Interference

In principle, if interference is not coincident in time, frequency, polarization, and direction with a celestial object, we should be able to observe in the presence of the interference. We should not forget that most of time-frequency space is still open to us in the skies above our observatories, even at frequencies that we think of as full of active signals (Gulkis et al, 1991 and Swarup & Venkatasubramani, 1991). Lest the active users of the radio spectrum conclude that the problem of interference suppression falls solely on the receiving antenna designer, I should point out that this same ``in principle'' statement implies that all active users could share a much smaller fraction of the present radio frequency spectrum than is currently allocated. The degree to which we can approach this ideal depends on economics and the state of the electronics and antenna arts.

Isolation of celestial signals in direction means very low response of our antennas to signals well away from the main beam. Even the best antennas have some response in their ubiquitous far sidelobes, so nulls must be steered onto the sources of strongest interference. My guess is that the extremely low signal-to-noise levels of radio astronomy present unique challenges to adaptive nulling that haven't been considered in the literature. This implies that we need to add more information to the signal processing such as a priori knowledge of the antenna phase and amplitude response in the direction of interference. The introduction of a cancelling signal to the receiver means added noise and gain instabilities which must be minimized by having considerable reception gain and stability in the auxiliary interference sampling antenna.

Temporal rejection of interference can involve time scales from microseconds, e.g., radar pulses, to many tens of seconds. Most of the integration times that we currently employ in our observations are much too long for effective interference rejection. Strong bursts of interference are easy to recognize and excise, but most intermittent interference has an amplitude distribution that resembles white noise too closely to remove it with post-observation data selection. Again, added information in the form of a separate interference monitor and known periodicities and duty cycles must be added to the process.

As the radio spectrum gets more crowded (Thompson et al, 1991), rejection of interference in the frequency domain will require filters closer to the front end of our receivers. Satellite down-link signals already pose some severe dynamic range problems for receivers designed for adjacent radio astronomy bands. Our present optimization of gain distribution for the system temperature and tunability must be tempered by the need for large signal handling capacity.

Every radiated man-made signal is completely polarized, by the very nature of a transmitting antenna, even if it is a random collection of wires in an incidental radiator. Hence, for any given interference signal there should be an orthogonal received elliptical polarization in which there is no interference power. To test this I made a series of measurements in the 60 to 80 MHz range with the 140-ft telescope with a polarization-sensitive spectrometer. Most of the received signals were from broadcast TV stations. By subtracting the polarized component of the measured spectra the expectation was that we would be left with only the unpolarized natural radiation.

Of approximately 400 spectral features measured, a few were suppressed by 20 dB or more, but most were suppressed by an average of about 8 dB and some were reduced very little. The distribution of suppressions is shown in Figure 1. This tells me that we really don't understand the nature of interference signals as propagated to our radio telescopes, and, until we do, our efforts at rejection will have limited results. Nevertheless, a significant suppression of interference can be achieved by recording all polarization information in our spectra.

  
Figure 1: Distribution of suppression achieved on 392 interference features in spectra between 60 and 80 MHz by subtracting the polarized flux.

© Copyright Astronomical Society of Australia 1997