To test the idea under real conditions, the SKA Observatory built a new prototype station, AAVS3, with 256 dual-polarized antennas arranged in the sunflower pattern. The team tested it both as a signal-combining beamformer and as an imaging array, pointing it at the sun, bright cosmic radio sources, the galactic plane and several pulsars.

The researchers also developed two ways to calibrate the station — that is, to correct the raw data so they accurately reflect the sky. The simpler method used the sun as a known reference point. The more thorough approach combined the sun with a sky-brightness model and detailed antenna response profiles, and unlike the sun-only method, it also worked at night.

What went wrong at 125 MHz?

Simulations run while AAVS3 was being built predicted a loss of sensitivity directly overhead at around 125 MHz. The team traced the cause to repeated antenna spacings of about 2.4 meters (8 feet) throughout the sunflower pattern. These repeating distances create interference patterns that reduce sensitivity. The older, scattered AAVS2 layout showed no such problem.

Observations confirmed the prediction. When the galactic plane passed directly overhead, the images clearly showed the same dip in sensitivity at 125 MHz. That confirmed the problem was a built-in consequence of the antenna geometry, not a flaw in the equipment or in the analysis method.

A separate check of sensitivity at 230 MHz showed the more thorough calibration method worked well: nighttime measurements closely matched predictions. Daytime measurements showed larger gaps, most likely because unusually high solar activity during the observations made the sun's brightness harder to predict accurately.

What happens next?

Because of the confirmed sensitivity loss, engineers ruled out the original sunflower layout for the finished telescope. Further simulations led them to a modified version, called "Perturbed Vogel," which keeps antennas spaced far enough apart to avoid mutual coupling while removing the repeating pattern that caused the 125 MHz dip. Testing that new layout falls outside the current study.

Even so, the AAVS3 results give SKA-Low's teams practical guidance for building and calibrating stations during construction. Combined with lessons from AAVS2, this work has directly shaped the observatory's testing process and supported a key construction milestone, known as AA0.5, which marks the first stations coming online.

The study demonstrates a broader point for instrument builders: a design that looks optimal in theory and in simulation still needs on-sky verification before it is locked into a multi-year construction campaign.