Australia Telescope National Facility

Base Band Data for Testing Interference Mitigation Algorithms

Jon F. Bell, Peter J. Hall, Warwick E. Wilson, Robert
J. Sault,
\\ Rick J. Smegal, Malcolm R. Smith, Willem van Straten, \\
Michael J. Kesteven, Richard H. Ferris, Frank H. Briggs, \\
Graham J. Carrad, Malcom W. Sinclair, Russell G. Gough, \\
John M. Sarkissian, John D. Bunton \& Matthew Bailes, PASA, 18 (1), in press.

Next Section: Properties of Interfering Signals
Title/Abstract Page: Base Band Data for
Previous Section: Base Band Data for
Contents Page: Volume 18, Number 1

Introduction

Radio astronomers make passive use of wide spectral bands (sometimes hundreds of MHz) outside the much smaller bands allocated for passive use. The prime motivation for this is achieving greater sensitivity, since sensitivity improves as the square root of the bandwidth for broad band sources. The wide spectral bands inevitably have other legally licensed emissions, which are typically much stronger than the desired astronomical signals. Without the means to suppress unwanted signals the utility of wide band astronomy systems may be limited. Furthermore astronomical spectral line sources are at very specific frequencies. If interference is also present in these frequency channels, we need a suppression approach which makes these frequency channels useable. We have already reached the point where it can be difficult to obtain good quality data for particular experiments. In the next 10-20 years, when next generation radio telescopes (Butcher 2000) such as the Square Kilometre Array (SKA) and Low Frequency Array (LOFAR) are built, suppressing interference will be essential. There is no silver bullet for mitigating against interference. A successful mitigation approach is most likely to be hierarchical or progressive through each of the telescope, signal conditioning and signal processing systems (see Ekers & Bell 2000 for a summary). The techniques of blanking time samples or frequency channels that are affected by interference are already used extensively for observations of astronomical sources. In this paper we focus on digital signal processing solutions for tackling the suppression problem and present a data base to use in testing techniques. In many communications systems the modulation or coding of the desired signal is known. As shown in Figure 1a the modulation or coding signal can be used at the receiver as a reference signal for adaptively selecting the desired signal in preference to the interfering signal. For most radio astronomy this is not possible because there is no coded or modulated signal, just band-limited or frequency-dependent noise. As a result radio astronomy is forced to try to find a reference signal for the interference (see Figure 1b), use that to adaptively select the interference, then cancel it. There are many well-known adaptive suppression techniques (Haykin 1995, Widrow & Stearns 1985, Ellingson 1999) ready for testing. The key outcomes of those tests are what level of suppression they can achieve (up to 80dB may be necessary), and how harmful or toxic they are to the weak astronomical signals. Some have already been successfully tested with astronomical data (Leshem and van der Veen 2000, Leshem, van der Veen, & Boonstra 2000, Barnbaum & Bradley 1998, Ellingson & Hampson 2000, Sault 1997, Kewley, Sault, & Ekers 1999, Briggs, Bell, & Kesteven 2000, Kesteven et al. 2000, Kewley et al. 2000) .

**Figure 1:** (a) Left diagram: Typical communications system where the known modulation or coding a_m (applied to the data signal a gives the transmitted signal a_s) is used in a black box to select the desired signal in preference to the interference, giving the detected data signal a_d. H_A and H₁ are the unknown transmission functions of the desired and interfering signals respectively. (b) Right diagram: Typical radio astronomy system in which the properties of the desired signal a are unknown and a reference signal for the interference must be obtained instead via some other receiver (to the right of the dashed line).
$\begin{figure} \centering\leavevmode \epsfxsize=0.65\columnwidth \epsfbox{intmit_diag.ps} \end{figure}$

Finding a high quality reference signal for the interference is of key importance. There are several ways of devising a suitable reference:

1.: A prior knowledge of the modulation or coding properties of the interference. Using the data presented here, Ellingson, Bunton & Bell (2000) have demonstrated a parametric cancelling technique that uses the known coding sequence of GLONASS (the Russian global positioning system) signals to cancel them.
2.: A specially designed separate reference antenna that optimises the interference to noise ratio, and does not detect the astronomical signals. A simple horn was used as a reference antenna to obtain some of the data in this paper. This data has already been used to demonstrate effective cancelling of interfering signals from a point-to-point microwave link (Briggs, Bell & Kesteven 2000). Barnbaum & Bradley (1998) also used such an antenna with a real-time least-mean-square (LMS) based adaptive canceller to remove frequency modulated (FM) transmissions.
3.: Use of multi-feed receiver systems, such as the Parkes 21 cm multibeam receiver (Staveley-Smith et al. 1996) allowing a reference signal to be obtained by cross-correlating signals from groups of receivers to form a reference for another (Sault 1997, Kewley, Sault, & Ekers 1999) . The dataset described in this paper includes data from the multibeam receiver. Tests by Briggs, Bell, & Kesteven (2000) and Barnbaum & Bradley (1998) suggest that this is not as effective a reference as a separate antenna.

In this paper we summarise data in which all the above methods for obtaining reference signals are used. The interfering signals and how they affect astronomical observations made at the CSIRO ATNF Parkes and Narrabri observatories are described in Section 2. The properties of the reference antenna are described in Section 3. The available data, including formats and sample spectra are presented in Section 4. Matlab scripts for reading the data are available at http://www.atnf.csiro.au/SKA/intmit/.

Next Section: Properties of Interfering Signals
Title/Abstract Page: Base Band Data for
Previous Section: Base Band Data for
Contents Page: Volume 18, Number 1

ASKAP

Public