Digital modelling

On 21 November 1956, Captain J.M. Armstrong, Defence R&D Representative in Washington, wrote to Lonergan about U.S. sources for some of the computer components which would be needed to expand the SAD computers for the Bloodhound and Blue Jay modelling projects. He concluded; "My visit to White Sands was to attend a three day symposium on Simulation. It developed into a slogging match between the proponents of Digital and Analogue computers and as most of the arguments were highly mathematical, I was able to catch up on my sleep. (We had to get up at 5.30a.m.)!

I gathered that the Analogue computer was looked on askance because you could not guarantee its exactness and the errors were liable to be cumulative and the digital computer was still thirty times too slow to (allow) real time simulation which was thought to be five years off. Some hope was seen by the people who advocated combining the two."

Armstrong may have snoozed through the more mathematical sessions of the symposium but he had got to the nub of the simulation problem as it existed in 1956 and for some years thereafter. Analogue computers proved to be powerful tools for understanding the behaviour of missiles but when it came to really accurate measurement of performance then the greater accuracy of the digital machine was needed. Since the analogue machines worked in parallel even inaccuracies of 0.1% to 0.3% added up as as problems became larger and involved more and more equations.(23) It should perhaps be emphasised here that checking involves two stages.

The first is ensure that the computer is computing the problem as accurately as possible, ie that the machine has no internal correctable errors. The second is that the model, i.e. the mathematical concept that is supposedly analogous to the real system is correct. Bloodhound 1 policy was that only this second was checked against expensive missile trials, cheaper methods were used to ensure maximum machine accuracy. It was this second check though that made modelling such a powerful investigative tool at WRE. The opportunity to check the model for mathematical accuracy against a digital computer never really existed as the simulation of runs on WREDAC; even allowing the much greater run time, was not feasible because of that machine's complete inability to cope with even essential trials data processing.

Better times for digital modelling?

With the new IBM7090 in sight, SAD began to make plans to experiment with digital simulation. L.A. (Les) Nicholls joined Bill Watson's Assessment Trials Group in January 1961 just before the 7090 was officially handed over. He had been working on simple digital simulations of the ballistic trajectories of near vertical launch upper atmosphere sounding rockets such as 'Long Tom' and the computation of missile and target trajectories and other derived quantities from instrumentation at Woomera for input to the analogue computers in SAD. His first job was to write FORTRAN code for a program which would replicate the fairly modest analogue simulation of the Infra Red Homing Missile Blue Jay.

David Strahle and his team had just finished a massive program of analogue computer simulations of the missile to allow RAE to assess the lethality of the weapon. Whereas the analogue computer had had difficulty in reproducing results in runs where inputs were the same because of drifts within the amplifiers, which affected the accuracy of the whole, the digital machine did not have the same problem. Nicholls wrote the code and was producing results for comparison with those obtained by the analogue machine in about three months. However, he found difficulty in achieving accurate solutions for some of the system differential equations having markedly non-linear coefficients with step-by-step Runge Kutta integration methods unless very small-time steps were used. This stage was reached just as the overall Blue Jay modelling work was being wound up but the results had clearly shown the potential value of digital simulation.

To get some idea of the enormous number of calculations performed during the process of modelling the larger weapon simulations it is worth quoting the Brisbane Telegraph which carried a small story when the IBM 7090 was officially opened. The Telegraph said "On a typical mathematical problem - -solving three simultaneous equations each with ten variations -- a university graduate would have to labour two years to do it entirely by hand."

The Telegraph then went on to point out that the new computer did the same amount of work in one second. The amount of digital computer time spent on modelling work varied of course but, at its peak, went as high as 400 hours a month of 7090 CPU, or Central Processor time. The idea that any one group could have available some three million mathematician years of calculating power every month must have seemed like a fantasy to the guided missile designers of two decades earlier. Just how the modellers went about using this enormous new resource is told in the section dealing with the modelling of Bloodhound 2 and Thunderbird 2.

The next generation - - Bloodhound 2 and Thunderbird 2

Bloodhound 2 and Thunderbird 2 were Britain's second-generation surface-to-air guided weapons ordered by the Air Force and Army respectively. While they were superficially similar to their first generation predecessors, inside they were very different. For example, Bloodhound 2 had a much longer range and had several different modes of operation to cope with different tactical situations. Their guidance systems were no longer classical pulse radars, but were now Continuous Wave (C.W.) Doppler, giving them far better performance against low altitude targets and much better immunity to Electronic Countermeasures (ECM). Indeed they were designed to operate in an Electronic Warfare (EW) environment.

Electronic Warfare was first used extensively in World War 2 when allied bombers dropped aluminium foil, called window or chaff, to confuse the German air defence radars. EW has many facets and by the early sixties a whole new military discipline had grown up around it. This postulated a continuous battle for tactical advantage between ECM and Electronic Counter-Counter-Measures (ECCM). Thus the new weapon systems had built-in ECCM features. Notice that the emphasis was now on the weapon system - not just the missile.

Alex Biggs still led the Bloodhound team. "We in SAD had to learn this new discipline fast and not just superficially but in sufficient depth to write accurate mathematical models. Australian scientists were quite used to being isolated by geography from our Allies, now we were isolated by security in a way we had never experienced before. Again we were quite used to safeguarding the military and scientific secrets of our Allies, but now, even within our close-knit team, we had to learn to live with security barriers. Not much more can be said about EW - the weapon systems we are discussing are still in operational use in several nations.

"When we became privy to the details of these new guidance systems it soon become apparent that more research was necessary if we were to model them with some hope of success. A special guidance laboratory was set up within SS(1) group SAD and staffed jointly by the UK and WRE. Such was the importance to the UK of this work that they sent some of their foremost guidance specialists to the laboratory. Don Wooller, the team leader, from RRE, Wilf Robinson from Ferranti and an engineer from BAC Stevenage, while WRE provided John Highfield and some technical support. One Technical Assistant deserving special mention was Eric Durban - ex-RN but otherwise unqualified he was gifted with 'green fingers' in the laboratory. Everything he built worked and worked well.

"We also received support from the UK of a different kind. Tom Lawrence had said to the UK weapon contractors, perhaps to indicate we had nothing to hide from them, "You may look over our shoulders whenever you wish". They took him at his word and Frank Calcraft, who from UK had been a strong source of support to us during Bloodhound 1, and Betty Cook, both Engineers from BAC Filton, took up residence in SAD. Their terms of reference, which were accepted by WRE management, certainly gave them the right of access to our files and staff but the arrangements were by no means reciprocal and they were responsible not to WRE management but to BAC Salisbury. While the presence of people of such calibre certainly boosted the total effort at WRE we lost the advantages of a closely knit team which had previously been a source of strength. We also had to get used to seeing accounts of our latest research in BAC reports before we had published them ourselves.

"Although we were not so sure about it at the time, we were about to enter the age of digital simulation. The IBM 7090 computer acquired in 1961 was 70 times faster than WREDAC. However, it was still much slower than AGWAC which simulated missiles in real time. We could live with computations performed at up to ten times real time, but not much slower. Initial estimates of computing speed and problem size placed us tantalizingly close to this goal. On the other hand the new missiles were so complex we would need to more than double the size of AGWAC and ARTVS. We decided to hedge our bets and the two analogue computers were expanded as described in the following section.

"To get the highest possible speed out of the digital machine we launched a new program of research. With the 7090 we had acquired a marvellous new high level language called FORTRAN. But even after we became proficient in FORTRAN the programming of the 700 odd simultaneous equations was a huge task. Remember that professional programmers hardly existed in those days. A small team in SS(l) group, led by Peter Benyon with Jim Henry and Stan Parkhill, developed one of the world's first languages designed specially for simulation and control engineering. It was called SIMTRAN. But the speed of the machine depended critically on the algorithm and step length used for integration. Recall that in the real world motion is usually smooth and continuous - the digital computer has to approximate this with lots of tiny steps. If you make the steps too long you get errors and even instability. If too short the computation is too slow. In 1962 Les Nicholls investigated the stability of key integration methods, such as the Runge-Kutta, using the stability map approach of the U.S. mathematicians Gray, Gurk and Rubinoff. Peter Benyon investigated dozens of different algorithms, so we knew which were the best for our purposes. Benyon subsequently published his findings in the US Journal "SIMULATION" (24).

"One of the good features of analogue computers was that they were very "user friendly"; that is to say close human interaction in the problem solving process was easy to achieve. During model validation it was usually necessary to alter the parameters of the model frequently in order to match flight trials results. This was a trial and error process in which all the knowledge and skill of the modeller could be brought to bear to achieve quick and accurate results. The turn-round time was usually only a few minutes. With the 7090, however, one had to punch a new set of cards, hand them into the computer centre where they were processed in batches and collect the results hours later or even the next day. Harry Newbigin, the digital modelling team leader sought ways of getting the digital machine to do more of the trial and error work. A young Queensland graduate, Scot Allison had recently joined the team. Scot soon came up with an elegant solution to the problem using 'Parameter Influence Coefficients' (25) and Rolf Seecamp wrote the necessary programs.

"One particularly intractable problem was to determine certain in-flight gains in the Bloodhound guidance receiver. Things were happening very fast in that part of the model and it could be checked only on an analogue computer. For the same reason the input from trials was recorded in analogue form. Another recent Queensland graduate, Bill Dickson, conceived and developed a technique which made the determination of the gains a routine affair.

"When a radar tracks a target it does not see the target as a simple point source of reflection. In fact, continually changing interference patterns result from radiation scattered from various parts of the target. These interference patterns, known as glint, affect the accuracy of a homing missile. In a CW system the guidance receiver is also affected by different Doppler shifts. John Moll put order into this chaos with a new mathematical model which correctly reproduced the most important features of the radar echo (26).

Expansion of the machines for Bloodhound 2 and Thunderbird 2

Planning for modelling work on Bloodhound 2 and Thunderbird 2 posed problems. The amount of work that could be carried out on the IBM digital computer was an unknown quantity both from the point of view of the machine's utility and the amount of time that would be available on it for modelling work. As it turned out, after the cancellation of Blue Streak, there was ample computer time available. Had the missile not been cancelled the situation would have been much different. SAD had little choice but to plan an expansion of the analogue machines that would be sufficient to cope with the expected maximum workload. They would need to be increased in size to cope with the more complex models but there was also a need to make them more flexible and to take advantage of new technology, partly new equipment but also that device that was revolutionising electronics, the transistor.

Much of the planning and design work was centred on John Dunne's Development Laboratory within SAD but Dunne's friend and fellow chorister Don Bennier led an engineering group attached to SAD. There, a young electronics engineer G.V. (Graham Dunne) and a drawing office under the supervision of Don Ward undertook much innovative design and development of circuit boards and other items and oversaw workshop manufacture of the equipment. The arrangement worked well, some of the items needed were built in the laboratory with the assistance of the engineering group design and drawing office staff. Much, including the new kinematics and the dozens of amplifiers and their associated equipment, went through workshops under the scrutiny of the engineers closely associated with the design of the expansion.

ARTVS doubled in size from 120 amplifiers to 234. Rather than discard the 120 Gilbert designed amplifiers they were refurbished and installed in the 'A' side of the computer. Most of the work was to go into the completely new 'B' side. This used the standard WRE amplifier design but fitted into a chassis which attached to standard PMG shelf jacks, chosen because they offered a sturdy and reliable, if bulky, connector at a time when connectors were often both bulky and unreliable.

Two module types were designed, one for adding, one for integrating but both capable of plugging into any position. As well as on-board computing components each amplifier was fitted with a socket connected to all inputs and outputs and supplies so that it might be used for the direct simulation of complex transfer functions. Additionally, on the adder modules, two further sockets were fitted so that etched circuit diode function generators could be used. Only relatively low impedance points appeared on the patch boards which enabled the use of relatively cheap unshielded boards.

The modules were designed for efficient workshop production with the minimum of metal work and arranged so that wiring was assembled in looms before fitting to the units. (photos of various units, racks etc. may be found in SAD 141) During the earlier expansion it had not been possible to buy much of the necessary equipment, particularly from non-dollar sources but now suitable commercial equipment was available.

ARTVS was fitted with Solartron servo-multipliers and EMI Ltd electronic multipliers. The amplifiers were fitted with the normal electro-mechanical choppers and drift stabilisers connected to an overload system which would show both general and particular amplifier overloads, something that was essential when an unforeseen overload could possibly invalidate results. The designers took advantage of the new transistor technology to eliminate most of the slow mechanical relays from the machine. Those on the modules were retained to provide isolation but digital transistors replaced them in the timing and control circuitry with consequent improvement in accuracy and repeatability.

Such improvements were not achieved without teething troubles. The sharp rise times of digital pulses means they may easily find their way into very sensitive analogue circuits, so particular attention had to be paid to isolation and earthing. Here the close liaison between scientists and engineers proved invaluable and it is a fair comment to say that SAD led the rest of WRE in the successful introduction of transistor technology at the time. It is probably also fair comment to say that the designers may also have been in advance of their colleagues who were to use the machines. Extensive digitally controlled monitoring and measuring facilities were also installed, to be abandoned when only part of ARTVS was required for modelling work. The refurbished Gilbert amplifiers were unneeded for instance because much of the work could be performed on the IBM 7090.

Improvements were also made to the Elliott built AGWAC, primarily in building a similar controller and in installing new and more compact Hollerith patch boards to replace the old telephone jack system. As problems grew in size it became increasingly necessary to be able to set up interconnections on removable boards so that, once connected and checked, they could be maintained without alteration while other parts of the model, or even different models could be worked on without the risk of introducing connection errors.

The old WREDAC to Analogue Converter was replaced by two IDACs, IBM Digital to Analogue Converters. These read a different tape format and provided 20 analogue channels or could even be used directly as digital multipliers or dividers when fed into the amplifiers. They provided the essential link between trials and the analogue modelling work. Dunne and Bennier would have liked to have introduced a digital computer to allow hybrid computing as foreshadowed in the 1956 white Sands symposium but these were still very expensive and normally worked in batch mode, unlike today's multi-task, multi-user machines. The IDACs were the next best way of using the precision of the digital computers coupled with the speed of the analogue machines.

The rebuild of ARTVS and AGWAC would also have allowed them to be combined, together with the three dimensional kinematics equipment in the electro-mechanical annexe, to operate as one large installation. For those interested in such things the statistics of the complete installation were:

Amplifiers: 485 general purpose, 200 axis transformation

Electronic multipliers: 56 "x" inputs, 92 "y" inputs, giving 92 "xy" outputs

Coefficient potentiometers: 736 10 turn helical type

Servo-multipliers: 18, driving a total of 194 linear potentiometers.

Servo-resolvers: 7 sine and cosine resolvers driving a total of 84 sine and 84 cosine potentiometers

Input equipment: Two 20 channel digital-to-analogue converters.

Output equipment: Four 8 channel hot-stylus recorders, two 4 channel ink-on-paper recorders, 2 single gantry two pen plotters

TTCP and SAD morale

The Tripartite Technical Co-Operation Program is one of the world's more exclusive clubs. Originally set up to promote interchange of technical information between defence scientists in the United States, UK and Canada it offers a chance for them to communicate with professional colleagues in allied countries and to bridge some of the restrictions normally brought about by the secret nature of their work.

When the first meeting of the TTCP Panel dealing with "Guidance and Control Techniques for Battlefield Air Defence Missiles" took place at RRE Malvern, between 13 and 21 July 1966, no Australian scientist was present. WRE first learnt officially of the existence of Working Panel D-9, as it was called, later that year when Arthur Wills asked Director Woods if he wished to nominate anyone to attend its meetings. Woods inquired and was told there was no one working in that area. It was thought the talks would be hardware oriented and of no particular interest. There had been some disillusionment with other TTCP panels because of severe restrictions in the interchange of information brought about by commercial and military restrictions.

When extracts from the minutes of Working Panel D-9 became available at WRE, Alex Biggs, Principal Officer of System Studies (1) Group in SAD wrote a long memo pointing out that Panel D-9's meeting had covered many areas of interest to Australia and, while the fact was un-mentioned, much of the work presented had been carried out by British scientists while they had been working at WRE within Systems Assessment Division.

UK members A. Smart, F.H. Scrimshaw and J.E. Twinn and A.R. Cawthorne had all been closely concerned with modelling SAGW (Surface to Air Guided Weapons) at WRE and Dr D. Wooller of RRE and W.L. Robinson of Ferranti had been members of SS (1) Group's receiver team. E.G. Hayman and D.G. Strahle, Principal Officers of SS (2) and SS (3) Groups respectively, strongly supported Biggs in his attempt to gain membership of the new Working Panel which was reported as wishing to hold its next meeting in Australia.

Biggs' memo had the desired effect, the previous decision was reversed and he was sent to the next meeting of the Panel held in the USA in March of 1967. The first meeting had agreed to change the word 'battlefield' in the title to 'tactical' to eliminate discussion of defence against ballistic missiles from the objectives of the Panel. Biggs took with him a paper, describing in unclassified terms, SAD's work in the 'Guidance Modelling and Weapons Evaluation' field which included references, among others, to a paper by J.S. (Scot) Allison which had been published in IEEE Transactions in January 1967 (27) and another by J.W. (John) Moll which was subsequently published by IEEE (26).

Biggs reported that the overall picture in the United States was of a great variety of activity in the the air-to-air and surface-to-air field with improvement programs being considered for all missiles and new developments, like phased array radars and fluidics, being actively pursued. He was not impressed by their simulator, modelling and studies work. Their missiles had unexpectedly large miss distances which they could not explain satisfactorily.

They had no counterpart of the flight trial/simulation technique used at WRE for missile evaluation and consequently models were unvalidated and did not really explain flight behaviour. Biggs reported with some amazement that his frequently asked question, "How many millions of data points from flight trials did you use in the validation of your model?" was always answered by, "None". The Americans had recognised the problem and had made attempts to bridge the gap with expensive and elaborate simulations involving actual missile hardware.

Biggs pointed out that the inclusion of missile hardware in guidance simulation had been proposed five years previously by the British and had been rejected by SAD in favour of mathematical modelling. The setting up of a receiver laboratory and study group had been the more difficult option but the decision was now seen as correct and had led to a much greater understanding of the many factors involved in guidance accuracy than the Americans appeared to possess. It was obvious that Australia had much to offer the Panel and Biggs proposed a meeting dealing with simulation be held in October of 1967. Other meetings of specialists on target reflecting properties and control would follow elsewhere.

The meeting of specialists in simulation duly held in Australia from 24 October to 3 November 1967 boosted SAD morale considerably. Those involved in the various modelling groups had the opportunity to present their work personally. Dr William C. McCorkle, Director, Advanced Systems Laboratory, Research and Development Directorate and the US National Leader was particularly impressed and enthusiastic about the Australian work. His comments indicated just how well the SAD teams, in cooperation with RAE, RRE and the contractor firms, had managed to achieve the objectives set out by Chacksfield in 1956.

US security procedures prevented McCorkle sending his visit report to Biggs but he managed to make known his view that SAD had brought the techniques of automatic parameter optimisation to a very high level. This had resulted in simulations which reproduced flight test results in a remarkably accurate and detailed way. The models thus produced enabled the determination of performance boundaries of the systems simulated with a far greater accuracy and degree of completeness than would be possible in any other way except a very large and expensive flight test program.

He was impressed by Dickson's paper which described a novel technique used to avoid the bandwidth limitations inherent in computers to simulate the radio frequency portions of receivers and commented that John Moll's target echo model (footnote 26) was considerably more realistic than those currently in use in most simulations in the USA. McCorkle too, was impressed by the point of Ted Hayman's paper, that flight tests were specifically designed to collect data to validate the model.

During the meeting John Clayton, the UK representative, remarked that the British now required detailed simulations meeting certain standards to be part of contractual obligations entered into by missile system development contractors. McCorkle enthusiastically endorsed this view and said that a properly validated mathematical model should be regarded as important as drawings or specifications and felt that similar obligations should be put on US contractors.

The Americans were eager for some members of the SAD teams to work in US laboratories but senior management people in Australia took a much more jaundiced view. They regarded, no doubt quite rightly, the offers of jobs and opportunities to complete Ph.D. studies as very much a one way street.

The SAD Superintendent, P.M. (Peter) Twiss, said bluntly; "I am not very sympathetic towards assisting the Americans in this or any other area because they have shown no desire to assist us in areas where we know they can -- you will appreciate, of course, that I am speaking solely from the viewpoint of this Division. We have had virtually no response to our many requests for information under our Data Exchange Agreement on Air Strike Analysis, even though we have responded to the limited American requests we have received. There seems to be such an enormous gulf between the energetic good will of high level U.S. Department of Defence officers who have visited us, and the useful access which we subsequently achieve at working level that it is not likely to be bridged by our assisting the U.S. Army Missile Command in one of its establishments, viz. the Redstone Arsenal. So, I find myself asking, "what is the payoff for this Division in assisting the Americans at working level?"

Twiss went on to canvas all the aspects of the matter in his memo to Deputy Director, Space Physics Wing, but it was obvious that there would be nothing much to be gained by sending people to work in the USA and the real risk that staff could be poached although, he added, "I have great faith in the clogging ability of the American security system when it comes to integrating foreign nationals into their defence establishments."

The real payoff came only in later years. Bloodhound and Thunderbird models included Electronic Counter Measures (ECM) and Electronic Counter Counter Measures (ECCM). SAD had ridden successfully with the hounds. when the time came to appreciate the side of the hares as well and establish an electronic warfare capability within WRE/DRCS their skill ensured their entry to further TTCP groups where they could learn as well as contribute by scientist exchanges both ways, but, since that was not part of Joint Project work, it is outside the scope of this chapter.


While the evaluation of Bloodhound and Thunderbird was proceeding the British Aircraft Corporation was developing a low altitude surface-to-air weapon system. Once again SAD was to be deeply involved with the trials and the evaluation of system performance. M.L. (Max) Possingham directed a group within SAD, with key members Brian Ford, Graham Readett, David Lloyd and John Gardner, all of them with experience of weapons systems and aeronautics.

Ford had worked on 'Blue Sky', Readett was familiar with the early VR725, Gardner, like Readett was an aeronautical engineer and Lloyd had been involved with the Blue Steel trials. BAC provided support in the persons of Frank Calcraft, Eric Smith and Betty Cook, all of them experienced modellers. Previous modelling work had been concerned only with the flight phase of missiles. The Rapier Evaluation Study required an assessment of the whole system from the initiation of target intercept by surveillance radar through visual acquisition and target tracking by human operator to the launch and guidance phase culminating in lethality assessment.

This was quite as large a job as the Bloodhound simulation and validation and really deserves a chapter of its own. The involvement of a human operator considerably complicated modelling. BAC at Filton worked on modelling the human acquisition process and John Curtin at Salisbury did some similar work. Optics Group built a tracking simulator and this was integrated with the new 'B' side of ARTVS.

AGWAC was scrapped about 1969 because it was unreliable and the original ARTVS amplifiers were hardly used but the new part of the machine was still useful in spite of the now almost complete reliance on digital computers for modelling work.

The validation of the model from the extensive recordings of missile and the associated ground equipment performance via the telemetering of some hundred channels of internal parameter values during missile firings at Woomera again presented a massive data collection and handling problem. Here the IBM 7090 played a key role in association with the WRE built Data Converters. The collaboration between the UK and Australia was probably at its highest in that the Rapier Weapon System was manned and operated at Woomera by 21 JSTU. This unit included officers and men from the Royal Australian Artillery and the Royal Australian Mechanical and Electrical Engineers in support of their counterparts from the British Army and Royal Air Force. Additionally, Les Nicholls from SAD, had been seconded to RRE in the UK to participate in research and develop of Rapier and the planning of the Evaluation Trials and had then become the Officer in Charge (Scientific) of 21 JSTU to oversee the scientific aspects of the Rapier Evaluation Trials at Woomera. By now it was known that the acquisition radar could not be adequately modelled without accurate data concerning the echoing characteristics of target aircraft. The EMIE Company carried out these measurements at their radio range facility in the UK and the data were analysed, mainly by Vic Sobolewski at Salisbury, to put them in a form suitable for modelling.

Laurie Henschke had begun planning for the new Rapier simulator in 1965. Rapier was, at first, a manually controlled missile, similar in principle, if little else, to the earlier Project E anti-tank missile.(28) Henschke obtained the specifications for the missile performance and essential transfer functions which controlled the behaviour of the target tracker under the guidance of a human operator and proceeded to build, with the assistance of technicians Peter Dadswell, Fred Buttignol and Ted Mahony, a flight control simulator which would look and feel the same to a tracker controller as a real firing at a real target.

A controlling 'joystick', similar to that used to guide the missile, was used to move an optical image of a missile inside the simulator. Tiny models of both planes and missiles were made and by the time-consuming process of single shot photography, movie films were created which closely simulated the presentation and appearance of targets against the Woomera sky. Trials began in 1968 and did not go well at first, but later were very successful. Modelling was, by now, seen as an established aid to design but the results from the model firings and the missile trials did not agree. No simple pattern appeared. The operator success rate for 'hits' was initially poor and, for a time, it appeared that the British Aerospace missile trials might be cancelled, and the resulting savings used for further development to improve reliability of the ground equipment. Army and Air Force alike needed a missile which could be prepared and fired quickly and reliably by no better than average gunners if Rapier was to be an effective service weapon.

But even the best initially seemed to have little. success with Rapier, because of the initial teething trouble with sensitive electronics having to operate faultlessly in difficult conditions. At Salisbury, Henschke arranged for one of the best operators to try to simulate a recent track on the SAD simulator. The combination of transfer functions which involved position, velocity and acceleration, controlled by the 'joystick', was critical to proper performance of the weapon and he ensured that these conformed exactly to the maker's specifications.

The 'joystick' which controlled the tracker during firings was similarly supposed to be calibrated before every firing. This unfortunately was a lengthy process which involved partly dismantling the equipment and hanging weights on the stick to simulate the forces involved and making adjustments until all the control forces were to specification. They made the first firing run. Did it feel right? No, lighter the stick in the real firing. Henschke adjusted the amplifier gains to increase the feed-back forces. Another run, more adjustment until the operator said the stick now, 'felt about right'.

Henschke sent a telex to the firing team leader at Woomera suggesting he recalibrate controls as they were 40% stiffer than laid down in specifications. Subsequently Henschke designed a simple and easy method ofcalibrating the controlling joystick. The result was a shaped lead weight they called the 'Henschke mass' which slipped over the 'joystick' and enabled all the required calibrations to be simply and easily made. Like modelling itself, it was simple and effective once it had been thought of and applied but by no means obvious to those first involved in missile development.

That was not the end of the Rapier story by a long way. Many thousands of simulator runs followed and these were analysed in detail with the 7090 to derive the operational envelope for the missi1e and to enable all its performance characteristics to be tabulated. Overall human engineering aspects were supervised by Ron Speight, a British psychologist attached to SAD.

Postscript (1997)

Computer simulation is now a powerful technique for development and understanding of all kinds of complex systems, for example; climate change, the design of cars, design of aircraft, the study of the economic system, to name a few. But in 1950, when the work described in this paper commenced, such possibilities were unknown. The Cold War, at its height, threatened a third World War in which no part of the planet would be safe; it was as much as anything else, a battle of technologies. This paper describes how a new concept was developed between the early 50's and the late 60's, a concept with enormous ramifications. At the beginning of this period it was thought that Australian scientists would not be involved in the developments of rockets; at the end of the period they had developed a powerful mathematical modelling and computer simulation technique that affected the development of all future weapons and probably many other complex systems as well. This was a period of world-wide developments in computer applications, not the least for simulation. What was different in the Australian work was the attention paid to model validation; the model not only had to be valid, but be seen to be a valid representation of the real system.

Of course, in 1997, we can look back at all that effort and wonder whether the expenditure of money and talent was justified. At the time, we all thought that Stalin, or his successors, would launch World War III, if they thought they could win. With this planet lacking an all-powerful world government there seemed to be no alternative to working with our allies to defend ourselves. In the light of history, nations continue to improve their weapons. Is this progress? In a sense it is. In world war II Germany rained rockets indiscriminately on Britain, and the allies pattern bombed German cities, while the war with Japan ended with frightful nuclear destruction. Contrast this with the Gulf war where the use of precision guided weapons achieved victory over the Iraqi invaders with remarkably few casualties.

23. Digital computers also have errors due to truncation or rounding off but these can be reduced by making the digital word longer so that the least significant bit is a tiny percentage of the number. Also the finite integration intervals in the solution of differential equations result in errors.
24. Benyon, P.R. A review of numerical methods for digital simulation SIMULATION November, 1968
25 Allison, J.S. The Parameter Estimation Problem in Model Checking Proc. Third Australian Computer Conference Canberra, May 1966 also Allison, J.S. On the comparison of two methods of off-line parameter identification. Jour. Mathematical Analysis and Applications Vol. 18, No 2, May 1967
26. Moll, J.W. Calculation of Radar Reflecting Properties of Jet Engine Intakes using a Waveguide Model IEEE Trans. on Aerospace and Electronic Systems. Vol.AES-6, NO. 5 September, 1970
27. J.S. Allison. " The Linear Rectification of Two Amplitude Modulated Signals and Noise." IEEE Transactions on Information Theory, Vol. IT-13, No1, Jan 1967.
28. The Rapier system, as originally designed could only be used in good weather conditions. A blind fire capability was later added through the addition of a differential tracking radar which took over the role of the operator. This system was known as DN181 and WRE also participated in its evaluation. The work on the different Rapier versions spread out over 14 years and the conclusion of the DN181 work brought an end to the association between RRE, now renamed the Royal Signals and Radar Establishment, the private firms and Systems Assessment Division at Salisbury, now renamed the Electronic Warfare Division.