On Sunday, June 1, 2014 11:21:24 AM UTC-4, gghe...@gmail.com wrote:   
   > On Saturday, May 31, 2014 11:05:34 AM UTC-4, haitic...@gmail.com wrote:   
   >   
   > > FIX CR's   
   >   
   > >   
   >   
   > >   
   >   
   > >   
   >   
   > > Suppose we have two channels, a + B, which are neighboring wavelengths in   
   >   
   > > spectrum. Suppose there is a signal buried down in channel B but not in A   
   >   
   > > Say the signal is 1/1000 the strength of A-B.   
   >   
   > >   
   >   
   > Are you assuming that the noise in A is the same, (correlated) with that in   
   B?   
   >   
   > In general that is not the case.  (maybe only when the source is laser   
   noise?)   
   >   
   >   
   >   
   > There are some neat tricks to get rid of uncorrelated noise in two channels.   
   >   
   > (But this assumes there is also a signal in both channels.)   
   >   
   > So imagine two opamps looking at the jonhson noise of a single resistor.   
   >   
   > If those tow signals are multiplied together then the uncorrelated opamp   
   noise goes away (in time) and the correlated johnson noise in each channel   
   remains.   
   >   
   >   
   >   
   > George H.   
   >   
   > >   
   >   
   > > So we just add the signals in A,B 10,000 times. Then the buried   
   differential   
   >   
   > >   
   >   
   > > signal in B can rise above the concomitant noise in A,B. We see the signal   
   by subtracting the value in A.   
   >   
   > >   
   >   
   > >   
   >   
   > >   
   >   
   > > The question is, can this work under some circumstances? Can it work as a   
   blind   
   >   
   > >   
   >   
   > > heuristic where we don't know where a signal is, nor where A,B are? This   
   can be   
   >   
   > >   
   >   
   > > run in  "supervised" pattern recognition mode, where we are looking for   
   >   
   > >   
   >   
   > > correlations to some external thing. It could also be run in "unsupervised"   
   >   
   > >   
   >   
   > > PR mode, where a signal emerges by itself, and you say "this is   
   interesting.   
   >   
   > >   
   >   
   > > It might mean something."   
   >   
   > >   
   >   
   > > >   
   >   
   > >   
   >   
   > > > "If you subtract the killings, DC has one of the lowest crime rates in   
   the   
   >   
   > >   
   >   
   > > >   
   >   
   > >   
   >   
   > > >  country." - Mayor Berry, DC.   
      
   Interesting. George I was thinking in general of "hidden signals." My thing is   
   non invasive medical diagnostics. (among other things.) So you postulate there   
   is a signal in a sensor input from a patient that is buried in other signals.   
   If you look at an NIR spectrum, for example, reflected from a person's skin,   
   you speculate there's a pony in that mess. :)   
   jb   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)