Low-Frequency Aperture Array


[NOTE: This contribution was sent before the LFAA CDR, that was held at SKA HQ on 11-13 December]

The consortium together with the SKAO completed the Critical Design Review material. The CDR will be held in Dec 2018, the final milestone of the consortium. In the meantime engineering work continues.

Aperture Array Verification System 1, AAVS1: initial evaluation

Steady progress is being made to characterise and verify the performance of the SKA-low prototype system Aperture Array Verification System 1. AAVS1 is a full-scale prototype system consisting of 256 “SKALA2” log periodic dipole antennas arranged in a pseudo-random configuration with diameter 35 metres.

Figure 1. A real shot of the AAVS1 station

AAVS1 is primarily a test and verification system aimed at:

  • providing on-the-ground experience in deploying, commissioning and maintaining SKA-low hardware and software
  • verifying the proposed station-level commissioning and calibration plan
  • verifying the performance of station-level beamforming against SKA sensitivity specifications
  • shaking down core station-level software and firmware for calibration and beamforming

The existing system has been in the field at the Murchison Radio-Astronomy Observatory (MRO) for over 12 months now, showing very little issues. The field hardware, as well as the hardware located at the CPF, has showed reliable operations.

This report is a summary of work that has formed part of the large documentation set submitted for LFAA CDR, specifically the LFAA Demonstrator Test Report.

Initial bootstrapping of AAVS1

While the primary outputs of SKA-low stations as part of the SKA telescope will be station beams, AAVS1 is a standalone system that is being used for verification, so most of the characterisation work for AAVS1 thus far has been performed by using the station as a standalone 256 antenna interferometer. The activities have thus been following a series of steps to “bootstrap” the system from initial deployment (where we know nothing about the system) to full station beam.

Using debugging and/or calibration modes of the station firmware, standard radio interferometric cross-correlation products (the “visibilities”) can be generated between all antennas for a single (~1 MHz wide) coarse channel using the internal GPU-based correlator. Alternatively, raw voltage data for a single coarse channel for all antennas can be written to disk for short (~2 seconds) bursts.

The bootstrapping process broadly consists of:

  1. Checking signal level and normalising the power levels for each antenna (as described in the previous edition of SKA eNews)
  2. Verifying that there are no large unanticipated/uncorrected delays between signal paths in the station. (This is an assumption/requirement associated with forming station beams at the coarse channel level such that there is no bandwidth decorrelation over the channel.)
  3. Verifying the mappings between antennas (including location and polarisation labels), through the analogue and digital systems, to voltage and visibility data
  4. Verifying that the station can be calibrated using a sky model that incorporates the sun and diffuse background sky

The second and third items in the bootstrap list above were achieved by manually forming visibilities with 32 “fine” spectral channels from a few seconds of captured (single coarse channel) raw voltage data, then converting the visibilities to a standard radio astronomy data format. This dataset was collected in April when the sun was the single dominant radio source in the daytime sky. The data showed:

  • Signal path mappings were mostly correct
  • Some antennas had their polarisations swapped
  • There were no large uncorrected delays in the system, which would have manifested themselves as a large phase slopes over the coarse channel
  • The digital coarse channel bandpass qualitatively looks as expected

A “first image”

The data described above where used to create an all-sky image with the Sun as the dominant compact source at the phase centre.

Figure 2. The image above shows the first image made by AAVS1 using approximately 2 seconds of data centred on 160 MHz. The image is the XX polarisation using robust -0.5 weighting, all baselines and light (niters=2000) cleaning. The sun is at RA/DEC 1.4/9.3, which is the phase centre of the image. (Note for scale, the sun has been assigned as a 10000Jy source. The intensity scale at the bottom deliberately burns out the sun so that other features can be seen.) Since the image is a snapshot, is has a slant orthographic projection such that the coordinate system is accurately represented by ds9 in this view. As well as the sun, two other radio sources are clearly visible: Fornax A and the LMC. Eagle-eyed readers will also spot the SMC and PKS2356-61.

Towards quantifying calibration accuracy and stability of AAVS1

After the initial bootstrap, the internal station GPU-based correlator was used to generate visibilities from a single coarse channel for many hours to test both the stability of the system and the ability of various calibration schemes to use the diffuse galactic sky to calibrate the station. Daytime data from the southern autumn were particularly useful for diagnostic use. Example calibration solutions from a 2-hour dataset using the sun as a calibrator are shown below.

Figure 3. This image shows example phase solutions from a 2-hour 160 MHz dataset using antenna 2 as the reference. The solutions are generally flat with small deviations on ~20 minute timescales.

24 hour datasets were collected in May, June and July. Using the calibration solutions from daytime data, the full 24 hour dataset was imaged at 1 minute intervals. An example snapshot image is shown below.

Figure 4. A zenith-pointed all-sky image at 160 MHz from AAVS1. The compact station is sensitive to the large-scale diffuse emission from the Galactic plane, which is clear in this image. Also visible are several “A-team” radio sources including CenA, VirA, HydA and TauA.

In order to estimate the stability of the station calibration over time, a statistical analysis of the calibration solutions was performed for different datasets (daytime, nighttime) and calibration schemes. Each solution dataset was divided in 10-minute bins, which is the timescale employed as update rate in the SKA-low station calibration.

For every sample, the variation of the phase and amplitude solutions, relative to the beginning of each 600-s interval, was calculated for every antenna. The figure below shows an example of phase difference distribution calculated for the 2018-11-08 solar observation, in particular the channel corresponding to 160 MHz of observed frequency, in the XX polarization. Here the calibration solutions were obtained using a sun-based calibration scheme.

Figure 5. This plot shows the drift in the phase solutions in 10-minute bins calculated for the 2-hrs observation carried out on 2018-11-08 centered on the solar transit.

The plot of the amplitude variations of the same complex solution set is shown below.

Figure 6. Drift in the amplitude solutions in 10-minute bins calculated for the 2-hrs observation carried out on 2018-11-08 centered on the solar transit.

The variations of the complex calibration solutions in each bin were investigated using a number of statistical parameters. A preliminary Kolmogorov-Smirnov was applied to the calibration solutions per antenna and it confirms that their variations follow a Gaussian distribution in both phase and amplitude, as expected.

A statistical parameter, useful for the stability analysis, is the maximum standard deviation of the solution variations among all samples contained in each 10-minute bin. It allows the comparison with the SKA LFAA stability requirements.

For example, the values of that parameter calculated for the phase solutions of the 2018-11-08 solar observation is illustrated in the following plot, for both the XX and YY polarizations at 160 MHz.

Figure 7. Maximum standard deviation of the phase cal solutions at 160 MHz within 600-s bins measured during the 2018-11-08 solar observation for the XX (blue diamonds) and YY (red dots) polarizations.

Also for other observations and frequencies, as in this case, the calibration solutions in YY polarization seem to be less stable than the XX one, in both amplitude and phase. This behavior needs to be further investigated.


Figure 8. Prototypes of the SKALA4 antenna under tests at the MRO (SKALA4 UCAM), Western Australia (left) and during wind tunnel tests (SKALA4 INAF)

Two versions of the new LFAA antenna SKALA4 have been realized and tested, a steel version by UCAM and Cambridge Consultants Ltd. and an aluminum version by INAF in collaboration with CNR-IEIIT and the industrial partner Sirio Antenne. Two SKALA4-AL prototypes have been built, one of them used for tests in Italy, while the other antenna has been shipped to the Murchison Radio Observatory (MRO) site in Australia.


An option under consideration is an alternative location for the RFoF units. In Figure 9. A prototype of the so-called Smart box is presented.

Figure 9. Smart box combining 16 Front End Modules

Unmanned Aerial Vehicle (UAV) for verification

New-generation radio telescopes such as the SKA’s aperture array require advanced techniques to confirm the desired performance and calibrate the instrument. At low frequencies, the strong mutual coupling between the array elements and the interaction of the antennas with the environment can drastically alter the expected response of the elements and compromise the calibration process.

In this context, an antenna measurement system consisting of a radio-frequency signal source mounted on an Unmanned Aerial Vehicle (UAV) has been developed in order to perform tests at both subassembly and end-to-end level. Measurement campaigns have been successfully carried out on two prototypes of the SKA low-frequency instrument placed in United Kingdom and on a station of the LOFAR radio telescope in The Netherlands. Recently, near-field verification strategies have also been proposed.

Figure 10. UAV drone used for measurements of the SKALA4-AL (INAF)

Report provided by the LFAA consortium