CFH12K: the new CFHT wide field CCD mosaic camera
|
Jean-Charles Cuillandre (1), Gerard Luppino (2), Barry Starr (1), Sidik Isani (1)
1: Canada-France-Hawaii Telescope Corporation
2: Institute for Astronomy, University of Hawaii
From CFHT Bulletin 40 (August 1999) with extra figures.
Table of contents
Abstract
1: Camera
a: CCDs and Focal Plane
b: Cryostat
c: Shutter
d: Filter wheel
e: Auxiliary Control Electronic
|
2: Data acquisition system
a: CCD controllers
b: Acquisition host
c: Archiving
|
3: Prime focus upgrade
a: Reducing scattered lights
b: Improving optical transmission
c: Bonnette rotation
|
4: Observing with the CFH12K
a: User's interface
b: Observing tools
Focus
Telescope offsets
Dithering patterns
Taking twilight flat-fields automatically
|
c: Filters
d: File format
|
5: Performance
a: Mosaic organization
b: Optimization and characterization in laboratory
Linearity, gain and readout noise
Full well capacity and anti-blooming
Charge transfer efficiency
Quantum efficiency, brick wall pattern and dark current
Residual image
Crosstalk
Cosmetics
|
c: Characterization on the sky
Mosaic alignment
Mosaic geometry
Image quality
Photometric performance
Photometric zero points
Sky brightness
Brick wall pattern impact
Short exposure time
|
Flat-fielding, brick wall pattern and fringes
Cosmic rays
Telescope noise
|
d:First light
|
Conclusion
Acknowledgments
Note
References
Abstract
CFH12K, the new CFHT wide field CCD camera, a 12,288 by 8,192 pixel mosaic
covering 42 by 28 square arcminutes (1/3 of a square degree) with 0.2
arcsecond per pixel, saw first light on the sky at prime focus early
January 1999.
This instrument, a collaboration between University of Hawaii and CFHT,
benefits from a number of major technological improvements.
These improvements include twelve 2k x 4k pixel back-side illuminated
CCDs providing significant increase in quantum efficiency, in particular
in the B band. In addition to improved CCDs, a new generation of data
acquisition system has been developed. It is optimized for mosaics
and reduces the readout time and data delivery to less than a minute.
To facilitate observations, a new modular software user
interface has been implemented. This system allowing the observer to interact
with the camera and the telescope through either a graphical interface or a
command line based interface. Significant features include the support of
scripting a set of tools to prepare and assist observations in order to
optimize the use of telescope time.
The CFH12K is scheduled for a very high fraction of the telescope
time. The instrument is also physically located in a relatively remote
location in the prime focus cage atop the telescope.
To facilitate efficient handling by the CFHT technical staff
and to optimize on sky reliability, steps have been taken to provide for
the remote monitoring and control of the entire system including the
prime focus environment.
In support of overall system performance, the 20 year old prime focus was
revisited to improve sensitivity and baffling. This article includes
the following sections: camera description, data acquisition system,
prime focus upgrade, how to observe with the CFH12K and finally the
performance of the camera.
1: Camera
The CFH12K camera was primarily designed and built by Gerry Luppino (Principal
Investigator of the project) from the Institute for Astronomy, University
of Hawaii, with substantial assistance of CFHT staff.
Such collaboration was not new since both parties have been working on
several CCD mosaic projects for CFHT over the past years (see Luppino et
al. 1998 for a review of CCD mosaics built around the world in the last
decade).
Barry Starr from CFHT (Project Manager/Project Engineer), handled the design
of auxiliary control electronics unit (ACE) which controls the filter wheel,
the shutter, the temperature regulation and various remote sensors.
a: CCDs and Focal Plane
The CCDs used in the CFH12K focal plane originate from
the CCD consortium led by Gerry Luppino, grouping several facilities
(ESO, UH, Keck, CFHT, Subaru, AAO) to fund the development
of devices by the MIT/Lincoln Laboratory (MIT/LL) (Burke et al.
1998). CCID20 is a thinned 4096x2048 15 micron square
pixel CCD that can be organized in a "2xn" mosaic since
is abuttable along 3 sides (figure 1). There are two types
of CCDID20 devices: those made out of high resistivity bulk
silicon (HiRo) and those made out of the standard resistivity epitaxial
silicon (Epi), and the CFH12K contains both types. HiRo devices
have a higher quantum efficiency in the red part of the spectrum
(figure 7) and a bit less fringing in the near-infrared than
epitaxial devices.
|
Figure 1 (left): The CCID20 CCD from MIT Lincoln Laboratories, a 2048 by 4096 15 micron
pixel thinned device (3.5 cm x 7 cm). Twelve of this model form the CFH12K
mosaic.
Figure 2 (right): The CFH12K focal plane: 12,288 by 8,192 pixels, over 100,000,000 pixels!
The whole surface (21 by 14 square centimeters) is flat within 40 microns.
|
The CFH12K houses twelve devices arranged in a 2x6 mosaic (figure 2),
giving a total of 12,288 by 8,192 pixels: more than 100 million pixels!
The CCDs, which are flat to within 20 microns, are mounted on a custom
package with three shims which were machined to within 2 microns.
When mounting the twelve CCDs+package on the extremely rigid focal
plane plate, the overall flatness of the mosaic was kept within 40
microns. This flatness is well within the depth of field at CFHT
prime focus: 60 microns at f/4 with a 0.2"/pixel sampling.
Flatness was the first priority to ensure the best image quality
over the whole field, no guarantee was given to a perfect alignment
between the devices along the lines and columns due to increased
complexity. However, as described in the "Performance" section, the
devices are well aligned thanks to a careful focal plane manual
assembly.
|
b: Cryostat
The focal plane is kept cold at -85 degrees Celsius using liquid
nitrogen. The 8 liters nitrogen tank keeps the CCDs cold for more than
20 hours in a "down looking configuration" (figure 4) with a
typical vacuum in the cryostat better than 5E-6 Torr.
The CFH12K has successfully been run for weeks while preserving its
vacuum integrity.
The diagram on figure 4 shows the CFH12K mounted in the prime focus
cage, the ring on top of the cryostat provides multiple attachment
points for the various cables connecting to the camera and the
prime focus bonnette (power supplies, control cables, fiber optic
cables,...).
|
c: Shutter
The shutter, a double blade system, designed and fabricated by
Gerry Luppino and refined by the CFHT mechanical group is an evolution
of the UH8K shutter. It is far more reliable and more compact. The
opening and closing blades being made of two parts sliding on top of
each other. The overall dimension is 17 by 14 inches, for a
open area of 9 by 8 inches. The whole unit is removable
from the CFH12K stack (filter wheel, cryostat) in a
matter of seconds, improving the maintenance/check of this
very critical part of the camera.
A second "back-up" shutter is under development and is scheduled
to be completed before the end of the year. Since the shutter
is such a key part of the system and historically has proven
to be one of the most unreliable, efforts have been taken to
improve performance, reliability and diagnostics.
A set of hall sensors monitored
by the ACE allows the system to constantly check the shutter's status
and report failures if the opening/closing/re-cocking
sequence fails at any point. In the case of failure, it is the
responsibility of the master CCD controller
to report the failure to the acquisition
host which takes action based on standard failure scenarios
(e.g. reset opening blade if shutter did not close at the end
of the exposure). These efforts are taken to preserve integrity of
data still on the detectors, since certain known failure modes
may allow us to preserve science data.
Since the opening blade travels across the focal plane
at an increasing speed, the closing blade has to travel
the same direction at the same speed to ensure uniform
illumination. Two independent springs pull the
titanium blades out/in the bean, hence the ballistic of the
blade is slightly nonuniform. The time delay (difference
in exposure timing across the focal plane) is approximately 50
milliseconds (gravity influences this timing depending on the telescope
position), hence for exposures of 5 seconds, the effect is only 1
part for a thousand, totally negligible for the type of photometry
precision one can hope getting from such camera (standard photometry
precision is approximately one percent). The opening and closing
times of each exposure are measured by the ACE/CCD controller and
saved in the file FITS header.
|
d: Filter wheel
The filter wheel, based on a Geneva mechanism, hosts four filters
at once. The ACE controls the wheel motion and monitor its
different states (moving, on position, ...) through hall effect
sensors. It takes less than 4 seconds to move from one position
to the nearest one, each position being secured by a pin ensuring
a positioning repeatability of approximately 10 microns: good enough
to ensure that flat-fielding patterns due to the filters (dust)
will repeat themselves in the course of the nights through various
filter changes.
See below in the "Observing with the CFH12K" section for information
on filters.
|
e: Auxiliary Control Electronic
This complex unit based around two FPGAs (Field Programmable Gate
Array) handles the shutter control, filter wheel control,
focal plane temperature regulation, and the monitoring of various
sensors distributed in the CFH12K system.
The ACE not only controls actions from these different elements,
its behavior is based around a state machine where any evolution
from one state to another is conditioned by given values for
a set of variables defining the system. In particular, this
allows effective handling of unexpected events in a normal
sequence to raise a flag.
The ACE receives simple orders from the CCD controllers through
a reduced number of lines (high and low level), these orders
being the interpretation of the commands received by the CCD
controller from the data acquisition host (e.g. "FLT 1"). In
return, the ACE sets input lines for the CCD controller that
get interpreted to identify the current state of the system.
The ACE can also be controlled manually from a well furnished
front panel easily accessible in the prime focus cage. This
allows control and check of the auxiliary functions without
the use of the whole data acquisition system.
|
2: Data acquisition system
The CFH12K data acquisition system was a complete in-house CFHT
development effort led by Sidik Isani (data acquisition software,
global software architecture) and Jean-Charles Cuillandre (observing
session & tools, CCD controllers & CCDs optimization, also camera
characterization) with system engineering support from Barry Starr.
The archiving was handled by the CFHT software group.
a: CCD controllers
A new controller generation was needed to match the performance
of the MIT/LL CCDs: two separate Generation II SDSU CCD controllers
(Leach et al. 1998) handle the CFH12K mosaic, each reading a bank
of six CCDs.
The SDSUII system was still under development at the time the CFH12K
project started. This resulted in some extra effort from the
CFHT side to setup properly two controllers, each of them
dealing with six channels running in parallel and both controllers
synchronized from a similar 50 Mhz clock. DSP code development
and system analysis eventually led to a readout time of 58 seconds
for the whole mosaic with the two controllers synchronized within
6 nanoseconds over the whole duration of the readout (essential
to avoid pickup noise from one controller to the other, a problem
that forced for instance the two UH8K controllers to be read
out sequentially, hence multiplying the readout time by a factor
of two). A 58 seconds readout time results in a data rate
of 1.8 megabytes per second per channel, both along the
fiber optic cables leading to the IfA custom serial to
parallel boards that connect to SBUS interface cards
into a single computer, a Sparc20. The system could
run faster but some CCDs require low serial transfer
rates to ensure better charge transfer efficiency.
The CCD controllers are also in charge of sequencing the
exposures (timing, readout) and also interpret the filter
wheel and shutter commands to set orders for the ACE.
In the course of an exposure, the state of the system
is checked every millisecond to see if the ACE raised
an error flag. In that case a message is sent down to
the acquisition host which reacts based on foreseen
scenarios.
|
b: Acquisition host
A Sparc20 hosts the program called "12kcom", the main
element for taking data with the CFH12K. It is a
client running inside a wrapper called "director" providing
extensive features to ease interaction between a human
being and a piece of software (commands/status/feedback)
(figure 5).
"12kcom" incorporates a set of features already extensively
used by DetI (used on KIR and EEV).
A great feature of "12kcom" is its ability to handle several
tasks in parallel in order to reduce the overhead when taking
data on the sky and hence maximizing the time spent collecting
photons. Reading data from the detectors down to memory takes
58 seconds and then the write to disk still has 40 seconds to go,
but the next exposure is actually already starting (shutter open).
With typical exposure times of 5 to 10 minutes, the CFH12K routinely
reaches an efficiency on the sky of more than 90% over the
whole night!
The observing session with "director" runs on a Sun Ultra2 equipped
with 100 gigabytes of disk space. DLT (35 gigabytes) and DDS3 (12
gigabytes) tape drives are available to the observer to save their data.
"director" can wrap several clients: they all run
in parallel with "12kcom". One handles the connection
(through an RS232 link) to a set of power supplies placed in
the prime focus cage to power the ACE, for example.
|
Figure 3: CFH12K system architecture. Blue labels indicate
the data rate for the various paths the
data taken on the sky go trough. The calibration and data processing hosts
are related to off-line
activities. Refer to text for further details on each element of this diagram.
|
|
A remote 110 volts AC power strip can also be turned on and off to
power, for instance, the SDSU switching power supplies and provides a
"hard way" to reset the system. Extreme care is taken
to protect the devices, however
not only the DSP code is setup to
always gently set on and off the CCD voltages, but an internal
power control board inside each CCD controller takes
care of switching off the voltages if the input voltages
enter a forbidden range.
|
c: Archiving
Data archiving needed an upgrade to handle the flow of
200 megabytes files, each night of observing producing
typically 20 gigabytes of data. The archiving support
is now DLT tapes with a 35 gigabytes capacity.
Fortunately, CFHT now has a DS3 link between its headquarters and the
summit, able to handle a data rate of 5.6 megabytes
per second. Images actually get to be archived as fast as
they get acquired since the writing data rate for
DLT tapes is 4.7 megabytes per second!
A duplicate of the tapes is then sent to CADC for
the official data archiving.
|
3: Prime focus upgrade
The 20 year old prime focus which had been designed for the
photographic plates (total field of view available is 1 square degree),
showed its limitation for high sensitivity cameras used for
deep imaging observations as outlined in Cuillandre et al. (1996).
Scattered lights in particular were a major concern since light
reflections from nearby bright stars were a real concern, as well
as overall scattering. The issues make the flat-fielding of the large mosaic
images a real challenge (MOCAM, UH8K). The prime focus upgrade conducted by
the CFHT optic and mechanical groups now makes this focus a very
clean and efficient light collector for the CFH12K.
|
Figure 4:The CFH12K camera in the observing configuration.
|
a: Reducing scattered lights
Light scattering reduction was achieved using simple techniques
such as implementing a special black velvet in the entry cone
and on all the metallic internal surfaces in the prime focus
bonnette and the wide field corrector (see figure 4).
|
b: Improving optical transmission
The central lens of the wide field corrector (made of 3 lenses)
was removed and sent to the Dominion Astrophysical Observatory
(Canada) to get a special anti-reflection coating applied on
its surfaces, providing an improvement of 8% of the
transmission over the whole optical spectrum. The total
efficiency of the primary mirror and the wide field corrector
is now 73%.
|
c: Bonnette rotation
Bonnette rotation is disabled for the CFH12K: the camera occupies
so much room in the cage that only +/- 20 degrees rotation is
supported (XY axis aligned with the WE/SN axis). Analysis was conducted
by the TCSIV group to efficiently map the guide star area
to improve pointing efficiency and be able to obtain a guide
star right at the first pointing (the guide camera pick-up
mirror catches a part of the beam close from the edge of
field to avoid vignetting the CFH12K field, hence it is suffering
from the wide field corrector non linear distortion).
|
4: Observing with the CFH12K
Piloting the CFH12K is straightforward since the instrument complexity
is totally hidden to the user. But observing, even for such direct
imaging camera, can be specially tricky at a time the goal is to
be the most efficient on the sky. For that purpose, a versatile
user's interface was developed as well as tools to prepare and
assist the observations. Refer to the "CFH12K Observer User's Guide"
document for further information about observing with the CFH12K.
|
Figure 5: the director window: a command line based interface. The
main control and status
window of the CFH12K session.
|
a: User's interface
The CFH12K can be controlled either from a Graphical User Interface
(GUI) (figure 6), a Command Line Interface (CLI) (figure 5) or
from scripts (any shell).
The central element of the CFH12K session is the director window which
acts both as a CLI and a status window. The main GUI window serves
as both the camera status display and the control interface. This
GUI reproduces only a reduced set of information,
elements such as progress bars and detector
temperature are indeed only appearing in the director window.
Here is the elementary set of commands one gets to use when
observing with the CFH12K, all of them are accessible from a GUI,
at the CLI or can be combined in a script (on top of these,
commands such as "abort", "stop", "break", "stopscript" can
be entered when exposures are running).
------------------------------------------------------------------------------
go n -> take "n" exposures ("go" for a single exposure)
snap -> take a quicklook exposure with current parameters
flux -> exposure (small raster) to measure sky background
etype t -> set exposure type (o=object, f=flat, d=dark, b=bias)
etime n -> set exposure time to "n" seconds
raster full -> set raster readout mode (full, full bin2, full bin4)
filter 0 B -> select filter 0 (here with B filter at that position)
fits observer CFHobs -> set the FITS OBSERVER header for the next exposures
fits object NGC3486 B -> set the FITS OBJECT header for the next exposures
fits comment skyphoto -> set the FITS COMMENT header for the next exposures
offset a b -> offset the telescope by a" East and b" North
pfocus Z -> set the Z of the bonnette (-1.0 < Z < +6.0)
------------------------------------------------------------------------------
|
b: Observing tools
The following tools are available either from the GUI, the CLI
or from scripts:
Focus
Take automatically a sequence of several exposures
at different Z values on the same frame and offsetting
the telescope between each position (only one readout,
see figure 11).
The image can then be quickly analyzed with standard tools
to find out the best position.
|
Telescope offsets
Small telescope offsets (< 400 arcseconds, depends
on the initial position of the guide star in the
guiding field though).
|
Dithering patterns
A set of dithering patterns are proposed from 2 to 8
positions. The patterns are set so that none exposure
has a similar X and Y coordinates value to allow
removing detectors artifacts during data reduction.
With such patterns, the CFH12K can run for hours without
any human intervention. The observer can also build
his own set of scripts if he isn't satisfied with the
ones made available.
|
Taking twilight flat-fields automatically
Twilight flat-fields appear to be extremely efficient to
flatten CFH12K scientific data both at large scales and small
scales, hence it is important to acquire the maximum
number of such frames.
With the limited time to take flat-fields at twilights,
an automatic tool has been developed to set up properly the
exposure time from one exposure to another to keep
the flux constant on the images while the sky brightness
is changing rapidly. The input from the user is the initial
exposure time (defined with the use of a "check flux" exposure),
the number of exposures to acquire, the day period
and the current filter.
|
|
Figure 6: the graphical interface window (GUI). The various options
are described
in the text.
|
|
c: Filters
Here are the available filters available for the CFH12K. Transmission
curves are displayed on figure 7. An automatic
focus adjustment will be provided soon, as well as a focus
adjustment between exposures based on dome temperature changes.
CW = Central Wavelength (nm)
CN = Cut on wavelength at 50% (nm)
CF = Cut off wavelength at 50% (nm)
T = Transmission (%)
BW = Bandwidth (nm)
Broad band filters: Narrow band filters:
----------------------------- -------------------------
Name CW CN CF T BW Name CW T BW
----------------------------- -------------------------
B 431 380 475 85 95 H-alpha OFF 645 92 9
V 537 501 595 90 94 H-alpha 658 95 7
R 658 590 720 85 130 TiO 777 90 18
I 822 725 930 80 205 CN 812 XX 180
Z XXX 850 XXX 95 Open
----------------------------- -------------------------
|
d: File format
Two file formats are proposed to the users at the moment:
the Multi-Extension FITS format (MEF) where all 12
CCD images get stacked into a single FITS file, and the
SPLIT format where the 12 CCD images are saved in
individual standard FITS files in a dedicated subdirectory.
For both modes, an independent binned by 8, overscan
and relative gain corrected image is generated to allow
quick and efficient display of the whole field of view.
This quicklook (automatic display) allows the observer
to check in one eye blink if something is wrong in the
image (guide probe vignetting, scattered lights,...).
|
|
Figure 7: Quantum efficiency of high resistivity bulk silicon CCDs
(dashed line) and
standard resistivity epitaxial silicon CCDs (solid line). Optical transmission of broad
band filters available for CFH12K observations. The cut-off of the Z filter is set
by the fall of the CCD
quantum efficiency.
Narrow band filters are pointed on the lower axis.
|
5: Performance
The CFH12K was on the telescope for five dark periods out of the first
seven months of the year 1999. This is by any measure high use, especially
for a new instrument. Prior the first light on the sky in
January, an in depth characterization of the camera had taken
place in CFHT's CCD laboratory in Waimea. The amount of information
collected is enormous and shortly reported here. Since an upgrade
of the focal plane will take place in August 1999 to replace two defective
CCDs (early draft engineering grade arrays), it is pointless to
provide precise information on the current mosaic, hence only
general values valid for the whole mosaic are given. When the
revised focal plane will see its first light in September 1999, a complete
report will be posted on the CFH12K web page. Still, apart of a better
cosmetic for the new CCDs, the general performance will be very
similar as mentioned hereafter.
|
Figure 8:Mosaic geometric configuration and naming convention.
|
a: Mosaic organization
Figure 8 provides a complete mapping of the CCDs within the mosaic.
|
b: Optimization and characterization in laboratory
The CCD consortium provided a set of characteristic precise
enough to assert the quality of each device (CTE, cosmetic, QE),
but it was in no means supposed to deliver optimized running
parameters. This optimization work had to be conducted for
the CFH12K in the particular case of a direct imaging application.
Priority was given to linearity over the whole
range of the converters and charge transfer efficiency (CTE),
then quantum efficiency, then the full well capacity, then readout
noise (usually quickly dominated by the sky background photon noise,
in particular with broad-band filters), and finally dark current.
All CCDs were tested together within the CFH12K cryostat. A special
optical bench was set up to accommodate the illumination of the
very large focal plane (21x14 square centimeters) and special
automatic softwares developed to handle multiple detectors at once in
order to be the most efficient and parallelize the optimization.
Linearity, gain and readout noise
Linearity was measured and optimized using the photon
transfer curve procedure allowing also a measure
of the conversion gain and the readout noise with a
reasonable accuracy.
The CCDs voltages were delicately
tuned to get a linearity better than +/- 0.5% over
the whole range of the 16 bits analog-to-digital
converters, which for the gain of 1.5 electron
per ADU gives a digital saturation at 100,000 electrons.
The gain is set so that readout noise gets sampled on
at least 2 to 3 ADUs: for most of the devices the noise
is close to 4 electrons (at 150 kilopixels per second).
|
Full well capacity and anti-blooming
Dithering anti-blooming was investigated in depth but was
abandoned for an on-sky use due to the large generation of
pocket pumping sites (see Tonry & Burke 1998 about pocket
pumping).
Fortunately the very high full well capacity of the
CCID20 devices, typically around 400,000 electrons
(Wei & Stover 1998), limits efficiently blooming from bright
stars to a reduced number of pixels.
|
Charge transfer efficiency
CTE was measured on all CCDs using a Fe55 source illuminating
the devices through a custom beryllium window.
Particular care was given to tune the serial register
transfer speed to ensure at least a 0.99999 CTE, which
is good enough for our direct imaging sky background limited,
application with a 0.2 arcsecond/pixel sampling and a typical
image quality (seeing) of 0.7 arcsecond. Parallel CTE is at
least 0.999999.
|
|
Figure 9:The two types of brick wall pattern found on
six of the CFH12K CCDs (easily
corrected with proper flat-fielding). Peak to peak is about 10%.
|
|
Quantum efficiency, brick wall pattern and dark current
The quantum efficiency (QE) is an intrinsic property of each
device depending
on the building process (thinning, coatings), however, the
sensitivity in the red part of the spectrum can be increased
(up to 20%!) by running the CCDs warmer than usual, say
-90 C versus -120 C. Running the CCDs warmer also presents
the great advantage of reducing the amplitude of the
brick wall pattern (there are actually two types of BWP
as shown on figure 9), a feature from early type
CCID20 devices that has been corrected in the later versions
(only half of the CFH12K CCDs suffer from this effect).
This pattern is corrected with proper flat-fielding but since
the peak to valley amplitude can be up to 20% in the B band
(less important at longer wavelengths, i.e, R band), this means
a substantial change in sensitivity across the devices
(physical scale is approximately 150 pixels). Reducing it
as much as possible is a major concern and the CFH12K now runs
at -85 C, high enough to maximize the QE and minimize the
amplitude of the brick wall pattern, even to the cost of a
potential dark current requiring an image processing correction.
The dark current appears to be very low even at -85 C and
can be neglected in broad-band filters applications. In the
other case, since the dark current appears uniform over the
whole devices, hence a simple constant subtraction takes care of
removing this additive signal.
|
Residual image
Residual image is a major concern for the CCID20 devices
(Wei & Stover 1998) specially when observing with narrow
band filters (figure 10). The signal generated from the release
of charges trapped in the Si-SiO2 interface is a function
of time and varies from one device to another. In sky
background limited observations, this extra signal gets
lost within the photon noise though. Special cleaning
schemes have been investigated to eliminate as many
trapped electrons as possible in the short amount
of time available between two exposures. Dithering
anti-blooming mode (with shutter closed) was actually found
to be an efficient way to reduce the residual image effect
by massively bringing up holes from the substrate up
to the Si-SiO2 interface where they recombine with
trapped electrons.
|
|
Figure 10 (left):Residual image on a dark exposure: residuals from two
saturated stars on the previous exposure. With broad band filters, this
signal is quickly dominated by the sky photon noise.
Figure 11 (right):An example of a "focus plate": several images
of the
same stars at different "Z" focus values appear in a sequence (double
offset for the last exposure for easier identification).
|
|
Crosstalk
Crosstalk between the video channels is negligible (8E-5)
within a common video board (hosting 2 channels) and
not detectable from one board to another.
|
Cosmetics
Several CCDs from the mosaic can be qualified as "perfect"
with only a fraction of a bad column. These CCDs appear
also to be the ones with top QE, low BWP and low noise.
They are positioned at the center of the mosaic.
|
|
c: Characterization on the sky
Mosaic alignment
The long axis of the mosaic can be precisely aligned along the
East-West axis to within 0.1 degree by letting the telescope
drifting for a few minutes to let stars cross the whole
mosaic (iterative process once at the beginning of an
observing period). See figure 12. A mechanical pin should be
installed soon to facilitate the bonnette alignment.
|
|
Figure 12 (left):The central 8kx8k field of a 3 minutes "drift"
exposure.
Figure 13 (right):Typical cosmic ray hits. Image size is 250x250 pixels.
These hits cover only 5 to 7 pixels.
|
|
Mosaic geometry
Astrometric measurements were conducted on the current mosaic,
the largest tilt angle between two CCDs within the whole mosaic
is 0.4 degree. The gap between the CCDs both along the X and
Y axis is close to 6 arcseconds (#30 pixels or 450 microns).
|
Image quality
We were blessed with a few nights of 0.5 arcsecond seeing
during the first light (spatial sampling is 0.2 arcsecond
per pixel) and were able to determine immediately
that image quality is uniform over the whole field, with only
a 0.05 arcsecond degradation from the center to the
edge, most certainly the result of an optical aberration
rather than a tilt of the focal plane since the same effect
is present in all directions. Since then, 0.45 arcsecond
seeing images have been obtained several times (R, I and Z
bands)!
|
Photometric performance
Photometric zero points
The CFH12K is an extremely sensitive camera, the zero points
surpass all the previous detectors ever mounted at any
CFHT foci (except for RCA4 in the B band back in the
late 80s, but it was "only" a 320 by 512 pixel CCD).
This is the result of the wide field corrector upgrade
(coating) with the combination of higher efficiency
filters (interference filters) and thinned CCDs.
The parameters for the photometric equation are:
magnitude = -2.5log[counts(e-/sec)] - a*X + b*Col + Co
with "a" the airmass term, "b" the color term (%) and
"Co" the photometric zero point at zero airmass (e-/sec).
--------------------------------------
Filter a b Co
--------------------------------------
B 0.17 +5.7% 26.07
--------------------------------------
V 0.10 +0.5% 26.60
--------------------------------------
R 0.07 +4.0% 26.57
--------------------------------------
I 0.05 -0.1% 26.15
--------------------------------------
(to be compared to the MOCAM or UH8K
zero points in V:25.4 and I:25.4)
|
Sky brightness
Sky brightness (in magnitude per square arcsecond)
was measured at zenith in photometric dark time.
----------------------------
B V R I
----------------------------
22.5 21.6 21.2 19.8
----------------------------
|
Brick wall pattern impact
There were some concerns about photometric accuracy for
objects on peaks versus valleys of the brick wall pattern.
This was carefully tested by moving a set of objects across
the detector and checking for color evolution (the data
having been flat-fielded). No color effect could be detected
at a level of a fraction of a percent (i.e. the QE curve shape
in the peaks is similar to the one in the valleys). See
figure 9 where the two types of brick wall patterns encountered
on some CCDs of the CFH12K are shown.
Flat-fielding corrects properly the brick wall pattern.
|
Short exposure time
As mentioned in the "Shutter" section, exposure times
less than 5 seconds are not recommended if a contribution
from the shutter ballistic in the photometric error budget
of no more than one part over a thousand is required.
|
|
|
Figure 14:Fringing. Image of a full CCD (2k x 4k) fringe frame.
Peak to
peak values change from device to device but is typically 10% to 15%.
|
|
Flat-fielding, brick wall pattern and fringes
With the prime focus optical upgrade, flat-fielding became
much easier than for UH8K thanks to the overall reduction
of scattered lights. Twilight flat-fields appear to work
very well and provide easily a flattening close to
the percent over the whole field. Building superflats
is also much easier. Dome flat-fields work poorly due
to the mismatch of the lamp spectrum and the dome paint
with the night sky spectrum. CFHT will soon evaluate the
use of a special screen and more appropriate lamps.
Fringes appear in the frames when observing in the I
and the Z bands. One needs to acquire twilight
flat-fields in the given filter to flatten the
scientific images from which a fringe frame can
be extracted. This fringe frame needs then to
be properly scaled for each exposure and then
subtracted. See figure 14.
|
Cosmic rays
About five hits per minute per square centimeter
(around 100 events per minute per CCD). Cosmic
rays hits are much more punctual on thinned
CCDs compared to thick CCDs and are much easier
to identify (figure 13).
|
Telescope noise
To optimize the telescope time, some operations
on the telescope can be conducted while the
camera is reading out:
- pointing to a new field
- setting a new Z for the bonnette
- rotating the dome
What must not be done (pickup noise would result):
- activating the prime focus guide probe
|
|
d: First light
The first light took place on the 9th of January. The first
field observed was the Horsehead nebulae. In the course of
the following nights, a huge amount of calibration data was
obtained along with scientific data and other objects for
the promotion of this new CFHT instrument. The picture on
figure 15 is Messier 81, observed in B, V
and R. The data have been reduced with the FLIPS package.
Twilight flat-fields were used to process the images (four
dithered 5 mn exposures in each filter). Other images will
be presented soon in the CFHT image galleries.
|
|
Figure 15:Messier 81: the central 8k x 8k area of a composite color
image from multiple bands CFH12K exposures. 20 minutes in B,V
and R using a dithering pattern to "erase" the gaps between CCDs and
bad cosmetics. Diffuse structures at the bottom are residuals from an
interaction with the nearby galaxy NGC 3077.
|
Conclusion
The CFHT community now has a new visible wide field camera available at
prime focus. The CFH12K covers one third of a square degree with high
sensitivity CCDs providing a spectral range access from the B band up to
the near infrared. The 12K by 8K pixel mosaic provides
excellent image quality over the whole field with a 0.2 arcsecond
per pixel sampling. Images down to 0.5 arcsecond are now common beyond 600 nm.
Readout time is less than a minute, resulting in very high observing
efficiency (up to 90%). The observer can interact with the camera and the
telescope from a graphical interface or a command line interface, scripting
is also fully supported. A set of observing tools, such as automatic
twilight flat-field sequencer and a variety of dithering patterns, is
available from any of the user's interfaces. Broad-band filters (B, V, R,
I, Z) and narrow-band filters (Halpha, HalphaOFF, TiO, CN) are available.
The upgrade of the prime focus greatly improved flat-fielding efficiency
by reducing the amount of scattered lights. Photometric performances
are impressive: never CFHT has had such a sensitive instrument at one
of its foci. Remote system control, reset and diagnostic functions
serve to provide the most reliable system possible.
In only five observing runs, the CFH12K has covered tens of square
degrees on the sky and stacked more than than one terabyte of data.
The CFH12K is scheduled for more than 70 nights for semester 99II,
40% of that period, and the user's community pressure on this
instrument still keeps building.
Likely developments closely related to the camera include
a dome flat-fielding facility and a dedicated data processing
pipeline to support the user's community by providing optimal calibration
frames. In the near future, queue scheduling mode will be implemented
for the first time at CFHT for use with this instrument.
Acknowledgments
The authors, who have worked on this project for more than 2 years,
wish to greatly thank the CFHT staff for its giant effort
during the Christmas and New Year's day period to allow the CFH12K getting
its first light on the sky early January 1999, and its contribution
to make this instrument at CFHT prime focus this outstanding scientific tool
now available to the entire CFH community.
We wish to thank John Tonry from IfA for his constructive discussions,
criticisms and advice provided all along the instrument integration.
Note
Visit the CFH12K web pages at: http://www.cfht.hawaii.edu/Instruments/Imaging/CFH12K/
Comments and requests are welcome, you can contact the author at:
jcc@cfht.hawaii.edu
References
-
Burke, B., Gregory, J., Mountain, R., Kosicki, B., Savoye, D., Daniels, P., Dolat, V., Loomis, A., Young, D., Luppino, G., Tonry, J., 1998, Proceedings of ESO conference on Optical Detectors for Astronomy, 19
-
Cuillandre, J.-C., Mellier, Y., Dupin, J.-P., Tilloles, P., Murowinski, Crampton, D., Wooff, R., Luppino, G., 1996, PASP, 108, 1120
-
Leach, R., Beale, F., Eriksen, J., 1998, SPIE, 3355, 512
-
Luppino, G., Tonry, J., Stubbs, C, 1998, SPIE, 3355, 469
-
Tonry, J., Burke, B., Schechter, P., 1997, PASP, 109, 1154
-
Wei, M., Stover, R, 1998, SPIE, 3355, 598
|
|