Showing posts with label imaging. Show all posts
Showing posts with label imaging. Show all posts

Monday, 9 December 2013

Six-element non-redundant array

We've been discussing array layouts recently, with regard to riometry and outlines for LOFAR superstations. There are lots of possibilities of course, but there is one which is rather elegant as well as being practical. Today, we are featuring it here.

The imaging or beamforming properties of an array of antennas is related to the baselines that the array possesses. A baseline is the vector between any two antennas in the array. If an array has 3 antennas, it will have 3 baselines: the baseline between antennas 1 and 2, between 2 and 3 and between 3 and 1. A 4-antenna array will have 6 baselines, a 5-antennas array will have 10 and so on. The number of baselines is related to the number of antennas N by the expression B = N (N-1) / 2.

Although more antennas means more baselines, these baselines may not necessarily be unique. For example, let's imagine 3 antennas, equi-spaced on a line. The baseline from 1 to 2 will be exactly the same as the baseline from 2 to 3. These are referred to as redundant baselines. Although the extra antennas add sensitivity, the fact that the baselines are identical mean they do not add to the imagine quality (i.e. the beampattern).

Finding an array layout with no redundant baselines can be tricky. Even harder, is finding one whose baselines are neatly distributed, thus sampling as much of the baseline parameter space as possible. This parameter space is typically plotted on a graph referred to as a uv-diagram (the differential spatial axes are u and v).

However, there is one which actually has a uniform distribution for a relatively small number of antennas. It is the next logical extension of the 3-antenna equilateral array. The layout of the antennas looks like this:

A manx-array antenna layout (Image: D. McKay-Bukowski)

In the diagram the antenna positions are labelled as 0, 1, ... 5. The antennas themselves are not necessarily hexagonal, but it does fit nicely on a hexagonal grid, thus making it easy to visualise. The baselines for this grid look like this:

The uv-coverage of a manx array (Image: D. McKay-Bukowski)

In the uv-diagram, the baselines are labelled XY, where XY is the baseline for Antenna X - Antenna Y. As you can see, the baselines are equispaced with no redundancies.

I was introduced to this array by Björn Gustavsson (UiT), but I don't have a reference for who came up with it originally. We tend to refer to it as a Manx Array, owing to the similarity between the antenna layout and the triskelion that appears on the brattagh Vannin (the flag of the Isle of Man). But if anyone has any original references, please let us know!


PS:  It is delightful to work with hexagonal grids. It takes me back to the old days of playing with hexgrid-based games.

Monday, 2 December 2013

SFIR coverage

Hmm... data Monday again. Where did that weekend just go?

We had some visitors from Linz, Austria visit the KAIRA site on Friday. It was great to show them around as they asked lots of interesting questions (and I had a great chance to practice my now-very-rusty Deutsch). Most of the questions were about the imaging, which I've shown countless times on this weblog.

However, there was one screen which went uncommented...

Imaging coverage (Image: D. McKay-Bukowski)
This screen uses old-fashioned ASCII-art to show the coverage of one of our imaging riometry projects (SFIR). The lines refer to the day of the observation (e.g. 20130920 = 20-Jul-2013) and the columns are the hours of the day.

Then, for each position, there is a number from 0 to 9, or a blank or a hash (#). The blanks are where no data has yet been processed. The has shows the hours for which complete processing has occurred. The others show a fraction where 0 = 0-9%, 1 = 10-19%, 2 = 20-29% up to 9 = 90-99%.

As the processing continues, this chart will slowly fill up. However, with huge transforms, massive data sets and a lot of time, this will still take some four to five weeks of processing time on one of the powerful KAIRA computers.

However, it is nice to still see old-fashioned ASCII displays. Reminds me of the old days...

Thursday, 16 May 2013

Recent Results on Radar Coding and Priors


Two papers by KAIRA-team have been published in the May issue of Inverse Problems and Imaging:

L. Roininen, P. Piiroinen and M. Lehtinen, Constructing Continuous Stationary Covariances as Limits of the Second-Order Stochastic Difference Equations, Inverse Problems and Imaging 7 611-647 (2013).

L. Roininen and M. S. Lehtinen, Perfect pulse-compression coding via ARMA algorithms and unimodular transfer functions, Inverse Problems and Imaging, 7 649-661  (2013).

The first paper discusses strong-weak convergence of probability measures and a certain class of discretisation-invariant Gaussian Markov random fields. These random fields can be used as a priori distributions in Bayesian statistical inversion. Developed formalism can be used whenever regularisation techniques are needed. We use these priors for ionospheric tomography and radar imaging.

The second paper focuses on a control-theoretic study on amplitude and polyphase radar coding. In collaboration with our MIT colleagues, we are working on actual implementation of these coding methods at Haystack Observatory. The methods will be used operationally in the future EISCAT_3D incoherent scatter radar!

Both papers are open-access!

Inverse Problems and Imaging: http://aimsciences.org/journals/home.jsp?journalID=11