# How do ASKAP's focal plane phased array feeds interact with the entire array phasing?

Sky & Telescope's ASKAP Joins the Hunt for Mysterious Bursts says:

A new telescope, the Australian Square Kilometre Array Pathfinder (ASKAP), has joined the search for energetic and elusive fast radio bursts. And in just a few days of looking, it’s already had success!

A new player is now on the scene, however, and it’s already had huge success. ASKAP is a wide-field radio telescope made up of an array of 12-meter antennas. Using phased-array-feed technology, ASKAP is able to instantaneously observe an effective area of 160 square degrees — an enormous field compared to Parkes’ 0.6 square degrees! This capability significantly increases our chances of being able to detect fast radio bursts.

The article links to The Detection of an Extremely Bright Fast Radio Burst in a Phased Array Feed Survey which is not paywalled and contains some discussion of the array.

This sounds really interesting! It is a phased array of dishes, but at the focal plane of each dish there is also a phased array of receivers.

This is different than just adding an array of feed horns and independent receivers as I've discussed in my surprisingly robustly unanswered question What is the highest granularity focal-plane array on a dish radio telescope? Or is this the ONLY ONE? where receivers measure intensity as pixels, in a similar way to a camera (thus the counterintuitively placed astrophotography tag there).

Here, each focal plane is equipped with a true phased array, where if I understand correctly, phase relationships between elements in the focal plane are maintained and recorded.

According to Wikipedia and CSIRO's Australian Square Kilometre Array Pathfinder – Fast Facts there are 36 separate 12-meter dishes in the entire array, and the focal plane of each dish is equipped with a 188 element focal-plane phased-array. 188 $\times$ 36 = 6768 channels, not counting polarizations because I don't know if they are separate channels in this case.

That would require quite a hefty correlater if the analysis was done flat, without any heierarchy in the calculation. This is a pathfinder for the SKA so pushing limits is important, but I'd like to know how this large number of signals is handled. Is there pre-processing?

Question: How do the phased array feeds on each dish of the ASKAP interact with the entire array phasing?

I'm not sure of the "focal plane array" is located precisely at the focal plane or not. If it were, then a given source in the sky would produce signals in only one, or a small group of receivers, and not the whole focal plane array! One normally uses a phased array instead of an imaging system so there must be something more interesting going on here, perhaps only nearest and next-nearest neighbor correlations within the focal plane array but full correlation between the dishes?

How does this thing really work?

The phased array feeds (PAFs) on ASKAP are indeed located at the focal plane. The ASKAP PAFs have 188 receivers, and from this digital beam-forming will create 36 independent beams on the sky. Some more information can be found from the CSIRO. Imaging is then done by scanning the dish (or letting the sky roll overhead). The scanning can then be done much faster than with a single receiver.

Its important not to confuse these independent beams with purpose of the array itself. The main reason for using an array is interferometry. We use an array for this purpose to have multiple baselines (the distance between the antennas). Having multiple baselines allows the telescope to 'see' structure various angular resolutions. It's true that you can use the array in a 'fly's eye' mode, and then every dish is providing 36 independent beams. In this mode even larger areas of the sky can be scanned quickly. This is useful for searching for transient events such as fast radio bursts.

So, the phasing within the PAF needs to be correlated and the beams formed digitally, and then the signals from each dish need to correlated. A huge task indeed! The computing breakdown for ASKAP is given here. A new supercomputer (Pawsey) has been built in WA for the purpose of processing ASKAP data.

• Thanks for your helpful answer! I'm still trying to understand how the phasing is actually done. Phasing within a focal plane array for beam forming and for each given beam direction, phasing between antennas within the whole array. I'm following the links you've listed so far and reading further.
– uhoh
Jun 29 '17 at 5:07
• No problem! I wasn't sure what level to pitch my answer at. If there's anything I mentioned that you'd like me to expand on, or if you want some additional references, just let me know. Jun 29 '17 at 22:47
• Great, thanks! After I wrote the question I read further and realized that the early work was done at a relatively low frequency of about 900 MHz, where the 1st minimum of the Airy disk is at a whopping 2°. This is why — even at the focal plane — an incident plane wave will still excite enough feeds for focal plane phasing to be useful. Synthetically shaped beams are then formed by phase apodization.
– uhoh
Jun 30 '17 at 0:57
• By "...focal plane phased array feeds interact with the entire array phasing?" I mean for example, are beams in a given antenna synthesized first, then separately the array interferometry provides resolution within each synthesized beam basically as two separate steps, or is the interferometry done in a more sophisticated, single-step process, combining multiple feeds from all telescopes simultaneously? The latter would be more interesting mathematically, but also potentially much more demanding on the fiber bandwidth and correlator computation.
– uhoh
Jun 30 '17 at 1:01
• There's a beautiful and thorough discussion in the newly published The Rapid ASKAP Continuum Survey I: Design and first results
– uhoh
Dec 2 '20 at 1:47