I would like to design a signal processing system that will extract a
signal from noise. The target signal is a sinusoid at some frequency
between DC and 50 kHz. The signal is sampled at 100 kHz. Processor
bandwidth and memory are, of course, at a premium. The strength of
your signal may be small compared to ambient noise, including near
frequency jammers.

Can any one throw some ideas of how to solve this? Thanks

On Wed, 26 Mar 2008 17:57:49 -0700, itsh11 wrote:

> I would like to design a signal processing system that will extract a
> signal from noise. The target signal is a sinusoid at some frequency
> between DC and 50 kHz. The signal is sampled at 100 kHz. Processor
> bandwidth and memory are, of course, at a premium. The strength of your
> signal may be small compared to ambient noise, including near frequency
> jammers.
> 
> Can any one throw some ideas of how to solve this? Thanks

Is the frequency of the signal known before hand, or does one have to 
figure it out, in the presence of jammers?

What intelligence are you trying to get?  Just the presence of the 
signal?  Some message that's modulated on?  How often does it turn on and 
off, or change frequency?

If the signal is changing rapidly in an unknown way, then you're probably 
out of luck -- what distinguishes the signal from noise?

-- 
Tim Wescott
Control systems and communications consulting
http://www.wescottdesign.com

Need to learn how to apply control theory in your embedded system?
"Applied Control Theory for Embedded Systems" by Tim Wescott
Elsevier/Newnes, http://www.wescottdesign.com/actfes/actfes.html

On Mar 26, 9:54 pm, Tim Wescott <t...@seemywebsite.com> wrote:
> Is the frequency of the signal known before hand, or does one have to
> figure it out, in the presence of jammers?
>
> What intelligence are you trying to get?  Just the presence of the
> signal?  Some message that's modulated on?  How often does it turn on and
> off, or change frequency?
>
> If the signal is changing rapidly in an unknown way, then you're probably
> out of luck -- what distinguishes the signal from noise?
>
> --
> Tim Wescott
> Control systems and communications consultinghttp://www.wescottdesign.com


I need to find the frequency of the sinusoid, so I do not know the
frequency of the signal beforehand. The noise is wideband and might
have more power than the target signal (in this case the sinusoid at
some unknown frequency between 0 and 50 kHz).

The most logical way to do it would probably be with FFTs. You would
calculate an FFT of appropriate length for the signal resolution you were
looking to achieve, and then say that your signal was whatever the
highest-powered frequency might be (perhaps with some minimum floor below
which you assume there is no signal present). And that is about the only
way you could do it - you might be able to extract a sinusoidal signal but
your chances of getting a voice signal or other arbirary function out of
there are practically nil.


That or something similar is about the only way you could 

>On Mar 26, 9:54 pm, Tim Wescott <t...@seemywebsite.com> wrote:
>> Is the frequency of the signal known before hand, or does one have to
>> figure it out, in the presence of jammers?
>>
>> What intelligence are you trying to get?  Just the presence of the
>> signal?  Some message that's modulated on?  How often does it turn on
and
>> off, or change frequency?
>>
>> If the signal is changing rapidly in an unknown way, then you're
probably
>> out of luck -- what distinguishes the signal from noise?
>>
>> --
>> Tim Wescott
>> Control systems and communications
consultinghttp://www.wescottdesign.com
>
>
>I need to find the frequency of the sinusoid, so I do not know the
>frequency of the signal beforehand. The noise is wideband and might
>have more power than the target signal (in this case the sinusoid at
>some unknown frequency between 0 and 50 kHz).
>

On Mar 27, 5:57=A0am, its...@yahoo.com wrote:
> I would like to design a signal processing system that will extract a
> signal from noise. The target signal is a sinusoid at some frequency
> between DC and 50 kHz. The signal is sampled at 100 kHz. Processor
> bandwidth and memory are, of course, at a premium. The strength of
> your signal may be small compared to ambient noise, including near
> frequency jammers.
>
> Can any one throw some ideas of how to solve this? Thanks

distance between peaks of autocorrelation function might come handy

important question, Tim also asked for it : is the frequency of the
sinusoid steady or does it change? With what precision do you want to
identify this frequency?

If the frequency is steady (constant) try averaging the spectrums of
short-time-fourier-trasforms over approprietly small selected windows.
This might reveal your frequency.

Another question : do you want to do that in real-time or you can do the
processing off-line, for example in MATLAB?

Manolis

On Mar 26, 10:54=A0pm, its...@yahoo.com wrote:

> I need to find the frequency of the sinusoid, so I do not know the
> frequency of the signal beforehand. The noise is wideband and might
> have more power than the target signal (in this case the sinusoid at
> some unknown frequency between 0 and 50 kHz).

Sounds like a textbook application for an adaptive line enhancer.
Adaptation is based upon the difference between the signal and a
delayed version of the signal.  The noise is decorrelated by the
delay, while the sinusoid remains correlated even with the delay.

Search for "adaptive line enhancer", "adaptive noise cancellation",
etc.

Greg

On Mar 26, 7:57 pm, its...@yahoo.com wrote:
> I would like to design a signal processing system that will extract a
> signal from noise. The target signal is a sinusoid at some frequency
> between DC and 50 kHz. The signal is sampled at 100 kHz. Processor
> bandwidth and memory are, of course, at a premium. The strength of
> your signal may be small compared to ambient noise, including near
> frequency jammers.
>
> Can any one throw some ideas of how to solve this? Thanks

Is the signal from the jammers always wideband and uncorrelated?

I agree with Greg,

ADAPTIVE LINE ENCHANCEMENT is the best solution.

Another idea is based on the fact that the autocorrelation of a periodic
signal is periodic with the same period. So, you could compute the
autocorrelation sequence of your signal and measure it's period. The
inverse of this period will be the frequency of your sinusoid.

Manolis