18 AUSTRALIAN SKY & TELESCOPE July 2017
LEAH TISCIONE / S&T, SOURCE SKA ORGANISATION (2)
destroy the agricultural economy.
Then there’s the technical challenge. Building the dishes,
pedestals and sensitive receivers is one thing, but hooking
every single antenna up to a central supercomputer through
optical fibre connections is another. SKA1-mid’s total raw
data output amounts to 2 terabytes per second, or over
60 exabytes (60 × 1018 bytes) per year — more than 5% of
the total internet data traffic in 2016. This all has to be
processed in real time by dedicated correlators and data
processors to produce the final high-resolution radio images
that astronomers are after. The required number-crunching
power is on the order of 350 petaflops — 350 thousand
trillion calculations (or floating point operations, hence ‘flop’)
per second; the expected yearly output of archived science
data products would fill 7 billion DVDs. As Australian
astronomical computing expert Andreas Wicenec (University
of Western Australia) says, “The deluge continues”.
Outback proving ground
I met Wicenec in June 2016 in Perth, Western Australia,
where he’s involved in the Australia SKA Pathfinder (ASKAP).
Nearing completion at the Murchison Radio-astronomy
Observatory, some 800 kilometres north of Perth, ASKAP is
another SKA precursor telescope. It’s a six-kilometre wide,
36-antenna array. The 12-metre dishes are each equipped
with phased array feeds, receivers capable of detecting multiple
radio beams simultaneously, providing the observatory with
an unprecedented 30-square-degree field of view — 150 times
the apparent area of the full Moon in the sky.
According to Wicenec, ASKAP produces some 250
terabytes per day of raw data, more than 15 times the
nightly data forecast for the upcoming Large Synoptic Survey
Telescope. A dedicated supercomputer called Galaxy at the
Pawsey Supercomputing Centre in Perth processes the flood.
At present, it’s the fastest radio observatory in the world
in terms of survey speed. Astronomers expect ASKAP to
eventually map some 70 million radio sources — a thirtyfold
increase over the current number.
Like MeerKAT, the project was largely developed to
support an SKA bid — in the first decade of this century, both
South Africa and Australia were hoping to host the future
array, back when the plan was to have it in only one place.
But contrary to MeerKAT, ASKAP will not become part of
the SKA: SKA-low will be at the same site, but it will use a
completely different kind of antenna. Instead, says CSIRO’s
Astronomy and Space Science director Douglas Bock, ASKAP
will be a great observatory on its own. Bock expects the array
to be completed in 2018.
SKA Precursors
Name Frequency range
Australia SKA Pathfinder (ASKAP) 700 MHz–1.8 GHz
Hydrogen Epoch of Reionisation Array (HERA) 50–250 MHz
MeerKAT 300 MHz–3 GHz
Murchison Widefield Array (MWA) 80–300 MHz
SOUTH
AFRICA
NAMIBIA
Cape
Town
Port
Elizabeth
80 km
8 km
Perth
WESTERN AUSTRALIA
80 km
3 km
SON THE MAP Shown are the preliminary locations for SKA1-mid and SKA1-low. Each South African icon represents a single dish; each
Australian icon represents a station of 256 individual antennae. The insets give a slightly clearer picture, although the sheer scale is still hard to
grasp. Designers chose the spiral configuration because it provides a variety of baseline distances and angles between antennae, enabling very
high-resolution imaging with interferometry. (The best layout would be random, but construction considerations make that undesirable.)
Antenna type
1 2-mdish
1 4-m meshdish
13. 5 -mdish
dipole antenna
Location
Australia
South Africa
South Africa
Australia
RADIO REVOLUTION