Transcript Testing and characterizing the CCDs of the OmegaCam wide-angle camera F. Christen2, C.

Testing and characterizing the CCDs of the OmegaCam
wide-angle camera
F. Christen2, C. Cavadore1, B. Gaillard, O. Iwert1, K. Kuijken2, D. Baade1, S. Darbon1
1 : European Southern Observatory
2 : Kapteyn Instituut (Groningen)
OmegaCAM will cover the 1deg x 1deg field of view of the ESO VLT Survey Telescope with thirty two 2k x 4k CCDs. Four additional chips of similar type will be used for the
auto guiding and active optics control. Since the replacement of a unit detector or a re-arrangement of the mosaic after commissioning are not an option, every detector needs to
be fully characterized in advance. These tests need to be carried out under realistic astronomical operating conditions.
Summary of the results obtained to date
Quantum Efficiency
12
Non Linearity
10
0.8
8
0.7
0.6
Rms (%)
6
4
2
680
760
850
940
1040
320
420
520
620
720
820
920
Wavelength (nm)
Figure 4 : Quantum efficiency of 19 CCDs
Figure 5 : Photon response non-uniformity of 19 CCDs
Quantum Efficiency is measured at an operating temperature of -120°C. These measurements are
regularly checked by recalibrations. The repeatability is very good (Error : ± 0.2% max). The QE
dispersion from device to device is increasing towards shorter wavelengths. The QE figures in the
range of 300- 400 nm are higher than the minimum specification. PRNU is increasing in the red due to
fringing effects, and in the blue due to the backside laser annealing (diamond pattern, see Fig. 6)
OmegaCAM CCD Camera
Figure 14 : Peak-to-peak nonlinearity= 1.5%, Rms= 0.4%
Figure 15 : Linearity for 19 CCDs,
Mean= 0.27% Rms= 0.15%
After flatfielding ( the test is performed at 600nm where PRNU effects are the lowest), the
columns are averaged, applying a median filter. The residuals from a linear least-squares fit
to the result represents the non-linearity (Fig. 14). Non-linearities are typically less than 0.3%
(Fig. 15).
Cosmetics of the CCD “Indus II” :
The next images are three examples of 2kx4k full frame flat fields. The full test report includes 40 flat
field images at different wavelengths, light levels and gains.
Electrical Parameters
Conversion Factor (Gain) : derived from the measurement of the mean signal of flat field and
its variance.
Read Out Noise : spatial rms of the bias measured.
0.7
8
0.65
7
6
0.5
2
1
0.4
0
Capabilities :
• Long dark exposure time (several hours)
• Several modes and readout speeds driven by FIERA CCD controller
• Uniform illumination up to a field of 20cm diameter provided by the integrating sphere
• Large wavelength coverage (300-1100nm)
• Good spectral sampling (1nm)
• All devices can be controlled remotely by software
Left Port 225k
lia
At
Right Port 225K
Left Port 50k
Right Port 50k
Left Port 225k
Right Port 225K
Figure 17 : Read-out noise for 19 CCDs,
Mean= 2.9 e- Rms= 0.4 e- at 50kpix/s
Mean= 4.4 e- Rms= 0.7 e- at 225kpix/s
Conversion factor and read-out noise are measured for the two ports and at different speeds
(50 and 225 kpixels/s). The test bench is actually not read-out noise optimized because the
objective is to measure relative differences only.
Six different kinds of cosmetic defects :
1.0
Total non
column
defective
pixels
Sum of defects
1.2
Total
defective
pixels
Traps
Contract specification
Figure 9 : Total balance of cosmetic defects
With a mean per CCD of 66 hot and dark pixels (rms: 72) , 5 very bright pixels (rms: 4), 3 traps (rms:
3), 1 very large trap (rms: 2) and 2 bad columns (rms: 2), the overall cosmetic quality of the
Omegacam device is very well within specifications.
2
1.5
1
0.5
0
1.000000
0.999999
0.999998
0.999997
0.999996
0.999995
0.999994
0.999993
0.999992
0.999991
0.999990
Figure 18 : Dark current for 17 CCDs at -120°C,
Mean= 0.7 e-/pixel/h Rms= 0.6 e-/pixel/h
ae
l
an um
is
M
aj
or
C
ar
in
C
a
en
ta
ur
C
ha
us
m
ae
le
on
C
irc
in
us
D
or
ad
o
Fo
rn
ax
G
H
or rus
ol
og
iu
m
H
yd
ru
s
Le
pu
s
M
en
sa
Pa
vo
1.2
2.5
C
10.0
Charge Transfer Efficiency
C
34.6
Dark Current
a
100
Ar
100
Ap
us
100
At
l ia
100.0
CTE
Cosmetic defects
us
A
C ra
C ae
an lu
is m
M
C
en ajo
t r
C
ha aur
m us
ae
le
C on
irc
in
D us
or
ad
Fo o
rn
ax
H Gr
or
ol us
og
iu
H m
yd
ru
Le s
pu
Lu s
pu
M s
en
sa
Pa
vo
• Hot pixels : provide a signal of > 60 e- /pixel
/ hour.
• Dark pixels : have 50% or less than the
average output for uniform illumination at a
flat field level around 500 photo-electrons.
• Very bright pixels : provide a signal of >
200000 e-/pixel/hour.
• Trap : captures more than 10
electrons, measured with a flat field level
around 500 photo-electrons. Pocket pumping
acquisition technique is used to trace them.
• Very large trap : captures more than
10 000 electrons, measured with a flat
field level around 90% of the full well
capability.
• bad column : 10 or more contiguous hot or
dark pixels in a single column or a very
bright pixel or a very large trap.
Dark Current : A median-filtered stack of five 1-hour dark frames is used (temperature :
-120°C, read out speed : 50kpix/s, high gain mode : 0.55e-/ADU) is computed. The horizontal
and vertical over scans pixel are considered to determine the dark current. It is mostly less
than 1 e- / pixel / hour (see Fig. 18).
Charge Transfer Efficiency : Two methods have been used, one based on the extended pixel
response through the image overscan area (EPER), the other used the standard Fe55 method.
Ap
The following measurements
are routinely performed :
• Quantum efficiency (QE)
• Photon Response
• Non Uniformity (PRNU)
• Dark (short and long exposure)
• Bias
• Readout noise, conversion
factor
• Linearity
• Dark current
• Charge Transfer Efficiency
(CTE)
• Cosmetic defects : (hot pixels,
very bright pixels, dark pixels,
traps, very large traps bad
columns and coating blemishes)
Cosmetic Defects :
Right Port 50k
Figure 16 : Conversion factor for 19 CCDs,
Mean= 0.53 e-/ADU Rms= 0.03 e-/ADU
lia
In 1996, Amico & Böhm[1] designed the new ESO testbench. After several improvements
(automatization, high-level scripts) this system is now optimized for mass testing. The
main components are shown in the next picture (Fig. 3).
Left Port 50k
At
The Test Bench
Figure 8 : Flat field ( λ= 900nm,
bandwidth= 5nm, 1500ADU)
Between 420 and 870 nm the PRNU is photon noise limited. The acquisition of these images has
three goals : general appearance of the images, identification of unexpected/expected defects (bright
spot, patterns, 512x1K block stitching, scratches) and traps. Finally, all the defects are recorded for
compliance with the contract (see Fig. 9).
Logarithmic scale (%)
Figure 2 : Two Marconi 44-82 CCDs
Figure 7 : Flat field ( λ= 600nm,
bandwidth= 5nm, 1500ADU)
Ap
lia
0.45
At
Figure 6 : Flat field ( λ= 350nm,
bandwidth= 5nm, 1500ADU)
4
3
Ap
0.55
5
us
A
C ra
C ae
an lu
is m
M
aj
o
C r
a
C
e rin
C nta a
ha u
m rus
ae
le
C on
irc
in
D us
or
ad
Fo o
rn
ax
H G
or ru
ol s
og
iu
H m
yd
ru
Le s
pu
Lu s
pu
M s
en
sa
Pa
vo
Rms (e-)
0.6
us
A
C r
C ae a
an lu
is m
M
aj
C or
C ari
e
n
C nta a
ha u
m rus
ae
l
C eon
irc
in
D us
or
ad
Fo o
rn
ax
H G
or ru
ol s
og
iu
H m
yd
ru
Le s
pu
Lu s
pu
M s
en
sa
Pa
vo
CF (e-/ ADU)
• Thinned back-side illuminated devices 2kx4k 15μm
pixels.
• Single-layer Hf02 anti-reflective coating insures optimal
sensitivity in the blue and near UV.
• Two serial read-out registers
• On-chip Pt100 temperature sensor
• Invar package providing high flatness level
• Required flatness : ± 10 mm
Read Out Noise
Conversion Factor
Figure 1 : The OmegaCAM Camera
Marconi CCD44-82 1-A57
Figure 3 : The ESO CCD Testbench
Figure 13: Illumination for
computing the non-linearity
DC (e-/pixel/hour)
• 1 deg x 1 deg field of view
• Sampling 0.2’’ / pixel
• 32 CCDs mosaic 16k x 16k (CCD EEV 44-82 1-A57)
• 40 Science grade CCDs ordered
• Contract with Marconi : CCDs to be tested by ESO
• 4 auxiliary CCDs (two for guiding, two for image analysis
for the active optic control for the VST)
• Mosaic filling factor 93 %
• Operating temperature -120°C
0
1020
us
Ar
C
Ca ae a
ni lum
s
M
aj
Ca or
Ce ri
n
Ch nta a
am uru
ae s
l
Ci eon
rc
in
Do us
ra
Fo do
rn
ax
Ho G
ro rus
lo
gi
u
Hy m
dr
u
Le s
pu
Lu s
pu
M s
en
sa
Pa
vo
600
0.1
0
lia
520
0.3
Ap
440
0.4
At
370
0.5
0.2
Wavelength (nm)
Netherlands : NOVA, Kapteyn Instituut Groningen, Leiden
Germany : Universitäts-Sternwarte München, Göttingen.
Italy : Osservatorio Astronomico di Padova, Capodimonte.
ESO : European Southern Observatory.
The method consists of reading the CCD while it is illuminated with a
light source (assumed to be constant on such short time scales), see
Fig. 13.
14
100
90
80
70
60
50
40
30
20
10
0
320
OmegaCAM Consortium
Linearity
Photon Response Non-Uniformity
Rms (%)
The test bench of ESO's Optical Detector Team ( ODT ) provides a platform well suited for
characterizing CCDs. However, to make easier the test of more than 40 CCDs for
OmegaCAM (including additional CCDs for other instruments), several enhancements were
implemented in this set-up to increase its throughput. The last hardware improvements
optimized the turn-around time and the precision of individual CCD characterization. Two
new detector heads accommodating two Marconi CCDs each have been assembled and
adapted. Scripts for a largely autonomous data acquisition and reduction have been written.
The implementation of these high-level procedures permits essentially an un-supervised
reduction of a complete data set. Finally, comprehensive test reports in HTML and PDF
format enable a convenient sharing and comparative evaluation of the results.
QE (%)
Introduction
Horizontal Transfert
Vertical transfert
Figure 19 : CTE for 17 CCDs at -120°C,
Horizontal CTE : Mean= 0.999997 Rms= 0.000003
Vertical CTE : Mean=0.9999995 Rms=0.0000002
Example of an unexpected defect : Parasitic light injected by on-chip ESD protection diodes.
Conclusion
• 88 CCDs ordered
40 science grade CCDs
- 29 received, 29 characterized
16 engineering grade
- all CCDs delivered and tested
32 mechanical samples received
• Streamlined CCD test procedure
• Improved reliability
• Accuracies have been assessed
The procedure
After the delivery of a CCD, the data sheets and the CCD are checked, and the ESO
database is updated. For the ease of reference each CCD is internally assigned a name
of a stellar constellation; they are also used below. A standard test cycle is executed and
some complementary measurements are carried out to check any particular parameters.
Comprehensive test reports are generated semi-automatically. The full report includes
tables with the values measured, comparisons with the Marconi measurements and basic
statistics. Compliance with the contract is checked for each parameter.
Figure 10 : Bias 50kpix/s,
binning 15x15, color scale in
ADU, gain= 0.55e-/ADU
Figure 11 : Bias 225kpix/s,
binning 15 x 15, color scale in
ADU, gain= 0.55e-/ADU
Figure 12 : Bias 625kpix/s,
binning 15 x 15, color scale in
ADU, gain= 0.55e-/ADU
• Time to test 2 CCDs decreased from 2
weeks to 3 days
• Time to reduce and analyze data
shortened from 1 week to 2 days
• Ability to test 2 CCDs simultaneously
• 3/4 of the OmegaCam CCDs have been
tested
• Results are generally very satisfactory
[1] Amico, P., Böhn, T. (1997) : ESO’s New CCD testbench. In J. W. Beletic and P. Amico (eds) : Optical detectors for
astronomy, Astrophysics and space science library, Vol. 228, Kluwer Academic Publishers, Dordrecht, Page 95-114
[2] : http://www.eso.org/~ccavador/testbench/Prism/CCDtest-US.html