e- / DN Uncertainty

[Harry Marsh]

Posted to CCD-world:
-+-+-+-
Jim--
I think what Paola is trying to state is that when the signal is reduced,
there is more error in the encoding of the signal [if one gets down in the
<1000DN range), and other errors (like the CTE floor creeping up to the
noise floor).

I found that e-/DN calculations were fairly accurate down to about 1/8th
scale on the ADC, or about 1000 e- CCD-wise.

	HHM

----------
> From: MY PIXEL <MYPIXEL at aol.com>
> To: CCD-world at cfht.hawaii.edu
> Subject: e- / DN  Uncertainty
> Date: Saturday, January 24, 1998 3:48 PM
> 
> Posted to CCD-world:
> -+-+-+-
> In a message dated 98-01-23 03:32:00 EST, pamico at eso.org writes:
> 
> << Subj:	 A question....
>  Date:	98-01-23 03:32:00 EST
>  From:	pamico at eso.org (Paola Amico)
>  To:	MYPIXEL at AOL.COM
>  
>  Dear Jim,
>  I take advantage of the offer you made us at the CCD Camera course last
may
>  for asking you a question. Hope it is still valid.........
>  
>  I am reviewing your notes concerning CCD characterization & testing
(Section
>  I) and I am puzzled by the result on page 2-12 concerning the Camera
gain
>  constant uncertainty. Without entering into formulas, I concluded that
the
>  error affecting the conversion factor should be inversely proportional
to
>  the total signal received on your test array and not only to the total
>  number of pixels. Is there some kind of "normalization" I am missing? 
>  
>  Thanks a lot,
>  
>  all the best....................paola amico
> 
>  
> ***************************************************jj
> 
> Paola,
> 
> Given . . .. 
> 
> K (e-/DN) = Signal / (Shot Noise)^2
> 
> Variance in K (e-/DN) = 2 K^2 / Np   where Np is the number of pixels
sampled
> 
> 
> I looked all over for my calculations using the "propagation of errors"
> formula for the variance in e-/DN (performed 22 years ago). I have one
more
> place to look. I sure would like to see your calculations.
> 
> For sure the number of pixels is important . . . sampling the signal and
shot
> noise for one pixel would be very uncertain. Statistics is improved by
pixel
> count (averages the shot noise by the square-root of the number of pixels
> sampled). As for your question about signal . . .  shot noise is going up
by
> the square-root of signal .. . K is equal to the Signal / (Shot Noise)^2.
So
> the numerator and denominator are tracking one another. 
> 
> K (e-/DN) is also in the equation. A lower voltage camera gain produces a
> higher K (e-/DN)  value with less DN count variation (for a fixed average
shot
> noise).  So I think that is right.
> 
> Let's look into those calculations further.
> 
> Jim
>   
>  **********************************jj
> -+-+-+-
> For information about CCD-world, send email to
owner-CCD-world at cfht.hawaii.edu.


-+-+-+-
For information about CCD-world, send email to owner-CCD-world at cfht.hawaii.edu.