The order for the telescopes was placed with Wilton-Feyenoord in October 1965. During the winter months of 1965 - 1966, Dr. Gijs van Herk, the senior atrometric scientist of the Leiden observatory, spent many nights carefully establishing a series of reference points along an East-West line.
An assembly-hall with templates for the dishes and the declination/hour- angle axes was erected in the summer of 1966. In March 1967 the foundations for the ten fixed telescopes were completed. The contract for the 300-metre long high-precision rail track was signed in the fall of 1967.
The first telescope was mechanically completed in August 1967, the twelfth one in November 1968. The production process of the telescopes included a number of novel techniques, never before used in the construction of radio telescopes. One of these was the use of epoxy-glue not only to fix the wire mesh onto the surface panels, but also to connect the panels, set to millimeter accuracy on a template, with the backing structure which had been welded together carefully, but not to millimeter accuracy.
The general lay-out and the mechanical aspects of the WSRT are described by Baars and Hooghoudt (1974). When the first parabolic dish was completed it was carefully surveyed by Jaap Baars and one of the consultant engineers. Their measurements showed that the paraboloidal surface of the dish was considerably better than specified. Since this standard was maintained during the construction of the other eleven dishes, the telescope could ultimately be used at wavelengths considerably shorter than 21-cm, the wavelength for which it had been designed.
The electric drive system of the telescopes was lagging behind the mechanical construction by almost a year. By the middle of 1969 it was virtually complete.
Simultaneously with the construction of the telescope, electrical and coaxial cables were installed. In the entire design of the electronics a very careful, conservative approach was taken. In order to avoid uncontrolled phase-changes the six coaxial cables connecting each of the twelve dishes to the control building were all cut to the same length.
Also, since it was unavoidable that certain lengths of cable, connecting the two movable dishes, could not be buried and would be exposed to the outside temperature, equal lengths of cable were exposed at the telescopes with fixed positions. This conservative approach certainly turned out to be extremely effective, especially in the early days of the operation of the WSRT.
Obviously, less conservative methods to calibrate phase-changes resulting from asymmetry were available some years later. In order to compensate for the difference in path length of a celestial signal to the different telescopes varying as a function of the instantaneous position of the source in the sky, adjustable lengths of delay-cable had to be put in the signal-path. For this purpose 24 sets of delay cables were to be installed in the basement of the control building. Each of these sets of cable had a total electrical length of 1600 metres (the length of the array) and was switchable in increments of 10 metres. The unit of 10 metres was equivalent to one full wavelength at the intermediate frequency of 30 MHz, so that any change of delay-length would cause a negligible phase change in the interferometer signal.
Development of the entire electronic system of the new telescope was done partly by the group at the Leiden Observatory under direction of Jean Casse, originally one of the engineers appointed by Belgium, and partly by the staff of the laboratory in Dwingeloo under direction of Lex Muller. The Leiden group developed the IF-system and the correlator, the Dwingeloo group the 21-cm parametric amplifiers for the frontends and the LO-system.
Casse and Muller published an article on the 21-cm receivers in 1974. The production of most of the electronics system was performed almost entirely by the Dwingeloo laboratory, the number of staff of which was rapidly increasing between 1966 and 1969. Completing the twelve frontend receivers, with uncooled parametric amplifiers, turned out to be a major effort: the laboratory staff was not yet used to `mass-production'. In March 1969 the first two telescopes, one interferometer, were equipped with receivers. It would take another full year before the entire array had receivers.
A contract was placed with Philips to deliver both hardware and software for controlling the positioning of the telescopes. Essentially, this meant modifying an existing system, designed for two-dimensional control of a lathe. Of course, it also implied controlling 12 two-dimensional systems (telescopes) simultaneously. In addition the control system would also control the switching of the delay-cables and the stopping of the interferometer fringes for the centre of the observed field, so that the sample rate of the data could be limited.
Reading out the 160 correlator channels (20 interferometers, 4 polarisations each with a cosine and a sine output) every 10 seconds was sufficiently fast. The 10-second samples were averaged over 30 second intervals before writing them to tape. Applying the control system, originally designed for a lathe no more than a few metres removed from the controlling computer, to radio telescopes, some of which were almost a kilometre away, turned out to cause some rather serious problems, which had to be solved after installation.
Also, designing and writing software for a fairly complex astronomical instrument turned out to be a major challenge to industrial software engineers. Fortunately the software, running on a Philips P9202 computer (a modified version of a Honeywell 816), was sufficiently well structured that even some of the fundamental design errors made could be solved during the debugging period. At this stage Raimond got intimately involved in the testing and commissioning process, which took rather longer than anticipated. Early 1968 he had returned from a period of 18 months learning the ropes of radio-interferometry using Caltech's two-element interferometer at the Owens Valley Radio Observatory in California.
In the nineteen-sixties a mainframe computer was required to process the data from the telescope, to calibrate them, and to perform the complex Fourier transform, required to turn twelve hours' worth of observations (about 0.5 Mbyte, a large volume of data in 1970!) of the WSRT into a sky image. Wim Brouw, astronomer at the Leiden Observatory, designed and wrote all software for this purpose for the IBM 360 computer of the Leiden University.
The principles of operation, the performance and the data reduction were later described by Högbom and Brouw (1974). Once the telescope was in full operation (spring 1970), a tape with data was mailed from Westerbork to Leiden every few days to be processed. In those days the processing of the Westerbork data took a major fraction of the available computer time in Leiden (cf. Brouw, this volume).
The total cost of telescopes, buildings, cables, electronics, computer hardware and software amounted to approximately 25 million guilders, not very much less than earlier estimates for the more ambitious Benelux Cross.