Atmospheric turbulence limits the resolution of ground-based high-resolution telescopes. In the optical and radio regimes, the seeing may be described in similar ways, although the parameters are different.

Once upon a time, there were radio and optical astronomers, each group with its own way of detecting and thinking about radiation. In spite of the shock treatment administered by Hanbury-Brown and Twiss more than thirty five years ago, this dichotomy persists to a surprising extent even today. As new techniques emerge, we clearly need to do better. Coherent and incoherent detection, heterodyning and interferometry are all situations where a full analysis involves quantum behaviour of the radiation. This informal review covers the general approach needed and goes over some points of principle which arise.

The University of California Infrared Spatial Interferometer (ISI) for the 10 μm wavelength region is briefly described along with results obtained on prominent stars and on atmospheric phenomena. The system has two movable telescopes of 1.65 m aperture. It operates in principle like a modern radio interferometer, using heterodyne detection, CO2 laser local oscillators, RF delay lines, and lobe rotation to maintain a fixed-frequency fringe rate.

Rather extensive measurements have been made on atmospheric pathlength or phase fluctuation characteristics which show substantial deviations from the Kolmogorov-Taylor model, fortunately in a direction favoring adaptive optics, long baselines, and the use of infrared wavelengths. Outer scales as small as about 10 meters occur under good seeing conditions. Visibility results on 13 stars show that 6 of them have dust shells rather far from the star and give evidence for episodic emission of gas. Others of the 13 stars also vary with time, but are characterized by more continuous emission and dust formation near the stars at temperatures as high as 1300 K.

The SAO Submillimeter Wavelength Array is under construction and is expected to be ready for observations in 1997. It will consist of six telescopes with baselines from 9 to 470 m and resolutions as fine as 0″.1. Eight receivers will cover all bands from 180 to 900 GHz in linear polarization; two orthogonal polarizations will be available at 350 GHz. Multiyear site testing on Mauna Kea shows that the precipitable water vapor is less than 1 mm for 17% of the time, and the rms phase fluctuations on a baseline of 100 m are less than 55 μm for 25% of the time. The configuration of the array will be optimized to provide nearly uniform coverage in the uv plane. The correlator spectrometer will have 105 spectral channels (8192 per baseline) and will be able to provide 0.6 MHz resolution for a total bandwidth of 4 GHz.

Our ideas of how to image objects have progressed by leaps and bounds in the last twenty to thirty years. We now have a sophisticated understanding of many clever and subtle approaches to imaging. There are a few outstanding contributions which have provoked flurries of development and acheivement in many different areas: top of my list would be the principle of aperture synthesis, Jennison's closure phase, Högbom's CLEAN algorithm and Labeyrie's speckles. In addition, we have benefited tremendously from developments in computing hardware, software and algorithms (the most spectacular being the Cooley-Tukey Fast Fourier Transform). Each one of these contributions did not so much spur development in existing areas as open up entirely new vistas of possibilities. For example, Jennison's closure phase is rarely directly used by radio-interferometrists now but it did show, particularly when developed in VLBI, that imaging in the presence of severe phase errors is possible. This success then encouraged the two separate developments. First in the more flexible selfcalibration routines in which closure is implicit and, second, in pushing imaging interferometric arrays to shorter and shorter wavelengths, now ending up in the infra-red and optical regimes. Following on from these great works, many people have made lesser but still vital contributions. The example of speckle comes to mind as one where many people have had a hand in determining what is now standard technique.

This lecture is being given as part of an international meeting on astronomical imaging. It is, in fact, the first major meeting of people with backgrounds in both optical and radio imaging, and has attracted 200 people from all around the world. One of those is a man who has been an active researcher in microwave and infrared spectroscopy for over half a century. It is my pleasure to introduce that man, Professor Charles Townes from the Space Science Laboratory, University of California, Berkeley. It is clear that throughout his life Professor Townes has been attracted by a series of fundamental challenges. He was already an acknowledged molecular spectroscopist when, in 1951, to solve the problem of short wavelength oscillators, he conceived a system for using excited ammonia molecules that became the ammonia beam maser oscillator. He followed this in 1958 by publishing a paper with his brother-in-law, Arthur Schawlow, that laid the foundations for the development of the laser. These two activities, flowing as they did from the pursuit of the most fundamental physics, paved the way for some of the key elements of modern communications.

The spherical aberration in the primary mirror of the Hubble Space Telescope causes more than 80% of the light from a point source to be spread into a halo of radius of 2–3 arcsec. The point spread function (PSF) is both time variant (resulting from spacecraft jitter and desorption of the secondary mirror support structure) and space variant (owing to the Cassegrain repeater optics in the Wide Field / Planetary Camera). A variety of image restoration algorithms have been utilized on HST data with some success, although optimal restorations require better modeling of the PSF and the development of efficient restoration algorithms that accommodate a spacevariant PSF. The first HST servicing mission (December 1993) will deploy a corrective optics system for the Faint Object Camera and the two spectrographs and a second generation WF/PC with internal corrective optics. As simulations demonstrate, however, the restoration algorithms developed now for aberrated images will be very useful for removing the remaining diffraction features and optimizing dynamic range in post-servicing mission data.

For several decades, lunar occultations have represented a powerful and productive technique for high angular resolution investigations of a broad class of astronomical sources: from stellar angular diameters to close binaries, from circumstellar shells to the Galactic Centre.

The resolution and sensitivity offered by lunar occultations still compare favorably with the capabilities of other more modern methods in the optical and near-infrared. A remarkable fact, when one considers that during the past 20 years practically nothing or very little has changed in the way we observe lunar occultations: a sharp contrast with the impressive technological and scientific advancements of other techniques.

The novel possibility of using small areas of fast-readout, low-noise array detectors to record lunar occultations, however, is about to bring radical improvements. We discuss the impact of such detectors on the efficiency of observation and the gain for what concerns the resolution and sensitivity of lunar occultations measurements. New applications such as wavelength-resolved observations and investigations of low-contrast extended sources appear particularly interesting.

Several limitations reduce the field of view in radio-interferometry. With an optical array, two of them can be overcome to some extent according to the beam combination method. A beam combination in the pupil plane can completely overcome one of them. In the image plane, a beam combination obeying the rules of geometrical optics can overcome both limitations in principle, but is difficult to achieve in practice. We discuss particularly the real case of a Michelson Stellar Interferometer where a periscope partially re-introduces these limitations, yielding a trade-off between the extension of the field of view and the use of the periscope.

The strategies for image reconstruction from optical synthetic-aperture data can be divided into two camps according to their historical legacies: those coming from the field of more conventional optical image processing, where the data are relatively complete in terms of Fourier coverage, have concentrated on trying to recover images using “brute force” methods. They make use of the massive size of the typical datasets to try and derive an image without having to use any a-priori constraints. The size of the datasets to some extent restricts the algorithms used towards being as simple and hence fast as possible. Radio-astronomical imaging strategies, on the other hand, have always been designed to cope with sparse and missing data, and many sophisticated algorithms have been designed to make maximum use of a-priori constraints with iterative global fitting techniques.

We discuss the distribution of Source- and receiver-noise in radio synthesis images of polarised sources. Analytical expressions are presented for the rms noise at any location in the polarisation images. We compare these results with those for the images of unpolarised sources and discuss the usefulness of deconvolution of snap-shot images in reducing the source-noise from the off-source regions in the images.

The Millimeter Array (MMA) is proposed to operate on baselines of up to 3 km at frequencies up to 350 GHz. Since no operating interferometer can indicate the expected phase stability on such baselines on high, dry sites, NRAO has undertaken a site survey, measuring the opacity and estimating the phase stability on baselines ranging from about 100 m to 5 km with a tipping radiometer. Radiometers have sampled the atmospheres at South Baldy, NM, and Springerville, AZ, and NRAO has recently started taking data at the 12,200 foot VLBA site on Mauna Kea.

Residual delay errors in VLBI data limit the image sensitivity and dynamic range attainable using current calibration techniques such as self-calibration. Applying phase corrections derived from regular observations of nearby calibrators (phase-referencing), the residual delay errors on a target source can be reduced, increasing coherent integration times. In this paper we present the results of a series of observations made with the VLBA to examine the effectiveness of phase-referencing for improving instrumental phase stability. Optimum observing strategies when studying weak (~ mJy) sources are discussed.

There are now approximately two dozen operating interferometers in astronomy, with more under construction or in planning. The wide distribution of interferometer arrays, and their acceptance as essential tools in astronomy are a result of the remarkable flexibility and imaging capabilities of modern arrays. Much of the success of these arrays is due to advances in data processing and imaging brought about by innovative and powerful algorithms.

The VLBA is a new radiotelescope dedicated to research in astronomy and astrometry/geodesy using the techniques of VLBI. The array comprises ten 25 m diameter reflector antennas located so as to give optimum coverage of baselines up to 8600 km. Instrumentation provided at each antenna includes receivers for nine frequency bands between 327 MHz and 43 GHz, a hydrogen maser frequency standard and a tape recorder system capable of a peak data rate of 512 Mb/s and unattended operation for a day recording at 128 Mb/s. The antennas are controlled and monitored by a single operator located in the Array Operations Center (AOC) in Socorro, New Mexico. The AOC also houses the 20 station correlator which processes the data recorded at the antennas.

The VLBA construction phase is now over and in the spring of 1993 it is hoped that full VLBA observing can begin. However the VLBA antennas have been operational for some time and several astronomy programs have been run on them. Below we shall present some of these latest results.

The recent upgrading of MERLIN has resulted in a much improved array with both increased sensitivity and higher angular resolution. At L-band and higher frequencies the array is now operated in a manner similar to the VLA with routine phase calibration every few minutes using a nearby compact source. An automatic data analysis “pipeline” has been developed using AIPS software. This enables MERLIN data to be routinely processed to produce images in all Stokes parameters and with positional accuracy much better than the size of the restoring beam.