you think they work, and how you think they're different from any other noise cancelling headphones, but only for entertainment value.
You suggested EvilBit was confusing these with something else, and said that's not how state-of-the-art active noise cancelling headphones work.
(Note that "state-of-the-art active" has no meaning there unless by contrast with some non-state-of-the-art and/or passive alternative, so you are implying that there is some such distinction. There is none. All noise-cancelling headphones are active. And state-of-the-art and non-state-of-the-art ones do not differ in the operational functionality, only in the types of circuitry used to accomplish it.)
EvilBit clearly was thinking of active noise cancelling headphones, and since you're claiming there's a relevant difference between state-of-the-art and non-state-of-the-art ones, which no one else seems to see, you should provide some back-up.
His description was correct insofar as required to point out the main difficulty -- unknown location, hence unknown phase difference. He was wrong, inasmuch as he claimed that the phase difference had to be positive, but you didn't even address that. And there are no non-state-of-the-art noise-cancelling headphones that have such a limitation, so what you said didn't even address the one issue with his comment.
Originally Posted by Saturn
Yes, you take a sinusoid of one frequency, add a sinusoid of the same frequency with a small phase shift, and, voila, you get a high-frequency hiss... Do you expect me to take anything you say seriously? I guess you're right, I'd be far better off reading online and in libraries and completely ignoring your take on things.
If it's not exactly opposite (i.e. if it has a small phase advancement or delay) it will still 'work' with a less perfect result. That is, it will remove the original noise but the remainder will introduce another noise, usually in the form of high-frequency hiss. Which is what exactly you'll hear more or less in all of this type of headphones currently in the market.
PS: Sorry if I sound like a smart-a55, actually I'm quite dump and it took me more than 3 years to get some of what that crazy professor was talking about.
Originally Posted by EvilBit
Actually, it's more helpful to put those in degrees, or radians.
Except aircraft cabin noise, for example (which is honestly going to be one of the lowest-frequency noises one might wish to reduce) carries a range of frequencies between 50 and 2000 Hz. Some of the more significant components of the vibrations fall between 40-100 and 100-200 Hz. At 100 Hz, the wavelength of the sound is 3.43m, of which 0.5m distance would be 15% out of phase. At 200 Hz, the wavelength is 1.715m, of which 0.5m is 29% out of phase. I'm pretty sure those variations aren't exactly insignificant.
at 100 Hz, it's 52 degrees, which is bad enough.
at 200 Hz, it's 104 degrees, it's actually amplifying the noise.
If you're trying to cancel out 20 Hz vibrations, then you MIGHT actually be getting into an acceptable range of error, but 29% isn't exactly "close enough" for active cancellation. The high-frequency hiss you hear in "more or less in all of this type of headphones currently in the market" comes from an error of millimeters.
But you're very much right here. 20 Hz gets 15 dB attenuation. At about 114 Hz, the noise gets passed with no attenuation; starting at around 340 Hz and going up, you get a series of peaks with over 6 dB of noise amplification. Not what I'd term state-of-the-art. I'm attaching a graph of the response.
And one thing not mentioned yet, even if you do put in a time delay, you're still toast if it's a foot from your ear. Because noise comes from different directions, you've violated the 1-d assumption we've predicated this whole discussion on. The only way to get back to the 1-d assumption in a 3-d noise environment is to collocate the sensor with the listener. That is to say, the microphone needs to be on your ear to deal with arbitrary (even if band-limited) noise.
Last edited by Benson; 2008-02-20 at 23:47.
| The Following 3 Users Say Thank You to Benson For This Useful Post: | ||
 |
|
2008-02-21
, 00:35
|
|
Posts: 126 |
Thanked: 23 times |
Joined on Jan 2008
|
#20
|
If it was not so sad I might consider this whole discussion as quite funny...
From electronic point of view it's rather simple,mic to pick up the ambient noise, low pass filter (R/C network) a Opamp who inverts the signal and a amp to bring it to a decent level for the speaker. You can then add things like feedback control by using second mic next to the speaker to measure the diff between the surround and the inverted signal to control the gain of the inverted signal.
Bottom line is that with all this stuff the real pain point is the placement of the mic(s) vs the output. you need to know the delay of the sound, etc...
If you don't know that then you might make a software that inverts some input (like from the stock N8x0 headphones with mic) and the a few sliders who control delay, cut off frequency and gain. But no success is warranted. And in worse case you can actually amplify (a part of) the noise you want to cancel and have a really load humm in your ear which may be painful or damaging even.