Presentation at the BCS Displays Group, 1991
The origins of computer graphics get less clear as each year passes. This paper will attempt to give some information concerning the early attempts to produce graphics using a computer showing when particular techniques were in use.
The subject of this presentation is pioneering images in computer graphics, by which I assume is meant the first appearance or early activities in particular categories or areas of what is now known as computer graphics.
An immediate problem is defining computer graphics. Looking at any standard text book today and you will read statements such as 'computer graphics can be said to have formally started with the work of Sutherland in 1963' [1]. There are two problems with such a statement. The first is that it is unlikely that the term computer graphics was in use in 1963. The second is that it ignores any activities that predate 1963.
In June 1958, the Engineering Drawing Division of the American Society for Engineering Education changed its name to the Division of Engineering Graphics and this started the ball rolling at least in terms of using graphics for what was being done rather than character, line, curve, graph drawing or plotting which tended to be the terms used previously.
I do not believe there is a definitive statement as to when the phrase computer graphics was first used. Both John Vince and I are of the opinion that it originates from the Boeing Corporation around 1964 or 1965. It is certainly no later than 1965 as Bill Fetter from Boeing had a book published in 1965 that has computer graphics as part of the title [2]. Much, if not all, of the literature prior to 1965 uses other phrases to describe the activity.
At a Computer Museum event in Boston in October 1983, Andy van Dam stated that computer graphics was such a new field in 1964 that it was difficult to find even 30 people to support a new computer graphics association called SICGRAPH.
In the event, the first SICGRAPH Newsletter appeared in November 1966, the first major international SICGRAPH Session at an ACM National Conference in 1970 and the first SIGGRAPH Conference in 1974. Consequently, to the uninitiated, it might be thought that computer graphics has existed for no more than 25 years. What I would like to show is that practically all the tools and techniques in use today were in active use more than 25 years ago and that SIGGRAPH was only formed after all the important activities had taken place!
When did the use of computers to draw graphics first appear is a difficult question to answer. We may never know but it is likely to be a great deal earlier than is anticipated. I can certainly find examples in the early 1950s. We may well be past the 40th anniversary of computer graphics already! For example, at MIT Lincoln Labs, the Memory Test Computer (MTC) [3] was in service in 1951. It fulfilled a number of roles including research into input/output devices including microphone and television input and was equipped with charactron tube displays. TX-0 was installed in 1955 with similar peripherals and had a lightpen attached to the CRT. Larry Roberts wrote a program for it to recognise hand drawn characters. Ivan and Claude Shannon wrote a maze searching program. Doug Ross was also involved. SAGE (Semi-Automatic Ground Environment) [4] was the pioneering computer-based air defense system that was initiated in 1949. Lincoln Labs was established in 1951 to undertake the long-range developments. Whirlwind and TX-0 were computer systems produced as part of this project. The initial aim was to install at each of 30 sector defense centres a pair of AN/FSQ-7 systems derived from the earlier prototypes. The AN/FSQ-7 generated 200 different displays every 2.5 seconds with about 100 operators able to select subsets of these displays for analysis. Each operator had a 19in Convair charactron tube and lightpen. The first centre was in operation in 1958. In total, 24 systems were installed on 8 sites. Of these, 6 were still in use in 1983! Each system weighed 250 tons and used 3000 kilowatts. The individual operators tracked aircraft movements across the USA air space and were the main defense system for many years.
In 1955 [5], the Australian Weapons Research Establishment in South Australia jointly developed with Elliott Brothers in the UK the 403 computer. Connected to this system, called WREDAC, were 4 digital plotters. These were modified fax machines producing 11 inch square plots with a resolution of 40 dpi and an address range of 1024 × 1024. This was the first use of reasonable quality raster graphics to my knowledge.
In 1958, there was a whole range of plotters on the market mainly controlled by paper tape or punched cards. In 1958, Burroughs announced an electrostatic plotter with a resolution of 360 across the page. The CALCOMP 565 drum plotter came out the same year. The speeds of these plotters were not always impressive with drawing speeds being in terms of single vectors per second. When I arrived at AERE Harwell in 1959, there was a Benson-Lehner plotter installed which required paper tape input. Lines could be drawn between points specified on the tape. The plotter wandered off course if the increment between points was large so that care was needed not to make too large a vector movement. The accuracy was not very good, drawing a circle with a number of line segments inevitably produced a circular curve where the first and last points did not coincide! It had no character font for it and I designed one in 1960 mainly so that I could get my name on the output to avoid it being returned to somebody else.
The early days of computer graphics was dominated by a single company - Stromberg-Carlson [6]. The number of plotters world wide prior to 1960 being used for computer graphics was probably a few hundred. Even by 1964 the number of CRT displays for interactive use was no more than 100. In this climate, Stromberg-Carlson had a range of products based on its unique CHARACTRON shaped-beam tube which produced almost all the world's graphical output from 1956 until the middle of the 1960s. The CHARACTRON shaped-beam tube had a set of character templates installed in the electron path between cathode and anode. The beam was extruded through the character template so that characters could be displayed on the phosphor at the rate of one per beam movement, it was not necessary to draw the individual character strokes. The quality of the output depended on the quality of the mask. It had a lot of similarity to the later shadow masks. The speed of character drawing meant that Stromberg-Carlson devices could generate characters much faster than conventional lineprinters. By 1959, Stromberg-Carlson had a whole range of products based on this one concept. The SC1000 series had a 19in monitor used mainly as the operator's debugging and control console. The SC2000 series of Bright Displays coupled to a Haloid Xerographies system gave large screen projection facilities. These were used in a range of Command-and-Control systems. The SC3000 series coupled to a Xerox printer provided printed output at the rate of 10,000 words a minute. For computer graphics, Stromberg-Carlson had the SC4000 series and, in particular, the SC4020 which could generate 20,000 characters or 10,000 vectors per second.
The SC4020 was introduced in 1956 and until the late 1960s dominated the computer graphics output world-wide. The SC4020 had a small high precision CRT with a 35mm camera mounted above and a paper camera mounted to one side. The beam could be split to produce output on both devices at the same time. A single SC4020 was capable of producing 1.5 million frames per year when the average plotter was having a job to produce 5000 frames per year. One SC4020 could produce more output than all the plotters in the world combined.
The interest in computer graphics and the number of companies that had installed SC4020s had grown by 1962 to such a level that a user group, the Users of Automatic Information Display Equipment (UAIDE), was established and in the 1960s its annual conference was the equivalent of today's SIGGRAPH, being the place where most of the new techniques were presented.
The first UAIDE Conference took place on October 2 to 4th 1962, with 79 people from 39 companies attending. The attendees were limited to users or owners of StrombergCarlson equipment so this give some idea of the number of installed systems, less than 39 but more than 20. As quite a few were installed in sensitive areas, the actual number installed was not easily obtainable. Even in 1962, there was a range of activities. Meissner Engineering, for example, had decided that only 5% of the company's design effort was involved in the creative design function while 85% was involved in the routine drafting activity of drawing lines on paper. The company spent the late 1950s automating that 85% [7].
Most of the software systems in those days expected the application to understand the coordinate system of the device giving, at most, some help in scaling the output to allow more application-oriented definition of the graphics. Even then, it was clear that some modelling of graphical output was useful and independent scaling of the background grid and related graphics was an innovative addition.
PERT networks were the major project management tool of the day and these suffered from the inability to produce significant networks automatically. An early facility was, therefore, to define a virtual image which was mapped on to a set of frames to be abutted to produce a much larger diagram than was available in a single output frame.
Graph plotting was a major part of the workload in the early days and there was good support for automatic graph production using a variety of axes (linear, log, polar) and with automatic scaling and axis annotation.
The Atomics Weapons Research Establishment at Aldermaston installed an SC4020 in May 1963. AWRE had acquired an IBM 7030 STRETCH computer, the world's fastest, to replace the earlier IBM 7094 and the need for an improvement in the output facilities was recognised at the same time. My job that year was to be one of a team writing an optimising FORTRAN compiler for STRETCH to allow the system to be used! The product provided by IBM turned out to be slower in both compilation and run time than the IBM 7094. I worked with the applications programmers most of the summer debugging the compiler and their programs. Animated films were being produced at Aldermaston around 1963 to simulate bomb blasts.
I returned to the Atlas Computer Laboratory in September 1963 and we installed two IBM-compatible tape drives partly to ensure that we could continue to access the SC4020 at Aldermaston. We fell in love with the device and in 1968, mainly through the work of Paul Nelson, we purchased it second-hand (and refurbished) when Aldermaston replaced it by an SC4060. It continued to provide good service for the next 5 years until it was replaced by an FR80 microfilm recorder from Information International Inc (III).
The year 1961 was the start of computer animation. The early SC4020 systems only came with a 35mm camera with no pin registration. This had two major problems. Firstly, not many people had 35mm projectors so that it was difficult to view the output. Consequently, it was necessary to reduce the 35mm film to 16mm before viewing or attempt to use a 35mm editing bench. The lack of pin registration meant that unless the camera was in excellent condition the film jittered up and down on projection, and this was the second major problem. The results tended to be admired less because of the quality than because they could be achieved at all.
Bell Labs was particularly active producing a range of animated films from satellite simulations to program visualisation. Ed Zajac [8] and Ken Knowlton were two of the enthusiasts in that period. Ken Knowlton was particularly interested in producing images with richer content than the line drawing films being produced by others. He invented BEFLIX, a language which simulated a 252 × 184 8-bit grey level raster image on the SC4020 [9]. The system had a set of processors that walked around the mesh making changes dependent on the contents of the pixel it was examining and neighbouring pixels. On this basic underlying architecture, a rich animation language was defined. The system is reminiscent of the current parallel processing shared memory systems, except that it was in use in 1962.
By 1965, establishments like Los Alamos were regularly using computer animation for CFD visualisations using many of the techniques in vogue today [10].
Good quality animation of line drawings had an essential requirement for good registration of images. Stromberg-Carlson eventually added a 16mm film camera to the SC4020, a Voigt camera, but, alas, this was not pin-registered. Consequently, setting up the SC4020 was a laborious process if good quality output was to be obtained. The answer was to mount a pin-registered camera. Sherrill Martin of Joseph Kaye and Company in Boston produced a modified camera mount that allowed a pin-registered Flight Dynamics pin-registered camera to be used and in the late 1960s he would mount this on your local SC4020 when you needed final output. Judah Schwartz of MIT produced a range of Physics and Calculus films for Harper and Row in the late 1960s using Sherrill Martin's camera.
Near the end of its life, Stromberg-Carlson finally added a pin-registered 16mm camera which cured the problem.
There was little colour output in the 1960s. What there was early on was produced by photographic processing and 3 separation prints. Penetron CRTs were in use in a few establishments but not very successfully.
In 1963, Stromberg-Carlson produced a special white phosphor tube and a red, green, blue colour filter system placed between the tube and the camera. This system was installed at Sandia Labs and used to produce the famous Sandia butterfly, a short film clip that was widely shown.
As mentioned earlier, the software available in the 1950s was quite primitive with differential scaling of grid and graph being the only enhancement from working in the device's raster coordinates. The use of homogeneous coordinates for modelling transformations is thought to have been introduced by Larry Roberts in his 1963 PhD thesis, Machine Perception of Three Dimensions.
In the late 1960s, GINO was still describing coordinate transformations by effectively specifying the scaling between user and device coordinates. In the period 1966/67, Frank Sarno et al at the Polytechnic Institute of Brooklyn developed a system called POLYGRAPHICS [11,12] which allowed separate areas of the screen to be identified as regions and named. Each region could be scaled to effectively achieve a set of user coordinates all mapping down onto device coordinates. The individual regions could act as clipping or shielding boundaries for output. GHOUL (the graphical output language for GHOST) started in 1964 [13] allowed a maths space to be defined in a manner similar to POLYGRAPHICS. GROATS in 1968 allowed recursive region definitions with the limits of each region being defined in the same way as a modern system would define a window/viewport mapping [14]. The difference was that new regions could be defined in terms of existing ones which current systems do not allow. Probably the most adventurous system in terms of coordinate definitions was SPROGS which allowed truly recursive definition of window/viewport mappings to any depth in 1971 [15]. No system since has had such a rich coordinate definition system if you discount the modelling transformations in such systems as PHIGS. SPROGS had complex clip and shield facilities far more sophisticated than current systems. It was also the first system to treat sections of animated films as files with the ability to merge two existing film sequences to create a third which communicated constraints on each sequence through an area of shared memory.
The time before the conventional window/viewport mapping was introduced is surprising. On the other hand the lack of progress in coordinate transformations since 1971 is also surprising.
The first use of graphics in particular areas is also difficult to define with any certainty. However, nearly all the main areas and types of output were probably in use before 1970. The examples below give some idea at least of the latest date when specific activities occurred:
These are just a few of the uses of computer graphics over 25 years ago. It was a mature subject nearly 10 years before SIGGRAPH had a Conference. Apart from UAIDE, probably the most important international gatherings were the Computer Graphics Conferences organised by Mike Pitteway and Bob Parslow at Brunel in 1968, 1970 and 1972. These attracted large numbers of attendees which SIGGRAPH did not outstrip until around 1977 or 1978.
One of the greatest areas of interest today is Virtual Reality with the range of anthropomorphic input devices such as glove, body suit and similar devices. The general view is that such devices are relatively new.
In 1967, Computer Image of Denver [19] had the dream of producing real time computer animation of the quality of Walt Disney. They built a number of hybrid Analogue/Digital systems aimed at this goal. An early system was SCANIMATE which allowed artists to directly control the imagery and motion of the animation in real time. The artist donned an anthropometric harness and his movements were followed by the character in the animation. Computer Image were highly successful in the TV Commercials area and produced an extremely interesting set of films for teaching in Navajo Schools.
One project was to be able to produce an animated political cartoon each evening on TV. The idea for the cartoon would be sent to Computer Image in the morning once the days main news was established. Computer Image would generate a 1 to 3 minute cartoon in time for the 6 o'clock news and transmit it by satellite to New York ready for transmission. The viability of the project was established and some pilot cartoons produced but unfortunately the TV companies did not go through with the project. I can remember an excellent one of President Carter 'Singing in the Rain' as the water (national debt) steadily rose until it was over his head.
Computer Image were a major presenter at UAIDE around the period 1968 to 1972 despite having no Stromberg-Carlson equipment as was Burtnyk and Wein from the National Film Board of Canada. There was great interest in these new techniques from the conventional animation industry who also attended. I can remember sitting next to an animator from Walt Disney who had spent the last 20 years drawing Mickey Mouse eyeballs!. He could see this new technology making him redundant yet was still enthusiastic about the potential. In many cases, it would be a year or so later before he saw the results of his work, if he saw it at all.
The ability to produce synchronised sound tracks was of great interest in the late 1960s. As the sound tracks on the film were just drawings on the sound track area in the case of an optical sound track, people soon started removing part of the face plate of the camera to enable drawing in this area. Los Alamos, Livermore, Sandia Labs and the Atlas Computer Laboratory all generated sound tracks in this way. The Atlas Laboratory also used a computer controlled synthesiser around 1970 to produce synchronised soundtracks. Today we have the recently announced Indigo from Silicon Graphics and the NEXT machine making a great deal of publicity concerning this new advance.
My aim has been to indicate that the history of computer graphics maybe a great deal older than is initially thought. Perhaps techniques have been in use longer than people believe. There may even be worth in looking at these early systems for ideas. Reinventing the wheel is as common in computer graphics as elsewhere.
