filtered - x - LMS

Hi,

some days before  I asked you something about the LMS and the NLMS.
Now I am would  take one step forward to the FXLMS. I have  achieved
very good results in signal attenuation with the NLMS (about 200db for
white and colored noise input signals),but with the fxlms the noise
reduction ratio decreased to 3dB.
I use a simple 1 sample delay as secondary path and the same for the
estimated secondary path. When I use a correlated input signal like a
sinus  the attenuation is similar to the simple NLMS but for white
noise the performance decrease.

Now may question, is there anybody who has some experience with the
FXLMS because may coworkers aren`t familiar with the FXLMS.
Is it a normal result or could there a mistake in my Matlab skript?
How can I improve the performance for uncorrelated signals?
I have also changed the secondary path to an highpass filter I think
its more  	realistic, but the result for uncorrelated signals didn`t
improve.

So I would be glad if I can found anybody for a discussion about the
FXLMS.

Thanks Markus

On Fri, 28 Apr 2005, Gie78 wrote:

> Hi,
>
> some days before  I asked you something about the LMS and the NLMS.
> Now I am would  take one step forward to the FXLMS. I have  achieved
> very good results in signal attenuation with the NLMS (about 200db for
> white and colored noise input signals),but with the fxlms the noise
> reduction ratio decreased to 3dB.
> I use a simple 1 sample delay as secondary path and the same for the
> estimated secondary path. When I use a correlated input signal like a
> sinus  the attenuation is similar to the simple NLMS but for white
> noise the performance decrease.
>
> Now may question, is there anybody who has some experience with the
> FXLMS because may coworkers aren`t familiar with the FXLMS.
> Is it a normal result or could there a mistake in my Matlab skript?

Hi!
For filtered x-lms in the feedforward case, the performance of your system
mainly depends on some prior knowledge about your signal before sent to
the plant. Therefore I think you should have a delay in your primary path,
then I think you will see the results are getting better. The delay should
be equal to or larger than the delay in the secondary path to get good
results.

The results also depend on the crosscorrelation between the disturbance
signal and the filtered reference signal.

I recommend you to look in Stephen Elliots "Signal Processing for active
control" or Kuo and Morgan "Active Noise Control Systems" for more
information on F-XLMS.

> How can I improve the performance for uncorrelated signals? I have also
> changed the secondary path to an highpass filter I think its more
> realistic, but the result for uncorrelated signals didn`t improve.
>
> So I would be glad if I can found anybody for a discussion about the
> FXLMS.
>
> Thanks Markus
>

Gie78 wrote:
> Hi,
> 
> some days before  I asked you something about the LMS and the NLMS.
> Now I am would  take one step forward to the FXLMS. I have  achieved
> very good results in signal attenuation with the NLMS (about 200db for
> white and colored noise input signals),but with the fxlms the noise
> reduction ratio decreased to 3dB.

200 dB is a voltage ratio of 1 to 10,000,000,000. How do you measure it?

   ...

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

"Jerry Avins" <jya@ieee.org> wrote in message
news:d8qdnarlmuDJEe_fRVn-hg@rcn.net...
> Gie78 wrote:
> > Hi,
> >
> > some days before  I asked you something about the LMS and the NLMS.
> > Now I am would  take one step forward to the FXLMS. I have  achieved
> > very good results in signal attenuation with the NLMS (about 200db for
> > white and colored noise input signals),but with the fxlms the noise
> > reduction ratio decreased to 3dB.
>
> 200 dB is a voltage ratio of 1 to 10,000,000,000. How do you measure it?


Well, if he's using floating point numbers (this is digital after all), it might
be possible, though still not likely for an adaptive filter algorithm.  But I am
guessing this is a typo and maybe he meant 20dB?

-- 
Jon Harris
SPAM blocked e-mail address in use.  Replace the ANIMAL with 7 to reply.

Jerry Avins <jya@ieee.org> wrote in news:d8qdnarlmuDJEe_fRVn-hg@rcn.net:

> Gie78 wrote:
>> Hi,
>> 
>> some days before  I asked you something about the LMS and the NLMS.
>> Now I am would  take one step forward to the FXLMS. I have  achieved
>> very good results in signal attenuation with the NLMS (about 200db for
>> white and colored noise input signals),but with the fxlms the noise
>> reduction ratio decreased to 3dB.
> 
> 200 dB is a voltage ratio of 1 to 10,000,000,000. How do you measure it?
> 
>    ...
> 
> Jerry

I usually use my 34 bit ADC. Of course, the Johnson noise of my 1 ohm 
resistors are a problem.

-- 
Al Clark
Danville Signal Processing, Inc.
--------------------------------------------------------------------
Purveyors of Fine DSP Hardware and other Cool Stuff
Available at http://www.danvillesignal.com