Re: OT: Data Mining RBT - need a great sound source for new noise-cancelling headphones

On 2008-03-08, bjw@xxxxxxxxxxxxxxxxx <bjw@xxxxxxxxxxxxxxxxx> wrote:
On Mar 2, 2:29 pm, Ben C <spams...@xxxxxxxxx> wrote:
On 2008-03-02, Jay <jbol...@xxxxxxxxx> wrote:

I have just bought a pair of noise cancelling headphones, for the bus and
train legs of my daily commute. I really need to muffle the sound of the
chit chat, mostly on the bus, to preserve what is left of my sanity.

http://www.amazon.com/gp/product/B000OMKR8E/ref=cm_rdp_product

It will cancel the rumbling of the bus and the train but unfortunately
you will hear most of the chit chat loud and clear.

It can only work out the antinoise based on being able to predict what
will come next. That's easy for the continuous hum of an engine, but not
possible for the gibbering sounds produced by the other passengers.

If the earpieces were quite thick and you could put the microphone and
speaker 4cm apart then you'd have 0.0001s in which to generate the
antinoise. That's easily enough time for a computer to work out the
antinoise in, but actually getting the speaker to respond that quickly
may be more of a problem. As far as I know the technology to do this
doesn't exist yet.

I don't think that is really how noise canceling headphones
work. That is, I don't believe predictive power is the limitation
on canceling high frequency noise. It is decorrelation of
phase between the sensor and the ear.

You may be right-- I was only guessing at how these things work.

See
http://en.wikipedia.org/wiki/Active_noise_control

AFAICT, NC headphones have trouble with high frequency
noise in general, not just people's speech. Engine hum is
fairly low frequency. The electrical signals in a NC system
travel much faster than sound, which suggests that
predictive speed is not a problem, and the speaker does have
response at about the frequency you are trying to cancel.

One way of looking at this is that if you were trying to
generate antinoise with an 0.0001sec lag time, that
corresponds to 10,000 Hz, which is fairly far up in the
human hearing range. The power in most human speech
is lower frequency, so it isn't changing rapidly at 10 kHz,
and you shouldn't need a system with quite that high
bandwidth.

In the hypothetical you mentioned, with a sensor 4cm
away from the ear, the sensor will hear some combination
of sound waves, which arrive from several different
directions. Consider 2000 Hz background sounds. The
speed of sound is ~344 m/s and the wavelength of a
2000 Hz signal is 17 cm. Thus, your ear is about 1/4
wavelength separated from the sensor. If you just
generated the opposite of whatever the sensor heard,
it would be 1/4 wave offset from what you want. The
problem is that you can't predict whether it's ahead
or behind, because you don't know which direction the
sound is coming from, and it probably is several signals
from multiple directions.

Why does it matter which direction it came from? That gives you a delay
and phase difference between your two ears, but I'm assuming you have an
independent sensor and anti-noise generator on each earpiece.
.

Relevant Pages

  • Re: cross correlation sound
    ... My name is maz and I am working on simple sound source localization. ... Perhaps I can help with the cross correlation. ... sensor is spaced the same distance either side of the vertical north/ ... the waveform will hit both sensors at the ...
    (comp.dsp)
  • Re: cross correlation sound
    ... My name is maz and I am working on simple sound source localization. ... Perhaps I can help with the cross correlation. ... sensor is spaced the same distance either side of the vertical north/ ... the waveform will hit both sensors at the ...
    (comp.dsp)
  • Re: Hornby BR A4 DCC Sound - hows it done?
    ... I thought it went out of sync. ... the DCC software without a local sensor?? ... Most UK RTR locos with sound use ESU LokSound chips. ... drive component of the chip). ...
    (uk.rec.models.rail)
  • Re: OT: Data Mining RBT - need a great sound source for new noise-cancelling headphones
    ... of sound waves, ... wavelength separated from the sensor. ... sound is coming from, and it probably is several signals ...
    (rec.bicycles.tech)
  • Re: The Overtone series
    ... sound must go through the ear and therefore sound must "go through" ... You are assuming that Fourier analysis is always applicable. ... If it is discrete, the Fourier transform becomes a DTFT. ... No one has claimed that the ear does not ...
    (rec.music.theory)