The WASP Project and the SuperWASP Cameras
ABSTRACT The SuperWASP Cameras are wide-field imaging systems sited at the Observatorio del Roque de los Muchachos on the island of La Palma in the Canary Islands, and the Sutherland Station of the South African Astronomical Observatory. Each instrument has a field of view of some ~482 square degrees with an angular scale of 13.7 arcsec per pixel, and is capable of delivering photometry with accuracy better than 1% for objects having V ~ 7.0 - 11.5. Lower quality data for objects brighter than V ~15.0 are stored in the project archive. The systems, while designed to monitor fields with high cadence, are capable of surveying the entire visible sky every 40 minutes. Depending on the observational strategy, the data rate can be up to 100GB per night. We have produced a robust, largely automatic reduction pipeline and advanced archive which are used to serve the data products to the consortium members. The main science aim of these systems is to search for bright transiting exo-planets systems suitable for spectroscopic followup observations. The first 6 month season of SuperWASP-North observations produced lightcurves of ~6.7 million objects with 12.9 billion data points. Comment: 42 pages, 2 plates, 5 figures PASP in press
arXiv:astro-ph/0608454v1 22 Aug 2006
The WASP Project and the SuperWASP Cameras
D.L. Pollacco7, I. Skillen3, A. Collier Cameron8, D.J. Christian7, C. Hellier4, J. Irwin1,
T.A. Lister8,4, R.A. Street7, R.G West5, D. Anderson4, W.I. Clarkson6, H. Deeg2, B.
Enoch6, A. Evans4, A. Fitzsimmons7, C.A. Haswell6, S. Hodgkin1, K. Horne8, S.R.
Kane8, F.P. Keenan7, P.F.L. Maxted4, A.J. Norton6, J. Osborne5, N.R.Parley6, R.S.I.
Ryans7, B. Smalley4, P.J. Wheatley5,9D.M. Wilson4
1The Wide Field Survey Unit, Institute of Astronomy, Madingley Road, Cambridge, CB3
2Instituto de Astrof´isica de Canarias, C/V´ia L´ actea, s/n, E-38200 La Laguna, Tenerife,
3Isaac Newton Group of Telescopes, Apartado de Correos 321, E-38700 Santa Cruz de
La Palma, Tenerife, Spain
4Astrophysics Group, Keele University, Keele, Staffordshire, ST5 5BG, UK
5Department of Physics and Astronomy, University of Leicester, Leicester, LE1 7RH, UK
6Department of Physics and Astronomy, The Open University, Walton Hall, Milton
Keynes, MK7 6AA, UK
7Department of Physics and Astronomy, Queen’s University of Belfast, University Road,
Belfast BT7 1NN, UK
8School of Physics and Astronomy, University of St Andrews, North Haugh, St Andrews,
KY16 9SS, UK
9Department of Physics, University of Warwick, Coventry CV4 7AL, UK
– 2 –
The SuperWASP Cameras are wide-field imaging systems sited at the Obser-
vatorio del Roque de los Muchachos on the island of La Palma in the Canary
Islands, and the Sutherland Station of the South African Astronomical Observa-
tory. Each instrument has a field of view of some 482 square degrees with an
angular scale of 13.7 arcsec per pixel, and is capable of delivering photometry
with accuracy better than 1% for objects having V ∼ 7.0 − 11.5. Lower quality
data for objects brighter than V ∼ 15.0 are stored in the project archive. The
systems, while designed to monitor fields with high cadence, are capable of sur-
veying the entire visible sky every 40 minutes. Depending on the observational
strategy, the data rate can be up to 100GB per night. We have produced a ro-
bust, largely automatic reduction pipeline and advanced archive which are used
to serve the data products to the consortium members. The main science aim of
these systems is to search for bright transiting exo-planets systems suitable for
spectroscopic followup observations. The first 6 month season of SuperWASP-
North observations produced lightcurves of ∼6.7 million objects with 12.9 billion
Subject headings: instrumentation: photometers — techniques: photometric —
(stars:) planetary systems
– 3 –
In recent years, interest has grown in relatively small aperture and inexpensive
wide-field imaging systems, essentially composed of large CCDs mounted directly to
high-quality wide-angle camera optics. The first prominent success of such an instrument
was the spectacular discovery of the neutral sodium tail of comet Hale-Bopp (Cremonese
et al. 1997), with a temporary purpose-built camera system (CoCam). Since then, similar
cameras have resulted in imaging of a gamma-ray burst during the burst period (Akerlof
et al. 1999), and the first detection of the transits of an extra-solar planet in front of its
parent star, HD209458 (Charbonneau et al. 2000). Such instruments are ideal for projects
requiring photometry of bright but rare objects (see Pinfield et al. 2005).
1.1. Extra-solar planetary transits
The first extra-solar (exo-) planets were discovered in 1992 by pulsar timing
experiments (Wolszczan & Frail 1992). Whilst this technique is sensitive to the detection
of terrestrial-sized planets, its limited applicability has restricted its use to just a few
objects. Mayor & Queloz (1995) discovered the first exo-planet, 51Peg, from optical radial
velocity studies, and since that time the field has been dominated by this technique. One
of the surprises of these surveys (e.g. Marcy et al. 2005) is the existence of a significant
population of solar-type stars accompanied by relatively rapidly orbiting Jupiter-sized
planets. However, spectral measurements alone cannot determine unambiguously the true
mass or radius of the planet as the orbital inclination is unknown. These surveys have
discovered Jupiter-sized objects in orbits out to 3AU around 6% of the nearby Sun-like
stars surveyed. Of these, some 30% are Hot Jupiters situated in ∼4-day 0.05AU orbits,
where the equilibrium temperature is ∼ 1500K. About 10% of Hot Jupiters in randomly
inclined orbits will transit their host star. Therefore, in random Galactic fields, roughly 1
– 4 –
in every 1000 solar-type stars should exhibit transits lasting roughly 2 hours with a period
of a few days.
Given that the radius of Jupiter is ≃ 0.1R⊙, these transits should result in a dimming of
the parent star by ≃ 0.01 mag. The first transits were discovered in late 1999 (Charbonneau
et al. 2000). The V = 7.7 star HD209458 is dimmed by 0.016mag for 2 hours every 3.5
days, hence both proving the existence of the planet detected by a radial velocity search,
and resulting in a precise measurement of its radius, mass and bulk density. Because of
the multiplexing advantage of imaging, this technique promises to be the fastest way of
detecting exo-planets, and could over the next few years dictate which candidates are
followed up by radial velocity studies (rather than vice-versa as at present).
Initially, groups trying to find transits of exo-planets reported disappointing results.
For example the Vulcan Project (Jenkins et al. 2002) searched some 6000 stars, finding
only 7 transit like variables. Followup observations of these showed them all to be stellar in
origin. More recent surveys have have had more success with published transit detections
reported by the OGLE survey (Udalski et al., 2002a/b), TrES-1 (Alonso et al.2004),
HD189733 (Bouchy et al. 2005) and most recently, XO-1 (McCullough et al.2006).
Part of the reason for the apparent lack of transits stems from the difficulty in obtaining
photometry of sufficient numbers of solar and late type main sequence stars. To increase
the numbers of stars sampled wide-field surveys have often concentrated on low galactic
latitude fields. However, while the number of observable stars is undoubtedly increased,
we suspect the stellar population in such surveys is dominated by more distant K giants.
Brown (2003) showed that the number of bona fide exo-planet transits (as opposed to
stellar impostors) is consistent with expected numbers of binary and multiple stellar and
Horne (2003) lists some 23 photometric transit projects either in operation or under
– 5 –
construction at that time. While many of these are pencil-beam surveys from traditional
telescopes, a number, no doubt encouraged by the relative cheapness of the equipment, are
employing novel wide field cameras.
1.2. The WASP Consortium
The Wide Angle Search for Planets (WASP) Consortium was established in 2000 by
a group of primarily UK-based astronomers with common scientific interests. In order to
reduce the development cycle time and be on sky rapidly, we use commercially available
hardware and hence limit development work as much as possible. Our ethos is therefore
quite distinct from other apparently similar projects e.g. the HAT Project (Bakos et al.
The WASP Consortium’s first venture was the production of the WASP0 camera. Our
experience with these types of systems stems from the CoCAM series of cameras at the
Isaac Newton Group of Telescopes from 1996 – 1998, one of which was responsible for the
discovery of the so-called Sodium Tail in Comet Hale-Bopp 1995 (Cremonese et al. 1997).
WASP0 is composed completely of commercial parts, and utilizes a Nikon 200mm, f2.8
telephoto lens coupled to an Apogee AP10 CCD detector. WASP0 was used in 2000 and
2001 in La Palma and Kryoneri (Greece) respectively, and has been shown to easily detect
the extrasolar transit of HD209458b, amongst other variables (Kane et al., 2004, 2005a/b).
On the strength of the WASP0 success the Consortium was able to raise sufficient
funding for a more ambitious project – the multi-detector SuperWASP cameras. The limited
development required is reflected in the aggressive project timescale: for the La Palma
instrument funding was approved in March 2002 and first light achieved in November 2003,
while for the South African Astronomical Observatory (hereafter SAAO) sited instrument
– 6 –
funding was secured in April 2004 and first light in December 2005.
In this paper, we describe the SuperWASP facilities at the Observatorio del Roque
de los Muchachos on La Palma (SuperWASP-N) and the recently commissioned system at
the Sutherland Station of the SAAO (SuperWASP-S). Along with the hardware and data
acquisition system, we outline the SuperWASP reduction pipeline and archiving system for
the data products.
2.The Hardware System
In outline each SuperWASP instrument consists of an equatorial mount on which up
to eight wide-field cameras can be deployed. Each is housed within a two-roomed enclosure
which incorporates a roll-off roof design. Fig. 1 and 2 shows the enclosure at SuperWASP-N
and detector system at SuperWASP-S. For both facilities all observatory functions are
under computer control, including data taking.
2.1. The Robotic Mount
Both systems employ a traditional equatorial fork mount constructed by Optical
Mechanics Inc. (Iowa, USA; formerly Torus Engineering). The mount is manufactured
within their Nighthawk Telescope range. When properly configured, the mounts give a
pointing accuracy of 30 arc seconds rms over the whole sky, and a tracking accuracy of
better than 0.01 arc seconds per second. The mounts are easily capable of slewing at a rate
of 10 degrees per second. On site the mounts are attached to a concrete pier.
For our project we do not have a conventional optical tube assembly but instead we
employ a cradle structure to hold the individual cameras. The cradle allows limited camera
– 7 –
movement in 3 dimensions for balance and alignment purposes.
2.2. The Enclosure
The rapid slewing of the mount and large field of view make a traditional dome
impractical and inefficient, hence we have a custom roll-off roof structure. In most
deployments of this design the roof is moved on to rails overhanging the building, however,
in our design the space under the rails is utilized as a fully temperature-controlled
control and computer room, with the rails integrated into the roof. The building itself
was constructed by Glendall-Rainford Products (Cornwall, UK), and is manufactured in
laminated fiber-glass strengthened with wood, making the structure extremely rigid. The
likely absence of a crane during the building erection meant that the size of the roof panels
was optimized to be liftable by 3 people. The moving roof is controlled by a hydraulically
operated ram and associated electrics. In the case of SuperWASP-N, to fully retract the
roof takes ∼19 seconds, and ∼54seconds to fully close. The modular design of the building
meant the enclosure could be prefabricated by the manufacturer and then re-erected on site.
With the enclosure roof fully retracted, objects with declinations −20 < δ < 55 degrees
are visible for the entire period when their altitude is > 30 degrees. For δ > 55 degrees the
movable roof may obstruct visibility at some hour angles. For the SuperWASP-S instrument
we designed a longer enclosure to give better southern access.
2.3. The CCDs
The SuperWASP CCD cameras were manufactured by Andor Technology (Belfast, UK)
and marketed under the product code DW436. The CCDs themselves are manufactured
by e2v and consist of 2048 × 2048 pixels each of 13.5µm in size. These devices are back
– 8 –
illuminated with a peak quantum efficiency of >90%. Andor use a five stage thermoelectric
cooler to reach an operating temperature of -75C. At this temperature the dark current is
∼11 e/pix/h – comparable to cryogenically cooled devices. As our exposure times are only
30 seconds we do not require this level of performance, and hence we cool the devices to
-50C at which the dark current is a ∼72 e/pix/h.
Andor Technology also provide a 32-bit PCI Controller card that is used to control
all CCD functions. These cards (one per detector) allow the devices to be read out at
mega-pixel rates so that even after all overheads (e.g. header collection, disk write etc),
a new image can be initiated within ∼5 seconds of the commencement of readout of the
previous image. Even at this speed the 16-bit images have good noise characteristics (gain
∼2, read out noise ∼8-10 electrons and linearity better than 1% for the whole of the
dynamic range). To simplify operations, we have not tuned detectors to controllers in any
The original design of SuperWASP-N conceived of using a renovated existing enclosure
with instrument control occurring from a nearby building. Hence, our shielded data cables
are 15m in length. Exhaustive testing showed that at this length data collection was
reliable and mains pickup rarely seen. The detector power supplies are stored next to the
2.4. The Telephoto Lenses
In common with other similar projects, the SuperWASP cameras use Canon 200mm,
f/1.8 telephoto lenses. These lenses are amongst the fastest commercially available and have
excellent apochromatic qualities. Funding constraints dictated that we initially purchased
5 lenses from a local supplier before this format became obsolete. We subsequently used
– 9 –
www.ebay.com to track down the remaining units needed for both instruments. With the
above detectors they give a field size of ∼ 64 square degrees and an angular scale of 13.7
arcsec per pixel. In the first year of operations for SuperWASP-N our observations were
unfiltered (white light) with the spectral transmission effectively defined by the optics,
detectors and atmosphere. Subsequently we have deployed broad band filters at both
facilities which define a passband from 400 – 700nm (see Fig. 3).
2.5. The Data Acquisition Computing Cluster
The easiest method to accommodate the high data rate from 8 cameras is via a
distributed data acquisition cluster, with each detector controlled by a dedicated DAS
(Data Acquisition System) PC with local storage disks. Data taking itself is initiated by
a central machine called the TCS (Telescope Control System), which also controls more
general observatory functions such as pointing the mount and roof control. The TCS
machine also has serial interfaces to a time service (supplied from a GPS receiver) and
weather station. The DAS machines synchronize time with the TCS through the Network
Time Protocol daemon. Overall the relative time on the cluster is accurate to <0.1 second,
while the GPS system ensures that the absolute time is accurate to better than 1 second.
As the operation of the TCS is vital to the running of the instrument, a heart-beat system
continually monitors the machine with any break in communication initiating a close down
of the enclosure.
During the night data are stored locally on each DAS machine. At the end of observing
the data are compressed and moved to a RAID system ready to be copied to tape (LTO2)
for transportation back to the UK (recently SuperWASP-N has begun sending data back
via the Internet).
– 10 –
The weather station is provided by Vaisala (foreground in Fig. 1) and has sensors for
internal/external temperature and humidity, wind direction and strength, precipitation and
pressure. A cloud sensor (IR activated) is also utilized.
3. Data Acquisition Software
High-level software control of the entire SuperWASP system (robotic mount, CCD
cameras and roll-off roof) is provided by a modified version of the commercial Linux
software Talon (now Open Source), produced by Optical Mechanics Inc. (hereafter OMI)
for use with the Torus mount. Extensions to the software include support for multiple
CCD cameras (developed by OMI) and some in-house modifications to add a command-line
interface to supplement the standard graphical interface.
Talon supports two modes of operation: one for manual control with an observer
present, using the graphical interface (or the new command-line interface), and the other
for automatic observing, where a dynamic scheduler takes control of the telescope and
performs observations from a pre-defined queue.
In the first season of operation, the observer was responsible for taking bias and dark
frames, opening and closing the dome, and taking twilight flat fields, automated using
a driver script for the command-line interface. Science observations were taken using
the telrun daemon within Talon, driven from a Perl script. A new dynamic scheduler,
waspsched, developed in-house by the Consortium, and using the command-line interface,
has recently been commissioned. This has increased observational efficiency by allowing
continuous operation of the equipment (previously a number of delays were required during
observing to synchronize the interactions with the existing Talon scheduler, allowing about
one 30 second integration per minute despite the 10 degree per second slew speed of the
– 11 –
mount). The dynamic scheduler also allows us to intersperse all-sky survey fields with the
exo-planet fields. Support for alternative observing modes may also be added, in particular
the ability to interrupt the scheduled observing and follow up transient events (e.g. gamma
ray bursts) without user interaction.
Data from the site weather station is fed into the software. Talon has a number
of configurable conditions on which a weather alert is issued, for example, high wind or
excessive humidity. On triggering a weather alert, the telescope is immediately slewed
to a predefined park position (to avoid collisions with the roof), and the roof is then
closed. After the alert condition finishes, the software waits for a short period (typically 20
minutes), and then opens the roof and continues observing if it is still dark. Other alerts are
generated by the cloud and lightning sensors. In the event of failure to operate the roll-off
roof, an alert condition is reached sending a radio signal to a receiver in a neighboring
operator attended telescope dome.
As Andor Technology provides a Linux-based software development kit all aspects of
the CCD control and data collection are integrated within Talon.
4. The Reduction Pipeline
The SuperWASP data analysis pipeline employs the same general techniques described
by Kane et al. (2004) for the prototype WASP0 project. We use the USNO-B1.0 catalog
(Monet et al. 2003) as the photometric input catalog. We carry out aperture photometry
at the positions of all stars in the catalog brighter than a given limiting magnitude. This
has two important advantages for subsequent data retrieval and analysis; all photometric
measurements are associated with known objects from the outset, and the aperture for
every object is always centered at a precisely-determined and consistent position on the
– 12 –
Bias frames, thermal dark-current exposures and twilight-sky flat-field exposures are
secured at dusk and dawn on every night of observation. The pipeline carries out a number
of statistical validity tests on each type of calibration frame, rejecting suspect frames before
constructing master bias, dark and flat-field frames.
Master bias frames and thermal dark-current frames are computed by taking iteratively
sigma-clipped means of the ten to twenty frames of each type taken on each night. The
master bias frame is subtracted from all thermal darks, flat-field frames and science frames.
Temporal drifts in the DC bias level are removed using the sigma-clipped mean counts in
the overscan region. The overscan strip is then trimmed off the bias-subtracted frames.
The thermal dark frame is scaled according to the exposure time and subtracted from all
flat-field and science frames.
The twilight sky flats are exposed automatically in a sequence of fifteen pre-programmed
exposures ranging in duration from 1 to 30seconds. They are timed so as to be uniformly
exposed to a maximum of about 28000ADU in the frame center. The mount is driven
to slightly different positions on the sky between exposures to facilitate removal of
stellar images when the images are combined. The flat-fields show a circularly-symmetric
vignetting pattern, caused by a sequence of baffles of similar size within each lens. Gradients
in the sky brightness across each flat-field image are removed by rotating each image
through 180 degrees about the center of the vignetting pattern, subtracting the rotated
image, and performing a planar least-squares fit to the residuals. The gradient is then
divided out from each flat field exposure.
– 13 –
The sky brightness distribution on the short-exposure flat fields is slightly distorted by
the finite travel time of the CCD camera shutters, which are of the five-leaved iris type. At
each pixel position, we must determine both the correction δt(x,y) to the exposure time
and the normalized flat-field value N∞(x,y) that would be obtained for an infinite exposure
time. The normalized counts N(x,y) in an image with exposure time texpare modified by
the shutter travel correction δt(x,y):
N(x,y) = N∞(x,y)
At each pixel position (x,y), we determine the combined flat-field and vignetting map
N∞(x,y) and the shutter time correction map δt(x,y) via an inverse variance weighted
linear least-squares fit to N verses 1/texp. An iterative rejection loop eliminates outliers,
which are usually caused by a stellar image or cosmic ray falling on the pixel concerned in
one or more of the frames. We then smooth the map of δt using a two-dimensional spline
fit and use this to recover an improved map of
1 + δt(x,y)/texp
for each individual exposure. We then average these corrected exposures, again using
iterative sigma clipping to eliminate stellar images in individual frames. The shutter
correction is applied to the science frames as well as to the flat fields. In general, for a 30
second integration the shutter corrections are between +0.02 and -0.01 and with a sigma of
0.006. While flat fields are obtained on a twice daily basis (weather dependent) we use an
exponential weighting scheme to produce a daily master flat field (Cameron et al. 2006).
– 14 –
In order to derive adequate astrometry, we must establish a precise astrometric solution
for every CCD image. The celestial coordinates of the image center can be established to a
precision of a few minutes of arc from the mount coordinates recorded in the data headers,
and from the known offsets of the individual cameras from the mount position. Subsets of
the TYCHO-2 (Høg et al. 2000) and USNO-B1.0 catalogs are made and retained for every
pre-programmed pointing of the mount, and for every camera. The sub-catalogs cover a
slightly larger region of sky than the images with which they are associated, to allow for
We use the Starlink extractor package, which is derived from SExtractor (Bertin &
Arnouts 1996), to create a catalog of the 104or so stellar images detected at 4σ or greater
significance on each frame. We project the TYCHO-2 sub-catalog on the plane tangent to
the celestial sphere at the nominal coordinates of the field center. We attempt to recognize
star-patterns formed by the brightest 100 stars in both catalogs, and establish a preliminary
plate solution consisting of a translation, rotation and scaling. Further stars are then
cross-identified, and the solution is refined by solving for the barrel distortion coefficient
and the location of the optical axis on both the sky and the CCD. The RMS scatter of the
extractor positions, relative to the computed image positions of TYCHO-2 stars on the
CCD, is always close to 0.2 pixel.
Once the plate solution is established, the pipeline software creates a photometric
input catalog from the list of all USNO-B1.0 objects brighter than magnitude R = 15 (in
the USNO system) whose positions fall within the boundaries of the CCD image. Positional
and rough magnitude matching yields USNO-B1.0 identifications for all but a few dozen
of the objects found by extractor. These “orphan” objects, some of which are likely to
be transient variables or minor solar-system bodies, are added to the photometric input
– 15 –
catalog at their observed pixel locations. In addition, the positions of the bright planets are
computed and, if they fall within the image area, they are added to the input catalog.
4.3. Aperture photometry
The photometric input catalog gives the precise CCD (x,y) coordinates of up to 210500
objects, together with their USNO-B1.0 magnitude estimates. We create an exclusion mask
for fitting the sky background, by flagging all pixels within a magnitude-dependent radius
about every object in the input catalog. A quadratic surface is then fit to all remaining
pixels in an iterative procedure. On the second iteration, the fit is refined by clipping
outliers to remove cosmic rays and faint stars, and adding their locations to the exclusion
Gradients and curvature in the sky background illumination are removed by subtracting
the quadratic sky fit from the image. Images are rejected if more than 50% of the pixels are
clipped or have too high a χ2value - usually indications of significant cloud effecting the
observations. Aperture photometry is then performed in three circular apertures of radius
2.5, 3.5 and 4.5 pixels (these apertures were selected by inspection of images of known
blended and unblended objects, at this spatial resolution). Since the aperture is centered on
the actual star position, the weights assigned to pixels lying partially outside the aperture
are computed using a Fermi-Dirac-like function. This is tuned to drop smoothly from 1.0
half a pixel inside the aperture boundary to 0.0 half a pixel outside it. The weights of these
edge pixels are renormalized to ensure that the effective area of the aperture is πr2where r
is the aperture radius in pixel units.
The sky background is computed in an annulus of inner radius 13 pixels and outer
radius 17 pixels, so that the sky annulus has ten times the area of the 3.5 pixel aperture.