wiki:astronomy:observational_astronomy:snr

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision		 Previous revision
wiki:astronomy:observational_astronomy:snr [2024/11/04 20:20] – Roy Proutywiki:astronomy:observational_astronomy:snr [2024/11/04 21:06] (current) – Roy Prouty
Line 38: Line 38:
 What's worse is that there is no guarantee that each pixel must abide the same random distribution as its neighbors. There may be pixel-level or optical system-level defects ensuring this. So each pixel in each frame is a sampling of some unknown probability distribution. What's worse is that there is no guarantee that each pixel must abide the same random distribution as its neighbors. There may be pixel-level or optical system-level defects ensuring this. So each pixel in each frame is a sampling of some unknown probability distribution.
  
-=== Signals ===+=== Stacks of Signals ===
  
 The measurement of counts in the pixels of each light frame are clearly owing to a few separate phenomenon occurring in unison at the time of observation. The generation of photoelectrons from incident light, the liberation of electrons owing to their temperature, and the liberation of electrons due to electronic effects are all inherently random, [[wiki:maths:poisson|Possionian]] processes.\\ \\ The measurement of counts in the pixels of each light frame are clearly owing to a few separate phenomenon occurring in unison at the time of observation. The generation of photoelectrons from incident light, the liberation of electrons owing to their temperature, and the liberation of electrons due to electronic effects are all inherently random, [[wiki:maths:poisson|Possionian]] processes.\\ \\
Line 52: Line 52:
 And finally, considering that each pixel of each frame captured is a sampling of a probability distribution, we should sample that distribution $N$ times to generate a stack of frames. {{ :wiki:astronomy:observational_astronomy:pixelarraystack.png?400 }}\\ \\ And finally, considering that each pixel of each frame captured is a sampling of a probability distribution, we should sample that distribution $N$ times to generate a stack of frames. {{ :wiki:astronomy:observational_astronomy:pixelarraystack.png?400 }}\\ \\
  
 +If each pixel value (the counts) from the $i,j$-th pixel is a random sampling of a [[wiki:maths:poisson|Poisson Distribution]], then $L^n_{ij}\thicksim\mathcal{Pois(\lambda)} \rightarrow \mathcal{Norm(\lambda, \sqrt{\lambda})}$ if $\lambda$ is large enough (and it is for these purposes!). \\ \\
 +
 +So we can treat each "stack" of pixels as its own ensemble of randomly sampled events (identically and independently distributed random events). Each stack of pixels should therefore constitute its own $\mathcal{Pois(\lambda_{ij})}\rightarrow \mathcal{Norm(\lambda_{ij}, \sqrt{\lambda_{ij}})}$. **Perhaps this is the best demonstration yet of why we need to take multiple frames of the same type.**
  
 ---- ----
  
-If each pixel value (the counts) from the $i,j$-th pixel is a random sampling of a [[wiki:maths:poisson|Poisson Distribution]], then $L^n_{ij}\thicksim\mathcal{Pois(\lambda)} \rightarrow \mathcal{Norm(\lambda, \sqrt{\lambda})}$ if $\lambda$ is large enough (and it is for these purposes!).+==== Noise ==== 
 + 
 +For a Normal Distribution, we can call the $\sigma$ (i.e., the uncertainty or the standard of deviation) the noise in the signal ($\mu$--the average value measured in the stack). So then, any light frame pixel can be said to have an average value with some signal ($\mu$) along with some noise ($\sigma$). If there is high confidence in the measured signal (meaning low uncertainty), then $\mu \gg \sigma$. 
 + 
 +Further, for the sources of unwanted signal, they each come with their own signal and noise -- trusted to be Normally Distributed. So in the construction of the calibrated frame, only the *signal* from unwanted signal sources can be removed. The uncertainty associated with these signals cannot be separated -- they were measured at the same time! All one can hope to do here is either (a) not collect unwanted signal & noise in the first place or (b) acknowledge the uncertainty in the final result. 
 + 
 +{{ :wiki:astronomy:observational_astronomy:lightframe.png?600 |}} 
 + 
 +{{ :wiki:astronomy:observational_astronomy:calibratedframe.png?600 |}} 
 +In the above, despite the average Dark Signal being removed, the uncertainty associated with the collection of the Dark cannot be disentangled from the uncertainty in the source signal -- so it remains. The distribution of dark signal is therefore centered on $0$, but with some spread about $0$. 
 + 
 +==== Signal to Noise Ratio ==== 
 +In returning to the ultimate goal of claiming one hypothesis true over another ... The Null Hypothesis is that the signal is indistinguishable from the noise. The Alternate Hypothesis is that the signal is distinguishable from the noise. We can construct the statistical model of the Alternate Hypothesis with $\mathcal{Norm}(\mu, \sqrt{\mu^2 + \mu_D^2 + \mu_U^2 + ... })$. And the Null Hypothesis with $\mathcal{Norm}(0, \sqrt{\mu_D^2 + \mu_U^2 + ... })$ -- where the uncertainties are just the sum of the uncertainties in quadrature. \\ \\ 
 + 
 +So the figure of merit here is the Signal-to-Noise Ratio (SNR). In short, this quantity is the number of standard deviations the average calibrated signal is from the uncertainty in that average signal *collected*.\\ \\ 
 + 
 +**At the UMBC Observatory, we aim for an SNR > 10 to confidently reject the Null Hypothesis.**