Difference between revisions of "IEEE Circuits and Systems Society History"

Line 51: Line 51:
=== The Circuits and Systems Society: some historical remarks about the Original Core Subject Area <br>  ===
== The Circuits and Systems Society: some historical remarks about the Original Core Subject Area <br>  ==
(from CAS Society Directory, 2002-2003, written by Anthony C Davies)<br>  
(from CAS Society Directory, 2002-2003, written by Anthony C Davies)<br>  

Revision as of 11:27, 22 October 2012

March 20, 1951 - First meeting of the IRE Professional Group on Circuit Theory.

March 25, 1963 - Name change to IEEE Professional Technical Group on Circuit Theory

1966 - Became Group on Circuit Theory

November 2, 1972 - Name change to IEEE Circuits and Systems Society.

Are you a member of this society? Please help expand the article by using the edit tab.

Link to IEEE Circuits and Systems Society

Conferences of the Society: 

The first outside the USA

The IRE (then IEEE) Circuit Theory Group developed an international perspective at a very early stage.
Initially it supported several different conferences related to Circuit Theory, but then decided to concentrate support on one Symposium per year (the forerunner of ISCAS).

After holding this event a couple of times in USA, by 1969 there was a wish to hold it outside the USA. A bid in May 1969 from the Circuit Theory Chapter of the IEEE United Kingdom and Republic of Ireland Section to host it in London, England in 1971, was accepted in August 1969.
An announcement in the March 1971 Newsletter of the Circuit Theory Group and the Call for Papers are illustrated here.

                                                                            {to follow shortly}

Some concerns were expressed at the time about the effect of a ‘financial downturn’ on the attendance, and the registration fee was increased slightly to compensate. The editor of the same March 1971 Newsletter (Frank Boesch) wrote ‘….the economic condition of the engineering community in the United States is presently very poor……’
He also quoted some airline ticket costs: New York to London $210, and London to New York $255 – with a possibility of Charter Flights for $175 round trip. Numerically those are comparable to today’s costs, and considering the inflation since 1971, it is evident how dramatically the real cost of flying has fallen.

A postal strike in UK at the time of paper submission and review led to some temporary worries, but the event was judged to be a success and led directly to the policy of regularly holding ISCAS outside the USA which has been maintained ever since then.

Although the London Symposium was organised in the full knowledge of and with some cooperation from the National Society (IEE), some of the organisers feared that ‘retribution’ from IEE might follow afterwards. An indirect consequence was the initiation with the assistance of IEE of the ECCTD series of conferences, still held every two years in Region 8, which some staff and members of IEE hoped would keep IEEE out of Europe – in fact, there has normally been a cooperative association between IEEE CAS and ECCTD.

Anthony C. Davies
(Written in March 2003 for the CAS Society Directory)

The Annual Symposium of the CAS Society

The first annual symposium was held in Miami Beach, Florida, in December 1968. Prior to that the Circuit Theory Group had co-sponsored (apparently without financial involvement) several meetings in the Circuit Theory field: the Midwest Symposium on Circuit Theory, the Allerton Conference on Circuit Theory, etc. A decision was taken to concentrate on supporting only one event per year soon after the Miami Beach event.

The international perspective which characterized the Circuit Theory Group can be judged from the decision to hold this event in London, England in 1971 - a decision in principle to do so must have been taken not later than 1969.

Although the Society has sponsored and co-sponsored many different series of conferences and continutes to do so, the primary annual conferences is the the International Symposium on Circuits and Systems (ISCAS).

The following list gives the date, location and general chairman of ISCAS since its inception:

{to follow shortly}

The Circuits and Systems Society: some historical remarks about the Original Core Subject Area

(from CAS Society Directory, 2002-2003, written by Anthony C Davies)

A Tale of Long Ago

Many of the early members and contributors were passive-filter designers. The design of high-performance passive filters was a very specialized topic, understood by only a few experts, but it was crucial to the implementation of line-based telecommunications systems and quite important in radio communications.

For many years, industry generally used the ‘Image Parameter’ method, which provided an ‘easy’ design route (for which slide-rule accuracy was often enough) but which involved inherent approximations in terms of realization of the frequency characteristics of the ‘design’. The Insertion Loss method, pioneered by Cauer and by Darlington, was able to produce designs for which, with ideal lossless components, an exact synthesis of a prescribed transfer function could be obtained. However, the theory was not easy to understand at the time, accurate and extensive calculations were needed, and with only primitive mechanical calculating machines available, the method was laborious and did not find many supporters, and industry generally neither understood nor made use of the method.

Filter designers were thus often regarded as ‘a race apart’ - engaged in using abstract theories in an almost ‘black art’ of which most engineers had no understanding. Even when digital computers became available to assist in the design calculations, it was at first necessary to use triple-length arithmetic (or more) in order to obtain sufficient precision for useful designs.

Active filters were occasionally suggested, but never used in practice except for very low frequency (sub-audio) applications, such as mechanical servomechanisms - the only available active element was the thermionic valve (tube) which was expensive, unreliable, and required high voltage power supplies.

The beginnings of the Circuit Theory Group were also involved in educational aspects of Circuit Theory - with a strong mission to teach fundamentals of the subject, as opposed to the ad hoc approaches which characterized the circuit-teaching in much of the university electrical engineering curriculum. The subject offered some key advantages: the possibility of an axiomatic approach, rigorous development of a theory uncontaminated by the imperfections of practical components and experiments, and the prospect of formal synthesis - being given a ‘requirement’ and producing, by a step-by-step procedure guaranteed to succeed, a circuit implementation. Although the implementations were ‘theoretical’ ones, requiring idealized linear, time-invariant and often lossless components, there was a real sense in which this represented an alternative to much of engineering practice in other subjects. It also laid a pedagogical foundation, widely believed (at least by the Circuit Theory people) to be a strong contender to be the basis for an engineering education in all disciplines.

The concept of the ‘two port’ (or ‘four-pole, as it was more often called at the time) originally put on a systematic foundation by Feldtkeller in Germany in the mid-1930’s [see R. Feldtkeller ‘Einführung in die Vierpoltheorie der electrischen Nachrichtentechnik’ Verlag von S. Hirzel, Leipzig, 1937], became very important in Circuit Theory. It also laid the foundation for the treatment of transistors as circuit elements.

There was much hope from ‘analogies’ too. The realization that the successful field of electrical circuit theory could apparently be applied just as well to mechanical, thermal, acoustical and other dynamical systems seemed to suggest that electrical circuit theory could become the foundation for many branches of engineering and not only electrical engineering - unfortunately, the lack of good practical implementations of the ‘ideal linear lumped time-invariant’ circuit elements in non-electrical systems severely restricted the extent to which this hope could be realized.

The linear, lumped, passive, finite, bilateral, time-invariant assumptions limited the scope of much Circuit Theory in these early years but nevertheless provided a broad field in which significant fundamental research could be done, and provided the ‘training ground’ for many graduate students and professors.

There were some failures to adequately deal with electronic components - the practice of actual electronic circuit design (at that time involving thermionic valves/tubes) did not conform well to the formalized design processes advocated by many of the leaders in circuit theory, in many cases, it involved non-linearity in an essential way (so that a linear time-invariant assumption was simply not useful) and there was a lack of clear and agreed ideas about how to extend the set of ideal linear passive elements to take into account ‘activity’ in a suitably idealized way to make a formal extension of passive circuit theory. Active circuits were often simply defined as those circuits, which were ‘not-passive’, and little more was said about them.

What are the Real Fundamentals?

The driving point impedance of any linear time-invariant passive system / circuit / network is a positive-real function of complex frequency. Further, if a circuit is constructed from a finite number of linear lumped passive time-invariant components, e.g. from the familiar ideal {R, L, C, M, ideal transformer, gyrator} set, then this driving point impedance is a positive real rational function, for which a formal synthesis procedure is available. Brune [O. Brune ‘Synthesis of a finite two-terminal network whose driving point impedance is a prescribed function of frequency’, J. Math. Phys. 10, 191, 1931] showed that every such rational function could be systematically implemented by a systematic construction (though requiring, in most cases, inconvenient mutually coupled inductive elements or ideal transformers). Finally, Bott and Duffin [R. Bott and R.J. Duffin ‘Impedance synthesis without the use of transformers’, J. Appl, Phys. 20, 816, 1949] were able to use the Richards transformation [P.I. Richards ‘A special class of functions with positive real part in a half-plane’, Duke Math. J., 14, 777, 1947] to provide a transformerless (e.g. R,L,C) synthesis procedure.

These results appeared to be of great significance at the time - especially given the analogies with non-electrical systems - and the lack of practical utility of many of the synthesized circuits was more-or-less overlooked. However, it represented the achievement of an ideal missing in much of engineering then and today.

Starting with a precise, formal (mathematical) statement of the problem to be solved and achieving a realization by a systematic process guaranteed in advance by theory to succeed (in a finite number of steps) represented a major achievement, the importance of which can hardly be overstated, and it was also a philosophy which seemed ideal for the educational foundation of electrical engineers. (Would it not be nice if today’s office-PC software could be designed by such procedures).

Darlington made an outstanding contribution, which must have appeared to many at the time to be of no practical significance whatsoever. He was able to extend the synthesis methods for driving-point impedances to show that any positive real rational function could be implemented as a structure of lossless (e.g. L, C) elements and exactly one positive resistor. Despite the apparent practical uselessness of this theoretical result, relationships between the magnitude of scattering parameters of lossless two-ports enabled this to be related to the implementation of a prescribed magnitude-frequency response as the insertion loss of a lossless two port with resistive terminations. This led directly to a solution to the problem of designing the high-performance frequency selective filters upon which the whole of the analog frequency-division multiplex based line and radio communications industry depended at least until digital technology increasingly replaced them.

Note the absence of any consideration of non-reciprocity in the foregoing - Tellegen invented the Gyrator - a ‘missing’ element was needed to complete the theory for the ‘passive’ domain, in order to have non-reciprocity without activity. The ideal gyrator provided just such a passive device [B.D.H. Tellegen,’ The gyrator; a new network element’, Philips Research Report, 3, 81, 1948].

Subsequently, it became fashionable to try to devise ‘new circuit elements’, many of which did not survive or achieve importance. Among the more abstract and at first apparently useless concepts are a two-terminal element for which both the voltage and current are always zero and a two-terminal element for which both the voltage and the current are undefined. The description of such elements might cause the practically-minded engineer to check if the cover date of the publication he/she was reading was 1st April, yet these two elements, combined together as a ‘nullor’, represent a practically useful model of an ideal transistor and of an operational amplifier, and have found a permanent place in Circuit Theory.

Graph Theory. The concept of an electrical circuit as a linear graph formed the foundation for much of the theory and a basis for systematic methods of analyzing complicated networks, and as a result, Graph Theory laid the basis for the computer based analysis methods and simulators (such as SPICE) which we now take for granted, and which provide one of the foundations upon which the success of modern integrated circuit technology stands. Kirchhoff’s first and second laws were ‘graph theory based’ but what Weinberg called Kirchhoff’s third and fourth laws were almost unknown, yet provide a foundation for much of the network theory developed (using concepts of trees and cutsets, etc.) which is essential for the systematic analysis of large networks. For example, the determinant of the nodal admittance matrix is the sum of all the tree-admittance products - so enabling all numerical processing to be side-stepped and demonstrating immediately and dramatically the link between circuit topology and circuit transfer-functions. A book by S. Seshu and M.B. Reed (Linear Graphs and Electrical Networks) had a very strong influence on its readership (Addison Wesley, 1961)

Activity: Controlled sources were a natural way of introducing activity into otherwise passive circuits. However, since they were also used in the representation of passive circuits (for example in modeling mutual inductance, ideal transformers and gyrators) they did not offer the convenience of being a distinctive element to be added to the passive set to introduce the concept of activity. Initially, it seemed that the unlikely candidate of the negative impedance convertor (NIC) was going to fill this role.

The motivation for developing active filters was mainly the elimination of inductors (on grounds of their size, weight, and non-ideal properties).

Linvill [J.C. Linvill,’RC Active Filters’, Proc. IRE, 42, 55, 1954 ‘Synthesis of Active Filters’, Poly. Inst. Brooklyn, MRI Symposia series, 5, 453, 1955] showed that by using just a single voltage-inversion negative impedance convertor (NIC) as the active element, any rational function could be realized as the transfer function of an RC-active circuit, and shortly afterwards, Yanagisawa [T. Yanigasawa,’RC active networks using current inversion negative impedance convertors’, Trans IRE, CT-4, 140, 1957] provided a simplified synthesis procedure, using a current-inversion NIC. This stimulated intensive research into Active-RC synthesis using the NIC, and many ideas for implementing such an idealized component using transistors. However, the enthusiasm was soon dampened by the realization that the sensitivity of the circuits was so high that they were almost useless in practice. In the following decade of work with the NIC the principal value seems, in retrospect, to have been in the production of doctoral theses and the launching of academic careers. Very few systems went into actual production as a result of this work!
What was apparently not realized by many from a passive filter background was well-known to most practicing electronics engineers: to get a low sensitivity with only one amplifier, a very high loop gain is needed - and the many inventive schemes to make a highly accurate NIC with two or three low gain transistors were doomed to failure. It was not until the invention of the integrated circuit OP-Amp by Ralph Widlar at Fairchild that a cheap high-gain component became available to implement active-RC filters. It was then mainly the much older Sallen and Key structures [R.P Sallen and E.L. Key, ‘A practical method of designing active RC filters’, Trans. IRE, CT-2, 74, 1955], that survived the transition to engineering practice.

An important breakthrough came with the observation by Orchard [H.J. Orchard, ‘Inductorless Filters’, Electronics Letters, 2, 224, 1966] that the sensitivity to component tolerances in the classical doubly-terminated passive, lossless, LC ladder filters is exceptionally good, especially in the passband, because of the non-negative property of the insertion loss of such a passive structure, and this led to the understanding that imitating this behavior in active and digital filters was a route to getting the low sensitivity needed in practical circuits.
The large silicon area required for accurate resistors prevented successful single chip implementations of Active RC filters, and Switched-capacitor filters were the first practically successful approach to widespread implementation of high-performance filters in silicon monolithic form. Despite the need to distribute high-frequency clocks around the chip, they found their way into many real systems.

Digital filters were an inevitable development, although for a long time, their practical implementation was severely limited for real-time signal processing over the frequency ranges needed by communications systems. Although there were many attempts to implement digital filters in integrated circuit form, the development of the TMS 320 series of DSP chips by Texas Instruments was a major stimulus to converting these ideas to widespread use.

The usefulness of wave digital filters may sometimes be questioned, but the theory developed by Fettweis [A. Fettweis, ‘Wave Digital Filters’, Proc. IEEE, 74, 270, 1986] showed the unanticipated result that concepts from classical network theory (including passivity) could be transferred into the field of purely numerical processing of data, and that classical network theory did, after all, have something important to contribute to the emerging field of digital signal processing. In June 2001, the company Infineon Technologies AG celebrated the delivery of 50 million units of a subscriber-line filter product, each one of which contained several wave-digital filters

Non-linearity has not been mentioned so far - it was often considered unwelcome, to be avoided, either by pretending it was not there or by modifying designs so as to minimize its effects. A perfect world was often assumed to ‘linear, lumped, finite, time-invariant, passive, bilateral’, and anything falling outside this was regarded as unwelcome and harmful. It took the recent developments in dynamics and the discovery of chaos and fractals to demonstrate more widely that reality is non-linearity and that the real world is non-linear in a way that engineers need to understand and to exploit.

The rest of the story is not history, it is going on around us! A message embedded within the story is that what looks like useless theory today can often turn out to be an essential foundation for tomorrow’s technology.


to Ben Leon, for the useful material that I extracted from his ‘History of the Circuits and Systems Society’ published in the Centennial Issue (December 1983) of the IEEE Circuits and Systems Magazine (vol. 5, No. 4) - to Sean Scanlan for his observations about the technological significance of Darlington’s results on the synthesis of RLC networks, and to Tom Wehner for filling in a lot of data and putting it all together (without which it would have remained in a partially completed continue producing updated versions.

                                                                                                                                                                                     Tony Davies

More Historical Moments

An extract from the Minutes of a 22 March 1956 meeting of the AdCom of the IRE Circuit Theory Group:

“…. only by taking in more fringe areas (e.g. Transistor Circuits) can we really obtain more members..” “…consensus … not to have a membership drive…”

This extract appears to indicate that the AdCom members did not consider that the theory or design of circuits containing transistors was either important or within the real scope of Circuit Theory.
In view of subsequent developments in electronics, the description of Transistor Circuits as a ‘fringe area’ of Circuit Theory seems rather quaint, and certainly an indication of not foreseeing the future.