1 Introduction

The DRAO Synthesis Telescope (ST) consists of seven 9 m antennas having equatorial mounts that are situated along a 620 m east-west baseline. It operates at radio wavelengths of 21 and 74 cm. A total of 12 half-days of observing is required to completely sample the UV plane for a particular piece of sky. The small size of the antennas means that the array has a very wide field of view; the primary beam HPBW is 107 and 330 arcmin at 21 and 74 cm respectively. This wide field of view, coupled with a shortest spacing of 13 metres, makes the telescope an exceptionally useful instrument for studies of our own galaxy; the telescope is currently dedicated to a systematic survey of that part of the galactic plane above declination 30 degrees that is visible from the DRAO site. This project is called the Canadian Galactic Plane Survey (CGPS).

The telescope has been gradually built over a 25-year time period using components acquired from an eclectic variety of sources; some antennas were even obtained from a San Francisco junk yard (Galt, private communication)! In contrast, most well-known aperture synthesis radio telescopes, such as the Very Large Array (VLA) or Westerbork Synthesis Radio Telescope (WSRT), were designed and manufactured to some pre-defined performance criteria - e.g. all dishes should have the same primary beam shape, all dishes should have a well-specified RMS surface accuracy allowing observations to a given wavelength, etc.

This combination of eclectic mechanical and electronic components is coupled with a very large primary beam. Consequently the images produced by the DRAO ST often contain artifacts visible above the noise that cannot be easily removed by standard self-calibration algorithms developed for telescopes such as the VLA or WSRT.

In this Paper I describe the special image processing techniques that have been developed to remove these instrumental effects from DRAO images. The end result is a set of high quality images for the Canadian Galactic Plane Survey (CGPS). These techniques may have applications to other synthesis arrays, especially when these arrays are pushed to the limits of performance at, for example, very high frequencies.