[CCD-astrometry-photometry] [Aavso-photometry] SXV-H9 linearity

[Pedro Pastor]

Thank you very much Arne for your lecturing (as usual). 

(I have just received your book and I'm going to attach to it all the
printings from your e-mails ;-).

I wish I sent before this question to the list. I've been living in a big
mistake due to a long e-mail explanation from a "supposed-to-be" CCD guru on
a very popular discussion list. 

On the other hand, your explanation raises a new question (to me): Where the
increase of sensitivity comes from in a binned CCD??

(You even said -from the accumulated signal point of view- binning could be
equivalent to software pixel post-summing).

I would said the increase of sensitivity came (in a 2x2 binned CCD) from
having an equivalent photosite of 4 times the surface BUT with 4 times well
depth, that is to say, as if the chip "created" a new big well from the
union of the 4 individual ones. (In that particular case no individual pixel
saturation effect would appear).

And finally, that reasoning leads me to the point of asking myself what's
the definition of "Gain" for CCD's. From manufactures' specifications, it
seems like it is the electrons to ADU's ratio (a mere conversion factor).
How can I work out the gain of my CCD on different binning configurations??

Thank you very much, and sorry for pestering you (all).

Regards,

Pedro

-----Mensaje original-----
De: Arne Henden [mailto:aah at nofs.navy.mil] 
Enviado el: martes, 07 de diciembre de 2004 17:36
Para: Pedro Pastor
CC: CCD-astrometry-photometry at yahoogroups.com; aavso-photometry at aavso.org
Asunto: Re: [CCD-astrometry-photometry] [Aavso-photometry] SXV-H9 linearity

Pedro Pastor wrote:
> Hello,
> 
> I find this topic very interesting. I have always wondered how binning
works
> with respect to the main CCD sensor parameters (linearity, gain, SNR,
etc).
> And I end up thinking I have no idea how binning works.
> 
> Some time ago I was told a binned CCD camera is always linear. After Gary
> comment it seems to be not. How does it work? If I use 2x2 binning, what
do
> I get: a) The sum of the electrons on 4 different wells; b) Only 1 well
> where all 4 photosites pour electrons together (up to ADC range)??
> 
The exposure is always made with the entire array in unbinned mode.
It is only during the readout process that binning takes place.
   Charge is first shifted from row to row, with the first row
being shifted into the serial register.  Vertical binning takes
place in this step, as you can shift more than one row into the
serial register at one time, effectively summing the charge.
   Charge is then shifted along the serial register, each serial
pixel ending up in the output summing well and read by the electronics.
Horizontal binning occurs here, as you can shift more than
one serial register into the summing well, effectively summing
the charge.
   For this binning process to work properly, the serial register
"pixels" usually have twice the well capacity of a typical array pixel.
Therefore, two nearly saturated pixels can be summed together in a
serial pixel without any loss.  Then the output summing well has to
have 4x the capacity of an array pixel in order for a 2x2 binning
operation to not have any loss.  These increasing well capacities are
common in science-grade detectors.  No detector I know of has 3x/9x
capacities in the serial and summing well pixels, so anytime you
do 3x3 binning, you always risk saturation of the summing process.
   Now, what happens if an array pixel gets saturated, but its neighbors
do not (for example, if you are undersampled so most of a star's light
falls on a single pixel)?  The 4 array pixels comprising a 2x2 binned
pixel get summed together; the resultant signal is less than 4x the
capacity of a single array pixel, so you assume the binned pixel is
unsaturated.  This can be a false assumption, which is what Gary
found out.
   Usually commercial camera vendors keep the same signal-chain gain
(electrons
per ADU), no matter what binning takes place.  This means that summing
4 pixels, each containing 1/4 of their full well, can result in a
signal that pegs the ADC.  Many professional cameras have adjustable
signal-chain gain so that you can keep full-well performance if you wish.
   You can always do binning in software after the readout, of course,
and handle any kind of binning factor without saturation.  The
disadvantage is readout time, software time, and increased readnoise
since each array pixel is then read in unbinned fashion.
Arne