On 9/7/2011 11:44 AM, robert bristow-johnson wrote:> On 9/7/11 12:56 PM, Vladimir Vassilevsky wrote: >> > ... >> >> To my knowledge, there is no control loop in noise canceling headphones. > > you mean no loop at all? no LMS nor NLMS (or whatever adaptive filter)? > >> It uses simple analog subtraction. > > how does it know what to subtract? i would think that not only is the > subtraction done analogly, but it's done acoustically (actually with a > sign change and addition). that's the whole point. but i would think > that in the headphone, close to the orifice of the ear canal (the > closer, the better) is a tiny microphone that picks up the official > difference signal (that we're trying to minimize) to feed back into the > adaptive filter. > > but, in fact, i have never ever ever done diddly with noise-canceling > headphones, so i cannot claim to know diddly about it. i just have an > idea for how i would approach the problem. i might suggest to the OP > that adaptive filters usually work better on broadbanded signals and a > single and pure sinusoid is what i would call "quite narrow banded". >I always thought that adaptive filters work better on narrow-band signals. No? In an ANC, the filter adjusts to remove sinusoidal components and shuts off for broadband components lest the removal add broadband noise. So, there's no change in the broadband components at the output. In an ALE, the filter does the same thing but the signal output of interest is the "correction" component. Since the filter shuts off for broadband components, it becomes a bandpass for narrow-band components. So, in that case, the broadband is suppressed and the only output would be the narrow-band components present. Otherwise, one can do the same job on broadband with a switch OFF. But that's not what these ANR headsets do as Vlad mentions. They use external microphones to sense the external "noise". The waveform isn't important because all that's done is to drive the headset speakers with a suitably scaled inverted version. If the scaling matches the headphones outside-to-inside attentuation then the net sound into the ear canal is (ideally) zero. Then "desired" signals can be injected on top of that. In-the-medium devices like the headsets work up to some max frequency where the wavelengths get short enough such that inversion from mic to speaker isn't necessarily subtractive in the medium. Or, said another way, when the spacing between the mic and the speaker (as a simple model) gets to be a reasonable fraction of a wavelength then things start to fall apart. If the spacing is 1/2" then that means stay *well* below 24kHz like maybe below 2.5kHz? Just arm-waving numbers.... From the schematics I've seen I don't see where long time intervals could matter. Doesn't mean there aren't fancier circuits. So, the phenomenon described would be unexpected. Sounds like some kind of feedback to me but the wavelength at 80Hz is 13 feet! Not much chance for phase shift in the headphones or in the head! I'd go along with Jerry and suggest the things aren't working right. I have no idea what "the kind of music kids listen to these days" has to do with anything here. Fred
Interesting artifact of Bose noise cancelling headphones
Started by ●September 6, 2011
Reply by ●September 8, 20112011-09-08
Reply by ●September 10, 20112011-09-10
On Sep 8, 11:39�am, Fred Marshall <fmarshallxremove_th...@acm.org> wrote:> On 9/7/2011 11:44 AM, robert bristow-johnson wrote: > > > > > > > On 9/7/11 12:56 PM, Vladimir Vassilevsky wrote: > > > ... > > >> To my knowledge, there is no control loop in noise canceling headphones. > > > you mean no loop at all? no LMS nor NLMS (or whatever adaptive filter)? > > >> It uses simple analog subtraction. > > > how does it know what to subtract? i would think that not only is the > > subtraction done analogly, but it's done acoustically (actually with a > > sign change and addition). that's the whole point. but i would think > > that in the headphone, close to the orifice of the ear canal (the > > closer, the better) is a tiny microphone that picks up the official > > difference signal (that we're trying to minimize) to feed back into the > > adaptive filter. > > > but, in fact, i have never ever ever done diddly with noise-canceling > > headphones, so i cannot claim to know diddly about it. i just have an > > idea for how i would approach the problem. i might suggest to the OP > > that adaptive filters usually work better on broadbanded signals and a > > single and pure sinusoid is what i would call "quite narrow banded". > > I always thought that adaptive filters work better on narrow-band > signals. �No? > > In an ANC, the filter adjusts to remove sinusoidal components and shuts > off for broadband components lest the removal add broadband noise. �So, > there's no change in the broadband components at the output. > > In an ALE, the filter does the same thing but the signal output of > interest is the "correction" component. �Since the filter shuts off for > broadband components, it becomes a bandpass for narrow-band components. > So, in that case, the broadband is suppressed and the only output would > be the narrow-band components present. �Otherwise, one can do the same > job on broadband with a switch OFF. > > But that's not what these ANR headsets do as Vlad mentions. �They use > external microphones to sense the external "noise". �The waveform isn't > important because all that's done is to drive the headset speakers with > a suitably scaled inverted version. �If the scaling matches the > headphones outside-to-inside attentuation then the net sound into the > ear canal is (ideally) zero. �Then "desired" signals can be injected on > top of that. > > In-the-medium devices like the headsets work up to some max frequency > where the wavelengths get short enough such that inversion from mic to > speaker isn't necessarily subtractive in the medium. �Or, said another > way, when the spacing between the mic and the speaker (as a simple > model) gets to be a reasonable fraction of a wavelength then things > start to fall apart. �If the spacing is 1/2" then that means stay *well* > below 24kHz like maybe below 2.5kHz? �Just arm-waving numbers.... > > �From the schematics I've seen I don't see where long time intervals > could matter. �Doesn't mean there aren't fancier circuits. �So, the > phenomenon described would be unexpected. �Sounds like some kind of > feedback to me but the wavelength at 80Hz is 13 feet! �Not much chance > for phase shift in the headphones or in the head! �I'd go along with > Jerry and suggest the things aren't working right. �I have no idea what > "the kind of music kids listen to these days" has to do with anything here. > > Fred- Hide quoted text - > > - Show quoted text -Dave, do you mean that you are playing 50Hz through the headphones, or producing a 50Hz acoustic signal in the room and then seeing if the headphones cancel?
Reply by ●September 10, 20112011-09-10
> >Dave, do you mean that you are playing 50Hz through the headphones, or >producing a 50Hz acoustic signal in the room and then seeing if the >headphones cancel? >Simply playing 50Hz through the headphones...The result is repeatable, but you need to be wearing them so that the cavity is closed - or preferably the dummy is wearing them!
Reply by ●September 10, 20112011-09-10
On Sep 10, 10:58�am, "gretzteam" <gretzteam@n_o_s_p_a_m.yahoo.com> wrote:> >Dave, do you mean that you are playing 50Hz through the headphones, or > >producing a 50Hz acoustic signal in the room and then seeing if the > >headphones cancel? > > Simply playing 50Hz through the headphones...The result is repeatable, but > you need to be wearing them so that the cavity is closed - or preferably > the dummy is wearing them!Does this only happen when you turn them on? Most headphones have a soft-start circuit that can take several seconds to turn on.
Reply by ●September 10, 20112011-09-10
On 9/10/11 5:25 PM, Robert Adams wrote:> On Sep 10, 10:58 am, "gretzteam"<gretzteam@n_o_s_p_a_m.yahoo.com> > wrote: >>> Dave, do you mean that you are playing 50Hz through the headphones, or >>> producing a 50Hz acoustic signal in the room and then seeing if the >>> headphones cancel? >> >> Simply playing 50Hz through the headphones...The result is repeatable, but >> you need to be wearing them so that the cavity is closed - or preferably >> the dummy is wearing them! > > Does this only happen when you turn them on? Most headphones have a > soft-start circuit that can take several seconds to turn on.Bob, can you verify (or refute) what Vlad said? he said that there is no DSP going on with those noise canceling headphones. i sure thought that they had something like an adaptive filter going on. i know that they have microphones mounted outside the "cans" or earpieces. i just thought that there was more sophistication involved than just subtracting the microphone signal (with gain adjusted by trimpots) from the incoming headphone signal. if that were the case, this technology would have existed in the 70s, but i don't remember seeing any such products until the 90s. i would think that there would be some small amount of delay matching involved (but the optimal delay matching would be different for sounds coming from different directions, which is why i guessed that there was some sorta adaptive filter happening to adjust it). i have never owned such a product, but have seen them at the AES show floor and i surely thought that they were showing off some more sophisticated technology than what could have existed in 1980. -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
Reply by ●September 11, 20112011-09-11
On Sep 10, 7:01�pm, robert bristow-johnson <r...@audioimagination.com> wrote:> On 9/10/11 5:25 PM, Robert Adams wrote: > > > On Sep 10, 10:58 am, "gretzteam"<gretzteam@n_o_s_p_a_m.yahoo.com> > > wrote: > >>> Dave, do you mean that you are playing 50Hz through the headphones, or > >>> producing a 50Hz acoustic signal in the room and then seeing if the > >>> headphones cancel? > > >> Simply playing 50Hz through the headphones...The result is repeatable, but > >> you need to be wearing them so that the cavity is closed - or preferably > >> the dummy is wearing them! > > > Does this only happen when you turn them on? Most headphones have a > > soft-start circuit that can take several seconds to turn on. > > Bob, can you verify (or refute) what Vlad said? �he said that there is > no DSP going on with those noise canceling headphones. �i sure thought > that they had something like an adaptive filter going on. > > i know that they have microphones mounted outside the "cans" or > earpieces. �i just thought that there was more sophistication involved > than just subtracting the microphone signal (with gain adjusted by > trimpots) from the incoming headphone signal. �if that were the case, > this technology would have existed in the 70s, but i don't remember > seeing any such products until the 90s. > > i would think that there would be some small amount of delay matching > involved (but the optimal delay matching would be different for sounds > coming from different directions, which is why i guessed that there was > some sorta adaptive filter happening to adjust it). > > i have never owned such a product, but have seen them at the AES show > floor and i surely thought that they were showing off some more > sophisticated technology than what could have existed in 1980. > > -- > > r b-j � � � � � � � � �r...@audioimagination.com > > "Imagination is more important than knowledge."Robert Yes, Vlad is correct; as far as I know, 99% of these devices contain fixed analog filters. Bob
