From: jeroen@nospam.please   
      
   On 2014-01-24 14:43, haiticare2011@gmail.com wrote:   
   > On Friday, January 24, 2014 3:26:37 AM UTC-5, Jeroen Belleman wrote:   
   >> On 2014-01-24 05:58, ticare2011@gmail.com wrote:   
   >>   
   >> [Snipped the mangled text ... {Snipped more mangled text...]   
      
   >> Jeroen Belleman   
   >   
   > Hi Jeroen,   
    >   
   > To be fair, you are correct, the signal IS growing faster   
   > than the noise. But what allows Messr. Horowitz to see it is that the   
   > 'standard deviation' of the noise goes way down, more than the SD of   
   > the signal. So if you do Y measurement episodes of X measures each,   
   > the variation in the noise (in the reference beam) will be tight, and   
   > the signal (in the measurement beam) will almost always be greater   
   > than that tight noise floor.   
      
   Different terms, but same argument: If you think of the measurements   
   as the sum of a coherent signal and an incoherent noise part, the   
   SD of the signal grows with the square of the number sweeps,   
   while the noise grows linearly.   
      
      
   > To be frank, I am a software person, and to be put algorithms in   
   > code, you have to know exactly how the little gears turn.   
      
   OK, here is a description of an algorithm that you can play with   
   in software:   
      
   Take a floating point random number generator producing values   
   on [-1:1]. Fill an array of size 100 with random numbers. That's   
   your noise.   
      
   Add, say, 0.01 to one of the elements. That's your signal. If   
   you plot the array, you'll see nothing of that small addition.   
   The signal is well below the noise.   
      
   Now add, say, a million of these arrays, with the signal always   
   added to the same element of the array. The signal will now have   
   height ~10000, give or take a few thousand, while the noise band   
   is much lower, and centered on zero.   
      
   (You may learn a thing or two about your random number generator   
   in the process, too.)   
      
   Jeroen Belleman   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)