Confusion may have two effects on the interpretation of a synthesis image:
The best way to handle confusion is to learn about it in the context of your particular target field. Before you observe, consult published radio surveys, or low resolution images of the field around your target, for possible confusion. If you know you will be observing near a bright confusing source, you may consider either of two strategies for reducing its effects on your final images.
If the confusion is relatively nearby, plan to look at it--to make wide-field images containing both the target and the confusing source. Then subtract or deconvolve the confusing source and its sidelobes from the interesting region. The ungridded subtraction technique (Lecture 9--coded in the NRAO AIPS package as the programs MX and IMAGR) helps this strategy, as only the parts of the wide field that contain significant emission need to be computed and deconcolved (CLEANed). This approach can work well if the confusing source is only one or two times as far from the target as the radius of the field of view that you would otherwise have imaged. If the confusing source is strong, consider displacing the delay tracking center away from the target toward the confusion, to minimize distortions of the response to the confusing source by chromatic aberration and other effects.
If the confusion is relatively distant, try not to
look at it--choose your IF bandwidth and the delay, phase and
pointing centers to minimize the response to the confusing source.
Note that because the point source attenuation produced by chromatic
aberration and time averaging increases with baseline length
, these effects do not filter
confusing sources from the short-baseline data as effectively as they
do from long-baseline data. If a distant confusing source still
dominates the data after attenuation by the primary beam, it may
therefore produce wide-angle ``ripple" in the final image. In such
cases, the pointing center should be chosen to minimize the response
to the confusion rather than to maximize the response to the target.
A difficult case arises when the response to the confusing source is strong even after adopting this stratagem. Variable pointing errors and (for altitude-azimuth antenna mounts) the rotation of the primary sidelobe pattern of the antennas on the sky may then make the confusing source appear to vary throughout the observation. It is hard to make images with high dynamic range in these circumstances (see also Lecture 10). It is worth recognizing this difficulty before observing, if only because you might be able to select an alternative target that is equally interesting.
If a confusing source lies in the target field itself, no advance precautions are necessary, as the confusing source and its sidelobe pattern can be deconvolved from the image as part of normal data reduction. Note however that it is important to look for confusion after taking the data, e.g., by making a heavily-tapered low-resolution image covering as much sky as possible around your target. It is particularly important to do this before proceeding with computer-intensive processing such as deconvolution, or self-calibration at full resolution. Making low-resolution images to search for confusing sources early in your image processing may (a) improve the models used for self-calibration, (b) reveal unexpected large-scale structure around your target (and so help to define CLEAN deconvolution windows) and (c) lead to unexpected discoveries.
In detection experiments, confusion may make the interpretation of a positive detection questionable if a source is detected near, but not at, the target position. Source counts (e.g., Appendix: Confusion) should be used to estimate the probability that the detected source lies in the image by chance.