SNR Measurements for CCD vs CMOS sensors

[ellencglad]

Hi,
In the paper online "ASTROIMAGEJ: IMAGE PROCESSING AND PHOTOMETRIC EXTRACTION FOR ULTRA-PRECISE ASTRONOMICAL LIGHT CURVES (EXPANDED EDITION)," I found the equation used to calculate Noise (N) that is used to compute SNR in the Measurements table.

Why does N actually decrease when System Gain increases? Is this native to CCDs or does anyone have experience with this occurring in CMOS sensors as well?

Thank you,
Ellen

[SCO]

I cannot find the paper "ASTROIMAGEJ: IMAGE PROCESSING AND PHOTOMETRIC EXTRACTION FOR ULTRA-PRECISE ASTRONOMICAL LIGHT CURVES (EXPANDED EDITION)".  Could you please provide a link to it?

[ellencglad]

Hi,
This is the paper documenting AstroImageJ :
ASTROIMAGEJ: IMAGE PROCESSING AND PHOTOMETRIC EXTRACTION FOR ULTRA-PRECISE
ASTRONOMICAL LIGHT CURVES (EXPANDED EDITION)

Link: https://arxiv.org/abs/1701.04817

Authors: [KAREN A. COLLINS1,2,3 , JOHN F. KIELKOPF 3, KEIVAN G. STASSUN1,2, AND FREDERIC V. HESSMAN 4]

[karenacollins]

Hi,

I think the best bet to get the deepest understanding relative to CCDs is to check out the three references provided in the paper included below. My 1000 foot level response would be that generally when an application (Gain) is added to a system, noise is expected to be increased. However, the term "Gain" used relative to CCDs is really a loss. As stated in the same section of the paper, Gain is defined as G = electrons/ADU. So if your detector has a "Gain" of 1.5, that means when you have 150 electrons in a well, those electrons get converted to 100 ADU at the end of the chip (i.e. actually a loss).

I don't have any experience with CMOS detectors yet, but I expect similar conventions would be passed along. Maybe other users have more to add.

Also, for future reference, links to the AIJ papers are provided at the top of the main page on the forum.

Karen

-Mortara, L., & Fowler, A. 1981, in Proc. SPIE, Vol. 290, Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, 28

-Howell, S. B. 1989, PASP, 101, 616

-Howell, S. B. 2006, Handbook of CCD Astronomy (Cambridge, UK: Cambridge University Press)

[ellencglad]

Thanks Karen for the reply. It makes sense that increased Gain (increased electrons per single ADU) is a loss and that should increase Noise.

Using that CCD equation to calculate N, when I hold other values constant and increase G, N actually decreases. Do you know why that is? I've read that strictly speaking Gain is [ADU/e-] but people generally state its inverse [e-/ADU] so I wonder if I'm supposed to be inputting the inverse value.

Thank you,
Ellen

[karenacollins]

You'd need to input the gain number as G = electrons/ADU.

[ellencglad]

Hi Karena,
When I input a higher Gain for the Aperture Settings in AIJ prior to performing photometry, SNR values in the Measurements Table increase. Is this what's expected? I'd think that a higher gain would cause a decrease in SNR.

Thanks,
Ellen Glad

[karenacollins]

Since Noise has a factor of roughly 1/SQRT(Gain), with gain as defined above, when gain is increased, noise will decrease (with all other factors remaining the same). Since Noise is in the denominator of S/N, S/N should generally increase when gain is increased (with everything else remaining the same).

That's because when gain, as defined above, is increased, it implies you had more electrons that caused the ADU values in the aperture (since it is fixed in your test), which is F* in the equation. Since a higher gain implies that more electrons resulted in the measured ADUs, that means that more photons were incident on the detector, which would mean lower Poisson noise, and thus a higher signal to noise ratio.

Karen

[ellencglad]

That makes sense. Thanks for the response!

-Ellen