Telecommunications Research in the United States and Selected Foreign Countries: a Preliminary Survey. Report to the National Science Foundation (1973)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

152 COMPUTER COMMUNICATIONS NETWORKS I. Introduction The need for direct communication among computers is the most recent manifestation of society's demand for flexible communications services. In providing for this need, and in performing R&D that will lead to systems adequate for the future, the United States substantially leads the rest of the world. Electrical data telecommunication began with the demon- stration of the electrical telegraph by Wheatstone and Cooke in l837. Nationwide telegraph networks grew rapidly, and the first successful trans-Atlantic telegraph cable was completed in l866. Bell's invention of the telephone in l876 led, by early in the twentieth century, to a public voice telephone network that surpassed the telegraph network in size and use. In l93l AT&T began service on a TWX (teletypewriter exchange) network specifically for low speed (about l0 characters/second) data communications. Data communication by teletypewriter continues to play a significant role in business and, to some extent, personal communication. A revolutionary change in data communication requirements occurred in the late l950's due to improved electronic digital computer technology. With data processing systems able to generate and process vastly increased volumes of data, improvements were necessary in both the rate at which data could be transmitted and in the convenience with which this data could be gotten into and out of computers. Ubiquitous telephone networks were attractive resources for providing data communications services throughout the United States and the rest of the world. The voice network, however, was designed for the trans- mission of continuously changing "analog" waveforms spanning a frequency range from about 200 to 3000 hz. Data, with its binary nature, had to be transformed by "modems" (modulator- demodulators) from the digital form required by the data terminal and computer equipment to the analog form required for transmission on the telephone network. By the late l950's the Bell System's DATAPHONE service allowed terminals and computers to communicate through the switched telephone network or private lines at modest bit rates. During the l960's the variety of modems supplied by the telephone companies grew substantially, and independent suppliers entered the market. By l972 the number of modems in service was about 300,000, with about one-third of these provided by non-Bell System suppliers.

153 In the early l960's much of the interest in computer- computer communication was for load sharing-purposes. This communication was usually off-line and involved tape-to-tape transmission. Increased volumes of data required higher transmission rates, and data sets for "broadband" transmission at 40.8 Kb/s (later 50 Kb/s) became available. Another form of computer access also became popular. Taking advantage of the great speed of digital computers, systems were designed that allowed the computer to switch rapidly from serving one user to another, giving each the impression that he was being served continuously. Remote access to these "time-sharing" systems through teletypewriter and display terminals grew dramatically and commercial bureaus were established to sell such services. Primary access to these systems was made through the switched telephone network. Specialized data processing services were also established along industry lines, for example, airline reservations, hotel reservations, stock quotations, credit card checking, bank operations, and many others. These all make use of the public telephone network to link thousands of terminals to central computerized data bases which can be examined and updated appropriately for each transaction. Several vendors of commercial time-sharing services have established networks employing leased lines to link their computers. CDC's Cybernet, Tymnet, and the G.E. time-sharing network are examples. Efforts aimed at reducing communica- tions costs (i.e., line charges, data set charges, and inter- face costs) have led to extensive use of minicomputers as programmable communications "front ends" to larger machines, and as communications concentrators which obtain economies by multiplexing intermittent low-speed data from many terminals onto a smaller number of higher speed channels. Some of this work is indeed R&D currently in progress. Although there is much effort in this important and rapidly developing field that cannot be included here for lack of space and time, the next two sections will outline a few of the U.S. and foreign projects that relate to computer netting. II. Computer Communications R&D It should be noted immediately that the United States leads the world significantly in R&D on computer communications, and in applications of computer communications techniques. This is hardly surprising in view of our commanding position in the areas of computer and communications systems manufacture

154 and use. Relevant R&D is taking place in many places in the United States, including government agencies, universities, the communications carriers, the computer manufacturers, and computer users. The largest effort directed specifically at computer communications research is the ARPA network. R&D effort for it is being carried out at many university and industrial laboratories supported by ARPA funds. It is an example of what Clay T. Whitehead, Director of the Office of Telecommuni- cations Policy, has called a "value-added network", i.e., a network that utilizes common carrier transmission facilities, but adds its own switching, error control and other special service features. 50 Kb/s private lines leased from AT&T are used to link about twenty large computer installations in all parts of the United States. The ARPA network is making a pioneering attempt to interconnect diverse computer systems, despite the difficult problems caused by vastly different software conventions found in various computer operating systems. In addition, the ARPA network has developed a distributed store-and-forward message switching system to implement the necessary routing and switching func- tions between computers. Analysis and simulation techniques have been developed to optimize network topology and to model the performance of the system. The network design is con- servative in that it uses available technology in its hardware implementation (e.g., standard Bell System 50 Kb/s trans- mission links and Honeywell 5l6 minicomputers for the store and forward switches). IBM also does extensive R&D on computer communications. Several laboratories are involved, including the Yorktown Heights and Zurich research laboratories. Analytical network optimization studies are conducted, as are experiments with switched digital transmission loops. The Bell System is supporting substantial R&D with the objectives of improving performance and lowering costs of data communications services. A significant commercial step was taken in November l972 with the filing of tariffs for a digital data service, an all-digital private line data network that is expected to serve 96 major cities by l976. Substantial economic advantages result both from the elimination of modems and from improved utilization of digital trunks that will be used in the system. Digital transmission employing regenera- tive repeaters will also lead to greater reliability for data communications. Continuing R&D on digital switching and transmission techniques will enhance performance and provide economic savings for both voice and data traffic during the next decade.

155 Several research experiments in computer communications are being conducted at Bell Laboratories. In-house digital systems are being constructed to experiment with minicomputer networks. Both distributed control "Pierce loop" and common control switched systems are being investigated. These systems use packet rather than line switching techniques and are capable of carrying data at rates exceeding l megabit/second. Experiments aimed at developing techniques for effective resource sharing are in progress. Because of the rapid switching and high data communications rates that are possible, new forms of computer sharing and usage will emerge. III. Foreign Efforts Just as foreign countries are behind the U.S. in computer and communications technology, their efforts in computer- computer communications also lag the U.S. by several years. For example, one measure of the use of data communications is the number of data sets in service. At the end of l97l, there were some 300,000 data sets in use in the U.S., compared to about 40,000 in all of Europe. However, in specific cases, good work is being done that overlaps some of the work being done in the U.S. England has the most highly developed computer and com- munications capability outside the U.S. The British Post Office has supported studies of a possible separate digital network for data communications. An experimental service of this type may be operational in a few years. Both industry and national laboratories have been involved. A group at the National Physical Laboratory under D. W. Davies has an in-house experimental packet switching network in operation. They have long been strong advocates of packet switching and digital transmission for data applications. In Canada a limited data network is already in operation. Calgary, Ottawa and Toronto have been linked with a 56 Kb/s transmission facility specifically intended for data. Experi- ments with digital communications loops are also being con- ducted at the University of Toronto. In France a cooperative experiment including IBM and several French universities is being undertaken to develop a software system capable of solving some of the difficult problems that arise in computer networks. Even though the network will only link IBM/360 computers, substantial problems must be overcome. IV. Summary The United States, with its substantial lead (3 to 5 years) in both computer and communications technology, also

156 leads the world in efforts on computer communications R&D. However, the field is in its infancy, and great progress must be made to realize the potential of effective computer- computer communications. Problems introduced by a lack of standards in computer operating system software, in communi- cations conventions, in character sets and usage conventions, and other areas lead to enormous operational difficulties. The large data networks that currently exist usually involve man-computer communication. Examples include time- sharing systems, credit card validation, and airline reser- vations systems. Having a man in the loop means that relatively small data transfers are required and low data rates « 300 b.p.s.) are adequate. In addition, the moderately long time (l0-25 seconds) needed to establish a connection through the voice telephone network is frequently acceptable. Having computers at both ends of the communications loop changes the requirements dramatically. High data rates become desirable. Short-call set-up times (milliseconds) would be useful. Delay in getting messages between machines must be very small, if high performance processors are to be kept busy. End-to-end digital transmission becomes desirable to reduce error rates, to obtain rapid switching and to make full use of available transmission bandwidths. Demand multiplexing and packet switching must be thoroughly evaluated as alterna- tives to present techniques involving line switching and synchronous time division multiplexing. Major benefits to be realized from computer networks arise from the advantages of "sharing". Various forms of sharing can be expected to develop. Load sharing will permit a heavily loaded machine to get some of its work done by a more lightly loaded member of the network. Data base sharing will allow a single data base to be accessed by many machines, providing storage economies by reducing the need for duplica- tion of files. However, many difficult problems concerning privacy and controlled access to files must be solved before large scale data base sharing will be socially acceptable. Specialized processor sharing will allow a unique processing capability, for example high-speed display image processing, to be accessed by all members of the network rather than only by local customers. Program sharing will permit processes in one machine to utilize programs that run in other machines without the need for physically moving the program. Repro- gramming algorithms to make them run in several environments will not be necessary. Another important benefit, although possibly further from realization, will result from new hardware and software

157 architectures that will emerge when machines can communi- cate effectively at high data rates. Simplification of software in each machine may result if a given computer is responsible for providing only one type of service for the whole network. We may see a range of network service nodes developed, each with a single or small number of functions. To increase network processing capabilities, nodes will be replicated or replaced by faster technology. Reliability requirements will dictate the amount of redundancy designed into networks to maintain service. Specialized processors of great power and efficiency will be designed to perform specific functions. Special attention must be paid to standards at all levels. New national data networks should be mutually compatible. The growing number of private networks will one day need to be interconnected. Standards for error control, synchronization and end-to-end coordination are required. Differences in word size, character set, and communication protocol currently hinder useful communication between programs in computers that are already connected electrically, and these and other problems are magnified in networks. In many of these cases it is not simply a matter of legislation. R&D is needed before the options are properly understood. In the near future we can probably expect that new developments will continue to unearth more questions than answers in this rapidly changing and technology-dependent field. With persistent effort we may expect that new and more effective uses of computers will develop as an under- standing of network technology is achieved. E. N. Pinson Bell Telephone Laboratories Murray Hill, N.J.

158 RESEARCH DIRECTIONS IN NETWORKS OF COMPUTERS Most of the research in the area of computer networks in the past have been on the question of interconnecting large numbers of computers in remote locations. The research effort has tended to concentrate on the question of sending bits from one location to another. There are now many solu- tions to this problem, most of which appear to me to be satisfactory. As such I no longer find this problem interesting I personally believe that the most important question in this area is to get a good understanding of what the end users want. Most computer networks have found that their traffic level is substantially below that which was predicted before the network was built. I believe this is due to two things. First, there was insufficient thought given as to what people would really be doing with the network. Second, although the network provided good facilities at the bit level, it was very difficult for the users to actually make use of remote facilities. I think the most interesting question is that of dis- tributed data base. The requirements for this can best be shown by an example. Today IBM does payroll processing in many locations, and, if an employee transfers from one loca- tion to another it often takes a period of several weeks before the payroll records are actually moved to the new locations. If we had a network of computers, we can see this taking place automatically. In a primitive version one would just enter a move statement. For instance if Phil Dauber was transferred from San Jose to Yorktown Heights, a move statement would cause the appropriate payroll records to be copied from the San Jose file to the Yorktown Heights file. In an auto- matic system, even the move statement might not be necessary. All that would be necessary would be for San Jose to stop accessing the Phil Dauber record and the Yorktown Heights system to start accessing it. The algorithm would then deduce that the most appropriate location for the Phil Dauber records would be in Yorktown Heights and hence would copy the record from one place to another. Another question of interest is load levelling. In some hypothetical load level system, jobs would be moved auto- matically from one computer system to another in a network in order to balance the loads at all these locations. Due to the difficulty of moving large online files, I do not believe load levelling makes sense in most commercial appli- cations. However, if it were true that in some scientific job shops a significant percentage of the computer time was

159 spent on jobs which consist of only a small number of cards, then load balancing would make sense. I believe a small study making measurements of typical environments and doing analysis could determine which are the environments in which load levelling made sense. If there were a sufficient number of these environments, then it might be worthwhile developing more sophisticated load levelling technology. P. S. Dauber IBM Corporation Yorktown Heights, N.Y.

160 ARPA NETWORK RATIONALE: A 5-YEAR REEVALUATION I. introduction Since its conception in l968 the ARPA Network has pro- gressed rapidly from a totally undefined hope to a major national computer resource sharing service interconnecting over 34 computer centers. Besides serving as the vehicle for the development and exploitation of multi-computer resource sharing, it has demonstrated the effectiveness and economy of an important new form of communications technology, packet- switched communications. Within the next five years, the extension of packet communication concepts into the areas of radio and satellite communication should revolutionize as well the effectiveness and economy of both local and inter- national data communications. The current ARPANET was not designed for these extremes and is most efficient for interconnecting moderate to large computers or user complexes scattered at a fair density over a nationwide geographical area. Since the future will bring vastly improved techniques for both tying isolated small users to the network and interconnecting networks together worldwide, this evaluation will confine itself to delineating the domain within which the ARPANET is used and for which its design makes it most economic. II. Historical Development The first four ARPANET modes were installed in late l969 after a year and a half of study and development. It was in the initial period (l968) that most of the global design decisions were made; i.e., to use packet switching, inter- connecting small Interface Message Processors (IMPs) with 50KB lines, accepting messages of up to 8000 bits from capable host computers, and breaking these messages into packets of l000 bits or less for store and forward transmission. The initial analysis of network performance and cost, made in late l968 just prior to contracting for the development of the IMPs, has proven to be very accurate and the same numbers are still quoted for para- meters like the end-to-end packet delay (.l sec).

161 The second year (l969) was spent developing the IMPs and installing a four node test network. Since the initial network worked essentially as predicted and it was recognized that at least l5 nodes would be necessary to achieve adequate resource diversity and usefulness, the network was immediately expanded to include the l5 ARPA computer research centers most likely to contribute to the development of resource sharing software. The communication network tying these nodes together was com- plete by February l97l; however, the development of an agreed upon standard for host-to-host intercommunication and the related host software for all the computers was not complete until August l97l. III. Development of Network Usage Although very limited usage of the network existed through- out its early development, it was not until all l5 nodes were compatibly interconnected in mid-l97l that true user activity could usefully begin. Shortly after this point, in September l97l, another phase of network growth was begun: the addition of user nodes without their own major host computer through the use of a new device, the Terminal Interface Processor (TIP) — essentially a small IMP connected to its own minicom- puter terminal processor. The TIP permitted a whole new class of network use, i.e., a group or centers on the network could now look to the network for all its computation requirements instead of operating a local computer service. This strategy for obtaining the optimum mix and balance of computer capability from large, cost- effective computer centers without the many problems associated with running a local center has been so successful that about half the network nodes are now of this type. Also, during the past year several additional host computer service centers have been added to the network as the user requirements for quality computer service expanded. The entire expansion of the network beyond the original l5 research nodes has been in response to requirements (and full reimbursement) from non-network research projects, both in ARPA and other government agencies. Usage of the network has mainly been of three types: remote access to time-shared systems, subroutine-like use of large numerical machines, and file transfer. The remote access use is similar to the use of commercial time-sharing networks; users access the net through a TIP and use one of the many

162 time-shared host computers on the net. For reliability they usually keep their files at several compatible host sites and use whichever host is available. The software and file trans- fer protocol has now developed to the point that, if one is operating on one computer with files stored at another, it is both convenient and fast (seconds) to access the files. The second major usage of the net is to run large numerical processes on large, fast remote batch hosts while using a time- sharing computer for the interactive portions of the tasks (program editing, data input, output processing and display). For many applications, interactive manipulation of the data is essential, but the central task of numerical computation can be run far faster and far cheaper on a large number cruncher host. The cost and time savings are often both more than a fac- tor of ten. The third major network activity is file transfer. Files must be interchanged between host computers quite frequently for the previous two applications. But, in addition, there has been considerable use of the network for the bulk movement of large data and program files in preference to mailing tapes since the network is more reliable, less error-prone, usually cheaper, far faster and eliminates human handling. A six-month Air Force test of the network for the pure movement of data traffic showed that throughput rates of 20- 30KB could be maintained. (Forthcoming changes to the routing technique should more than double this.) The test results also showed a lower monthly cost, while providing at least ten times the throughput and responsiveness of the alternatives available. So it has turned out that this type of network usage (pure data movement) is both economical and attractive, even though the original design goals of the network were aimed at inter- active computing. Such data traffic is useful for the network since it provides a large low-priority background load which expands the instantaneous capacity available and helps main- tain efficient line utilization, thus reducing the cost to everyone. IV. Network Traffic Figure l shows the internode network traffic over the first l4 months of user access. Usage has increased exponentially at the phenomenal rate of 26% per month throughout this period, L

l63 10M r ARPA NETWORK TRAFFIC (INTERNODE) Sept 1971 -Oct 1972 CURRENT NETWORK CAPACITY 1 M 100K Growth Rate = 26%/Month or X 10 per 10 Months I _L JUL 1971 JAN 1972 JUL 1972 JAN 1973 JUL 1973 Figure l

164 while network size has only increased linearly by one node per month. Internode traffic in October l972 was l.36 million packets per day which corresponds to 9% of the computer power in the network being used remotely. In addition to this, the network also handled .45 million packets per day of local traffic, which means that l2% of the total computing power of the network's 22 serving host computers was distributed via the network. On an annual basis the value of this computer time would be $2.l million or slightly more than the network cost. However, at the current growth rate the network should be fully loaded by July l973, in nine months. Since the cost saving incurred by selecting the proper network computer for each problem is usually l00 to 300%, the network is already cost-effective. When it is fully loaded, the network costs only amount to l0-l5% of the computer costs. V. Cost-Effectiveness In order to fully support computer resource sharing reliably and responsively on a nationwide basis, the communications net- work must be approximately the size and cost of the current ARPANET (34 nodes and $2 million/yr). Based on recent measure- ments the network traffic generated by a fully loaded, moderate sized, time-shared computer is 720,000 packets per day. A minimal network such as the current ARPANET has a basic capacity of l0 million packets per day. Therefore, in order to fully utilize the basic capacity of the ARPANET would require l4 moderate size computers to be fully accessed through the network. Additional capacity can easily be added beyond this point, but with no great economy of scale. Figure 2 shows this effect for generalized national networks of ARPANET technology. By utilizing the above measurement of network traffic produced by a host and estimating its rental at $720,000 per year, total network traffic can be related to the total com- puter resource value accessed via the network (each computer dollar produces 365 packets of traffic). Thus, the relative cost of network communication can be related overall to the annual computer value used. Since there are some fixed costs associated with adding each additional node, the cost- effectiveness depends partially on the number of nodes, but the main effect is produced by the dollar volume of usage. This means that at least l0-20 million dollars of computer time usage must be expected before a nationwide network becomes optimally cost-effective. Once this activity level is reached, the main benefit from increased usage is improved reliability and increased peak throughput capability. Lawrence G. Roberts •

165 I? i° 5 # 83 100 80 60 40 20 10 8 COST-EFFECTIVENESS OF NATIONWIDE NETWORKS I I I I I 8 10 20 40 60 80 100 TOTAL ANNUAL VALUE OF COMPUTER RESOURCES SHARED VIA NETWORK (In Millions) Figure 2

166 RESEARCH IN TELECOMMUNICATIONS SWITCHING I. Introduction Within the time allotted a review was made of research work undertaken both in the United States and overseas in the field of telecommunications switching. This review included contacts with overseas associates, and reviews of overseas publications, the minutes of the United States Independent Telephone Association Manufacturers Sub- Committee, and papers presented at the International Switching Symposium in Boston and the International Conference on Com- munications in Philadelphia, both in June l972. In summarizing the results of this review, the remainder of this document will provide suggestions for further research, review the funding of research in telecommunications and switching overseas and in the United States, and review the relative positions overseas and in this country in five main sectors: public telephone switching; private telephone switching (common carrier and interconnect); message switching; telex switching; and data switching. II. Suggestions for Further Research Arising from this study are indications that further research would be desirable in the following areas: Electrical Interfaces Research is needed into miniature electrical interface devices so as to eliminate bulky items such as transformers and other electro-mechanical devices used mainly for isolation in line and trunk circuits. Man-Machine Interfaces Research is needed into application of electronic logic in simplification and functionalization of man-machine inter- faces for operator/attendants' work stations, so as to reduce their decision-making to its most functional form. Electronic Crosspoints Continued research is needed into electronic crosspoints for both space and time division organization of switches to supplant presently used techniques with electronic devices having lower working impedances, logic operating levels com- patible with those used in control logic, and lower power consumption.

167 Telecommunications Traffic Additional research would be desirable in stochastic processes viewed from randomness in time of arrival (as appropriate to switching systems), rather than randomness in level which has traditionally been studied for supporting transmission systems. Specific areas where additional research would appear productive include development and improvement of analytical formulas for dimensioning of common control equipment for switching systems; traffic overload analysis and simulation; and traffic of a wide variation of data rates vying for access to common facilities in packet switches. III. Overseas Research Funding The principal finding of this study has been that over- seas switching research is significantly funded by Government departments. This is in effect a natural outcome of the organization of telecommunications overseas where Government departments, or corporations with total or partial Government ownership operate the telecommunications services. Research leading to products covering all principal sectors of the telecommunications switching market are included in these countries' programs. As few of these Government departments own their own switching factories, there is a tendency for this research to be made freely available to national industries. These national industries also conduct substantial research in this field with their own funding. It is quite clear that the intent of this Government funding and free interchange with industry has as its underlying motive the support of industry in export markets. For example, in France (2) in speaking about the reason for placing into production Time Division Electronic Switching, the motive was given as follows: "The principal advantage of a decision of this nature will be in the industrial sector; it will permit French industry to prepare its production and maintenance organizations and to look for foreign markets". In France, the CNET (National Center For Telecommuni- cations Studies) has a budget for externally conducted studies in electronic switching of $89,000,000 over the 5 year period l97l through l975 (l). This amount is listed as 33% of the total of such external study contracts. This is not only the largest single identified element in the CNET External Studies Program, but electronic switching research is listed as the first of four principal programs in the text of the CNET sub- mission for the French Government's 6th (five year) Plan (2). It is also very evident that the French Government is trying to promote the system developed by CNET by assisting the French

±68 company CIT in their sales activities abroad. An example is Poland where CIT was assisted by its Government in obtain- ing a recent order for time division PCM switching. In the United Kingdom, for the year ending March 3l, l97l, direct research in switching and signalling is listed at $6,500,000. In addition, a major share of the $2.l million on data and related sub-systems covers work on Data Switch- ing (3). These two items combined represent 28% of the British Post Office's (BPO) research funding. A large portion of this work is done in association with industry which provides com- plementary research funding. An example of the close colla- boration between the BPO and industry may be found in reference 6 In Japan, the Telephone Administration, the Nippon Tele- phone and Telegraph Public Corporation (NTT) has conducted a development which has involved all major Japanese manufac- turers (5). By keeping the Japanese market closed to foreign suppliers and paying very high prices on switching equipment, Nippon Telephone and Telegraph is subsidizing Japanese manu- facturers on the export market. NTT seems also to influence the Japanese marketing strategy, and has on occasion provided NTT engineers to assist Japanese manufacturers on turnkey export installations. While less comprehensive information was obtained, reference to joint Government-industry cooperation on Elec- tronic Switching programs was identified in Germany (data switching only), Switzerland (4), Sweden, Norway and Denmark. There is also' a tendency of a few administrations in different parts of the world to try to control not only the development, but also the manufacturing of switching equip- ment. The Swedish Administration has had its own factories producing switching equipment since the automation of the network began. In Italy, the Telephone Administration (SIP) now belongs to the same group STET as the manufacturing company, Siemens Italiana (SIT), which was formerly part of the West German Siemens company. STET belongs to the large Government- owned holding company IRI. In Spain, the Telephone Adminis- tration CTNE will control its own switching manufacturing company in ten years, having at present a 49% interest in that company. IV. U. S. Switching Research Funding Switching research in this country is significantly funded by the Bell Telephone System for public and private telephone switching and for data switching. By practice in accordance with judicial decisions this research serves only

as a basis for products of the Western Electric Company and is available to that company's licensees only incorporated in products subsequent to initial manufacture by the Western Electric Company. The Department of Defense funds significant amounts on voice and message switching for military purposes. Signifi- cant impact has fallen out of programs such as AUTOVON and AUTODIN into the commercial products of the manufacturers who have been awarded those specific military contracts. The military program for the Tri-Tac Switch should provide similarly for future impact on product development in the commercial sector by the successful contractors for the studies leading to development of that switch. It would appear that the most significant impact of Department of Defense (DoD) supported research work in the past has been in enhancing the competitive- ness of U.S. computer manufacturers in the message switching area. The Advanced Research Project Agency (ARPA) network (24) continues to provide leadership in this field. A review was made of the support by the National Science Foundation (NSF) in the areas of Information Sciences and Systems, and Network Theory and Circuits. None of the projects for the l97l-72 fiscal years (as derived from their titles) appear directly aimed at the support of circuit switching, and only a few on packet switching. While there will be some fallout from the substantial work in support of the computer area, which will have an application in communications switch- ing, by and large, these NSF programs are unlikely to have more than limited effect on communications switching. In the industrial sector, funding by the computer manu- facturers of work on message switching is a natural adjunct of the primary computer business and seems likely to maintain a leading U.S. position in that field. Limited additional switching research is funded in the United States by General Telephone & Electronics, United (North Electric), Stromberg- Carlson, and ITT. For competitive reasons the exact values of research funding in the industrial area other than the Bell Telephone Laboratories (BTL) are not readily available. What is clear, however, is that the share of the switching market to which these manufacturers have access is so much smaller than that of the Bell System that their combined efforts are at least an order of magnitude smaller than those of BTL. Ideally. R&D should not come from Government, but from the anticipation of profits in a competitive free enterprise system. In view, however, of the substantial levels of subsidy available to overseas manufacturers, it would appear attractive

170 to have some form of U.S. Government support of R&D activi- ties, such as tax rebates to industry. Support of research in telecommunications switching at universities or non- profit institutions would result in greater public aware- ness and an increased pool of the technical manpower needed by American industry to produce and use the telecommunica- tions switching systems of the future. However, support for telecommunications research may not, by itself, be sufficient to make a substantial contri- bution to export competitiveness of switching products. The real problem of export-import trade imbalance on switching equipment in the United States is that of compet- itive pricing rather than technical superiority. Trade incentives, such as Japanese Government support and the Spanish Government export tax rebates, widen the gap of competitiveness much more than any superior technology achieved through Governmental research participation in such countries. A specific case in point is that of Fujitsu PBX's imported from Japan and sold by United Business Com- munications in preference to their own North Electric PBX. The reasons for this action appears to be price and quality, not superior technology, particularly as the Fujitsu PBX has required modification for the U.S. PBX market. R&D incen- tives alone, therefore, are insufficient to offset the cost- price differences resulting from the higher U.S. labor and overhead costs versus the broader support and incentives granted by foreign countries. They could, however, be a useful step in the right direction, particularly when coupled with other appropriate steps on the trade and monetary fronts. V. Public Telephone Switching The BTL representative on the Panel has expressed the view that the research at BTL on public telephone switching is not surpassed anywhere overseas. A review of programs underway overseas tends to support this view. While the Japanese and many European Administrations have such programs, no single country was noted with a more advanced program for local public telephone switching than that of A.T.&T. as exemplified by Electronic Switching Systems #l and #2. In Sweden and in the Netherlands public toll telephone exchanges of the semi-elec- tronic type, and in France local telephone exchanges of the PCM type have been commissioned in advance of similar switches in the United States. Some ten other countries will get such exchanges within the next three years. However, the research work of BTL leading to the Electronic Switching System #4 appears to be substantially ahead of any published information on analogous Time Division Switching Systems in Japan, United Kingdom, France (8), Switzerland (4),or Italy.

l71 In the independent telephone market, United (North Electric) has introduced in the United States the above referred to Swedish designed toll switching with sub- stantial local adaptation(7). ITT has announced the introduction of a proprietary French design of a local public telephone switch. In both cases, the design had benefited through Government (PTT) funded trial exchanges in the originating countries (Tumba, Sweden and Roissy, France). In addition to the substantial research and development by these manufacturers leading to these sys- tems, North has developed the NX-IE and Stromberg-Carlson its "Cross Reed" System in the United States. G.T.&E. also have a similar advanced switching program for the EAX, TSPS, Crosspoint Tandem (a small 4-wire tandem switch), and the CI-EAX. Part of the latter has benefited from work in a Canadian affiliate. Northern Electric of Canada (through their U.S. marketing subsidiary Northern Telecom) have recent- ly introduced their second generation stored program control electronic switching system (SP-l) to the U.S. independent, i.e., non-Bell market (2l). This work has been mainly done by Bell-Northern Research and is essentially funded out of tele- phone operating revenues from Canadian subscribers. The Canadian Government does, however, have two programs for sharing research and development costs in industry called PAIT (Program for Advancement of Industrial Technology) (22), and IRDIA (industrial Research and Development Incentives Act) (23). The PAIT program normally allows a 50% reimburse- ment of research leading to advances of the nature stated to be incorporated in this switch. In addition, the IRDIA pro- gram provides for tax free cash grants, or credits against federal income tax liabilities, equal to 25% of all capital expenditures during a fiscal year for scientific research and development in Canada, and 25% of the increase in current expenditures for scientific research and development in Canada during a fiscal year over the average of such expenditures in the preceding 5 years. A company which has received assis- tance under PAIT may include in applying for an IRDIA grant that portion of qualifying expenditures incurred, but not reimbursed under PAIT. Time did not allow investigation to determine whether these programs were a significant factor in funding research leading to this project. In general a calibration of the relative competitiveness of products resulting from research overseas can be determined from the acceptance in the export markets of products result- ing from such research. The structure of the industry in the United States, and a consequent lack of meaningful export activities in communications switching does not permit a simi- lar comparison to be made in this manner.

172 In summary, therefore, BTL is taking care of its own research in this sector of the switching field. The Bell System appears satisfied with its in-house efforts. For the remainder of the domestic manufacturers supporting the independent telephone industry, it is doubted that any pro- gram short of support on the scale being given in individual countries overseas would make it worth while for these manufacturers to design systems of this type wholly within the United States, except in specific cases where they see a realistic market potential. This is principally a result of the size of market accessible to them as compared with their overseas counterparts, or Western Electric, rather than a lack of technological capability to design comparable products in this country. VI. Private Telephone Switching For The Common Carrier and Interconnect Market In the electro-mechanical switching area, the high cost of tooling piece parts has tended to make private telephone switches (PABX) reconfigured derivatives of the public tele- phone exchanges designed for the same manufacture. This was generally true both in the United States and overseas. A strong trend has developed recently towards the use of elec- tronics in exchanges designed for private telephone use. As tooling is not so important a factor in the use of electronics, a divergence from this former trend is appearing. While some electronic PABX designs are appearing as derivatives from public exchange designs (e.g., France), some completely new designs for PABX use are appearing. Reference l2 gives an example of a Government (PTT) funded joint effort with industry in Japan. Quite a number of company-funded programs have also created commercial fully electronic PABX's such as the ITT TE 400A in the United States (9), two products in Germany (l0,ll) and the IBM 2750 in France (l3). The area of electronic PABX's appears to be one where the United States has a poten- tially exploitable technological base in active electronic components which could lead to the development of a strong U.S. industry capable of exporting products and technology (l4) In addition, the character of the private telephone switching market is in the process of change as a consequence of the decision of the FCC to open this field to competition through allowing interconnection of customer owned or leased equipment to the public telephone network under certain conditions. This so called "interconnect" market comprises both private telephone switches(PABX), but also a large array of terminal apparatus for voice and data communications. The former is of primary interest for this section of the report.

173 The immediate consequence of opening the "interconnect" market to full competition brought on conditions analogous to that of the automobile industry with many imported products appearing on the U.S. market. The few U.S. PABX equipment manufacturers who had been serving the independent telephone common carriers stepped up their efforts, as was anticipated. On the other hand, the record of the "inter- connect" market to date has been one of substantial imports of technology. Of the 49 PABX types recently surveyed (l5) covering products available to the "interconnect" market (excluding those of the Western Electric Company), l7 PABX's came from the United States, 1l from Japan (2 assembled here), 4 from Canada, 4 each from the Netherlands (part U.S. adapta- tion) , Sweden (part U.S. adaptation) and Germany, 3 from France, and one from Spain. As a consequence, the United States in 197l, for the first time, had an unfavorable trade balance in the customs category of telephone and telegraph equipment. VII. Message Switching The term message switching tends to have two different meanings according to whether it is applied to services such as reservations, message telegraph, etc., or to data switching. For the purposes of this review the term "message switching" is applied to the former services and the term "packet switching" to the latter in accordance with present usage by ARPA and the CCITT. The term "message switching" is often also called store and forward switching. The principal finding of this review has been that American companies are dominant in this field because of efforts by the principal computer manufacturers and reflecting the early support given this class of service by the Depart- ment of Defense, particularly such programs as AUTODIN. Even overseas where the principal PTT's have required such services for message telegraph and telex retransmission, American computer manufacturers have won in international competitions. Specific examples of this are in Australia (Common User Data Network), Switzerland (ATECO), Sweden (ATESTO) and Spain (CTNE Interim Data Network). Because of this indication of American dominance many PTT's are deferring programs for message switching to dates when integrated data switching can be available from national manufacturers. This subject is covered in a subsequent section on data switching. In the commercial sector, predominant usage has been reservation systems and concentration of computer communica- tions. The U.S. computer manufacturers have obtained an

174 overwhelming majority of contracts in this area. It has been noted that even where overseas manufacturers have won message switching competitions there has been a strong tendency to use processors, and memories and/or peripherals of U.S. companies. The conclusion of this review has been that there is no element in message switching which has not already been covered by research supporting the computer industry. Therefore no special additional treatment needs to be taken for research in support of message switching per se. Fur- ther comments are made in a subsequent section on the data switching market. VIII. Telex Switching The review covered both direct circuit switching for telex use, and packet switching. Included in this market are successor switches meeting that portion of the market now covered by 45.5 and ll0 baud TWX, as well as 50 and 200 baud national and international telex. France, Germany, Italy and Belgium have inaugurated 200 baud telex services and a commitment is noted by the balance of the Common Market countries to follow suit most- ly in the period between now and l975. It can be antici- pated that this speed of service will spread to the North American markets as a minimum to service requirements of international companies. Traditionally telex switching has been a direct cir- cuit type of switching and has been derived from public telephone exchange designs for the reasons discussed above under Private Telephone Switching. In the international electro-mechanical telex switching market, Siemens A.G. has been the dominant worldwide manufacturer in the past. Some years ago when it became apparent that semi-electronic telex switches would become practical, an early lead in this class of exchange developed in the United States. Major factors in this lead were the system installed in the mid-l950"s for the General Services Administration, and switches from two U.S. manufacturers for Western Union (domestic). This same class of system was also installed in Canada by Canadian National/Canadian Pacific Telecommu- nications. However, for want of continuing large scale market requirements and because of changing economics this lead was dissipated. Large semi-electronic direct circuit switching systems have been installed overseas recently in Denmark (l6) and

175 in the Netherlands. More modest computer-controlled switches have been installed in Italy, Hong Kong, Portugal and Panama. These modern designs appear more economical than the earlier U.S. designs and will presumably be the principal factors in the intermediate term export market. Only the small Portugal and Panama switches are products of a U.S. manufacturer (Fredericks). The development of packet switching discussed below has resulted in modern forms of semi-electronic direct switching being economical only in smaller sized telex exchanges and at higher speeds. This latter point results from the cost of direct circuit switching being affected in only a minor way by an increase in transmission speed, whereas cost of the packet switch's processor and memory increase in direct proportion to bit rate. Packet switching is a natural derivative of message switching in that the equivalent of circuit switching is obtained by providing sufficient throughput capacity achieved by fast links and short message units (packets) to reduce the queueing time to a transition, bit, or character (typically about l0 ms). Here again in the United States an early lead was obtained as a derivative of defense research and turned into successful products by three U.S. manufacturers (Control Data, Astrodata and Data Trends) who provided switches under contract to the three international record carriers (the Control Data Switch was, however, limited to outbound calls only). Overseas Siemens A.G. and N.V. Phillips have developed comparable and larger machines with indirect German Government support going into the Siemens development which also serves as the base for the German Post Office's data switching discussed in the next section. Based on this lead derived from Government support, Siemens has entered the U.S. market and sold its modern packet switching system to the domestic record carrier and to one of the international record carriers based on price and physical size. N. V. Phillips has provided its system to a second international record carrier. IX. Data Switching A review of progress in the data switching area shows that an unquestioned technical leader is BTL (l7). In Europe, considerable effort is being undertaken to provide comparable facilities, a description of which (with supporting additional references) is found in Reference 20. Japan, Canada and Australia are also planning to implement public switched data communications systems.

176 In support of this statement on BTL leadership, a state- ment has been made by a Japanese official that Japan lags by two to three years in this area behind the United States. As a consequence the Japanese Government, in cooperation with the four main manufacturers, intends to make an investment of $l60,000,000 (exclusive of transmission lines) to bring Japan to the forefront in this area (l8). A corresponding position has been expressed for France, indicating that by l975 that country expects to lag 5 years behind the U.S. in data switching as against a 3 year lag in the computer field (l9). The French Government is, therefore, supporting a program of external studies in this area of almost $30,000,000 over the next five years in an attempt to maintain its industry capa- bility in this area (l). It has been previously noted that German Government indirect support has been placed on elec- tronic data switching, leading already to the success in export by the German firm Siemens. In the United Kingdom, British Post Office support has gone to two British companies for work in this area (6). In Switzerland a total integra- tion of all forms of switching, including data, is receiving support from the Government (4). On the other hand, no U.S. manufacturer other than the Western Electric Company is likely to have modern data switching products on a tine scale matching overseas manufacturers, unless some special steps are taken. The reasons don't seem to be a consequence of lack of research in the U.S. in this area, which, as noted above, is considered to be ahead of the world. One reason may be ascribed to the previously referred to present policy on access of other manufacturers to research in BTL. However, as so much appropriate technology is available as a fallout from the ARPA network, this alone does not seem to be a compelling reason. A more likely reason is the lack of willingness of manufacturers' managements to rcake such large investments of R&D funds (including field trial exchanges) when they observe their overseas competition being partially financed by Government agencies (PTTs). Thus, to the U.S. domestic record carrier, the three international record carriers, and the one data oriented miscellaneous common carrier, the choice is currently left to looking for foreign products, or of themselves financing a U.S. development directly, or indirectly, and including it in the price of equipment. As the recent trends show in telex switching, the choice of foreign products is more probable in the absence of incentives to U.S. manufacturers to get them to a comparable level, including initial working switching exchanges. To permit fair competition between manufacturers sup- porting various classes of common carriers, this support

177 should be structured in such a way as to give parallel benefits to all industries for comparable R&D via such means as shared cost reimbursement for R&D, or double deduction from income taxes for R&D in this field. Even with this type of support to industry, it might be worth while to provide an additional incentive by having the General Services Administration and/or Department of Defense call for competitive bids for modernization of their own administrative systems in a direction which would benefit U.S. manufacturers by proving the results of R&D through construction of comparable exchanges to those being financed by overseas governments. In this manner, U.S. manufacturers would operate on a comparable basis with foreign suppliers when bidding on the future requirements of the record carriers. Most of the research necessary to support data switching is of a type analogous to that supporting the computer industry. Nevertheless, additional grants by the National Science Foundation to universities and non-profit corporations would help to stimulate additional research, and graduate level work in engineering on data switching. X. Summary As this study of telecommunications switching has shown, many other countries prepare their national manufacturers for exporting and for meeting import competition through a multi- plicity of incentive schemes. ATT has provided the level of funding necessary for the Bell Telephone Laboratories to meet and surpass in technological achievement the rest of the world in those specific areas destined for potential use in the Bell System. This has, however, left certain product areas where Bell does not operate without adequate R&D support to U.S. manufacturers. It has also left the U.S. without exportable switching products. And finally, in those areas where compe- tition has been opened up by recent FCC decisions, U.S. manufacturers, other than the Western Electric Company, have been left at a disadvantage with respect to foreign manu- facturers . As many foreign manufacturers receive Government incentives and the U.S. manufacturers do not, it is concluded that it would be desirable to establish incentives in selected areas of telecommunications switching. To be effective, this support for R&D should begin at universities with specifically directed NSF grants, extend through industrial research and development (with the Canadian PAIT/IEDIA programs forming an attractive pattern to emulate), and include provision for lease or purchase by Government agencies of working advanced switching systems.

178 Such a program would place a number of competing U.S. switching manufacturers on a par with their overseas counter- parts, with the attendant potential benefit to the U.S. bal- ance of payments of reducing imports and fostering exports. It should be emphasized that in the limited time allowed for this study, only a superficial look could be taken at the area of telecommunications switching. In any extension of the work of this Panel, more complete coverage could be made of this field. Lynn W. Ellis ITT New York City

179 References l. "Program Document No. 5 of the 6th (five year) Plan", CNET (National Center for Telecommunications Studies), Sept. 8, l970, p. 5. (In French) 2. Ibid. p. 7. 3. "Post Office Report and Accounts For the Year Ended March 3l, l97l", p. l4l. 4. K. E. Wuhrmann, "Systems IFS-l, An Integrated PCM Telecommunications System", l972 International Zurich Seminar on Integrated Systems, March, l972 (to be published). 5. K. Habana and M. Mitsigi "Remote Control Electronic Switching Systems", International Symposium, l972, Boston, June, l972. Conference Record, IEEE Catalog No. 72 CHO 6l7- l-Com. 6. T. W. Adam and A. G. Orbell "A Digital Data Exchange" International Switching Symposium, l972, Boston, June, l972. Conference Record, IEEE Catalog No. 72 CHO 6l7-l-Com. 7. S. J. Long and E. Aro "Structure and Characteristics of the ETS-4 Toll Tandem Switching System", International Switching Symposium, l972, Boston, June, l972, Conference Record, IEEE Catalog No. 72 CHO 6l7-l-Com. 8. A. E. Pinet "Introduction of Integrated PCM Switching in the French Telecommunication Network", International Switching Symposium, l972, Boston, June l972, Conference Record, IEEE Catalog No. 72 CHO 6l7-l-Com. 9. J. Reiner & S. E. White "ITT TE 400A Full Electronic PABX", Paper 70-CP-2l6-COM, l970 IEE International Conference on Communications. 10. K. L. Plank "A Fully Electronic PABX With PAM Speechpath-Switching" International Switching Symposium, l972, Boston, June, l972, Conference Record, IEEE Catalog No. 72 CHO 6l7-l-Com. ll. J. G. Schosnig "PAM-Speechpath-Switching in the III 6030", International Switching Symposium, l972, Boston, June, l972, Conference Record, IEEE Catalog No. 72 CHO 6l7-l- Com. l2. K. Ozeki, et al. "Time-Division Digital Electronic Switching System", NEC Research and Development, October l97l, No. 23, pp l5-22.

180 l3. R. Reynier, "Electronic Switching Network of the IBM 2750", IBM J. Research Development l3,4l6 (l969). 14. E. H. Gatzert, "The Silent Revolution" Telephony, Vol. l81, No. 23, Dec. 6, l97l, pp 29-32. l5. A. R. Meier, "Telephony's Survey of PABX's Now Available in the U.S.A.", Telephony, Vol. l8l, No. 23, Dec. 6, l97l, pp 33-34. l6. J. Jensen, "Siemens Crosspoint System in the Danish Automatic Telegraph Network", Siemens Review, Vol. 39, No. 4, April, l972, pp l48-l55. l7. L. R. Pamm, "The Bell System's Digital Data System", a talk to the International Communications Association, May ll, l972. l8. Statements by Dr. T. Kohima, Director, Integrated Electronics Development Division, Nippon Telephone & Telegraph Public Company, as reported in The Electronic Engineer, April, l972. l9. Electronic Actualities, May l5, l972. 20. J. Atkin, "Evolution of Public Data Networks", Electrical Communication, Vol. 47, No. l, l972, pp ll-l4. 2l. A. R. Meier, "'Springtime in the Rockies' Seminar Signals Step-up of Canadian marketing effort in U.S.", Telephony, June l9, l972, pp 55-56. 22. "Program for Advancement of Industrial Technology", Dept. of Industry, Trade and Commerce, Ottawa, Canada, Queen's Printer for Canada, l970. 23. "Industrial Research and Development Incentives Act", summarized in "Quick Reference on Incentive and Development Programs for Canadian Industry", Queen's Printer, Ottawa, Canada, l970. 24. L. G. Roberts and B. D. Wesslor, "Computer network development to achieve resource sharing", l970 Spring Joint Computer Conf., AFIPS Conf. Pres., Vol. 36, Montvale, N.J.; AFIPS Press, l970, pp 454-549.

l81 RESEARCH ON THE COMMAND AND CONTROL OF RAIL TRANSPORTATION I. Introduction A preliminary investigation has been carried out on the command and control of rail transportation, both in the U.S. and overseas. While this is only one narrow segment of transportation research all of which poten- tially merits similar investigation, it was chosen for review to be consistent with the focus of the Panel's parent Committee on Telecommunications as the latter expressed thoughtfully in the following passage: In its report "Communications Technology for Urban Improvement", the NAE Committee on Telecommunications saw the role of communication technology primarily to be supportive of that mode of urban transportation now desperately in need of special emphasis-mass transit. Further enhancement and even subsidization of private individualized transportation were felt only to aggravate current problems. The Telecommunications Committee came out strongly in favor of using communications to provide the urban public with improved accessibility to mass transit systems (20). As a general observation, research on this subject is at a technological level significantly below that encountered in other areas of telecommunications. While research in telecommunications is eagerly studying new materials technologies (e.g. optical fibers) and advanced systems architectures (e.g. inter-computer communications), research worldwide in command and control of rail transport appears to be cautiously grafting simple hierarchical control systems based on current technology computers on top of automatic block systems constructed of electro- mechanical components. Apparently, this cautious approach to research is traced to the twin considerations of safety and levels of R.&D. funding. An overriding consideration of rail transportation is safety, both for reasons of preservation of human life and for the high cost of a failure in potential damage to property. As such, railway signaling has evolved not only in a "fail-safe" mode, but in one where innovation must pass a series of slowly ascending levels of trial before being generally accepted in the industry. Under these conditions innovation has tended to be built on thoroughly proven component technology. The low mean-time-between- failure of many modern electronic sub-systems is an ana- thema to the railway command and control engineer. In

182 practice he insists on analytical proof that any failure, no matter how small, will lead to neither loss of human life, nor major property damage. An excellent justifica- tion for such a conservative attitude can be found in the accident which occurred October 2, l972, on the Bay Area Rapid Transit (B.A.R.T.) system with damage to a car and injury to 5 people (2l). Its cause was attributed to a malfunctioning quartz crystal (22). A result has been a leaning of such research as is done towards systems architecture of a straightforward nature. The second consideration is that of low R.&D. funding levels. In recent years gross investment in rail transpor- tation has been low as the high technology fields of air and space have preoccupied governments, as the poor finan- cial condition of the U.S. railroads and metropolitan sub- ways has kept down the incentive to invest, and as the high tax revenues derived from vehicles and the petroleum products they use have tended to be plowed back into roads, thus worsening the competitive position of the railroads and their capability to invest. For every dollar invested in rail transport, only a limited share can be invested economically in command and control equipment. In practice, priority of investment is often given to civil work on right of ways, and to train equipment, leaving com- mand and control equipment to be procured belatedly against a limited fixed budget. Yet without a command and control system, the right of way is worth no more than a hole in the ground. Limited investment in command and control equipment has led to limited incentive for the companies who manufacture such equipment to spend heavily on R.&D. in this field. Despite this condition, a number of manufac- turers described subsequently have maintained a technologi- cal capability which could be expanded readily at any time the potential appeared for increased levels of R.&D. funding. II. Suggestions For Future Research Arising from the investigations made in this area are a limited number of suggestions for future research. Some of this research could be directed towards university laboratories, while other items may need the specialized expertise of industrial laboratories to provide effective answers.

l83 A. Automatic Train Operation and Control System Organization (Linear Routes) 1. Study train location via block organization as used in the U.S. as against transposed wire schemes under trial in Europe. Technical (speed and headway limitations, availability of adequate bandwidth for control and for reverse channel capability to indicate control has been implemented correctly), economic (first and operating costs) and safety comparisons are all involved. 2. Application of advanced computer techniques (e.g. micro-processing, micro-programming, etc.) to meet the need for distributed "fail-safe" direct computer or logic control of switches, signals, etc., now carried out by electromechanical relays. 3. Systems architecture alternatives to the present trend toward simple hierarchies. Processor to processor communications security and telecommunications traffic flow in a distributed control structure. Flexibility for continued growth of served right of way in various control structures. B. Command and Control for Two Dimensional Route Structures j The advent of "People Movers" strengthens the need for traffic flow research into effective sharing of common route sections under criteria of human dissatisfaction with delays substantially different from normal telecommunica- tions traffic. Even on linear route structures such research would be helpful to devise more effective strate- gies for handling the merging under abnormal conditions of serious delays on one of the linear branches. C. Mathematical Modeling of Train Networks "Railroad operations can be likened to a set of inter- acting control loops (e.g. passenger and freight traffic) which are multi-variable, non-linear and with time-variable parameters, and whose system time constants range from minutes (response to failure conditions) to many months (major review of passenger time tables)" (28). Research is needed to seek effective mathematical system designs and algorithms to model such systems easily and economically on a computer and thereby identify where effective developments can be introduced to improve system performance.

184 III. Summary of Significant Programs A. Definitions For the purposes of this summary, it is convenient to categorize rail command and control systems into those involving control only versus those involving automatic operation, and also into applications to various route structures. The following definitions are applicable to control versus operations (3): "Automatic Train Control (ATC) "ATC is the term for various safety systems that have been in use since about l922. One common type forces the driver to operate safely and will bring the train to a complete stop if he exceeds a speed limit or fails to apply brakes after receiving a speed decrease signal. "Automatic Train Operation (ATO) "ATO is a term used to mean full "hands off" control. It includes a number of functions: 1. Speed command recognition 2. Speed regulation 3. Position stop - used for transit to make a con- trolled deceleration to stop at a specific plat- form location. 4. No motion - used to prevent opening doors in motion. 5. Safety functions analagous to ATC" The application of command and control systems on various rail route structures can be categorized as follows: Category A. High Speed Interurban Systems Modern rail transport systems employing heavy trains with substantial headways at speeds up to 200 km/hour (l20 MPH) over linear route structures comprising arrays of loosely interacting lines with infrequent intersections. Category B. High Density Linear Route Urban Systems "Modern rail rapid transit systems employing heavy trains operating on the order of l00 second intervals at

185 speeds up to l30 km/hour (80 mph)* over linear route struc- tures comprising arrays of loosely interacting lines." (8-Cat.I) Category C. Urban Distribution Systems "Distribution systems for major activity centers employing vehicles that are much lighter and smaller than rapid transit cars operating as one or two car trains at headways on the order of 20 seconds at speeds up to 50 km/hour (30 mph) over essentially linear route structures." (8-Cat.II) Category D. Personal Rapid Transit Systems Also called "People Movers" (l,23) and "People Moving Vehicles" (PMV) (1). "Personal rapid transit systems that approach the operational flexibility and ready availability of the private automobile. These systems employ automobile-size vehicles operating at headways on the order of five seconds at speeds up to 30 km/hour (20 mph) over route structures that are two dimensional networks with strongly interacting segments." (8 Cat.III) B. Country Programs The research work on these categories will be reviewed in the following study for the U.S., West Germany, Japan,Belgium, France, United Kingdom, and Sweden to the extent that information was readily obtainable. In reviewing the published papers a dramatic contrast in quality was noted between those by the authors in the U.S. and overseas. U.S. authors tend to be terse, write in a qualitative nontechnical vein and be affi- liated with manufacturers. Overseas authors tend to be more communicative, more technical in level, and be affiliated with the Government-owned railway administrations. This may be interpreted, to some extent, as a reflection of the com- petitive procurement policies in this country where manu- facturers are unwilling to share the fruits of their own research with competitors versus the tendency in overseas countries for long-term joint assocation between the railroad administration and national manufacturers, often on a territorial allocation basis. *Speed limits raised from the original text to reflect working limits on B.A.R.T.

l. United States The most advanced work on rail transport is concerned with People Moving Vehicles (PMV) which are seen as the future rail systems that approach the operational flexi- bility and availability of automobiles over two dimensional route structures (l,8,23). These are Category D systems involving automatic operations. Substantial technical problems will be involved in the command control systems associated with such systems which exceed those of more traditional systems by the need to select between routes and merge from several routes at convergence points. None of the traffic problems associated with this switching/ merging function appear to exceed, for example, those already long encountered by ATT in the routing of tele- communications traffic. Nevertheless, there is a substan- tially different impact of a lost PMV as compared with a lost telephone call, so that the system architecture must be designed for reliability. To date experimental systems in Morgantown, West Virginia and planned for Denver, Colorado work initially on linear routes and have not yet faced the two dimensional control problem. This type of product has a high potential for international exploitation, with an initial installation at the International Ocean- ographic Exposition on Okinawa. (36) Category C (Urban Distribution) systems are largely in the proposal state (8) and provide for vehicles much lighter and smaller than Category B (High Density Urban) systems over essentially linear routes. In common with other linear route systems, technology is required for longitudinal control, automatic protection, and automatic operation. Lateral control is normally achieved by physical restraints, but some proposals have been put forward for electronic control with respect to an electrical guidance wire (l). The Tampa Airport system (ll) has some of the features of this category, but without the full range of control needed for multi-vehicle operation on the same right of way. Category B (High Density Urban) systems of recent vintage include the PATCO system (3,l0), the Bay Area Rapid Transit (B.A.R.T.) system in California (4,5,8,37,38,39) and the Washington Metropolitan Area Transit Authority (W.M.A.T.A.) system under construction in the District of Columbia (8). From the limited published information on this latter system, it seems to follow some of the philosophy of B.A.R.T., with the additional innovation of CRT displays instead of mimic diagrams for the man-machine interface.

187 The B.A.R.T. system has clearly the most comprehensive command and control system in current use today, and as such serves as the benchmark against which to calibrate the work in other countries. As will be seen below, several other countries have railroad command and control systems in design or use which rival that of the B.A.R.T. system, and in some areas show signs of leading it. With the designers of B.A.R.T.'s command and control system (Westinghouse) having obtained an order for a similar system in Sao Paulo, Brazil, it is clearly a factor in U.S. export markets, but has no clear lead as, for example, does the U.S. computer industry. Its automatic operation is carried out by a number of integrated sub-systems (designed by Westinghouse Electric, except where noted): - An automatic vehicle control system in each car provides the functions of decoding the speed command sent from the way- side, automatically maintaining speed at the level set by the central computer, protecting against over speed, braking, generating identification signals, and stopping at the midpoint of each station platform. - Wayside controls set the speed of trains by means of a block system with up to 31 blocks per station. They also pro- vide information on block occupancy, and transmit control signals to the trains. - Station controls set basic operating conditions for blocks in their area (such as maximum speed in each block), pick up identification signals from trains and forward them to the central computer, and provide safety of trains on the route by sending diminishing speed signals to trains following each operating vehicle in proportion to the amount of headway. - Interlocking control of the traditional gravity relay type protect branching and merging points where the present routes intersect. - Automatic yard control in each yard identifies vehicles, assembles trains, and tests the operation of brakes, doors, etc., before trains pass on to the main line (Philco- Ford and Wismer & Becker). - A digital transmission (of wired logic at l200 bauds) system (DTS) interconnects the central computer and each station.

188 - An automatic platform display system in the distributed computer (PDP-8) control presents news and advertising subject to a priority interrupt for train destination taken from the same information sent by the* train to the central computer (Stewart Warner). - An electronic computer center in Oakland monitors all train movement from the identification signals sent from each station via the DTS, compares the arrival time to the scheduled arrival and sends back via the DTS signals to speed up or slow down the train. These adjustments can achieve a l0% reduction in travel time between adjacent stations (with the station control ensuring no speed limits are violated), or up to a 50% increase. These adjustments apply only in the four linear segments of the route. If a train is so delayed that it cannot make its schedule "slot" at a branching/ merging point, the computer throws control to the system manager to determine which of several recovery strategies is to be employed. From the standpoint of guiding research in this area, it is important to note that despite its complexity the B.A.R.T. system has the following limitations. - It has limited backward information flow, covering only 6 bits of destination, 4 bits of train serial number and 4 bits of train length, and this only from immediately before each station, not from each block. In effect, therefore, all safety is based on redundancy in the forward direction only. - Being based on audio frequency track circuits, it has limited forward transmitting bandwidth to the vehicles. - Train detection is only by shunting the tracks. An insufficient shunt will not show a train's presence to the station control. A fault could indicate a train where none exists with accompanying inconvenience to passengers on following trains which are forced to slow down. - Despite the complexity of the computer center, it optimizes four linear routes separately and does not attempt to manage the entire system from a traffic flow aspect under major disruption of schedules. Enumeration of the points immediately above is not a criticism of the B.A.R.T. command and control system, nor of its designers. Rather, it is a synthesis of features which other railroad administrations overseas have insisted upon

189 for their own systems. Only detailed study by competent researchers can adequately assess the pros and cons of the above competing alternatives. Although it is clear from a review of the B.A.R.T. system that it has many problems still to resolve (38), most of which are derived from too much innovation in a single system, and too many interfaces without a single overall contractor, it is one of the best urban transit systems in the world today from the standpoint of customer satisfaction and has no "bugs" as yet identified which cannot be cleared by the determined effort which is being applied. Category A (High Speed Interurban) systems commissioned recently include the Muskingum Electric Railway (3,9) which is operated fully automatically and a number of automatically controlled systems (3) . The most advanced control systems include continuous inductive loop communications and inter- mittent control from passive wayside devices, but still have much traditional technology (e.g.,redundant relays) for reasons noted in the introduction to this study. Some com- puter technology is used in some classification yards (2) and microwave technology is proposed for rail-highway grade crossing control systems (6). For these more cost effective systems to be applied widely, proof of reliability will have to be demonstrated. 2. West Germany In categories D (Personal Rapid Transit) systems and C (Urban Distribution) systems, respectively, technical pro- posals have been put forward by a national manufacturer for conveyor belt type systems, and linear route systems using magnetic cushions. The former needs limited control, but has the potential for high risk to passengers should a con- dition result in a departure from zero relative velocity at any transition point between belts, or between a belt and a rotating platform. The second will need as sophisticated a control system as for similar proposals in the United States. However, no reference was obtained on research work of an advanced nature in this area. The German manufacturers, AEG and SAG are proposing designs, and SEL (ITT) in its current installations, apply the same control systems to both Categories A (High Speed Interurban) systems and B (High Density Urban) systems (l2). That is, modern automatic control is designed for inter- changeable use on high speed intercity routes and high density (low headway) urban routes, involving similar

190 hardware and considerable commonality of software. As recommended by the Union Internationale des Chemins de Fer (UIC) (an international organization of railways in Europe), data transfer is via high-speed data transmission over an inductive loop laid between the rails along the line which is periodically transposed so that it also pro- vides train location information. In a typical Category A (high speed interurban) application between Hamburg and Bremen, Germany, a central control system of triplicate minicomputers (l6 K memory) interrogates all trains in a 40 km (25 mile) section of the route in a time sharing mode, each at intervals not longer than 800 microseconds. Train- borne equipment announces its location and speed and receives speed instructions from the computer which are evaluated to initiate speed changes. In addition, the computer interrogates all existing computerized district control centers and electromagnetic component interlocking equipments on the route for status information, and provides orders for control of signals, level crossings and switches. The transposed loop (l3) has the technical advantage over the control system used on the Muskingum Electric Railroad (9) of being balanced both against electrical interference (as it is transposed) and of being at ground potential with respect to the traction power line. In addition, it functions as a quasi-continuous block system with positive train detection down to at least the interval of transposition by simple means such as counting phase reversals. Because of its transposition scheme, it has the usable bandwidth of a telephone open wire carrier system, or well in excess of l00 KHz, with equally available capa- bility of transmitting to or from the train. In its inher- ent capabilities, the transposed loop meets three of the four limitations indicated above for the B.A.R.T. system. In effect, B.A.R.T. uses these capabilities for station platform approach (flare) control, but not elsewhere on the route. Finally, as the transposed loop (in several variations of transposition distance) has been standardized by the U.I.C., it may be specified by railways in developing countries, with the result of placing U.S. manufacturers at a competitive disadvantage, if they do not also study and master this technology. It should be noted that the approach in West Germany is for indirect computer control through existing signal towers. It has not yet been determined whether direct computer control of signals without local interlocking equipments is possible with safety, or if it is economical.

191 Both the German Railways (Bundesbahn) and the Ministry of Development are interested in research on even higher speed trains with the latter reportedly providing 30-80% support to industry programs. A Government trial track is to be installed not far from Augsburg, Bavaria, with a total length of 72 kms to provide for trials of new trains at speeds of up to 350 km/hour (220 mph). Support may be expected to include the development of command and control systems adequate for this velocity. 3. Japan In categories C (Urban Distribution) systems and D (Personal Rapid Transit) systems a slow speed (30km/hour monorail has been proposed for Tokyo (32). Also limited information was published on work being done by Mitsui for the "VONA" (Vehicle of the New Age) stated to be a 6 to l2 car, 40 km/hour train with computerized ATO with the novel feature that "when the train arrives at the terminal, the circular platform turns with it at a rate of 2 km/hour"(33). Notwithstanding these national efforts, import of U.S. Boeing technology is contemplated by the Kobe Steel Company (36). Category B (High Density Urban) systems are under active development in Japan. Specific installations have included the Monorail at the Osaka Exposition (Hitachi), the Sakaisuji Line (Toshiba) (l4), Sapporo (Hitachi) (l5), and the world's most widely used monorail between Haneda Airport and Tokyo (Hitachi). The Sakaisuji is a computerized automatic train opera- tion which controls train schedules as well as signals and switches. The Sapporo command and control system comprises separate computer sub-systems for train operation control, power supply control and supporting business operations. Category A (High Speed Interurban) systems include those for the New Tokaido Line and New Sanyo Line (7). The control system is based on ATC, and centralized train control (CTC). In order to operate the whole line efficiently over l000 km (625 miles), a computer-aided traffic control system (COMTRAC) has been developed. The functional operation of this system is analogous to that of the German Railways in that the computer control is indirect via the CTC system. The essential difference between the Japanese and German systems

192 is that in Germany distributed control is used for each 40 km (25 miles) of route, while in Japan the control system is centralized in Osaka. COMTRAC's sampling speed is about one half that used in Germany. COMTRAC is based on a 32K memory, 2 micro-second access duplicated computer with a l6K, l.6 micro-second satellite for controlling and adjusting schedules, and is thus by computer standards not a very complex system. The CTC system is modern, but still based on relay technology. Because of the extreme distances over which control is exercised, primary responsibility for safety rests in the CTC system. In ancillary systems, the Japan National Railways (JNR) has experimented with a micro-wave (Gunn diode) doppler radar speedometer in marshalling yards and a laser detector based automatic car identification system. Basic research is beginning on ultra-high speed railway systems at JNR's research laboratories planning towards a l980 service date. JNR has a very modern outlook on the use of computers and by the end of l97l had about 80 computers of various sizes in use in its operations (34). 4. France The Paris Metro has recently built a new Category B (High Density Urban) line at l00 km/hour (62 mph) with conventional block control for traffic safety, but with a centralized control for train traffic, passenger traffic measurement, passenger information and communications with sub- systems provided by CII, CGA, AMIGA, HALBERTHAU and CGCT (ITT). The automatic control computer exercises control via a dual com- puterized remote indication and control system (2 x l2K of core plus 2 x 2 MB of drum memory) handling secure telegrams over 2400 baud data links, and providing l6,000 indications and 6,000 controls. This remote indication and control system provides a func- tion analogous to the digital transmission system in B.A.R.T. The essential differences are that the French system is processor controlled (while B.A.R.T.'s digital transmission system is wired logic), that the French system's speed is twice as high, and that its total information handling capacity is appreciably greater, particularly in the train to control center direction. This remote indication and control system appears to be the most advanced in use in rail transportation today. Nevertheless, systems of this technological level are available in a number of countries generally having been applied first to the remote con- trol of electrical power transmission.

193 In Category A (High Speed Interurban) systems, the French Railways (SNCF) has been a leading experimenter on high speed trains. Tests in l954 and l955 went as high as 33l km/hour in one section of route (35) and the French were among the earliest to adopt 200 km/hour regular train service. Publications through l969 (30, 3l, 35), however, do not disclose more modern signalling methods than block operation (with look ahead) and speculative comments on transposed wire systems as described above under West Germany. Time and effort did not permit a more careful review of research in this area in France. 5. Belgium The application of electronics to centralized train control in Belgium, is described in Reference l6. This is wired-logic technology applied to a transposed inductive loop as described above for West Germany. 6. United Kingdom The most recent Category B (High Density Urban) system in the United Kingdom is the Victoria Line in London (l7, l8). This is a wired logic automatic operation system comprising a safety sub-system and a train command sub-system, transmitted via the running rails. Speeds are low, 80 km/hour (48 mph maximum); however, trains are separated by a close headway of only 82 seconds. Investigation by the British Railways Board (l9) has shown that up to 200 km/hour (l25 mph) their existing signaling systems would be satisfactory on Category A (High Speed Interurban) systems, but new and additional command and control systems would be required for speeds of 240 km/hour (l50 mph). Work does not appear to have gone as far as in West Germany and Japan, although towards the end of l972 a trial of a computerized CTC unit was scheduled to begin in Glasgow, and dual miniprocessor control (aboard the train) of speed and braking is being evaluated as a pilot scheme in another location (29). A longer range study of railway freight operations is being conducted by the research department of British Railways (28). This program covers modern additions to the present network (as described immediately above), a computer-based overlay to the present interlocking systems, and looks ahead to development of a comprehensive manage- ment information system (called TOPS-Total Operations Processing System) to be operational by mid-l975.

194 7. Sweden Although no direct information was obtained from Sweden on research in this area, an indication of its present status may be obtained from the installation by Sweden's largest electrical manufacturer (ASEA) being planned for a mining railroad in Colorado (27). This is a slow speed (40 km/hour or 25 mph) rail- road with automatic train operation, automatic train control, and data transmission systems. The ATO control system is basically a speed control system with three speed references received from the ATC system. Individ- ual slip control is incorporated on each locomotive. The ATC system (developed by SAAB-SCANIA) is a block- signaling system. It has, however, finer control than a normal block system as a number of antenna loops are used within each block for transmitting the programmed speed command. The last antenna loop in each block pro- vides occupancy status of each block. The data trans- mission system is over coaxial cable with 400 channels in the band from 200 to 300 KHz. Each channel has a maximum speed of l5 bauds. Although each of these sys- tems has novel features which would merit further study as to economic effectiveness, they are in the nature of engineering developments rather than the results of advanced research. IV. Comparative Assessment of Research Work The research overseas on command and control of rail transportation appears most advanced in West Germany and Japan. In these and other countries, a general innovation lag with respect to computers and telecommunications is evident for reasons covered in the introduction. In all countries, including the U.S., R.&D, appears to suffer from funding at too low a level to match the pace of the evolution of the computer and public telecommunications industries, let alone their counterparts in the defense/ space areas. Where comparative statistics could be obtained, these have been included in comparative charts attached for Categories A (High Speed Interurban) system and B (High Density Urban) system. A tentative conclusion is that a potential lead in this area may be being built up in West Germany and Japan through their administrations'

195 aggressive efforts in speeding up service in Category A and the design of command and control systems potentially applicable to both Category A and B use in those countries, No comparative statistics were obtained on systems in Categories C (Urban Distribution) system and D (Personal Rapid Transit) system, but continued monitoring of this area would appear prudent. Lynn W. Ellis ITT New York City

196 Comparative Chart Category A - High Speed Interurban Systems U.S X U.K. France West Germany Japan x Indirectly Computer Controlled Systems in use or trial Maximum Speed (km/hour) in use or trial Maximum Speed (km/hour) planned Computer core memory size Computer access time (ft see) *l955 tests went to 33l km/hour (35) Comparative Chart Category B - High Density Urban Systems l60 200* 200 2l0 240 300 350 500 l6K 32K l6K & l.5 2 & l.6 U.S. (BART) U.K. France West Germany Japan Indirectly Computer Controlled Systems in use or trial X XXX Maximum Speed (km/hour) in use l28 80 l00 70 Minimum Headway (sec.) Computer Core memory size Computer access time (ft sec) *Plus 375K of disc. The standby unit which handles housekeeping duties and simulation when not on-line service, has 52K of core plus l.5M of disc memory. X X 128 80 100 90 82 l38 40K* l2K** **Remote control and indication system only.

197 References l. T. T. Trexler, "Control and Communications Systems for People Movers", IEEE Convention Record, March, l972, Paper 3B.1. 2. P. Delvernois, "Automatic Control of Railroad Classification Yards", ibid, Paper 3B.2. 3. P. T. Ryan, "Automatic Controls for Locomotives and Trains", ibid, Paper 3B.5. 4. C. K. Bernard, "BART-Concept and Evolution", ibid, Paper 3G.l. 5. W. A. Bugge, "BART-On the Frontiers of Urban Tran- sportation Technology", ibid, Paper 3G.3. 6. J. B. Hopkins and F. R. Holmstrom, "Cost-Effective Microwave Systems for Railroad and Automobile Safety Applications", ibid, Paper 6CK.4. 7. T. Matsuo, "Technical Achievements in the Tokaido Shin Kansen and Electrical-Electronic Techniques Contemplated for Shin Kansen", ibid, Paper 7E.2. 8. J. E. Freehafer, "Modern Trends in the Command and Control of Mass and Personalized Transportation in the U.S.A.," ibid, Paper 7E3. 9. L. E. Ettlinger & P. T. Ryan, "Automation of the Muskingum Electric Railroad", IEEE Paper 70 CP803-PWR. l0. H. H. Nennell et al, "Propulsion and Automation Equipment for PATCO's Rapid Transit Cars", IEEE, lGA Paper TOD-7l-35. l1. "Tampa International Airport Passenger Shuttle System", Westinghouse Electric Corp. Brochure WTD 7l-52 lA. l2. W. Koth, "Vergleich der Systemmerkmale verschiedener Zugbeein flussumgseinrichtungen" (in German). ETR-Eisenbahn technische Rundschau, Aug. l97l., pp.326-336. (Comparison of Systems Parameters of Various Automatic Train Control Systems). l3. G. Lentz, "The Automatic Train Control System for High Speed Trains in use on The Deutsche Bundesbahn", Technical Meeting of the IEE,.December l5, l966. l4. "Automatic Train Operation Put to Practical Service", Look Japan, May l0, l97l, p. 2l.

198 l5. "Sapporo's Futuristic Transit System", from Tomorrow's News (a publication of Hitachi, Ltd.). 16. M. M. Demeur and Schoonheyt, "The Development of Electronics in Signaling on Belgian Railways", Technical Meeting of IEE, Nov. 5, l969. Proc. l969/70, Inst. Rly. Sign. Engrs, pp. 74-96. l7. V. H. Smith "Victoria Line Signaling Principles", Meeting of IEE, Nov. l6, l966. Proc. l966/67, Inst. Rly. Sign. Engrs, pp. 76-l08. l8. F. G. Maxwell "The Victoria Line in Operation", Meeting of IEE, Feb. l7, l970. Proc. l969/70, Inst. Rly. Sign. Engrs, pp. l44-l70. 19. J. F. H. Tyler, "Signaling for High Speed Trains", Meeting of IEE, Jan. 7, l970. Proc. l969/70, Inst. Rly. Sign. Engrs, pp. ll8-l43. 20. Comments made by the NAE Committee on Tele- communications on the R&D program of the Department of Transportation at a meeting of the Committee Sept. 27, l972. 21. "5 on Bay Area Train Hurt As Automated Brakes Fail", New York Times, Oct. 3, l972, p. 89. 22. "Oscillator Triggers lst BART Mishap", Electronic News, Oct. l6, l972, p. l8. 23. "'People Movers' Set to Come to Denver", Wall St. Journal, Oct. l6, l972. 24. G. D. Friedlander "BART's Hardware-from Bolts to Computers" IEEE Spectrum, Vol. 9, No. l0, October l972, pp. 60-72. 25. G. D. Friedlander, "Railroad Revival: On the Right Track", IEEE Spectrum, Vol. 9, No. 8, Aug. l972, pp. 63-66. 26. G. D. Friedlander, "Bigger Bugs In BART", IEEE Spectrum, March l973, pp. 32-37. 27. L. G. Eriksson and A. Bjorklund, "Automatic Electric Railroad for Ore Haulage at the AMax Henderson Mine, paper No. C 72 938-4-IA, presented at the l972 Joint IEEE/ASME Railroad Conference. 28. E. E. Gelbstein & W. T. Parkman, "Railway System Installations which will Improve Performance", Electronics Weekly, May 31, 1972. 29. L. L. Alston and J. W. Birkby, "Developments in Train Control on British Railways", paper presented in London, October l3, l97l.

199 30. "Le Probleme des Grandes Vitesses", entire issue of Revue Generale des Chemins de Fer, March, l966. 3l. "Informations Techniques", S.N.C.F., No. 6, Oct. l969. 32. "Circular Monorail Proposed as City Traffic for Tokyo Met". Technocrat V. 3, No. 2, Feb. l970, p. 24. 33. "Dream Train 'VONA' (Vehicle of the New Age)", Technocrat, Vol. 4, No. 2, Feb. l971, p. 24. 34. "The JNR Computer Center", Technocrat Vol. 4, No. l2, Dec. l97l. 35. M. J. Michaux "La Signalization Speciale aux Grandes Vitesses Sur Lignes Existantes", Revue General des Chemins de Fer, Nov. l969, pp. 609-6l7. 36. "Boeing Pact Near for Use of People Mover in Japan", Wall St. Journal, Nov. l0, l972. 37. "World's Most Advanced Control System Operates BART", Passenger Transport, Oct. 20, l972, pp. 22-23. 38. "Not-So-Rapid Transit", Wall Street Journal, Nov. 7, l972, p. 32. 39. "Welco Working on BART's Signaling", Railway Signaling and Communications, Dec. l967.