We describe an image dissector photoelectric system that is used for acquisition, guiding, focussing, and photometric monitoring in telescope control. A similar system is used for area scanning and spectrophotometry.

Astronomical spectra are, with few exceptions, still being recorded on photographic emulsions. These generally have the advantages of better spatial resolution and larger available area than existing photoelectric imaging detectors. On the other hand they do suffer from a small dynamic range (15 or 20: 1), reciprocity failure, and a quantum efficiency of less than 1 per cent. Image intensifiers and image orthocons can be used to improve sensitivity, but where photographs are used for the record the other problems tend to remain. Calibration and extraction of the appropriate astronomical information in digital form involves a considerable amount of time and careful work, as one can appreciate from many of the papers presented at this colloquium.

A system is being developed at the Rome Astronomical Observatory for real-time detection of solar spectral information. The scheme and the philosophy of the system are here described. Finally, the first program to which it will be applied is presented.

Improvement of astrometric accuracy is a major need of astrophysics today. A significant increase in the amount of information about the masses of stars, parallaxes, and proper motions is to be gained. Some improvements have been accomplished in the last few years such as the astrometric reflector like the 61-inch at Flagstaff, (K. Strand, 1967), and the McCormick 40-inch (under construction). Another direction of improvement has been automation of measuring devices such as the Strand Measuring Engine and the Grant Measuring Engine; but one of the large improvements of the future ought to be observations outside the Earth’s atmosphere to the level of diffraction limited images instead of seeing limited images. The concentration of light in a smaller image will then mean both a fainter magnitude reached and a high astrometric accuracy.

The requirement for a machine of the GALAXY type arose from the large amount of information made available by the photographic plates from Schmidt telescopes. Even a small telescope of this type will produce many thousands of well-defined images per square degree on a single plate. Previously only a few selected images could be measured for position or magnitude in order to keep pace with the production of plates.

The GALAXY measuring engine was designed for the automatic measurement of glass photographic plates taken with Schmidt telescopes. These plates contain the negative images of tens of thousands of stars, with image sizes from 15 μm upwards. To carry out this task GALAXY works in two modes: search phase and measurement phase.

In search phase, GALAXY scans preset areas of the plate with a spot of light from a CRT in a linear raster scan. The transmission through the plate is monitored and each time the light spot crosses a star image, the reduction in transmission is recorded. The coordinates of the star are derived to a nominal accuracy of 16 μm and these coordinates output to paper tape.

The automatic measuring machine acquired by the U.S. Naval Observatory in 1966 was built by Nuclear Research Instruments (NRI) in Berkeley, California, based upon design specifications by the Naval Observatory and upon a considerable amount of innovative engineering by NRI. The performance of the machine and the results obtained over the past four years clearly show that a reliable high speed measuring system has been achieved. Hence, when requirements developed for a second machine it was decided to copy the earlier machine, incorporating modifications which the present state of the art permit and which were not available when the machine was designed in 1963.

The Cambridge automatic plate measuring project is an attempt to develop a reasonably cheap system capable of measuring photographic images to sub-micron accuracies at high speed, typically 10 images/sec.

In cooperation between the European Southern Observatory and the Göttingen Observatory the design of an automatized iris photometer has been developed. The principal idea, based on the classical iris photometer as described by Haffner (Veröff. Göttingen, Nr. 106, 1953), may be described here. (Fig. 1).

A method for precise centering on photographic star images is proposed. The image is moved only once across one scanning slit for the first coordinate and then across another slit for the orthogonal coordinate.

The photometric values are stored and treated in an on-line computer. The relative opto-mechanical simplicity suggests the application of the method to existing comparators.

The instrument described in this paper has been developed as the result of close collaboration over a number of years with the University of Utrecht.

It features a combination of advanced facilities which have not previously been available for such work.

Despite the great progress which is being continuously made towards direct photoelectric recording, it is almost certain that for many years to come photographic plates will remain the most common means for detecting images of stars and stellar spectra, whether directly or through the intermediary of an image intensifier.

The photographic image of the spectrum of a star contains an enormous wealth of information whose utilization is one of the most important tasks of observational astrophysics. To achieve this task many difficult problems must be solved, the first being to express the information in a form which lends itself to further processing.

This is usually done by means of two different kinds of instruments.

Firstly I should like to apologise for Mr. Loebl’s inability to come here today. Because of illness, he has been asked to rest and since he and I are working together on the same problems, he has asked me to present this paper.

It must be obvious to all concerned in the evaluation of photographic images, that microdensitometry has rapidly changed with the advance in technology. Improvements have taken place over the years in the use and refinement of double beam recording instruments. Present trends indicate that little further can be done regarding optical resolution and diffraction limited systems can now be built. The quantitative evaluation of images based on a line integral has not changed in principle for many years. A scanning spot of the requisite size is passed across, for example, a spectral line, and the densitometric profile line gives the information required, either by its height or its area.

My background is neither astronomy nor instrumentation but rather image analysis. Hence, I am not going to describe new equipment nor new astronomical findings, but rather some difficulties which have arisen in other fields where images must be interpreted and manipulated, and which are sure to arise in astronomy as well.

I shall limit my attention to problems arising from coherence of the radiation in the analysing instrument and attempt to isolate their source and give guidance on how to recognize them and avoid them.

At optical wavelengths the solar flare is a very complex phenomenon, and one of the major problems in the study of solar flares has been to find a consistent and reliable method of classifying them. A survey of the published data will quickly show that a solution has still to be found. A second problem is to find some parameter which will accurately represent the temporal variation of the flare. The relatively small amount of photometric measures available shows that there is a need for a rapid method of characterizing the flare’s intensity and areal fluctuations. In comparison to other astronomical measurements, the techniques used in measuring and classifying flares are primitive and highly subjective. The present system of classification differs only slightly from the pioneering efforts, and little attempt has been made to take advantage of the technological developments to remove the observer’s subjectivity—in fact, the latest classification scheme actually increases it!

Measurements of the magnitude distribution and integrated color of stellar fields in many parts of the sky by the star-counting method are important for studies of the structure of the Galaxy and the distribution of interstellar matter. On the other hand, in quantitative studies of the airglow and zodiacal light, we have to know the integrated brightness of the star field in order to subtract the background light from the measured light.

Up to the present, these star-counts have been made only for limited areas of the sky, since these measurements were made by using a reading microscope, and they would require much time and labor to extend the measurements to larger sky areas.

Starting from a digitized and calibrated stellar spectrum—what can we most usefully do with it?

Traditionally a spectrum consists of two parts, a continuum and some lines. These have been variously defined in the past, so perhaps another definition may be allowed. Over large ranges the two components can be distinguished quite sharply in the spatial frequency plane. The continuum contains only low frequencies, whereas in the line the information is contained in the high frequencies. In the word “continuum” is included the instrumental absorption functions, emulsion sensitivity, wide interstellar and atmospheric bands, and even the wide wings of very strong lines. From the word “lines”, on the other hand, the strongest lines are excluded; these may contain intermediate spatial frequencies in their core regions. They are relatively infrequent and so may be treated as special cases.

Let me begin my description of how we have been doing computerized microphotometry with a brief history of the hardware and its evolution. Then I’ll get on to the more interesting topic of what we do with the data after we get them into the computer.

In 1966, the Shock Tube Laboratory at Harvard College Observatory took delivery of a David Mann microphotometer. Briefly, this $50 000 instrument can measure positions in an area 250 x 250 mm to an accuracy of about 1 μn. The limiting resolution of the measuring slit approaches 1 or 2 μm at the plate, and the drift in the photometer output is on the order of 1 per cent in 12 hours.

In the past several years many, if not most, observatory microphotometers have been made or adapted to give digital output. At the same time, astronomers have been faced with the task of specifying exactly all the folk lore and experience necessary to reduce microphotometer output, so that the task can be performed by a computer. In addition to the speed and accuracy gains therefore, we have also been obliged to separate the subjective from the objective in this processing, and this may be the most important improvement of all.

Knowing that many other observatories have developed similar programming systems, this paper presumably does not contain much that is startlingly new, but it is hoped that it may promote the exchange of ideas desirable to overcome common difficulties and avoid duplication (or multiplication) of effort.

This paper deals with the possibility of recovering the information contained in a spectrogram, in the case of line widths larger than the mean period of the granulation noise of the photographic plate. Suppose the spectrogram to be known by means of its microphotometric one-dimensional record, e.g. x(t), where x is the transmittance of the plate at abscissa (wavelength) t. The measurement of the position and shape of a line of the spectrogram, easy when the amplitude of the line exceeds the granulation noise, becomes very hard to do if the S/N ratio approaches unity. Nevertheless, as the power spectrum of the noise has normally a bandwidth larger than the spectrum of the signal, it is possible to reject all the noise power beyond the highest frequency in the signal spectrum, by means of suitable filters. This technique does not lose any useful information, since by definition the information is the signal, while on the contrary it betters the S/N ratio by diminishing the noise power. Of course, as both the signal and noise spectra are continuous, it is not practically possible to perform the filtering without signal losses. However, the technique may be advantageous in all cases, such as highly rotating stars, in which the spectrum of the signal has a cutoff much sharper than the noise cutoff. A simple criterion of applicability is the base width of the lines: if the base width is larger than the granulation noise period, the filtering may result in an improvement of the S/N ratio.