Various techniques for using character and line printers for graphics in which the placement of “X” or “0” or “.” characters could be used to produce very large characters had existed from the earliest days of computing. They were regularly used as the title page of program listings. The technique was extended to the printing of graphs in data visualisation (with for example Jenssen’s weather forecasting work in 1957 and the representation of graphs in the early 1960s) and later to the sophisticated use of overprinting as utilized by Head and further developed by Iain Macleod.
Banners, calendars and other images, “naked ladies”, Michelangelo’s David as the picture for a calendar, the Mona Lisa, etc, that were always in the engineers’ offices and often printed out for Open Days, were as near as most people got to one of the big machines. All the academic computing systems of those days featured strongly in the University Open Days held every year. The Computer was still seen as the “Electronic Brain” and they were still massive hulking machines: the most impressive outcome of modern technology that the average citizen could ever see. Moreover, they were driving and confounding major parts of most people’s lives with their adoption by the Commonwealth and State bureaucracies and large banks and other utilities, let alone the threatened loss of jobs through automation. The injunction accompanying the many IBM punched cards one received with various bureaucratic forms: “Do not Spindle, Fold or Mutilate” led to movies like Colossus: the Forbin Project1 and THX 1138.2
After CSIRAC was decommissioned, in 1964, CSIRO set up their own computing facility, the Division of Computing Research, initially using a CDC3600 housed in Canberra and subsequently developing a network of peripheral machines in the other capital cities over the late 1960s. An example of work done with the CDC3600 that used line-printer output will illustrate the next developments in scientific use of computer graphics. Alan Head, a materials scientist then with the CSIRO Division of Tribophysics,3 was studying the elastic properties of metal crystals, which tend to be stiff in some directions and soft in other directions, i.e., the elastic properties of metal crystals are anisotropic for direction. This applies both in the frictional properties of the surfaces of metals and in the behaviour of metal deformations, and has great importance in understanding metal fatigue and the growth of fractures in metals. Head developed a FORTRAN program called ANCALC (anisotropic calculation) for exploring these differences. He had used CSIRAC to develop the original ANCALC technique, but produced no graphics.4
In the mid-1960s, Head was investigating the electron micrography of defects in metal crystals in order to match the theoretical, calculated data with actual electron micrographs. In the study of metal fatigue in metallurgy, where one needs to know how a cut in a metal surface distorts the crystal arrangement of the metal, the usual procedure is to examine the surface through an electron microscope. However,the problem becomes: how do you know that what you are seeing through the microscope is not some sort of artefact of the electron microscope itself? For example, work by J.K. McKenzie (also of the Division of Tribophysics) had shown that through judicious tweaking of the controls of the electron microscope one could actually produce almost any desired image.5 This meant that a means of calibrating the electron microscope against the theoretical results had to be established. The ANCALC program was perfect for this calibration procedure and Head knew that it could give considerable detail in the numbers and these numbers could be best revealed as a picture.6
The usual method of presenting the results as a profile “of the variation of intensity along a line crossing the dislocation image” led the researcher to have to mentally compare the theoretically generated graph with “a visual assessment of the experimental picture”,7 but Head realised that a more useful aid to interpretation would be to present these profiles as pictures. To get these pictures he used the CDC3600 and its line-printer to produce pictorial output [Fig.1] which he could compare with photographs taken with the electron microscope8 finding that there was a very good fit between the calculated results and the actual micrographs.
In investigating the presentation of these theoretical results as grey-scale pictures Head extended the existing technique of character graphic printing. The character density-surrogate technique does not give a great contrast range for a grey-scale (or “half-toned” picture) unless one uses character overprinting to get a good range of blacks and darker tones, so he developed a technique for printing the data using multiple printed characters to give a greater range of density-surrogates. Head examined the possibilities provided by printing and overprinting the available, mostly non-alphabetic punctuation, line-printer characters and selected 11 different combinations (including a blank space) for his grey-scale. The printout routine was then set up so that the selected character sets representing image density matched the range of values generated in the original profile calculation. [Figs 1 & 2]
Image resolution was set by the number of profiles that could be calculated in an acceptable interval of computer time and was set to 65 for a calculation period of approximately a minute. Given that the line-printer used 132 column paper a resolution of 129 characters across the page was convenient. The extra 64 numbers were interpolated into the calculated 65. The line-printer paper page length allowed 60 rows of characters so this set the useful number of picture points to be represented by characters at 129 x 60. This gave a complete picture on a single page and avoided the gap between rows of characters that would have appeared if the picture were spread across more than a single page of printer paper. By using character density for the contrast, the line printer output gave a picture that, as Head has commented; “if reduced photographically to postage stamp size was very realistic.”9 [Fig.3] Head commented to me that this way of making pictures left the publishers of research papers with the problem of adjusting to the fact that the pictures had no need of further dot-screen half-toning. If the pictures were dot-screened horrendous moiré effects marred the image detail.10
The technique of character overprinting to get good half-toned images was further refined by Iain Macleod, who was a research fellow in the Engineering Physics department of the Research School of Physical Sciences within the Australian National University in Canberra. He was a member of a group studying the possibilities of artificial intelligence led by Professor Steve Kaneff. Among the range of topics they were investigating were a number of approaches to picture processing including picture digitisation, the automated analysis of structure in pictures,12 picture storage and picture printing.13 A major area of use of this work was in the analysis of satellite based remote earth-sensing images from the LANDSAT project.14 Almost as a by-product of this research, Macleod did some work on improving the quality of line-printer grey-scale pictures. He found that by increasing the number of characters overprinted at the same location on the character grid of the line-printer page one could get a significant improvement in the possible contrast range of the image. One of the problems was that, given the conventional character set, transitions from one grey level to the next could be quite noticeable. However, by using a statistical selection from the much larger set of density surrogates afforded by the set of over printed characters, the average density over the region could be made to give a smooth transition over the range of the grey-scale. The technique was, for any desired density in the image, to make a pseudo-random decision as to whether to use the density surrogate lighter than or darker than the average density of the two nearest density surrogate sets.15
The best example I have found of this approach is in a souvenir collection of portraits of Queensland politicians who opened the new Queensland State Government Insurance Office building in Brisbane in 1970.
Brought to my attention by David Jacques, who was the commissioning engineer who installed the ICL1904A computer for the SGIO, these are double sheet images that used the techniques developed by Macleod.16 The portraits were made with the character overprinting technique,while the building profile image uses the characters to indicate not a grey-scale but a straightforward black and white drawing of the SGIO Building. One of the portraits, of the then Premier of Queensland Sir Johannes Bjelke-Petersen, is shown in Fig.5, while the building profile is shown in Fig.4.
As the use of printers became inadequate to the requirements for detail in an image, especially in a graph of complex scientific data, other technologies had to be incorporated into the computing environment to get these higher resolution images. Most of the work simply required line drawings and so the standard input/output device used in the differential analyser18 was pressed into service. The input and output of the differential analyser was handled by a plotting table which, if removed from that environment, can be thought of simply as a plotter. Obviously the plotting table by itself, being an analogue device was not appropriate for the digital data coming out of the computer, but a digital-to-analogue converter could be made to convert the numbers into voltages to drive the motors that moved the pen around the page, or alternatively the numbers could be used to directly drive a stepper motor, which was probably the more accurate solution.
In Australia, at the end of the war, the CSIR Radiophysics laboratory, led by E.G. Bowen, began its transition to civilian research activities. A series of research areas were selected, ranging from radio propagation and radio assisted air navigation, to research in vacuum tube technologies and radar studies of cloud formation and the coalescence of raindrops. Each of these areas was seen to require a great deal of mathematical calculation and the Division thus also became involved in the development of the automatic calculating machine the CSIRO Mk 1 under Trevor Pearcey (which we have covered above). Among the work on radio propagation was research into cosmic radio noise and it was this work that led to the development of radioastronomy research within the division.27
One type of antenna that had been developed exploited a technique called interferometry, and it is this type of telescope that was used to record the data, the visualisation of which we are going to look at here. Since both light and radio waves are simply different regions of the electromagnetic spectrum, that property of waves which produces interference fringes will produce them in any observing instrument that gathers electromagnetic radiation via two pathways of slightly different length from the original source. The slight difference in the path lengths results in the waveforms of the signals being slightly out of phase and thus when they are mixed together again at the telescope “correlator” they will produce interference fringes which can be measured to provide directional information about the source.
The first use of interferometry in Australia utilised an antenna setup on the cliff at Dover Heights by J.L. Pawsey and his colleagues in 1947 to study solar radio noise. When the antenna was pointing at the horizon cosmic radio signals would travel both directly to the antenna and via reflection off the sea surface, thus being slightly different in phase due to the very slight extra distance they had travelled. Two of the junior members of the team, J.G.Bolton and G.J. Stanley, had been using the antenna to look at other regions of the sky as they appeared over the horizon and discovered the first extra-galactic radio sources.28 This was a significant event, establishing a major role in radioastronomy for Australian science, but not part of our story.29
In 1957, Bernard Mills of the CSIR Radiophysics group had caused some controversy by calling into question the accuracy of some of the position information supplied in the 2nd Cambridge Catalogue of radio sources30 and it now became necessary to increase the detail in the observations. Several new telescopes were proposed, the most significant of which were the cross-shaped interferometers proposed independently by Mills for galactic and extra-galactic work and by Chris Christianson for Solar observation. Having tried out a variety of observational antennas Mills realised that the collecting surface area was not so important “in comparison with the spatial extent of the array”31 and he proposed building a pair of long arms lying in north-south and east-west directions in the form of a cross.
“The key feature was that each arm of the cross would produce a fan-shaped beam which could be fed to the central receiver through a switch which, in turn, would join the two beams alternately in and out of phase. Combining the two signals in this way would lead to a relatively narrow ‘pencil’ beam which would be far more effective in resolving closely spaced radio sources than other types of interferometers. The great promise of the cross was that its resolving power would be roughly equal to a parabolic dish with a diameter equal to the length of the arms.”32
Mills got the opportunity to implement this type of observatory at a disused airstrip west of Sydney at Fleurs, and his group built the Mills Cross with 450 metre arms over 1953-54.33
Meanwhile Bowen had proposed a project for a giant parabolic dish, which received its first real support in May 1954, though design work didn’t begin until 1956 and that took more than 3 years. The building of the Giant Radio Telescope at Parkes in NSW was completed in late 1961.34 However by that stage discussions within Radiophysics as to what other radio astronomy work to pursue led to a decision to go into solar radio astronomy with the Culgoora Radioheliograph.35 Both Christianson and Mills could see that none of their style of cross-type extended antennas arrays would be funded for a decade or more and they left the division. Christianson became head of the Department of Electrical Engineering at the University of Sydney and Mills joined Harry Messel’s School of Physics.36
The Molonglo Cross
Messel had already established an astronomy department with Robert Hanbury-Brown, who had been enticed from the radio astronomy group at the University of Manchester. Hanbury-Brown had been interested in extending the interferometry technique for radio wavelengths into visible light optical wavelengths and established the Stellar Intensity Interferometer at Narrabri to measure the angular diameters of what are known as main sequence, or common, stars rather than the especially bright stars which to that point were the only ones for which the angular diameters were known.37
Messel then also took on Mills and they went about finding a site for, and the funding to build, a new Super-Cross telescope. Mills had been involved in the search for a suitable site for Bowen’s giant radio-telescope away from interference and away from Fleurs. In the site selection process for what became the Parkes telescope a number of sites had been investigated. There were three prime sites, one out from Picton, one at Parkes and another at Molonglo on the plain along the Molonglo river south-east of Canberra. When the CSIRO picked Parkes, Mills opted for Molonglo which he persuaded Messel to purchase and, with funding obtained from the US National Science Foundation,38 Mills set about building his new Molonglo Cross telescope which would have two arms each a mile in length, one running exactly east-west and the other running exactly north-south. [Fig.7]
The East-West arm has a parabolic reflecting strip that focuses radio waves received from the sky onto a large number (2816) of short “dipole” antennas set along the very long line of focus of the reflector (which looks not unlike a very long strip heater). The signals from these dipoles are connected to the actual receiver amplifier via cables of equal lengths so that all the signals from the dipoles arrive at the same time giving a point focus. The arm can be rotated on its long axis so that it can be pointed at any part of the sky greater than 35° above the horizon. This direction is known as Declination and corresponds to terrestrial latitude in the sky. The aerial is brought to the desired longitude (known as Right Ascension) in the sky by waiting for the earth to rotate to that point. So the telescope is very well suited to making scans of the sky.
The North-South arm is similar to the east-west arm cutting through it at its centre, however it lies flat on its back and cannot be rotated. Since it cannot be pointed mechanically its pointing mechanism has to be entirely electrical. The trick to focusing a collection of signals from different antennas is that they should all arrive at the receiver at exactly the same time. Since it takes time for signals to travel along a cable, by using different lengths of cable to convey the signals, the time that all signals will arrive can be adjusted. If those cable lengths can be switched between antennas then sets of different lengths of cable can be selected to adjust the timings of each of the signals from the antennas. Molonglo uses eleven sets of cable lengths on its north-south arm and thus can point at (i.e., receive radio signals from) eleven different strips of the sky at once. This gives a reception area made up of eleven beams having a fan-like shape and is known as a fan-beam.
The reception regions for each arm then have a shape as in the diagram with eleven small regions where they scan the same section of sky. By multiplying each
of the eleven north-south fan-beam signals with the signal from the east-west beam a very high sensitivity scan of the sky can be made over the eleven strips of the sky probed by each beam of the fan as the earth rotates. Each strip would be about 15 arc seconds so the resulting scan was of very high resolution. 39 [Fig.8]
Imaging from the Molonglo Cross telescope
The correlated signals were then recorded with a strip recorder developed specifically for the telescope from a fax recorder (as used by the newspapers to transmit wire photos from office to office). Each of the eleven beams, a total power signal and a contour diagram of the radiated power over the whole strip were recorded.[Fig.9]
As Bruce McAdam, one of the radioastronomers working with the Molonglo telescope, told me in an interview:
“So, as you scanned across [the sky], we wanted to draw the profile of that beam as a contour diagram and this is where, I think, our first imaging was performed, here. That was much more spectacular in the extended sources rather than the point source. You could get very complex ones and [when a] source is asymmetrical, so it’s [the stellar source is] elongated.”40
The main project for the Molonglo Cross was a survey of the southern sky, which would take several years to complete. A number of scientists worked on it including Mills, Large and McAdam. Malcolm Cameron did his doctoral thesis with it and I am going to use some images from his thesis to illustrate the process of correlating radio data with visible objects. David Crawford, who had worked at Physics on Messel’s Cosmic Ray Air Showers project and was now(c.1968) at Cornell, was invited back to Sydney by Mills to assist in the computing required to deal with the data coming from the Molonglo Cross. Les Sharp wrote the original radio-astronomy plotting program.41 Cameron wrote the programs for the “production of line profiles and contour maps” and Crawford wrote the programs for the analysis of the data in Chapter 6 of Cameron’s thesis. 42
The task one is confronted with in Radio astronomy is in correlating the locations of sources of stellar radio emission with actual visible objects that one can see in optical astronomy. The position in the sky of the centre of the signal strength contours in Fig.37 can be located from the combination of the elevation (the declination) of the east-west arm antenna and the steering of the fan-beam on the north-south arm antenna (the right ascension), but this does not give a picture of what is in the sky, just a radio-map. To correlate the radio-map with what can be seen with an optical telescope, the radioastronomers had to produce a very accurate contour plot from a large number of scans of some area of the sky. The data from the telescope was recorded to punched-tape and then processed and scaled, to match optical astronomy photographs of the same region of sky, on the KDF9 computer at Basser. The computer results were then plotted out, using the Benson-Lehner plotter, to transparent Mylar sheets with as much accuracy as possible. These transparent sheets could then be placed onto a photograph of the appropriate region of the sky, and the visible objects that corresponded to the radio source could be discovered. The photographs mainly came from the Schmidt telescope at Mt Palomar in California. I will illustrate with Fig.37, taken from Cameron’s thesis.
Some years later the technique was still being used in surveys of the southern sky, only now the resolution of the Molonglo cross was considerably greater and very accurate position data could be obtained. For example see Fig.38, obtained by Crawford in 1982.
Things got interesting if there was no visible source matching the radio emission contours or if they were extended in some way. Several of these cases led to the discovery of some very interesting objects such as pulsars and quasars, but it is not my task to follow that story here. I refer you to Robertson or any of the many other works on radioastronomy that will tell these tales.
Alongside computing, many other developments, from the telegraph through to television 43 and analogue computing (i.e., the representation of real-time processes by the changes wrought in some analogy, e.g., electrical voltages)44 have also contributed to the possibility of video art and video synthesis,electronic music and sound synthesis45 and various other aspects of art and technology such as interactive performance and installation work. Thus several other devices arise from this relationship between the movement of a pen and the x and y-axes of the drawing area that defines the plotter. Non-electronic versions include the Harmonograph and the Lissajous figure. 46 John Hansen was one Australian artist who experimented with harmonograph drawings in the early 70s, which were recorded by John Hughes on 16mm film.47
The other important outcome from the plotter was its electronic realisation in the vector, or calligraphic, display on a CRT. We will look briefly at the development of the CRT as a graphic display device in the next section.
1 [Pearcey, Trevor and Beard, Maston] (1952). Proceedings of a conference on automatic computing machines held in the Department of Electrical Engineering, University of Sydney, August 1951. Melbourne: Commonwealth Scientific and Industrial Research Organization, 1952.
2 Sargent, Joseph (director) (1969) Colossus: The Forbin Project, based on the novel Colossus by D.F. Jones (1966), USA: Universal.
3 Lucas, George (director/writer) (1971) THX 1138, USA: American Zoetrope.
4 Tribophysics is the study of lubricants and frictional factors in sliding surfaces, especially metals.
5 Head, A.K. (2000) “Some Echoes of CSIRAC: Using CSIRAC for Scientific Computation” in McCann & Thorne, 2000, op cit, pp.114-117.
6 McKenzie, J.K, (n.d.) “Materializing Ghosts from Random Noise”, in CSIRO Annual Report 19??, pp.34-37. I have a photocopy of this report which lacks a date and have had no luck in enquiries with the CSIRO on its publication date.
7 As Head wrote to me in a letter dated 4/9/01: “All a situation where, to misquote, “a picture is worth a thousand numbers”.”
8 Head, A.K. (1967) “The Computer Generation of Electron Microscope Pictures of Dislocations” Australian Journal of Physics, vol.20, pp.557-566. 9 Head worked at the Division of Tribophysics laboratory in Melbourne. The procedure for getting his calculations done offers a good example of the transport net used by CSIRO DCR. Data and Program card stacks were flown up to Canberra for running (overnight) on the 3600 and printing on a line-printer and returned the next day with results. A better than 24-hour turnaround condition under which a great deal of scientific computing was done in the mid 1960s.
10 Head, 2000, op cit, p.117.
11 Head, conversation at Melbourne, 9 July 2001.
12 Macleod, I.D.G. (1970a) “Picture digitization via a modified X-Y plotter”, Australian Computer Journal, Vol 2, 1 (Feb 1970), pp.14–15.
13 Macleod, I.D.G. (1970c) “On finding structure in pictures”, pp231–256 in S Kaneff (Ed.) Picture Language Machines, Academic Press .
14 Macleod, I.D.G. (1970b) “Pictorial output with a line-printer”, IEEE Transactions on Computers, Vol C-19, 2 (Feb 1970), pp.160-162.
15 Fryer, J.; Macleod, I. and Smith, D, (1973) “Techniques in automated cartography,” Proc. Seventh United Nations Regional Cartographic Conference for Asia and the Far East, Tokyo, Oct 1973, Item 10, 7pp.
16 Macleod, 1970b, op cit.
17 A note attached to the folder of printouts says “The printouts in this folder were produced on the ICL 1904A computer at the SGIO Computer Centre. We acknowledge the assistance of Dr. Iain Macleod of the Australian National University who provided the data for producing the portraits.” The implication here is that the scans to digitise the original photographs for the portraits were done using the scanner that Macleod converted from a Calcomp Plotter at the ANU [Macleod, 1970a] and that Macleod the prepared the scans using his overprinted character technique.
18 The best reference for graphic art from computers for the 1960s is the catalogue for the Cybernetic Serendipity exhibition at the Institute of Contemporary Art in London, curated by Jasia Reichardt and published by Studio International. [Reichardt, 1968]
19 Hartree, 1950, op cit.
20 Robertson, P. (1992) Beyond Southern Skies. Radio Astronomy and The Parkes Telescope. Cambridge: Cambridge University Press, p.10.
21 Robertson, 1992, op cit, p.10, and see also Ghigo, F. (2003) “Karl Jansky and the Discovery of Cosmic Radio Waves” National Radio Astronomy Observatory, NRAO-Green Bank, West Virginia. Website at http://www.nrao.edu/whatisra/hist_jansky.shtml
22 Robertson, 1992, op cit, p.10.
24 Jansky, Karl (1933) “Electrical disturbances apparently of extraterrestrial origin”, Proceedings of the Institute of Radio Engineers, 21, p.1387. See also Jansky, Karl (1933) “Radio waves from outside the solar system”, Nature, 132, p.66.
25 Ghigo, F. (2003) “Grote Reber and his Radio Telescope” National Radio Astronomy Observatory, NRAO-Green Bank, West Virginia. Website at http://www.nrao.edu/whatisra/hist_reber.shtml
26 Robertson, 1992, op cit, p.14.
27 Robertson, 1992, op cit, p.15.
28 Robertson, 1992, op cit, p.29.
29 Deane, John, (1985) A Picture History of CSIRO Radiophysics, Epping: CSIRO Radiophysics.
30 In more detail: Initial work at Radiophysics investigated radio emissions from the sun. The greatest signals came from the solar corona, which was shown to have a temperature of about 1,000,000 degrees in papers by Joe Pawsey, Ruby Payne-Scott and L.L. McCready covering the observational results and Martyn covering the theoretical study published in Nature in 1946 [Robertson 1992, op cit, pp.40-1]. Interferometry gave Pawsey, Bolton and colleagues the resolution to “see” the relatively small sunspots on the surface of the sun. During a period in which the sun was quiet they used their interferometer to gather data of the size of a radio point source in the constellation Cygnus, which turned out to have an angular size of 8 minutes (0.13 degrees) of arc. [Robertson 1992, op cit, p.46] This was followed by a preliminary survey of the sky which showed sources in six other constellations, but it was not known what the sources were [Robertson 1992, op cit, p.47]. On comparison with a standard star atlas some of the sources coincided with known nebulous objects, one being the Crab Nebula, the remains of a supernova recorded by Chinese astronomers in 1054, and two of the others (Virgo A or NGC4486, and Centaurus A or NGC5128) being extragalactic, however Cygnus remained optically unidentified. Two other researchers at Radiophysics, Bernard Mills and Adin Thomas produced a more accurate position for the Cygnus object and on comparing it with a star atlas found it coincided with a nebulous object, which they were told (by Minkowski at Mt Palomar in California) was extra-galactic and a vast distance away [Robertson 1992, op cit, p.52] For such a distant object to be radiating the energy it was it had to be something at that stage as yet unthought of. Further refinements to its position were made and then some time was spent observing it with the Palomar optical telescope. An analysis of the spectrum of its light showed it to be greatly red-shifted indicating that the Cygnus object was perhaps a billion light years away. For such an object to have the power that it did and to be so far away was considered incredible. This was the discovery of the first Quasar.
31 Conversation with Bruce McAdam at USyd, Physics, 17th Feb 2005.
32 Robertson, 1992, op cit, p.71.
34 Robertson, 1992, op cit, p.73.
36 Robertson, 1992, op cit, pp.189-90, and for a detailed to description see Wild, J.P. (1967) The Culgoora Radioheliograph, reprinted from the Proceedings of the Institution of Radio and Electrical Engineers Australia, Sydney: CSIRO Division of Radiophysics.
37 McAdam interview recorded at Roseville, 26th Feb 2005, and Robertson, 1992, op cit, pp.189-90.
38 Robertson, 1992, op cit, pp.191.
39 McAdam interview recorded at Roseville, 26th Feb 2005, and Robertson, 1992, op cit, pp.188-90.
40 Large, M.I. (1966) “The Molonglo Radio Observatory” in Butler, S.T. and Messel, H. (eds) Atoms to Andromeda, Shakespeare Head Press, Sydney, pp.141-158.
41 McAdam interview recorded at Roseville, 26th Feb 2005.
42 Conversation with Dave Crawford at Pennant Hills – 3rd Feb 2005
43 Cameron, Malcolm, “Radio Observations of Bright Galaxies” – A Thesis presented at the University of Sydney for the degree of Doctor of Philosophy, September 1970, p.iv.
44 See: Crawley, Chetwode (1931) From Telegraphy to Television: The story of Electrical Communications, London: Frederick Warne and Co.,
45 Moseley, Sydney A. and Chapple, H.J. Barton (1930) Television: Today and Tomorrow, New York: Isaac Pitman & Sons.
46 Hartree, 1950, op cit.;
47 Aspray, William (ed) (1990) Computing Before Computers, Ames, Iowa: Iowa State University Press.
Return to Contents