The detailed choice of observing frequency, especially below 2 GHz, is often constrained by known sources of radio frequency interference (RFI). For example, U.S. frequency allocations in the L Band around 20 cm include aeronautical radio navigation (including global satellite networks), meteorological aids, and fixed and mobile use. Many of the possible external interfering signals are time variable, so freedom from external interference can never be guaranteed at L Band outside the protected radio astronomy bands. The protected band from 1400 to 1427 MHz is undesirable for some continuum observations, however, because galactic neutral hydrogen emission can increase the system temperature significantly in this band.
External interfering signals are partially rejected by interferometers (e.g., Thompson 1982) because only the component of the signals that (a) varies at the sidereal fringe rate, and (b) correlates with the correct delay, will affect the output. In practice, this combination makes low frequencies and short-baseline synthesis arrays more sensitive to RFI than high frequencies (where radio frequency band allocations are less congested) and long-baseline arrays.
Data averaging reduces the response to interfering signals by the
factor sinc( 
) where 
is the natural fringe frequency
and 
is the time over which the data are averaged when computing
the visibility in each u-v cell before Fourier transforming. As 
, where 
is the angular velocity of
the Earth's rotation and 
is the declination, fringe rate
rejection is least for observations near the celestial poles and for
baselines near the v-axis, where fringe frequencies are near zero. In
a given RFI environment, long-baseline arrays collect more data at large
values of u than do short-baseline arrays, so long-baseline arrays are
less affected by RFI (unless the RFI is strong enough to degrade the
noise performance of the individual receivers). Short-baseline arrays
may also suffer more from locally-generated RFI or from ``crosstalk"
(correlations between signals radiated by the electronics of one antenna
directly into the feed or the electronics of another).
Delay rejection is a more complex function of the azimuth of the baseline and the position of the interfering source, as described in detail by Thompson (1982). Its effects cannot be stated as concisely as those of fringe rate rejection and are not a simple function of location in the u-v plane. Delay rejection is not usually significant for narrow-band interfering signals.
RFI may always be edited from the data after they are taken (with particular attention to data falling near the v-axis), but careful choice of observing frequency can minimize both the effort necessary for editing and the consequent loss of data.