Front-Vs Back-Illuminated CCD Cameras For Photometric Surveys: A Noise Budget Analysis

Astron. Nachr. / AN 000, No. 00, 1 4 (0000) / DOI please set DOI!
Front- vs back-illuminated CCD cameras for photometric surveys : a

noise budget analysis.
Crouzet N.
, Guillot T. , Fressin F. , Blazit A. , and the A STEP team
Observatoire de la Cote dAzur, Boulevard de lObservatoire, BP 4229, 06304 Nice Cedex 4, France
The dates of receipt and acceptance should be inserted later
Key words instrumentation : detectors methods : numerical technique : photometric
Exoplanetary transit and stellar oscillation surveys require a very high precision photometry. The instrumental noise has
therefore to be minimized. First, we perform a semi-analytical model of different noise sources. We show that the noise
due the CCD electrodes can be overcome using a gaussian PSF (Point Spread Function) of full width half maximum
larger than 1.6 pixels. We also find that for a PSF size of few pixels, the photometric aperture has to be at least 2.5 times
larger than the PSF full width half maximum. Then, we compare a front- with a back-illuminated CCD through a MonteCarlo simulation. Both cameras give the same results for a PSF full width half maximum larger than 1.5 pixels. All these
simulations are applied to the A STEP (Antarctica Search for Transiting Extrasolar Planets) project. As a result, we choose
a front-illuminated camera for A STEP because of its better resolution and lower price, and we will use a PSF larger than
1.6 pixels.
c 0000 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

1 Introduction
2 Noise analysis
2.1 Description of the cameras
The photometric technique allows a direct detection of luminosity variations. Several disciplines are therefore concerned. In asteroseismology, these variations are used to
identify stellar oscillations, giving an access to the interior
of stars. In planetary sciences, a decrease of luminosity caused by an extrasolar planet occulting its parent star during
a transit is used to characterize the planet (Charbonneau et
al. 2000; Moutou et al. 2004). In both cases, a very high
precision photometry is required, typically to a millimagnitude level. Challenging technical issues have therefore to
be solved (Rauer et al. 2004), and high accuracy algorithms
are needed (Irwin at al. 2007; Gillon et al. 2006; Magain
et al. 2007). The Antarctica Search for Transiting Extrasolar Planets (A STEP) aims to detect planetary transits and
stellar oscillations from Dome C, Antarctica (Fressin et al.
2005). The three months continuous night as well as a very
dry weather are extremely favorable for photometric surveys. A fully automatized telescope is under development.
We present here a first noise analysis of this telescope that
leads to the choice of the camera, but that applies to other
photometric surveys. The first part shows a noise budget
obtained with a semi-analytical model. A second part describes a Monte Carlo simulation of a front- and a backilluminated CCD camera.
e-mail: [email protected]
When using available commercial cameras and a limited

fund, a problem for photometric surveys is whether to use a
backward-illuminated camera, which maximizes efficiency
by having electrodes on the non-irradiated side of the CCD,
or a front-illuminated camera in which the quantum efficiency is limited by direct reflection on the electrodes, but
which is simpler to build and has thus more pixels for a similar price tag. More specifically, in the case of A STEP,
these two classes of cameras were led by:
The back-illuminated camera DW 436 by Andor, with
a CCD EEV 42-40, containing 2048x2048 pixels. The
quantuum efficiency peaks at 94 %, with a mean of 90
% in the spectral range 600-800 nm. The pixel size is 13
m, and the total CCD size is 2.7 cm. The same CCD
is used for the CoRoT mission, giving us some facilities
for its testing and characterization. The pixel response
non-uniformity is around 2 %.
The front-illuminated camera Proline by FingerLake Instruments, with a CCD KAF-16801E by Kodak, containing 4096x4096 pixels. Its quantuum efficiency peaks
at 63 %, with a mean of 50 % in the spectral range 600800 nm. Around 40 % of flux is lost with respect to the
back-illuminated camera. The pixel size is 9 m, and
the total CCD size is 3.7 cm. The pixel response nonuniformity is around 0.5 %. The front-illuminated camera has also the advantage to be much cheaper, allowing
us to purchase a backup one.
author: title
2.2
CCD transmission
In order to test the CCDs, we model a grid in which the

optical transmission of each pixel is randomly set, with a
standard deviation of few per cent. Electrodes are assumed
to cover
% of a pixel surface in the case of a frontilluminated camera, thereby explaining the relatively modest quantum efficiency. On the other hand, the back-illuminated camera is considered as ideal, with no loss due to
the electrodes. An example of a CCD transmission is shown
in figure 1.
lated as the variation of the flux hitting the electrodes.

Overflow noise : Because of light sources such as crowding in the field of view, sky brightness, etc., a photometric aperture is set around each target star. The flux
outside this aperture is eliminated, shoud it come from
the target star or another source. This results in a loss of
the flux from the star if the PSF overflows the photometric aperture. The loss depends again on the PSF position
inside this aperture, and varies due to the telescope jitter.
The noise is then calculated as the variation of the flux
inside the aperture.
Interpixel noise : Each pixel has its own optical transmission, which vary from one pixel to another by typically 1 per cent. This PRNU (Photo Response Non Uniformity) is taken into account defining an equivalent
number of pixels under the PSF. The resulting noise
is :

were is the standard deviation of the pixel transmission distribution.

Fig. 1 Example of the CCD transmission matrix of a frontilluminated camera. The interpixel variations are typically few per
cent. The electrodes cover a part of a pixel and their transmission
is very low.
2.3
Other noise sources : Other noise sources are implemented such as the photon noise from the target star, the
noise from the sky background (taken as 22 mag/arcsec
with a slow variation along the CCD), and the camera
dark current and read-out noise.
The sources of instrumental noise

2.4 A semi-analytical model
We perform a first analysis of the different noise sources.

A good understanding of noises is indeed necessary for the
choice of a camera. In a more general way, this is critical in
transit survey data analysis (Pont, Zucker & Queloz 2006;
Smith et al. 2006). All noises are calculated in a semi-analytical model using squared photometric apertures of 3x3
and 5x5 pixels. We use the A STEP instrumental characteristics. The telescope is a 40 cm with a F/D of 4. Stars
range from magnitude 11 to magnitude 16. The exposure
time is 10 seconds. The PSF (Point Spread Function) is a
two-dimensional gaussian function. These parameters are
preliminar and will have to be defined more precisely during
the telescope design phase, in order to optimize the survey
(Horne 2002). Noises correspond to one exposure, without
any image processing. This is therefore a worst case. We
consider the following noise sources :
Electrode noise : The bad optical transmission of electrodes leads to a loss of flux. This loss depends on the
PSF position with respect to the electrodes. The PSF
motion onto the CCD, due to the telescope jitter, leads
to loss variations. The resulting noise is therefore calcu-
The results of the analysis based on the semi-analytical model are represented in figure 2. The electrode and overflow
noises are clearly dominant. The interpixel noise is also dominant for bright stars. As supposed, the other noise sources
are not dominant for a 11 magnitude star. For a 16 magnitude star, the photon noise is dominant for full width half
maxima between 1.2 and 2 pixels, but does not change the
total noise curve shape.
The main observation is that the electode noise reaches a
level of for PSF full width half maxima larger than 1.6
pixels. This noise can therefore be overcome adjusting the
PSF size. (It can be noted however, that the noise strongly
increases when considering non-gaussian PSFs with sharp
interfaces, e.g. a top-hat function. We will not consider this
further.) We also see that the overflow noise becomes dominant for full width half maxima larger than 2 pixels, given
the aperture of 5x5 pixels we use. In a general way, for our
PSFs of few pixels, we find that the photometric aperture
must be at least 2.5 larger than the PSF full width half maximum.
www.an-journal.org
Astron. Nachr. / AN (0000)
Fig. 2 Noise budget as a function of the PSF full width half maximum, in a semi-analytical model, for a front illuminated camera
with a 5x5 pixels photometric aperture and 1 % interpixel variations. The results for stars of magnitude 11 (top) and 15 (bottom) are
represented. The CCD dark current and read-out noise are not plotted, and are always lower than the sky background noise.
2.5
Monte-Carlo simulations
In order to test our two cameras, we use a direct simulation

of the CCD that attempts to mimic real observations including jitter and interpixel noise. The simulations proceed as
follows : during a run, a gaussian PSF is moved along the
CCD in an arbitrary direction. For each position, the flux
inside the photometric aperture is measured. The resulting
noise level is estimated as the peak-to-peak flux variation.
This noise is calculated for several PSF full width half maxima, since this parameter can be chosen during the telescope
design phase. This simulation is performed for three interpixel values, 0, 1, and 5 % as the real value is unknown,
and for two electrode sizes, 50 and 0 % of a pixel, i.e. for
both cameras. The results of the Monte-Carlo simulations
www.an-journal.org
are shown in figure 3. Note this simulation does not account

for different quantuum efficiencies of both cameras. The advantage of the back-illuminated camera, i.e. no electrodes,
is valid only for PSFs with full width half maxima smaller
than 1.5 pixels. For larger PSFs, both cameras give the same
results, which means that the electrode noise is negligible.
This is in perfect agreement with our semi-analytical model.
These simulations imply that for well-sampled PSFs, the
noise difference is essentially due to the difference in quantuum efficiency, i.e. for the two cameras that are considered,
. This is to be compared to the fact that
the 4 times increase in pixel number allows having either
4 times as many targets with the same crowding, or, with
exactly the same field of view, a reduction of the interpixel
author: title
Fig. 3 Noise as a function of the PSF full width half maximum, for a 3x3 pixels photometric aperture. A front illuminated camera (with
electrodes), and a back-illuminated camera (as an ideal case with no electrodes) are represented. Three values of interpixel variations are
used.
noise by a factor 2 due to PSFs that are better sampled spatially.

Other advantages of the presently built CCD cameras is
that front-illuminated ones have less interpixel noise (1%
vs. 3%), and are generally cheaper (by
%). A frontilluminated CCD camera is therefore, in the case of A STEP,
more advantageous.
3 Conclusion
We performed a semi-analytical noise analysis to identify
the limiting noise sources in precision photometry, using a
gaussian PSF. The electrode, overflow, and interpixel noises
are dominant, as well as the photon noise for faint stars.
We showed that the electrode noise becomes negligible for
gaussian PSF full width half maxima larger than 1.6 pixels.
We also found that for PSFs of few pixels, the photometric
aperture must be at least 2.5 times larger than the PSF full
width half maximum.
We then compared a front- and a back-illuminated camera in a Monte Carlo simulation. For photometric surveys
for which the PSF is well-sampled (at least 1.5 pixels full
width half maximum), and limited in terms of budget to
existing commercial cameras, we found that a front-illuminated camera is a better alternative.
Gillon, M., Pont, F., Moutou, C., Bouchy, F., Courbin, F., Sohy, S.,
Magain, P.: 2006, AA 459, 249
Horne, K.: 2002, ESA SP 485, 137
Irwin, J., Irwin, M., Aigrain, S., Hodgkin, S., Hebb, L., Moraux,
E.: 2007, MNRAS 375,1449
Magain, P., Courbin, F., Gillon, M., Sohy, S., Letawe, G., Chantry,
V., Letawe, Y.: 2007, AA 461, 373
Moutou, C., Pont, F., Bouchy, F., Mayor, M.: 2004, AA 424, L31
Pont, F., Zucker, S., Queloz, D.: 2006, MNRAS 373, 231
Rauer, H., Eisloffel, J., Erikson, A., Guenther, E., Hatzes, A.P.,
Michaelis, H., Voss, H.: 2004, PASP 116, 38
Smith, A.M.S., Collier Cameron, A., Christian, D.J., et al.: 2006,
MNRAS 373, 1151
References
Charbonneau, D., Brown, T.M., Latham, D.W., Mayor, M.: 2000,
ApJ 529, L45
Fressin, F., Guillot, T., Bouchy, et al.: 2005, EAS Publ. Ser. 14,
309
www.an-journal.org

Front-Vs Back-Illuminated CCD Cameras For Photometric Surveys: A Noise Budget Analysis

Uploaded by

Copyright:

Available Formats

Front-Vs Back-Illuminated CCD Cameras For Photometric Surveys: A Noise Budget Analysis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Front-Vs Back-Illuminated CCD Cameras For Photometric Surveys: A Noise Budget Analysis

Uploaded by

Copyright:

Available Formats

Astron. Nachr. / AN 000, No. 00, 1 4 (0000) / DOI please set DOI!