Boris Georgievich Sergeev
Edition and translation: Alexander Nitusov
B.G. Sergeev – a talented designer of diagnostic control and modelling systems for digital devices, undoubtedly takes special place among those who contributed to the national computer development. Although computer designers were numerous not so many of them worked with the combined -hard and software- means of control, diagnostics and verification, he was probably the first one on the list.
Boris Sergeev first came to the Institute of Electronic Control Machines (IECM) as a student for diploma practicum when he was studying at the Moscow Institute of Telecommunication Engineers, which he successfully graduated from in 1960. That time the IECM was headed by I.S. Bruk and such outstanding scientists as M.A. Kartsev , N.Y. Matyukhin and some others. Enthusiastic and creative atmosphere of the institute made strong impact on the young man, so after the graduation he remained to work and stayed there until his last day.
It was the time when I.S. Bruk established diagnostics department at the IECM with the main task to develop new methods of control for reaching the highest possible quality level of the computers, designed and manufactured at the institute. The department was headed by Dr. of physics and mathematics David Matveevich Grobman – one of the first soviet programmers, a very talented and educated person. Grobman was also a man of perfect scientific vision, with systematic and analytical way of thinking. Later he was an active founder of the soviet scientific school of modelling, synthesis and analysis methods for discrete devices testing. Being also good at pedagogy he managed to form a team of able young people who successfully solved numerous scientific and technical problems under his guidance. He was very attentive to his younger collaborators and in case of need always provided efficient assistance. One could say that Grobman was helpful and rather soft man, however, he could never forgive an unfair work neither did he anything unfair himself.
Sergeev was probably the most talented and certainly most independent of all Grobman's ‘youngsters'. By the time of his joining Grobman's new department he had already had some practice experience, which he obtained by participation in designing of arithmetical and logical units of the IECM's new computer M-5. That was the reason why Grobman trusted him with development of computers' automatic control technical means, what was not an easy task.
That time the methods of control and failure diagnostics were mainly based on various software applications. However, their basic disadvantage consisted in usage of working electric and electronic circuits of the controlled computer. Therefore, processing of the obtained control data was performed by the controlled computer itself, what naturally lowered reliability of the final test results.
As the problem was obvious, at least for Sergeev, he immediately started development of a special electronic device, which should be directly connected to the controlled circuitry of the computer thus providing much better controlling possibilities. In a short time he designed and put into operation one of the first soviet functional control tester s – a device of control and the failure automatic location in logic circuits. Grobman with his collaborators worked-out set of algorithms and programs, using computer M-2 for that purpose – the second computer of I.S. Bruk and M. Kartsev. They were implemented for generating test signals directed to the controlled digital devices. Finally, both real response signals and etalon reactions obtained with the use Grobman's software, as well as special diagnostic tables (later named diagnostic dictionaries ), served as input signals for Sergeev's tester .
Naturally the tester was far from perfect. Both the low quality of available electronic components and limited possibilities of input-output devices (punch-tape input, small lamp indicator output) essentially reduced its efficiency. However, even then it was an advanced device with such new features as automatic data input/output channels commutation and automatic error code generation for searching in diagnostic tables. Having analysed its operation experience Sergeev formulated for himself two following problems or statements:
- further development -and reasoning- of basic design principles for more efficient computer control automatic testers,
- development -and reasoning- of methods and recommendations for computer designers on improvement of their developed hardware control possibilities.
Solution of these problems was given in his candidate dissertation, which he was working upon unusually long – about eight years. The reason lay in his desire to provide perfect solutions of all problems he set forth and also to prove them with continual working experience of his devices. Besides of that, he constantly implemented new ideas and results appearing and obtained in course of the research and finally, as a person, he never had ambitions just to “hunt for scientific degrees”. He had enough possibilities to do what was interesting for him without higher academic status.
Anyway, his final presentation of the thesis and results was triumphant. M. Kartsev was the chief scientific expert on the work and the numerous positive scientific reviews and expert conclusions were sent not only by various scientific organisations but also by some plants that were producing and distributing his testers. By keeping on with active work he collected impressive amount of research materials on combined soft-hardware control systems within the following years, however he never tried to receive one more degree.
New ideas, which Sergeev presented in his dissertation thesis were realised in his two testers of functional and parametric control. One was designed for industrial automatic control devices in the beginning of the 1970-s and the other one for an aggregate probing complex in the mid-1970-s.
The first of them (AUPK) was used for adjustment of standard computer replacement units. The other one (APK-1) was also used for those devices, as well as for adjustment of another computer family's (SM computers) units.
The both testers were implemented by the enterprises subjected to the Ministry of Instrumentation in Kiev and in Moscow , and also by some aircraft producing plants. Quite logically, they were in use at the IECM itself too.
Both AUPK and APK-1 belonged to the same generation and their architecture was basically similar. However, APK-1 had more advanced characteristics and modern electronic components. It was based on a processor specially designed for solving various kinds of control problems, which was able to directly communicate with diverse control programs sources (punch-tape, ROM, magnetic-tape memory, etc.). The processor controlled circuitry testing, input test signals generating, testing data processing, diagnostic error codes forming, test results indication.
Data exchange between the tester and controlled device was performed through special universal communication channels. Each one consisted of two sub-channels for test and response signals. Their number varied from 64 to 256, depending on the system configuration. The tester normally operated in autonomous mode but could also be connected to a computer.
That time some special tools appeared, which enabled access to all contacts of every integrated circuit on a digital board. Their usage essentially simplified diagnostics. The “needle adapter”, otherwise called “nail-bed”, was probably the most remarkable (or even curious) of them. In fact that was a device with multiple needle-shaped contacts to be connected to contact pads of all integrated circuits (IC) mounted on a board. Reliability of a tester's ‘needle' – IC pad pair contact remained the biggest problem, which -for some prejudices- was considered to be unsolvable in the USSR of that time. However, Sergeev and his assistant, a perfect master-mechanist, A.I. Lomov managed themselves to produce one in their laboratory.
As the rapid growth of industrial computer production sharply increased need in efficient automatic control Sergeev and Grobman suggested, in 1977, an original method based on the manufactured computer “self-control” principle. The work was done by the computer's operating system and special testing system provided with self-control programs. The system included already well-tested and adjusted computer of new release and a set of special controlling terminals, each of them made for testing produced units of certain type. That, basically comparative, method let do without additional development of special control technologies and also provided possibility of real-time (frequency) testing mode, thus saving great deal of time and money.
In 1979 one more automated control system was designed and produced. It integrated control computer SM-3 or SM-4 (‘mini'-machines) and up to 20 functional control terminals, each of them had up to 192 peripheral contacts. Necessary information was provided by a big computer M-4030 using automatic modelling and analysis testing system created by Grobman's team.
Generally speaking implementation of reliable and inexpensive computers essentially increased quality of numerous technological processes. That time it was a matter of special importance in the USSR , where technological development encountered many problems. Sergeev quickly estimated the situation and in the following years all his controlling systems included mini- or built-in microcomputers.
As the controlled digital objects grew more complex, Sergeev was improving his testing devices making them universal. Thus he designed new dynamic tester AMC-0555 for complex functional control, mainly of computer family SM electronic units. The tester was able for performing dynamic functional control with algorithmic and pseudo-random tests and for dynamic combining of those tests into testing sequences ‘individually' adoptable for each tested output channel. The tester performed bi-lateral data exchange with each output, realised every necessary algorithm of asynchronous and synchronous exchange according to type of the tested device interface. It also functioned in interactive mode when the tested object was ‘guiding' process of control and the tester was emulating its environment, realised various methods of the objects' behavioural control, etc.
Same as the previous models that one could receive signals through 192 parallel channels and functioned both autonomously and in the time-sharing mode under control of computer SM-4, when up to 20 testers could be working simultaneously.
The tester was a unique one that time. It was jointly produced by the industrial association including the institute VNITI ‘Instruments' in town of Penza ( 600 km South-East from Moscow ) and by the Special Design and Technological Bureau SA of the ‘Sigma' enterprise in Vilnius ( Lithuania ). The tester was predominantly used for control of electronic units of computers SM-1600 and SM-1700 and also of some special purpose computers.
Practically at the same time Sergeev designed smaller tester AMC-0561 for computer logic units control, which was a portable instrument intended for outdoor computer inspection and maintenance. In spite of its relatively small size it was a universal tool with extended testing possibilities that were not much inferior to the previous bigger models. The tester performed complete functional automatic control in static mode with random tests software-generated upon a given algorithm. The testing response analysis was basically performed by comparison with etalon signals. Software emulation of the controlled objects' interfaces was also implemented.
That small device had 128 communication channels working in both directions. Its operation was controlled by built-in 16 bit computer “Elrctronica–60”. Its software included control programs. It was stored on a magnetic tape and loaded into RAM from a built-in cassette record player, which was later replaced with mini-disc storage. The tester was mounted in a suitcase with sizes of 160 х 360 х 520 mm and weight about 15 Kg . Then that was the only soviet tester with so advanced characteristics. Its design and production was done jointly with the Penza institute VNITI ‘Instruments'.
It was also Sergeev's last control device, after that he turned to development of combined soft- and hardware instruments for digital systems modelling.
Broad implementation of integrated circuits (IC) in the 1980-s drastically limited software-based modelling methods. They turned to be excessively time consuming for the reason of both growing difficulty of obtaining precise information about modelled IC inner structure and extremely big size of the modelling software (e.g. a common model of a big IC could contain more then 10,000 statements). The big software models were also unreliable since the proving of their adequacy was too complicated. Often the information about functional structure was insufficient.
Considering those circumstances Sergeev developed his own –pioneer- method based on combined hard-and-soft modelling (which seems quite logical after its having been accustomed). His modelling system consisted of a real physical sample and so-called ‘programming shell', which replaced a software model of IC. It chiefly performed mediator's functions, providing adequate connection of the real (physical) model with all others both purely soft and combined, respectively to its interface and modelling algorithm. It also analysed authenticity of the physical model's input data. The ‘shells' could be used for storage and subsequent restoration of information about the ICs internal state, what permitted to implement only one physical sample for modelling a device with numerous ICs of the same type and thus saved much time. Generally speaking, creation of the ‘software shells' needed tens times less time/work expenses than making similar programming models of big ICs.
All physical models/samples were collected as special ‘library' of several boards. Both static and dynamic communication channels were engaged for work with static and dynamic big IC and the latter could also store series of input signals. For that purpose they were provided with special memory devices.
By the end of the 1980-s special adapter for the ‘hardware library' was designed and manufactured at the institute. It had 1024 channels for connection with physical samples and its maximum frequency in dynamic channels reached 10 MHz. Buffer memory unit at each channel had capacity of 64 Kbit. Adapter with a set of interface modules enabled its connection with computers of all models. Its basic software included data exchange drivers, a language and compiler for the software shells, special means for finding channel addresses and connecting to them, testing programs for physical samples control and for the shells adjustments.
Advantages of the new instrumentation created by Sergeev's teams attracted attention of many leading computer enterprises (e.g. ‘Almaz', ‘Kvant', etc.). Many production agreements were signed. Besides manufacturing Sergeev should also control the process of implementation of his adapters by the customers. As a new generation of the ICs was appearing at the market there were also plans to implement them for new testers and adapters, what should essentially increase data exchange frequency and also make the instruments still more portable. However, those were already the 1990-s. Political and economic crisis broke out in the country and destroyed all plans. Science and engineering were in depression and the investments were cut practically to the zero level.
Already in the 1970-s Sergeev made his first attempts to apply hardware technologies to intensify modelling of digital devices. Among others he suggested an heuristic method of digital circuitry tests synthesis. It was based on interaction of a proved test object (etalon) with both correct and defect software models of single ICs of that object.
Approximately at the same time he also suggested, and patented, mew method of digital circuits modelling, which was based on programmed commutation of the pre-installed changeable ICs of the tested digital object. Later he made more researches on the ICs automatic commutation based on implementation of integrated switches with electric or optic commutation, however he didn't possess enough time to complete the work.
The 1980-s were marked with intensive growth of development and production of big and super big ICs widely using basic matrix crystals. During the big and super-big ICs development their physical modelling was either limited of simply impossible, therefore modelling methods were efficient and in many cases the only possible proving tools (before the firs big IC of a series was manufactured). That time, modelling software complexes for complex circuitry verification and operation control, were processed by common universal computers. In spite of their comparatively high performance solution of the whole modelling/testing task needed enormously long time. Thus logical modelling (on conditional switches level) of a big IC, which contained up to 10 thousand switches, lasted up to several hundreds hours, and the super big ICs naturally needed much more.
Nevertheless, appearing of new components made possible creation of the super quick processors accelerating the modelling and test analysing hundreds times, already in the beginning of the 1980-s. Some American companies concentrated their efforts on the task and ‘Zycad' corporation became the recognised leading developer of high performance modelling devices.
Sergeev in his turn also launched a ‘research campaign' and, in 1984-1985, formulated basic statements on new soft-hard-ware ‘accelerated modelling' complexes. Among them there were: higher adaptability of the testing complex architecture to character of the solved problems, hardware realisation of all functions instead of the traditional software methods, conveyor principle of the data processing, usage of the algorithm most efficient/quick for conveyor operations, highly autonomous modelling and data transmission with minimum influence of controlling computer, etc.
In 1986 the first Soviet experimental multiprocessor complex for modelling acceleration was produced, which was entirely based on those principles. It was used as peripheral sub-complex controlled by (relatively small) computer of SM series (SM-1420, SM-1600, SM-1700). New complex integrated two hardware modelling conveyor processors with distributed memory: a gated modelling processor for binary modelling on AND-NOT and AND-OR-NOT level, and a functional one for binary eventual modelling both on switching and functional unit level. Experience proved that such complexes provide 100 times higher (!) performance, comparing with available at that time software systems based on big computers ES-1061 and ES-1065. As the Ministry of electronic industry was also among his customers Sergeev got additional possibilities to quickly improve quality of switching modelling processors.
By the end of the 1980-s he produced new instrument called ‘Logical Modelling Device' (LMD), designed as a part of an automatic design systems based on computers SM-1700, ‘Elektronika- 82' and VAX (of the DEC). LMD had modular structure, what enabled its usage as united system with up to 15 processors working in parallel. The circuit being modelled was divided into several sub-circuits, each of them simultaneously processed by an own computing device. Operating data exchange between the LMD's computing units and between them and computing complex of automatic design system was conducted through special processor. Each of the processors was intended for modelling of digital circuit representing 64000 LDM basic components. Those were switches, transistors, resistors, capacitors, flip-flops and logical elements AND – OR of the single or four-digit memory storages with capacity up to 128 Kb. The basic (elementary) functions were given by in-storage verification tables.
Among other, the device also incorporated hardwired algorithm of event-driven multi-valued timing modelling. Modelling alphabet consisted of four signal's meanings, each of four power grades, which characterised the output signal source's conductivity. The follow of events was processed by synchronous eight-steps operation conveyor.
Average speed of modelling reached 1million event/sec, what corresponded to 10millions statement/sec of a conventional universal computer.
For comparison, one could remember that that time an average speed of the event-driven modelling performed by conventional computer varied from o.5thousand event/sec at mini-computer up to 100thousand event/sec at a supercomputer.
Functioning of the LDM was supported by software (authored by M.A. Brodsky), which provided programming-interface between the device itself and an automated design system. It was constantly controlled and monitored by the testing-monitoring software system (author B.M. Basok). Generally speaking creation of the software was special fundamental large-scale project, which integrated efforts of numerous specialists.
In the second half of the 1980-s two new devices were industrially produced. Those were – SM 05.13 with own software and two gated modelling processors and SM СМ 05.13.01, with the software and one gated modelling processor (GMP). By the end of the 1980-s two more devices (LMD) appeared. That was a desk-top variant for interaction with personal computers. It was provided with a software and GMP with operational capacity for 32000 basic elements. Development of the LMD was recognised by broad range of specialists, e.g. by academician V.K. Levin and highly appreciated by the state commissions. In 1990 B.G. Sergeev and other main designers (among them collaborators of “Processor” enterprise in town of Voronezh ) were awarded the Premium of the USSR Council of Ministers.
By the end of the 1980-s some more scientific centres paid attention to the combined soft-and-hardware Automated Design Systems. However, Sergeev remained the recognised leader in that field deeply respected by all his colleagues.
In 1990 Sergeev had contacts with a representative of “Zycad” Co., who was very much surprised to know that such impressive progress had been reached in the USSR in the field where his own company was the world's leader. Besides, Sergeev's systems were entirely based on the soviet-made electronic components.
In 1990-1991 Sergeev was going to launch production of his systems of the next generation. Appropriate investments should be provided for that purpose, however, ruining of the soviet economic structure in the 1990-s, interrupted all his projects. Many progressive ideas of him remained only on paper, for example development of special multiprocessor complexes for digital devices' tests analysis and synthesis.
Sergeev was a person fully concentrated on scientific and engineering part of his work. Although one could never call him an ascetic person, he always despised taking administrative positions and paid no real interest to scientific degrees. The colleagues humorously remembered how much time and efforts it took to persuade him to accept official position of the head of his own department, when D.M. Grobman retired. He was a typical ‘man of faith' in his field.
Nevertheless his fame was constantly growing. A former colleague of Sergeev, who lives in the USA now, told that some few years ago, during a visit to an exhibition there, he noticed a modern special modelling complex. About 64 Gated Modelling Processors were mounted on the single board. When he mentioned similar Soviet development in the accelerated modelling in the 1980-s, the young man exhibiting the system refused to believe it and rejected any possibility of appearing things of such level in the USSR . All of a sudden an elderly American gentleman, who heard their talk, joined the conversation and himself confirmed the story of Sergeev and his achievements.
Differently to his teacher Grobman Sergeev had few pedagogic talents. Nevertheless collaborators loved him because he was friendly and delicate with people and also was skilled in optimal distributing the work among members of his team. Sergeev wanted his colleagues to develop as scientists and loved when they themselves found the work they were willing to do. He may be not always explained new ideas or other things as elaborately and sophisticated as Grobman but he was himself a superb example of creativity and devotion to the work they did. He was very modest in common life but had big scientific ambitions and ‘addiction' to higher standards. He never saw any problem in repeating the same many times, until it was really clear, but could equally easily demand them to re-start almost completed complex project from the beginning, if the results were not perfect or a better solution was found. That happened sometimes, to sincere disappointment of numerous participants who had done their best, especially of the programmers who normally “suffered” of Sergeev's perfectionism more then others. However, nobody was really insulted.
Unfortunately B.G. Sergeev did not live long – he died when he was not yet 59. Quite possible that the crisis that cut his plans and ruined both his country and his work became one of the reasons. All, who knew him, came to his funerals.
- “Virtual Computer Museum”: Russian biography of B.G. Sergeev