A massive eye on the universe
The Square Kilometre Array is an international initiative to build the most powerful radio astronomy telescope ever, with a million square metres of collecting area. E&T visits a South African demonstration dish which is bidding to site the massive telescope.
When Jocelyn Bell, then a University of Cambridge research student, discovered the first pulsar in 1967, she used the data from a radio telescope designed by Anthony Hewish that combined the inputs of more than 2,000 dipoles operating at 81.5MHz spread over an area of 1.8 hectares, the same as three football fields.
In his account for The Nobel Prize in Physics 1974, Hewish wrote: "This required instrumental capabilities quite different from those of any existing radio telescope, namely very high sensitivity at long wavelengths, and a multi-beam capability for repeated whole-sky surveys on a day to day basis.
"By a stroke of good fortune the observational requirements were precisely those needed to detect pulsars. Jocelyn Bell was quick to spot the week to week variability of one scintillating source which more detailed observations subsequently revealed the pulsed nature of the signal."
The Square Kilometre Array (SKA) uses the same core principles, but is to be built on a much larger scale. And its backers are able to deal with one of the problems of building a sensitive radio telescope near Cambridge. One of the problems with the original array was eliminating man-made RF sources. So the SKA needs to be sited in a remote area, preferably a desert, far from human populations and equipment such as television transmitters, microwave ovens, radar and mobile phones. Although the SKA will be largely controlled from Europe and its observations fed back to the continent, the telescope will be built a long way away.
Perfect SKA location
Early contenders for siting the SKA core included deserts in the US, China and Argentina among others. Argentina was eliminated because it is too close to the equator and therefore affected by the ionosphere for example. In 2006, the field was narrowed down by a panel of experienced astronomers to Australia and South Africa.
The international SKA office, is run by the University of Manchester, which operates the Jodrell Bank radio observatory in the UK and is responsible for picking the final location. The SKA Design Studies (SKADS) project focuses on the development of the Aperture Plane Phased Array which uses fast digital technology to make a flexible, multitasking telescope that will be able to perform many different astronomical observations at the same time.
SKADS is partly funded by the European Community Sixth Framework Programme, and includes partners from 26 institutes in 13 countries. Australia and South Africa have been making further preparations.
"Both countries have now enacted legislation to restrict the use of radio interference sources in these areas," says Dr Andrew Faulkner, European SKADS project engineer based at Cambridge University's Cavendish Laboratory.
"The square kilometre refers to the total collecting area of all the dishes and aperture arrays. In practice, the total physical extent of the array is likely to be around 3,000km. This provides better resolution on smaller objects. One of the main purposes of SKA will be as a discovery and surveying instrument to find a billion galaxies, which we know are there, statistically.
"Because of their distances, we need high sensitivity. We are trying to detect hydrogen, which is the most common element in the universe, and which emits at 1.42GHz. As the universe is expanding, the further away the object, the faster it recedes and the Doppler Effect serves to reduce the frequency of hydrogen [emissions].
"The upshot is that you can place these galaxies in 3D space. To survey the sky quickly, you want to look at a lot of the sky in one go, which is what Europe is concentrating on. The SKA frequency spectrum covers from 70MHz to 10GHz. For the phased arrays we are concentrating on frequencies below the magic 1.4GHz hydrogen emission."
SKA in Europe
As part of its involvement with SKA, Europe plans to build a phase array where thousands of 20 x 20 cm antenna elements are spaced at half-wavelength distances from each other.
The aim is to emulate the image formed by a single, steerable, huge dish. Although there is no problem with merging all the signals into one, the phases and path lengths recorded at each of the antennas need to be corrected. One can do the same with a phased array by using digitally controlled time delays between elements rather than steering a dish. With the correct delays, each element's signal will be in phase with the others on the array. By aligning phases for multiple receivers and then adding all the signals together, a phased array achieves electronically what a dish does by mechanical means with a single receiver and reflector.
Faulkner explains that once you introduce the idea of electronic correction, "there is a huge advantage: by using high-speed digital signal processing we can perform the same process again for a second beam and repeat many times to produce hundreds of beams on different parts of the sky. This is what gives aperture phased arrays their extraordinary survey performance. The signals from all the 250 phased array stations are combined in a central processing facility to provide exquisitely detailed images of large portions of the sky with great sensitivity.
"This is a flexible collecting area. Pulsars were detected with a relatively low frequency phased array. We are trying to operate up to about 1GHz with the SKA phased array, which has many more elements. As with all receivers, you still need analogue amplifiers for weak signals before processing in the digital domain. The sensitivity of the telescope is going to be very dependent on the design of the front-end receiver and low-noise amplifier. If you can double the sensitivity, you can halve the area needed.
"Conveniently, what radio telescopes receive is Gaussian noise - if you see anything else, this is interference. You take each element, which has two receiver chains for each polarisation, amplify the signal and digitise it. Each SKA phased array will be in the order of 60m diameter, comprising around 150,000 receiver chains. With a total of 250 arrays for the SKA, you can end up with 30 or 40 million receiver chains, but the flexibility and performance you will get is really quite astonishing. You can choose whether to see more of the sky in a narrow bandwidth, or less of the sky at wider bandwidth."
As one of the countries shortlisted to host the SKA, South Africa is building an SKA technology pathfinder telescope. The Karoo Array Telescope (KAT) will be sited in the Northern Cape Province on 14,000 hectares of land purchased in March 2008. Subsequently renamed MeerKAT when more funding was obtained (meer is the Afrikaans word for more), the Astronomy Geographic Advantage Act was enacted to protect minimum of 12.5 million hectares designated as astronomy reserve.
The Hartebeesthoek Radio Astronomy Observatory (HartRAO), located north-west of Johannesburg, constructed a prototype 15m dish optimised for thermal performance, low tooling cost and on-site moulding. Dr Michael Gaylard, programme leader for Single Dish Astronomy, says a 12m dish has now been specified by SKA for mass production.
"In addition to radio frequency interference, the presence of clouds also hinders satellite reception. Water absorbs the radio signals you want to observe and radiates at 22GHz itself, so you want as high an altitude as possible to reduce the amount of the Earth's atmosphere," Gaylard explains.
"You want as dry an atmosphere as possible, hence the need for a desert environment. Apart from the SKA core, the phased array will spread out over thousands of kilometres, with radiating arms and a logarithmic spiral. The long baselines from the core to remote stations will provide the high angular resolution for observing objects.
"If South Africa was to win, the core would be located in the Karoo deserts with arms radiating outwards, in some cases to neighbouring countries. Extensions could go out as far as Mauritius, Madagascar and Kenya and possibly even Ghana, which will give a wider east-west extent than South Africa alone.
"Australia has a wider east-west extent, but would probably also use New Zealand as an extension. Part of the logistical issue is actually connecting all these elements by fibre optic cable and supplying power to these remote sites.
"The low-band system is probably going to be fixed tiles or dipoles that sit on the ground, with beam steering performed electronically. When you go into the mid-bands with smaller wavelengths, the dipoles and the collecting area shrinks. At that point it is better to put a dipole at the focus of a steerable dish and then to replicate dishes."
Dish mass production
The issue is then one of how to mass produce thousands of dishes cheaply. In determining dish size in an aperture array, sensitivity has to be traded off against field of view (FoV), where the smaller the dish, the bigger the FoV. Sensitivity is provided by large N and proportional to the number of dishes multiplied by the square of their diameter, while FoV is provided by small diameter.
"The area of sky that you can observe is inverse to the square of the dish diameter, so if you double the diameter, the area of sky that you can see shrinks by a factor of four," says Professor Justin Jonas, director of Hartrao and head of the department of physics and electronics at Rhodes University.
Jonas explains: "There are two basic methods for carrying out observations on objects. One is pointed or targetted observations where the aim is to 'stare' at an object for a long time to improve the image, as long as the object is smaller than the FoV.
"The other is where the object is bigger than the FoV, and you need to put together a mosaic. For SKA to do all-sky surveys, the bigger the FoV the fewer the number of pointings it has to make, because it is making a mosaic with bigger patches. It is basically tessellating the whole sky with these beams.
"Therefore the dish diameter has to be large enough to satisfy a 20-wavelength criterion at low frequency but small enough to support high-frequency observations. In SKA terms, the lowest frequency translates largely into the red shift in hydrogen, which puts a lower limit on dish size: the 1,420MHz hydrogen line has a wavelength of 21cm.
"Even if achieving a certain sensitivity or survey speed at minimum cost drives the argument for small antennas, the limit is determined by Maxwell's equations. There is a crossover frequency for SKA at around 300 to 700MHz, where we would swap from using dishes to aperture arrays, which are more effective at lower frequencies."
SKA will employ a huge supercomputer software environment to digitally filter signals prior to sampling and cross-correlation, comprising a mix of FPGAs, ASICs and multicore processors. The full SKA will call for more than 200 petaflops of processing power.
"One of the advantages of using commodity processors is their graphics handling capability for games, which are well-matched to the algorithms we want to run, but they are not designed for radio astronomy so you need people to squeeze the best out of them," Jonas says.
"ASICs are good for power consumption but are not reconfigurable, while FPGAs are the opposite. Off-the-shelf COTS components will always be cheaper than special components. For the vector calculations, we will need the biggest supercomputer available, not only today, but in 2020."