Reply by sigmonde February 3, 20142014-02-03
>Forgot to mention, in a duct another possible cause of low coherence would
be turbulence, but that would really only apply to air-flow situations and not audio (I think ..)
> >Bob >
Thanks guys if did wonder if this was double the work. I'll look into using the impulse response directly. I'm sorry I can't reveal how I calibrate the impulse response on the channel as it's a key part of our IPR . Colin _____________________________ Posted through www.DSPRelated.com
Reply by sigmonde February 2, 20142014-02-02
>Forgot to mention, in a duct another possible cause of low coherence would
be turbulence, but that would really only apply to air-flow situations and not audio (I think ..)
> >Bob >
Thanks guys if did wonder if this was double the work. I'll look into using the impulse response directly. I'm sorry I can't reveal how I calibrate the impulse response on the channel as it's a key part of our IPR . Colin _____________________________ Posted through www.DSPRelated.com
Reply by February 2, 20142014-02-02
Forgot to mention, in a duct another possible cause of low coherence would be turbulence, but that would really only apply to air-flow situations and not audio (I think ..)

Bob
Reply by February 2, 20142014-02-02
It's always a good idea to compute the coherence function between the cancellation speaker and the microphone. Drive the speaker with random noise and capture a long data record of both the source and the microphone output, then let Matlab do the work to give you a coherence spectrum. If the coherence at any frequency is much less than 1 then you either have a lot of non-linearity or some non- stationary paths, and in that case cancellation with a linear filter of any length will not be possible.  

Bob
Reply by February 1, 20142014-02-01
Back to the original question.....
If you HAVE the impulse response from a measurement, then i think you CAN determinate the required EQ filter directly from the impulse response without adaptation.

If the impulse response changes and you measure it again, the you recalculate the taps again.

If you are both measuring the impulse response AND running an adaptive filter, that is double work.

An adaptive filter in effect is a measure of the ir.  if it is changing, then yes you need to make repeated measurements of the ir or run an adaptive filter, but not both.

True.????


By the way, how do you measure the impulse response?

Mark
  
Reply by sigmonde February 1, 20142014-02-01
>On Thursday, January 30, 2014 8:19:01 PM UTC-5, sigmonde wrote: >> >On Wed, 29 Jan 2014 04:53:49 -0600, "sigmonde" <10104@dsprelated> >> >> >wrote: >> >> > >> >> >>>>Is the echo path acoustic? If so, it's unlikely to be truly static. >> >> >>>Acoustic paths have a funny way of not being LTI even though we like
to
>> >> >>>think of them that way. Also the speaker will be affected by >> >> temperature, >> >> >>>aging, etc. >> >> >>>> >> >> >>>>Bob >> >> >>>> >> >> >>>The path is a steel conduit, it's impulse response is measured
every
>> >> >>wake >> >> >>>up cycle. >> >> >>>C >> >> >>> >> >> >>>_____________________________ >> >> >>>Posted through www.DSPRelated.com >> >> >>> >> >> >>Sorry should have added that it is an acoustic transmission system >> >> >>C >> >> > >> >> >Is the signal transmitted through the air in the pipe or through the >> >> >pipe itself? Not that it matters, but I'm curious. >> >> > >> >> >Also, if you measure the IPR whenever the systems wakes up, it may >> >> >benefit from an adaptive algorithm, and then just freeze the taps once >> >> >you're confident that it's trained. This is a fairly common thing to >> >> >do. >> >> > >> >> > >> >> > >> >> > >> >> >Eric Jacobsen >> >> >Anchor Hill Communications >> >> >http://www.anchorhill.com >> >> > >> >> Thanks Eric >> >> >> >> Yes I'll try locking the taps >> >> >> >> It's a through steel communications system where the channel has a >> >> transfer function very similair in structure to a comb filter. >> >> >> >> Colin >> >> >> >> _____________________________ >> >> Posted through www.DSPRelated.com > >You can use the last locked values as the starting point for the next
training session. You should be able to see how much it (the filter) varies from session to session, and then you will know if adaptation is warranted.
> >Clay > > >
Thanks Clay That's an good idea , I can also try to develop a metric between changes in the measured impulse response and the changes in the filter taps post training. Yes Eric it is interesting- it's a very difficult application where the passbands have to be calculated in the impulse pulse on a regular basis , every day or so. The channel requires to have its impulse response calibrated in over 40 db on in band noise. Fortunately the noise is uncorrelated. Colin _____________________________ Posted through www.DSPRelated.com
Reply by January 31, 20142014-01-31
On Thursday, January 30, 2014 8:19:01 PM UTC-5, sigmonde wrote:
> >On Wed, 29 Jan 2014 04:53:49 -0600, "sigmonde" <10104@dsprelated> > > >wrote: > > > > > >>>>Is the echo path acoustic? If so, it's unlikely to be truly static. > > >>>Acoustic paths have a funny way of not being LTI even though we like to > > >>>think of them that way. Also the speaker will be affected by > > temperature, > > >>>aging, etc. > > >>>> > > >>>>Bob > > >>>> > > >>>The path is a steel conduit, it's impulse response is measured every > > >>wake > > >>>up cycle. > > >>>C > > >>> > > >>>_____________________________ > > >>>Posted through www.DSPRelated.com > > >>> > > >>Sorry should have added that it is an acoustic transmission system > > >>C > > > > > >Is the signal transmitted through the air in the pipe or through the > > >pipe itself? Not that it matters, but I'm curious. > > > > > >Also, if you measure the IPR whenever the systems wakes up, it may > > >benefit from an adaptive algorithm, and then just freeze the taps once > > >you're confident that it's trained. This is a fairly common thing to > > >do. > > > > > > > > > > > > > > >Eric Jacobsen > > >Anchor Hill Communications > > >http://www.anchorhill.com > > > > > Thanks Eric > > > > Yes I'll try locking the taps > > > > It's a through steel communications system where the channel has a > > transfer function very similair in structure to a comb filter. > > > > Colin > > > > _____________________________ > > Posted through www.DSPRelated.com
You can use the last locked values as the starting point for the next training session. You should be able to see how much it (the filter) varies from session to session, and then you will know if adaptation is warranted. Clay
Reply by Eric Jacobsen January 30, 20142014-01-30
On Thu, 30 Jan 2014 19:19:01 -0600, "sigmonde" <10104@dsprelated>
wrote:

>>On Wed, 29 Jan 2014 04:53:49 -0600, "sigmonde" <10104@dsprelated> >>wrote: >> >>>>>Is the echo path acoustic? If so, it's unlikely to be truly static. >>>>Acoustic paths have a funny way of not being LTI even though we like to >>>>think of them that way. Also the speaker will be affected by >temperature, >>>>aging, etc. >>>>> >>>>>Bob >>>>> >>>>The path is a steel conduit, it's impulse response is measured every >>>wake >>>>up cycle. >>>>C >>>> >>>>_____________________________ >>>>Posted through www.DSPRelated.com >>>> >>>Sorry should have added that it is an acoustic transmission system >>>C >> >>Is the signal transmitted through the air in the pipe or through the >>pipe itself? Not that it matters, but I'm curious. >> >>Also, if you measure the IPR whenever the systems wakes up, it may >>benefit from an adaptive algorithm, and then just freeze the taps once >>you're confident that it's trained. This is a fairly common thing to >>do. >> >> >> >> >>Eric Jacobsen >>Anchor Hill Communications >>http://www.anchorhill.com >> >Thanks Eric > >Yes I'll try locking the taps > >It's a through steel communications system where the channel has a >transfer function very similair in structure to a comb filter. > >Colin
Ah, interesting. Sounds like a fun application. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com
Reply by sigmonde January 30, 20142014-01-30
>On Wed, 29 Jan 2014 04:53:49 -0600, "sigmonde" <10104@dsprelated> >wrote: > >>>>Is the echo path acoustic? If so, it's unlikely to be truly static. >>>Acoustic paths have a funny way of not being LTI even though we like to >>>think of them that way. Also the speaker will be affected by
temperature,
>>>aging, etc. >>>> >>>>Bob >>>> >>>The path is a steel conduit, it's impulse response is measured every >>wake >>>up cycle. >>>C >>> >>>_____________________________ >>>Posted through www.DSPRelated.com >>> >>Sorry should have added that it is an acoustic transmission system >>C > >Is the signal transmitted through the air in the pipe or through the >pipe itself? Not that it matters, but I'm curious. > >Also, if you measure the IPR whenever the systems wakes up, it may >benefit from an adaptive algorithm, and then just freeze the taps once >you're confident that it's trained. This is a fairly common thing to >do. > > > > >Eric Jacobsen >Anchor Hill Communications >http://www.anchorhill.com >
Thanks Eric Yes I'll try locking the taps It's a through steel communications system where the channel has a transfer function very similair in structure to a comb filter. Colin _____________________________ Posted through www.DSPRelated.com
Reply by Eric Jacobsen January 30, 20142014-01-30
On Wed, 29 Jan 2014 04:53:49 -0600, "sigmonde" <10104@dsprelated>
wrote:

>>>Is the echo path acoustic? If so, it's unlikely to be truly static. >>Acoustic paths have a funny way of not being LTI even though we like to >>think of them that way. Also the speaker will be affected by temperature, >>aging, etc. >>> >>>Bob >>> >>The path is a steel conduit, it's impulse response is measured every >wake >>up cycle. >>C >> >>_____________________________ >>Posted through www.DSPRelated.com >> >Sorry should have added that it is an acoustic transmission system >C
Is the signal transmitted through the air in the pipe or through the pipe itself? Not that it matters, but I'm curious. Also, if you measure the IPR whenever the systems wakes up, it may benefit from an adaptive algorithm, and then just freeze the taps once you're confident that it's trained. This is a fairly common thing to do. Eric Jacobsen Anchor Hill Communications http://www.anchorhill.com