First-Hand:PDP-8/E OMNIBUS Ride

From ETHW

contributed by Remo J. Vogelsang, Life Member, employed by minicomputer pioneer Digital Equipment Corporation [short form DEC], Maynard, Mass., U.S.A. from 1968 to 1972


Please click Smiley's OMNI-Smile.png[index] to get to some background information or maybe a funny tale ! Then click ↑ to return.


The Highlight

1970 - WESCON exhibition Western Electronic Show and Convention, Los Angeles, August 25-28

A London double-deck "digital PDP-8/E OMNIBUSTM" toy - one of the 10'000 - helping to promote the new product at the 1970-WESCON show in Los Angeles. The author dug out this souvenir from one of his kid's play-bin he found in his attic.

I got dropped off the double-deck bus which had picked me up at the hotel that morning. The friendly hostess, dressed in matching red, said good-bye, and here I was right in front of the exhibition hall. Enjoying the morning rays of the August sun, I proudly glanced back at this red London double-deck bus with its "digital PDP-8/E OMNIBUSTM" label posted between the two window rows across the giant vehicle (a fitting promotional effort for the novel PDP-8/E minicomputer's I/O bus). Inside at the DEC booth I was struck. A crowd of interested visitors and arguing sales people circled the prominently placed PDP-8/E with its rattling Teletype printer leisurely churning out plots. OMNI-Smile.pngplot Next to it was a huge heap of piled-up Small Computer Handbooks, and then there was this big basket filled with red double-deck OMNIBUS toys as take-away gifts for the kids at home. With fascination and great satisfaction I kept watching visitors picking a couple of the OMNIBUS toys, then one or two of the handbooks from the pile and browsing in them before leaving the booth.

What an outstanding in-depth campaign to prove the PDP-8/E's distinguished new qualities. It also made visitors feel being most welcomed customers. I liked very much what I saw. And I was impressed by the fine job they had done, the people from public relations, marketing, sales, documentation. Having been permanently immersed in my engineering tasks and duties during the last months, I was not aware - until now - of all the preparations going on behind the scenes for a successful introduction of this product. As a prelude, they also had organized a PDP-8/E promotion for selected customers of the Bay Area, which took place at the Holiday Inn in San Francisco the week before. OMNI-Smile.png[tale 1]


The Beginning of a Long Journey

Education and Work in Switzerland

Born into a large family located in a small town in the German speaking part of Switzerland, I was fascinated by electricity as early in my life as I can remember. Any electrical installation I could spot would immediately draw my fullest attention. So from the moment I was told, that these things were done by electrical engineers, I wanted to become one of them. OMNI-Smile.png[tale 2]

Our father was a master mason, owning a small building company. Already as a first-grader, I studied all available blueprints at the construction sites and compared the building's progression against them. Any difference in detail I could spot, I would point to the crew, what they actually appreciated. To nobody's surprise, machines of any kind were also not safe from my explorer's delight. OMNI-Smile.png[tale 3]


Apprenticeship     At about fifteen, one had to decide which track to take for further education. With undisputed good performance at school, I was expected to choose an academic path. Moreover, I could count on the full support of my parents. Actually, this was what I wanted to do, although with a detour. I first wanted to get to know things, then accrue the skills to make them, and finally study to acquire the know-how to be able to create them. Against the recommendation of all my teachers and other people, I applied for an apprenticeship at Brown, Boveri & Cie (BBC) (since the merger with ASEA from Sweden in 1987 named ABB). Brown Boveri was a traditional Swiss electrical engineering company and with about 12'000 employees of impressing size and power. They sold worldwide well-known equipment and services for electrical power generation and distribution (including vapor turbines and gas turbines), as well as plant engineering for industries. An apprenticeship, lasting four years, is 'learning by doing' in a real company, supplemented by trade-related, focused studying in school - sort of an extensive work-study program. The majority of Swiss youngsters choose this track as the starting step of their professional career. In this model, craft and industry profit from a young, well educated workforce. For the country's economy it's a blessing, because this leads to very high employment of young people. At Brown Boveri, an apprentice was scheduled to change his workplace every half year within the company. This enabled him to gain insight and expertise in many different fields. The first half year was set up as an introductory course, organized in groups of sixteen apprentices for basic training in elementary metalworker skills. During the following years, the apprentices formed a close fellowship of teenagers spread across the vast company. On the one hand they carried quite a workload with responsibility at modest pay; on the other they were granted some fool's privileges as a balance. It was a place to grow-up. OMNI-Smile.png[tale 4]


Studies in Electrical Engineering     After successfully graduating from the apprenticeship program at Brown Boveri and after a two-year curriculum at a private school to attain the general qualification for university admission ('Matura') in 1959, I began my studies in electrical engineering at the Swiss Federal Institute of Technology in Zurich ETHZ. An excellent European university, hardly known in the US at that time, today is ranked as the best university in continental Europe. My preference was the electronics curriculum. Although profiting from a set of excellently lectured courses, semiconductor applications and digital circuit design were not yet among them. I enlisted in some voluntary up-to-date subject courses available at that time such as 'Transistor and active circuits'[1] lectured by John G. Linvill from Stanford University during his sabbatical year in Switzerland, Halbleiterbauelemente[2] lectured by Walter Guggenbühl and Digitale Rechenanlagen lectured by Prof. Ambros P. Speiser who founded and managed the IBM Zurich Research Center, the first research facility of IBM outside the US. My thesis partner and I had to develop the circuitry for a Tunnel-diode mixer as an experimental tool. The obtained results could then be compared with theoretical behavior[3]. OMNI-Smile.pngmixer Besides filter design, the challenging part of our task was to maintain stable operating conditions with the uncommon negative resistance property of the device, requiring a circuit layout conforming to gigahertz-frequency techniques, under the constraint of very limited availability of adequate instrumentation. This work was done within the department of 'Höhere Elektrotechnik' headed by Professor Max J. O. Strutt under the supervision of Hans Melchior. At the end of 1963 I graduated as 'Diplomingenieur (Elektroingenieur ETH)'.


My first Job as Development Engineer     I very much value my first job as a development engineer from 1964 through 1968 at the Oerlikon Bührle Company. My supervisor, a competent and experienced man, had - for good reasons - an affinity to US electronics. Books such as Radio Engineering [4], simply called the 'Terman' not knowing yet of his role in supporting the Varian brothers or Bill Hewlett and Dave Packard to start their companies and setting up the Stanford Industrial Park, then the Reference Data for Radio Engineers [5] nicknamed the 'bible', belonged to our arsenal, together with a number of pertinent US magazines. Besides, most electronic parts used in our products were of US origin. We also had efficient access to any new semiconductor part of interest, which showed up in the US market at that time.

An extensive range of applications from low-level/low-noise IR detection, charge amplifier design, magnetic sensing, 20MHz counters for velocity measurements (still done with discrete transistors), to servo-control systems, opened up a versatile engineering field for a novice. Starting with a technology of discrete semiconductors (even with germanium, before silicon took over) and proceeding in using small-scale (SSI) and medium-scale (MSI) integrated circuits, allowed us to stay ahead of the crowd. Besides, the equipment we built had to withstand harsh environmental conditions, most notably an operating temperature range from -55°C to 70°C and vibration-sustainability according to standards like MIL-STD-810A. Generally the name of the game was "design products based on straight-forward solutions and simple handling; apply unconventional methods to reach the goals if useful". Here an example of the latter: In order to cut costs and save time on the expensive use of vibration tables before the actual validation could take place, we shipped prototype equipment from Zurich to Amsterdam and back by rail. Unpacking the freight would instantly tell even to a novice what went wrong and where improvement needed to be done. Development work and evaluation of experiments required a good deal of computing. I had access to our in-house IBM-360 system using FORTRAN-IV language programming, doing debugging by dissecting memory dumps if necessary. Over time, I got ample hands-on experience to fulfill my computing tasks efficiently, thereby gaining competent programming skills and fairly good understanding of system use.

Anti Aircraft Tank Gepard with twin 35mm 600 rounds/minute guns each. Notice the two muzzle-velocity measuring bases integrated on the muzzle brakes at the ends of the gun barrels.

My most important and largest project was the development of the Muzzle-Velocity Measuring Equipment FMV312 designed for the Anti-Aircraft Tank 'Gepard' - a self-propelled anti-aircraft weapon based on the company's twin 35mm, 600 rounds per minute guns. Besides the actual magnetic stimulus/sensing circuitry including required power supplies, it contained a special-purpose processor with an embedded 15-stage high-speed counter/accumulator for clocking the temperature compensated fly-time[6] of the projectiles through the measuring base. It also included the computational means to check/arbitrate plausibility of the measured velocity and finally average the results per round. The result was fed to a ratio-metric analog/digital converter, imposing the velocity deviation onto the 400Hz signal of the fire-control system network. This equipment was entirely built from discrete components. The simulation and verification equipment belonging to the system was built using the newly available 5400-series TTL components.

I still feel privileged, that I could start my developer's career at - in hindsight - one of the most advanced electronics applications laboratory in Switzerland at that time. A footnote: The still compulsory military service at the Swiss Army was of some importance for me during the years 1957 through 1967, since I had to squeeze in most of my service time during vacations while studying. The some 690 days I spent on service duty, most of it as an officer, were useful nevertheless. I was trained to prepare an action, set up the necessary commands and then direct/control their execution.


Travel to the New World

Immigration and Work in the US

On June 6th 1968, I stepped out of the air-conditioned airplane in New York and found myself stuck in a stove - such felt the searing heat I had never before experienced in my life. "And I have come here to work?" was the anxious question I asked myself. Nevertheless, I felt happy, knowing that I would get picked up by a friend at Logan Airport in Boston. Our wives had spent time together in the same hospital room after delivery of their babies. This led to a loose friendship between us. My friend was just ending his employment at Polaroid Corp. in Cambridge - his family had left Boston already - to go back to Switzerland to join his father's business. I took over their rental apartment and bought their modest household inventory. I was lucky, now I had my place to stay, even including my own phone number.

In January 1968 I had applied for immigration visas together with my wife, our son of two years and our daughter of nine months expecting to wait in the pipeline for about one-and-a-half years to get the visas granted. We really got caught off-guard when we received our visas already in April of the same year. We also had expected some difficulties, not being able to name any relative or sponsor in the US, and having at our disposal only a tiny budget, which would barely allow covering all my expenses living in the US for four weeks plus buying a ticket back home, just in case. (For one dollar you had to spend 4.30 Swiss francs in those days.) My wife, born and raised in the same town as myself, had always loyally supported me. She consented to take the opportunity and face the risks involved by starting already the next day to organize whatever had to get done in time. And there was simply no time left worrying that I had rendered signature at the US Consulate General in Zurich, acknowledging that I could get drafted to serve in the US Armed Forces. (American involvement in the war in Vietnam had peaked that year.) And, of course, we felt relieved when I was assigned a draft status beyond any liability after I had been summoned by the Selective Service in April 1969.

After being settled - so to speak - I immediately began looking for a suitable employment, partially with the help of a placement agency. During the job search I got a pretty good overview of all sorts of electronics companies being located in the greater Boston area at that time. Although still alien to my new environment, I was really confident concerning my professional abilities. People appreciated that I took the risk of immigrating with my family and treated me favorably with openness, respect and trust. I was never ever subjected to any kind of harrowing tests or whatsoever. To give the interviewers some insight of my earlier work, I had the full set of schematic drawings and the detailed technical descriptions (authored by myself) of my FMV312 design at my disposal. I am still grateful about this most generous attitude by my former employer, granting me permission. After a three weeks run, I could choose among five offers. OMNI-Smile.png[tale 5] On June 27th I got formally employed (badge 5609) by DEC as design engineer reporting to Roger Cady of the PDP-8 engineering department. I had taken the first hurdle, and was relieved being able to inform my family to finally join me in our new life and place.

As a starting project, I had to design a PDP-8 interface to a commercial incremental tape (½") transport system. This gave me the opportunity to get to know all I/O-bus-related issues and the handling of PDP-8s. I also got familiar with all sorts of redundancy checks and error correction schemes. Following was a cocktail of smaller circuit-design tasks such as the development of a set of bus conversion modules, designing a very low-cost auxiliary power supply, hardening the design of disk read/write circuitry for production release, doing special debugging tasks on disk systems. OMNI-Smile.png[tale 6][7] Doing all this work I could expand my circuit design arsenal by progressively collecting know-how of the entire assortment of DEC modules. Meanwhile, I got quite familiar with the DEC organization in terms of knowing how to get things done efficiently. OMNI-Smile.pngmill Before I got involved with the PDP-8/E project, I was assigned the challenging design work of the read/write electronics and control unit for the TU10 Tape Transport (2'400 feet, ½", industry standard, 45"/s, 7 or 9 channel). I applied an advanced scheme of quasi-logarithmic sensing, thereby alleviating the effects of the dynamic vacillation of read signals, and also reducing data skew. The control unit, a master-slave concept, was built using a production- and maintenance-friendly packaging scheme.


The PDP-8/E Design Story

Development of the Basic Machine:   The roads were smooth, the days were bright, the spirits high !

The front of a PDP-8/E (PDP stands for Programmed Data Processor) is actually a programmer's console consisting of a switch register for input, an indicator bank to show the register contents selected by the rotary-switch, and a few switches like 'clear', 'deposit', 'examine', 'halt', 'continue' for control: A very simple man-machine interface, indispensable for initial hardware/software testing.

"Like many successful industrial designs, the PDP-8 was actually a redesign." This quote is from the well-balanced, surprisingly informative, fun to read, testimony "The PDP-8" by John Voelker, published in IEEE Spectrum's 25th anniversary issue. [8] (An article, enriched with the protagonist's quotes, still deserving top rank recognition.) Voelker's account gets to the point of our engineering task regarding this project. It praises the achievements of the people who had created, developed and launched the PDP-8 in the past. OMNI-Smile.pngpdp8 We were entrusted to refine and enhance it for the benefit of the present and near future. I felt honored to be part of such a team. The PDP-8 history released us from arbitrating architectural system varieties. Instead, we had the obligation to preserve and guard compatibility for previous users and to breed substantial progress in order to generate new sales. Interestingly, the public perceived it as a totally new machine.

A key factor for a rapid realization was the OMNIBUS concept, setting a base for truly "concurrent engineering", a buzzword of the 1980s. (Which could only draw a weary smile from someone who was witness to this PDP-8/E adventure.) Marketing was involved from the start, soon followed by production engineering, software diagnostics and finally maintenance regarding functionality, producibility, assembly, testability, and maintenance issues. And only by doing procurement chores first by engineering, before extending to manufacturing, was it possible to ship 900 machines after just one year from the start.

Starting with initial sales of 100 (PDP-5s) for 27'000 $ each in 1963, by the end of 1967 sales of the PDP-8 and predecessors summed up to about 7'500 (at a final price tag of 12'800 $). The rate of 900 PDP-8/E sales for the first three production months (price <5'000 $) was a promising start for the contribution to the final sum of some 190'000 PDP-8s sold.

Following is an overview to help classify the various types of this unique 12-bit word machines developed and built from 1963 till the end of the 1980s. (With the emergence of the 16-bit word machines the word-format of the PDP-8 - not conforming to the established 8-bit byte mainframe standard [1964] - was considered outmoded for tangible reasons.)

  • The original type PDP-8 was the first successful commercial minicomputer, and it was built using discrete components. The PDP-5 was its predecessor. The PDP-8/S was built as a compact tabletop unit with serially (vastly slower) working CPU.
  • The advent of Transistor Transistor Logic (TTL) allowed to implement the PDP-8/I with medium-scale integrated (MSI) circuits, featuring better performance than its predecessors at lower cost. But its design, still using small Flip Chip modules and a wire-wrapped back panel instead of a more adequate packaging scheme, did not fully exploit the new technology. Burdened with this handicap, it could only serve as an intermediate solution. The PDP-8/L was the compact tabletop version with similar performance but limited capacity.
  • Still using the same MSI circuits, a fundamental change was the introduction of a new I/O bus, the OMNIBUS, together with the striking modularity of the PDP-8/E packaging scheme. It allowed customers enormous flexibility in expanding their systems' peripherals as well as reducing cabling and packaging requirements. The PDP-8/M and the PDP-8/F OEM versions could be best described as a half-size PDP-8/E (one OMNIBUS backplane only) with identical performance. The later PDP-8/A model was equipped with semiconductor memory chips and its circuitry was condensed on larger boards.
  • With the CMOS-8 series, very-large-scale integrated (VLSI) single chips, the 6100 and 6120, were used to implement the PDP-8 architecture into hundreds of thousands of VT/78 and DECmate video data processors and workstations.

For general information about the PDP-8 family of computers and for detailed technical documentation of the PDP8/E design refer to: OMNI-Smile.png[tale 7] [9][10][11][12]


The Project     The Project started in February 1970 with a bang. Everybody was caught by surprise when Don White came up with a fully manufactured OMNIBUS assembly and a PDP-8/E specification. This 25-page document was written in a hands-on style for insiders, but was extensive enough that work could get started immediately. This boosted morale of the group tremendously - and this was good. Since our engineering manager Roger Cady had left to direct the development of the first PDP-11, our group lingered in some sort of an interregnum, and was gradually losing orientation.

The OMNIBUS was capable of accommodating up to 20, 38 ,56, or 72 modules depending on the number of the 20-slots assemblies used (an Expander Box was necessary if more than two were required)

In almost no time a mock-up of the new machine was posted. This quickly became sort of a forum for technical discussions among members of engineering and marketing, which was located on the same floor right next to us. The self-evident, self-explaining PDP-8/E concept, paired with its unique industrial design solution was an instant hit. Marketing people and their long-lasting customers were very much relieved at first; soon they got outright enthusiastic about their newly won potential to fulfill their needs. Only a short time before, they were stuck in a strategic deadlock. Attractive alternatives, from external competition as well as from the newly launched PDP-11s had begun to show up. But unattractive prospects of software porting for any of these alternatives clouded their future. With the then existing PDP-8/I models - although equipped with the up-to-date MSI circuits, but penalized by its poor packaging scheme in terms of size, cost and expandability - DEC was not in a position to counter some of the very efficiently built equipment from competitors. In other words, a large, loyal customer base felt abandoned. Most likely, many of them could have got lost as DEC customers for good. For them the sky cleared up; for us it rendered a foretaste of things to come in terms of mounting deadline pressure.

Development began with only Don White together with Lou Klotz working on the CPU and myself on the memory. It was a reflective phase with little distraction, rendering rich results. Just in time we got our new manager Dave Chertkow, hired from outside the company. Together with his assistant Ken Pierce and our indispensable, caring secretary Penny Smith, our group's top was complete again. Dave took the helm in no time. Soon he assigned fitting tasks to PDP-8 technicians and brought new people on board as required. He did a great job keeping our engineering potential at its best, by alleviating us from any chores which could get done by somebody else. He acted as a gifted promoter of the PDP-8/E OMNIBUS concept by actively supporting marketing to woo new customers. He then launched numerous OMNIBUS options, allocating whatever resources seemed fitting to him.

For us developers the overhead was kept absolutely minimal, there were no schedules, only deadlines ! Although we did not feel any budget pains, we sure had to pinch out any possible penny from our designs. Communication was simple and direct. I cannot recall a single design meeting during this period except the official Design Review. OMNI-Smile.png[tale 8]

Design Reviews were held by the design team under the auspices of the chief engineer. One had to win the attendance of persons from marketing, production, quality control, software diagnostic, and maintenance. The individual from production was to take minutes (for good reasons). Our Design Review was convincing and ended as a success. However, a cost reduction was requested, and we were given ten days to realize it. Not everybody was happy about that. Regarding the CPU it meant that they had to take out a few of the goodies they had proudly presented. I considered it a gift to have a few extra days available in order to get rid of e few irritating inconsistencies in my design and to have the legitimacy to further search for some better and cheaper components. OMNI-Smile.png[tale 9]

PDP-8 engineering was special: Practical expertise, team spirit, willingness to put in extra efforts, mutual recognition, maintaining fresh humor and be good sports were the name of the game. OMNI-Smile.png[tale 10]


CPU and Basic Peripherals     The design of the Central Processor Unit (CPU) consisting of Major Registers M8300, Register Controls M8310, Timing Generator M8330 was done by Don White in collaboration with Louis Klotz, a production veteran. The two perfectly complemented each others PDP-8 CPU know-how, warranting compatibility and still providing maximum possible functional extension of the basic machine. Certainly no easy walk - but Louis the "living simulation robot", playing-off his debugging virtuosity, made life a lot easier. Don also took care of the Bus Loads M8320 and Programmers Console with assistance by Greg Esser. The Teletype Control M8350, functionally a UART device built of MSI components, designed by newly hired John McNamara, completed the basic machine's board set (memory excluded).

Also under supervision of Don and Louis with the assistance of Larry Narhi, the high priority External Bus Interface Control Options KE8-E/KD8-E for Programmed Input/Output and for Direct Memory Access (Data Break) as interface to PDP/8I and PDP-8/L type peripherals (legacy type hardware) were realized. Besides - after he had already completed the schematic of a Switching Power Supply - Don prepared the technical part for the outsourcing of a Linear Power Supply and supervised the mechanical issues of the equipment box with the support of Paul Gardner.

PDP-8/E Timing Generator for processor and memory signals.


Timing     Working on the memory design I was compelled to define the machine timing in full detail. In his specification Don White had defined some very useful timing details, such as changing time state in the center of a 100ns pulse. He had suggested using delay lines for implementation as it was common practice. Rather than start a question/answer ping pong, I decided to communicate by sketching a possible circuit, reflecting the exact timing scheme I considered adequate. On this occasion I also designed the circuits I felt apt for properly driving the OMNIBUS signal lines including the appropriate termination. The kernel of the gear was a serial shift-register driven by a 20MHz crystal oscillator. OMNI-Smile.png[tale 11] Don White's response was simple. He pasted my timing proposal, unmodified, right onto his Timing Generator M8330 schematic and transferred the termination details onto his Bus Loads M8320 board.


Memory Expansion     When Dave Chertkow asked me to design the KM8-E Memory Extension and Time Share Control module I felt relieved. Development time was running out for this important option, which provides the CPU with necessary means to control memory access beyond its intrinsic 4k memory space, up to the maximum of 32k. Moreover, some minor logic for time-share systems had to be included. While designing memory and timing, I was forced to very carefully take all circuit delays into account which possibly could affect address and data signals. I had no choice but to mentally evaluate the appropriate signal paths required by this option, to make sure things would match. In other words, I preferred doing the design of something I knew already how it was to be implemented, rather than taking any chances. This piece of hardware is not too complex. The contents of two three-bit registers (for insiders: Data Field, Instruction Field), preloaded under program control, are to get directed through a machine-state controlled multiplexing scheme to the drivers of the three address lines required for the eight-fold extension of the address space. Transforming this logic into a schematic was business as usual for me at that time. Then, together with Dick Madden, we verified this partial design with that of the existing predecessor, the PDP-8/I, by marking up all the functionally matching circuitry on the relevant schematics. The remaining loose ends led to the modest arbitrating logic for directing address information in function of relevant machine states. We ported the identified arbitrating circuitry into our design without even bothering to check its functionality, completing our design in less than a week's time. Now Dick took over to have this option prototyped. Debugging it, he had to correct two mismatched connections of the arbitrating logic.


Memories

On the historical finish-line of a 20 years old technology or Core Memory's last battle, pushing its success to a veritable sales surge before surrendering to the semiconductor newcomer

The standard PDP-8/E memory is a non-volatile random access, coincident-current magnetic read/write core memory. Its cycle times are 1.2 µs for read-only access (fetch and defer cycles without autoindex) and 1.4 µs for read/modify/write access (all other cycles). Its storage capacity is 4096 (4k) 12-bit words per unit. The memory is expandable in 4k increments up to 32k words. The small .020" toroidal (ring-shaped) magnetic core storage elements are mounted in a 3-wire planar configuration.

Shown here is a dismantled 4k x 12-bit word MM8-E Memory consisting of the Sense Inhibit G104, Memory Stack G619 and XY Driver G227 boards. This souvenir set served as the author's personal production release reference system. Through all the years the author has proudly put up this set for display in his office.

Unlike the price/size/speed/capacity race for non-volatile bulk or mass storage which later took place between magnetic (disk) and semiconductor (flash) technologies, the race for work memory between magnetic (core) and semiconductor (DRAM) devices ended before it even had started. But due to the special circumstances, DEC enjoyed a last surge of a very profitable magnetic core memory production, which lasted for about four years. Reliability and outstanding low cost-to-sales ratio - one of the best guarded secrets at DEC which very few people knew about - enabled postponing the surrender to the semiconductor newcomer for the benefit of the company (not just in terms of profit). The first feasible semiconductor bulk memory chips, the 1103 1k-bit PMOS dynamic RAM IC was first available for prototyping in 1971. Reliability under required operating conditions got not reached until about 1972, and attractive pricing for DEC was not achieved until 1973. OMNI-Smile.png[tale 12]


State of the Art     At the time there was no feasible alternative to magnetic core memories for non-volatile, fast, random read/write access. With regards to packaging and testing - important aspects for efficient, reliable production - convincing implementations were still lacking. I was not enthusiastic about what I saw, including the new designs. Apprehensive about crosstalk, the common belief, that low-signal sense circuitry should better not be placed on the same board with 400mA current-drivers, led to unwieldy packaging schemes, including twisted-pair wiring. Testing and tuning had to be done by well-trained technicians in time-consuming procedures. Core memories appeared to be too costly, indispensable analog/digital devices, at odds with the digital world.


Theory and Research     Without any particular intention, since fall 1968 until the end of 1969, I had collected and partially studied every technical article about core and semiconductor memories which I ran across in magazines. I just wanted to be informed in a field, which seemed very important to me while working for DEC. When I was asked to design the new memory system, I immediately started to expand my research by ordering copies of all relevant literature I could identify (50 publications ! OMNI-Smile.pngcores) in the excellently researched and informative book 'Digitale Rechenanlagen'[13] by Ambros P. Speiser. I began with the systematic combing of these papers in search for fitting know-how to allow for simplifications on the design and also to enhance reliability. Accompanying the actual design work, I focused my studies on related topics available in the papers. And, of course, I did carefully study the just completed PDP-11 memory design. Combined with my knowledge acquired in my pre-DEC work, I indeed felt well prepared to reach the goals I had set myself.


Design     While doing design work, I made crucial decisions, and I also set up and refined my own planning:

  • Stay with the core size .020" for a 4k stack. Put aside the idea to go to the reduced .016" size for a possible 8k memory. Keep this tempting possibility for yourself in order to avoid stirring appetite in marketing.
  • Use as many parts from the PDP-11 memory fitting my design, notably the same core type. Using a proven core technology for operating a 4k by16-bit word memory (PDP-11) but applying it for 4k by 12-bit words (PDP-8) only, rendered extra solid margins (+25%), making the 'no-tuning' goal achievable.
  • Devise a scheme avoiding individual interdependence among the three memory boards. Interchangeability must be kept intact under all circumstances.
  • Devise a connection layout such that boards of a dismantled system open up connection loops, allowing a memory tester to interfere inside some specific circuitry. Enabling for instance varying xy-currents, sense thresholds and so on.
  • Do not tolerate the use of any potentiometers. None of that tweaking and tuning anymore ! (Some four to five trim elements were commonplace on core memories.) Tuning-potentiometers facilitated a hasty designer to hide incompetence, inexactitude and incompleteness behind it, shifting undue burdens to production and maintenance. (Also a manager's nightmare: 'Always almost finished' !)
  • Make provisions that initial CPU and memory testing could be done independently. Therefore, build two PDP-11-based memory units, one with an OMNIBUS-compatible interface reserved for CPU testing, and one with an interface, matching requirements of off-the-shelf memory exercisers.
  • Include as a design feature a concept of an automatic memory testing.
The author (left) in a memory test session assisted by Dick Madden. (Extracted from DEC Public Relations Department: "on line" monthly publication August 1970, p. 2[14])

This elaborate planning was a result of my permanent concern of how this piece of equipment would get produced, tested and how it would succeed in the field. Now I was ready for the actual design. The only design blip requiring academic level skills was the computation of the optimal read-wire termination using the Laplace transform method. The rest was doing puzzle work, minimizing parts count, slashing costs wherever possible, standardizing, harmonizing, figuring out testability, forming layout details, making it look good. A sort of 'form follows function' process, I loved to do.

What might appear a straightforward three-board memory set is by no means self-evident. The then common practice initiating layout work by 'Give them the schematic, they will do the rest for you.' would have been unrewarding in this case. Considering locations of signal lines on the bus, there existed one and only one solution for the proper placements of the components on the three boards. It boiled down to a three-dimensional puzzle only the design engineer himself could reasonably put together. Having done that, it was essential to prepare for a smooth course of the layout work required. I informed the supervisor of the model shop in advance, winning him over to schedule one of his top layout technicians to do the whole three-board memory set. The ability to place components in a layout advantageously - also an aesthetic matter - distinguished a top man from others. It was vital to instruct this man very carefully, making him forget that his expertise, which made him top, was not sought in this case, making him accept a set of rules for dealing properly with low-signal sense circuitry mixed with high-current drivers - and still arouse the necessary enthusiasm for achieving an impeccable result. All had to fit in a first pass. Underestimating these aspects - as it happened with the CPU set - could lead to disappointment, seriously endangering deadlines.


Prototyping and Test     My assistant, Dick Madden, had designed, built and checked the two temporary memory units, one with an OMNIBUS-compatible interface reserved for CPU testing, and one with an interface, matching requirements of the off-the-shelf memory exercisers in time. Besides, he helped me perform pretesting of some components and circuit parts. Once we had got the first prototype memory boards from the module shop, we could immediately start testing. Making steady progress, asserting expectations, was gratifying. For the sake of additional cost reduction we also tried an alternative scheme which would have substantially cut xy-driving circuitry. It essentially worked, but marginal recovery of the driver transformers, or alternatively requiring longer cycle time, urged us to abandon the alluring bonus.

This fully automated memory test station performs dynamic testing of the Sense Inhibit G104 / Memory Stack G619 / XY Driver G227 board kit as well as the assembled MM8-E memory system unit. Tests such as 'address', 'data' and 'worst case delta noise patterns' are run as well as typical operating characteristics under control of a PDP-8/I computer. On certain tests, xy-currents, read detection level and sampling time are varied within and beyond the defined operating field limits. The corresponding results are then churned out on Shmoo Plot charts as vivid certificates of the attained operating margins. This five minute per unit testing technique allowed no variation in quality; no marginal units survived these tests.

Later, one incident almost wrecked this success by forcing delays, setting unrest, anxiety and endangering our reputation. Our preproduction batch of memory stacks, some 25 pieces, were supplied by our preferred stack vendor, with whom I had successfully completed the stack design in collaboration. To maximize profits by enhancing yield at lower cost for prospective volume sales, they could not resist the temptation to recklessly change the production process without properly informing us. The result was a disaster. These stacks were tainted by magnetostriction, thereby generating not easily identifiable disturbances. Fortunately, we were already in the process of gearing up three additional sources, which allowed us to readily reduce the supply gap. OMNI-Smile.png[tale 13]


Parity Option     I did not mind when the politely formulated assignment to design this option fell into my lap. I knew I had to squeeze in this unplanned job somehow, but I also knew it was rewarding work I had superficially envisioned before. This three-board set would contain a regular stack board of which only eight of the twelve bit-planes would get used. Each of the bit-plane used would provide the necessary 4k of a 13th bit per 4k memory field. One MP8-E Memory Parity option would therefore serve the whole address range of 32k words. Since the field-select circuitry was located on the Sense Inhibit board, providing the resulting select signal to the xy-Drive via the top connectors, the xy-Drive board could also remain standard. Therefore, only the Sense Inhibit board needed modifications. Sense Inhibit drive circuits had to get curtailed from twelve to eight, providing ample space for parity and input-output logic. Proficient in drawing I rapidly set up a detailed schematic. Again, Dick Madden took over to have this option prototyped and tested.


Memory Tester     Bob Regan became responsible for the whole MM8-E Memory tester from design through build and test, as well as for introduction/training of production people. Concept and specification were defined in a few informal meetings with me. He was also encouraged to tap know-how from anybody he felt fit to help him get the work done. He sure delivered a fine result.

The tester had four distinct ports, three for each of the single boards of the kit and one for the whole system unit. Including the respective 'device under test', each of the three board-ports consisted of a full memory unit. In each case two of the boards were part of the tester. With a clever wiring/connection scheme these ports had simultaneous connections to the regular, tester-manipulated OMNIBUS, as well as to some distinct circuitry to control xy-currents, read detection level, sampling time and operating voltages. For the system-unit-port a regular OMNIBUS access was sufficient, since the complete unit was run under normal operating conditions but with all relevant diagnostic software. The tester-specific kernel actually consisted of slightly modified PDP-8/E memory components. The rest was a standard PDP-8/I equipped with some special interface hardware to operate the OMNIBUS plus some auxiliary off-the-shelf equipment.

This integrated test concept, combined with a restrained performance/capacity design policy, was a key factor for the outstanding reliability and the extraordinary low cost-to-sales ratio of this core memory.


The OMNIBUS Vehicle

What it's all about

The OMNIBUS - conceived in engineering team collaboration - consists of a number of connectors mounted on a printed circuit board (PCB), replacing the then commonly used wire-wrapped back panel with its thousands of wires. Each pin-number on any connector is associated with the same specific signal line. Thus, any module may get plugged into any of the connector-slots. If a functional unit requires more than one module (such as the memory), connectors on the top of the boards serve as internal links. Where required, connectors on the side of a module allow inserting cables to the "outside world". This scheme - adapted in countless equipment over the years - eliminated once and for all the thousands of wires previously required on back panels of preceding models. During implementation of this bus, there was plenty of headwind to be endured, coming from many directions, ranging from skepticism to outright disbelief, until hundreds of machines got out the doors. Don White knew I was his dauntless ally. He had been quite impressed at my job interview when showing him my own reference design, where I had already applied a partial bus system, substantially reducing wiring and enhancing functional modularity.

The 5'000 wires required in a PDP-8/I back panel (right) are eliminated in a PDP-8/E OMNIBUSTM back panel.

Already at the start of the project, Don had laid down the full definition of the OMNIBUS. Alterations were never required. And most important for a smooth project start, he had already prudently assigned the physical location as well as the logical functions of all bus lines. After I had taken care of the timing generation with the further electrical bus details, I felt being in charge of fostering any electrical issues regarding the OMNIBUS concept.

Its uncompromising concept of unrestricted free handling was of pioneering quality. It was also a proof that it worked with straight back-plane bus connections only. Moreover, because of the foreseeable progress in integrated circuits any future design could be regarded a subset of the OMNIBUS, in terms of the number of signal lines as well as its expanded length. The functional and logical organization was pragmatic, parts-saving, performance-squeezing and unabashedly application-friendly for programmed I/O applications. The set of the signal definitions used was not at all suitable for future standardization. It could only serve the then prevailing technology applied in a synchronous PDP-8 structure. Once medium-scale integration would get replaced by denser circuits, this approach would only survive during a transitory phase.


Electrical Characteristics     Against all skepticism due to the staggering number of signal lines (96), the OMNIBUS scheme was hard as a rock in electrical terms.

  • The OMNIBUS assembly as such, is of an electrically homogeneous, low-impedance design.
  • The interconnections of all ground pins (31) on a board complement longitudinal wiring with the appropriate cross-connections to form an almost ideal grid, growing in size simultaneously with a growing number of boards.
  • Clustering of all register signals of one sort in one place was diligently avoided. As a worst case up to 15 high-to-ground signal transitions could take place within a few nanoseconds time, forcing tremendous displacement currents to be absorbed without generating too much noise.
  • With the address-select scheme applied, signal switching would only take place on the selected module, the rest staying calm, therefore dramatically reducing switching noise. Example: A 32k memory size would occupy 24 slots, but on any access, possible switching activity would be curbed to only two slots (the stack board having no signal connection to the bus).
  • The load-relief technique applied by the device-select scheme kept DC-loads low even with numerous options installed.
  • The synchronous, determined timing scheme, providing well defined setup and clocking signal transitions, was averting any confusing or unpredictable signal patterns thereby banishing switching noise at incongruous times.
This comparison among the bloated OMNIBUS (96 signals) and a "global" type bus (some 36 signals) illustrates the effects of address/data multiplexing and the hiding of "internal" signals.

Diligence was required to maintain bus integrity. In case a designer used the interfacing circuits of one of our basic options as a template, most everything was ok. However, I felt obliged to observe designs in a watchdog-like manner in order to prevent compromising the OMNIBUS scheme in any way. Although I was keeping notes, it was not before the middle of 1971 that I finally got free time to write down the binding electrical OMNIBUS Specification. OMNI-Smile.pngspec


Functional Perceptions     There is no such thing as an ultimate bus system. The OMNIBUS was just one large, versatile, externally accessible data multiplexer system with inherent synchronous timing features. Some special control signals allowed also asynchronous operation for peripheral options but not for the memory. Besides carrying "global" type bus functions like doing address and data transfers, it provided all sorts of normally "internal" control signals for accessing CPU resources. Ultimately this assured warranty for full compatibility to any earlier internal application. In fact a bus participant was able to run the machine in "phantom" mode, autonomously forcing instructions. For normal applications, a designer needed not to understand those "internal" signals. In fact, he better disregarded them, so as not to get unnecessarily confused. The separate memory data signals MDnn, somehow redundant to the data line signals DATAnn, give rise for further explanation even in retrospect. Considering earlier PDP-8s (even PDP-5) - still built with discrete components - this scheme allowed to save expensive register-banks (flip-flops). It also rendered the benefit that the memory data MDnn - in metamorphosis - served as addresses for I/O instructions. At that time, a harmonized structure using memory-mapped I/O would have led to excessive costs. OMNI-Smile.png[tale 14]

With large-scale integration (LSI) being in sight, we put in efforts to be prepared regarding an adequate bus system detached from any PDP-8 structural handicaps. The preliminary bus system proposal OMNI-Smile.pngx-bus, employing address and data multiplexing as well as matching integrated circuit packaging requirements, offered a feasible scheme for implementing a CPU using large-scale integration (LSI). It pioneered the popular Q-Bus (also known as the LSI-11 Bus) first used on the LSI-11, then also on VAX systems.


The PDP-8/E Arena

About struggling to success:   Bumpy roads, heavy traffic, stormy weather, nothing would stop the trek to go on and on !

The days of more contemplative design phases were over for good. More and more time was spent in the lab in a busier, more heated (i.e. not air-conditioned) atmosphere. Gradually, the entire PDP-8 engineering crew got also involved: Al Czajkowski, Al Deluca, Ashwani Chadda, Remi Lisee, Owen Fisk, and Bill Hodgdon doing development of major peripheral controls. Then, versatile technicians like Mel Arsenault, Rod Sutherland, Bruce Hanson, Paul Kotschenreuter, Dave Adams, Mike Kozak were doing prototype checking, building and checking the seventeen preproduction machines scheduled for delivery during September 1970 to prime customers, and, if required, would take care of environmental testing and quality control. Finally, there were frequent guests like diagnostic programmer Ed Steinberger and system programmers Chuck Conley and George Berry or Stephen Kallis Jr. writing on one of his timeless, vivid technical booklets.[15] Necessary and increasing interaction among all these people became sometimes challenging. Uneasiness prevailed, that management had perhaps not fully anticipated this unavoidable change of the work pattern and style, and that appropriate provisions seemed to be lacking.


Test of the Basic Machine    The actual tests of the basic machine went according to plan. By the end of June, the bench prototype (using PDP-11 memory) was performing as expected and could soon be presented running its own memory. Next, those high-priority External Bus Interface Control Options, such as the interface to PDP/8I and PDP-8/L type peripherals were mastered. It was absolutely indispensable to perform diligent and comprehensive checks on all related peripherals still for sale in this new constellation. Things worked out to our satisfaction except for one peripheral device, which was using direct memory access in a most stochastic fashion. Every once in a while - hours, days - an error would show up on a fully loaded disk drive not directly related to the dubious equipment. What a nightmare ! It appeared as a "random" error, but nothing is random in computers. Any malfunctioning is due to flawed designs or marginal components. Catching this bug devoured most of our capacity for a long time, a too long time under the circumstances. While I was plotting trap programs, Louis was virtuously playing the programmers-console keyboard implementing and setting them up. Then, we were compelled waiting for a next error event. Working step-by-step, again and again, like hunters, until we caught the culprit. The designer of the faulty equipment had allowed himself to spare full decoding of memory addresses, inexcusably compromising common ground.


In Charge     Returning from the 1970-WESCON exhibition after a two weeks absence from the Maynard plant, I noticed a team with spirits as low as never before - the contrary what I had expected after this successful launch of the new machine. Of course, meanwhile the deadline pressure had built up. But the real cause for the black mood was that adequate supervising of the group by engineering management had ceased dramatically. Furthermore, due to some medical problem, Don White had to retreat from everyday business. In other words, the group was essentially missing their two leading individuals. Now, the alarm bells started to ring ! Alone the thought that all this great work could end up being compromised, and that instead of a humming production floor there would be doldrums, made me feel absolutely sick. Lamenting was not my thing. Avoiding any bad talk, I got to work even more focused than before. I actively began to support my fellow workers solving technical hitches thereby prodding them for crisp, system-consistent solutions. Gradually they cheered up.

Shortly, one day during September 1970, Dave Chertkow came to see me while I was working in the lab. He asked me directly, if I would be willing to take over as Project Engineer and serve as interim PDP-8 group leader, replacing him in his role. He himself would switch to PDP-8 marketing, a position - I knew - he was destined for. I gave him my immediate 'Yes'. I did not even ask on what terms my new position was to be compensated. As it was to be expected, DEC treated me very fairly. After this five-minute conversation, I was the appointed Project Engineer for the entire PDP-8/E development. In this position, I was responsible for getting the PDP-8/E into mass production. This included directing the development of internal options and peripheral controls from design through production release. That was a pretty hefty responsibility. In carrying out the above assignment, I had to directly supervise some 10 engineers and technicians. Then, I had to be accessible to the many persons of all disciplines which expected professional support when addressing their mostly technical issues. I did not feel this to be a burden. Having unrestricted control over a project you are one hundred percent convinced of, can be gratifying. Also, my professional and military background in dealing with all kinds of people, proved helpful. First, I had to straighten out a few personnel issues. Some members of our crew came under fire for justifiable reasons concerning results they delivered. However, it would have been unfair curtailing their professional careers, considering the apparent supervisory deficits reigning during that time.

During the fall I could attend a Supervisor Training Course given by Harvard's Larry Bennigson. In small, team-oriented case studies I could acquire some useful managerial skills. Also, I was introduced into the company organization by DEC management. To my surprise - against daily experience with rather chaotic organizational patterns - important rules and demanding processes existed as guidelines in a most detailed fashion. One was urged not to let subordinates know about these written rules. Nobody's individual initiative should get strangled; nobody should feel entitled insisting on rules rather than do what had to be done.


Getting it into Production     A few days after my designation I found myself, alone, answering questions to some 20 people in a production meeting. It sure was legitimate to exhaustively ask for first-hand information about the project at this point. Nevertheless it was a tough piece of work that morning. But I never blinked. I already had had the chance to build trust with some of the individuals present at that meeting. They knew me being always supportive, not letting them down. They knew I was sincerely proud of their PDP-8/E assembly line, anticipating lots of machines going out their door. The meeting got closed with a sense of mutual understanding, respect and helpfulness. OMNI-Smile.png[tale 15]

When building and checking the seventeen preproduction machines by engineering, we let Bill Miller and his people from production engineering participate in the process for early familiarization. Then, production began manufacturing boards before engineering had granted production release, thereby shouldering the burden of having to supplement some of the boards with wired fixes. They had no other choice. They needed production parts early enough to build up some inventory, check out production testing facilities and train their personnel in time to keep the monthly shipping deadlines from December 1970 through February 1971 with 200, 300 and 400 machines.

I was anxious to provide production with our fullest support possible. During the whole introductory phase of several months, I had situated my guys in fitting spots on the production floor. In a mix of training production personnel, helping them test first batches, upgrading test equipment and trimming procedures, they could ensure that learning was efficient and that communication with engineering was always possible. Dick Madden and Bob Regan were assigned to memory testing, Dave Sari for processor testing, Jack Grieve for automated product testing, and Tom Pittman for the acceptance line. During all these hectic months, a smooth, fruitful collaboration under a most trusted spirit took place between production and engineering.


Options, Options, Options     Options mushroomed, initiated not only by engineering but also by marketing to cover their customers' needs. The self-explaining PDP-8/E concept and its clever industrial design, offering common infrastructure as well, had reduced overhead and enormously enhanced capabilities for applications. Apparently, the developers' inventiveness and creativity got stimulated so intensely, that the total of standard peripherals culminated finally at over 70 items. Although some of them got rapidly obsolete due to accelerated progress in semiconductors.

At the beginning of December - the month during which the first 200 machines got shipped - the prospects for timely delivery of options were mixed. The four basic options, designed by the kernel team, had either been already released for production, or they were under final rework. Regarding the sixteen major peripheral controls, done by our crew, the delivery forecast varied between ready, timely, and two months.

A then unknown number of options were in the pipelines beyond the sphere of the PDP-8 product line. I was determined to funnel all of them through a technical release procedure controlled by PDP-8 engineering; nothing would escape the alert eyes of Louis Klotz. I let him build a lockable cage (one of the very few locks in the whole plant) with a complete PDP-8/E system inside it, and he possessed the key. Direct access to the cage was granted only to very few trusted individuals, among them Chuck Conley for testing of his OS/8 operating system. To grant production release I needed a positive oral report by Louis of his hardware compatibility tests, assurance by Chuck that OS/8 system tests still ran uncompromised and the certainty of a permanent installation of this new option in the "super system". Beforehand, the actual designs got approved by me informally. I usually went to see the developer at his place, letting him explain his design to me. I was fast in spotting any interface inconsistencies and demanded according changes if necessary. Concerning the design part representing the inherent functions of his device, I would nudge him to get rid of patterns and details I considered insufficient in terms of design, producibility or testability. Most often a module ended with fewer parts after my short visit. Generally, my interventions were welcomed, some even appreciated it. Like John McNamara who highlights in his communications book[16] a Transition Detector, an outcome of one of my purging actions. Although being defiant at times, I always was anxious having won a friend and not bred a foe. OMNI-Smile.png[tale 16]


Cost Reduction and further Development    In 1971 John Clarke, a likable personality with grips and experience, became engineering manager of our group. John would serve this engineering group, which later designed the various workstations, for many, many years to come. Complementing engineering capacity was John Kirk, who later designed the PDP-8/A. He joined us after having worked for DEC in England during several years.

There was plenty of 'meat on the bones' on a PDP-8/E with its large, expandable box for customers not needing full capacity. Therefore, the PDP-8/M as an OEM version was launched, which contained one single OMNIBUS backplane only. The design boiled down to a mechanical restructuring, therefore Paul Gardner, our mechanical designer, was assigned project engineer. We applied an adequately scaled-down, switching-mode power supply. A new supply-design group had been initiated, and they were eager to provide us with a fitting module according to our specifications. Novices in terms of predictable power-up/down behavior for uncompromising data retention under all circumstances, they needed my personal monitoring of their design. The machine was equipped with a one-switch one-light panel for starting its integrated Bootstrap Loader - a tiny program for loading paper tape software - implemented as a read-only memory consisting of a discrete diode array. This enabled an OEM customer to load his fully tested program without the need to buy the costly Programmers Panel and to even protect his equipment from undue manipulations by his end user. For test and maintenance purpose, the use of the common Programmers Panel was still possible. The PDP-8/F - actually a marketing distinction - was essentially of identical design.

During 1971, in a rather casually operating ad-hoc team, various feasible new architectures for a new machine were outlined. It was all brainstorming, very interesting, even though producing no tangible results. But it offered unbiased opportunities to peek into new technologies, detached from routine business. On several visits at the Los Angeles based North American Rockwell (now Rockwell International) Semiconductor Division, Louis Klotz and I got a first-hand introduction into microprocessor production. Not widely known in the scene, they were producing integrated microcontrollers for military applications, before similar commercial parts got even publicly announced. These insights provided the essentials for a bus system proposal employing address and data multiplexing, compatible with integrated circuit packaging. In a paper - still using the PDP-8 as a vehicle - I laid down a preliminary specification of such a bus system. OMNI-Smile.pngx-bus


Epilogue

Life after DEC

In April 1972 I left DEC to work for Balzers AG (nowadays Oerlikon Balzers), a leading company in ultra-high vacuum and thin-film coating technology, located in the tiny Principality of Liechtenstein, neighboring Switzerland. I was in charge of development of a novel digitally-controlled, computer-compatible Quadrupole Mass Spectrometer. First introduced at the 1973-Achema Exhibition in Esslingen, BRD[17]. Its concept of versatility and outstanding performance (of course equipped with its own special bus system) was well received, sales took off immediately, and its product life extended to well over a decade.

In 1974 I founded my own engineering company. Starting alone with a rather large real-time software design project, I soon could acquire customers for hardware too. The company grew rapidly to a well-balanced crew of about 10 people, a size I maintained over the years, a size allowing me to do some engineering work myself all of the time. My company became a DEC customer almost from its beginning. We used all sorts of LSI-11 and VAX hardware/software extensively in our special products and for in-house use as well. During the late 1970's and the 1980's, DEC regularly commissioned my company in helping to take care of numerous compliance issues with European Safety Standards for their power supplies regarding the PDP-8/A line and all sorts of DEC workstations. Over the years we had the pleasure to host several PDP-8 group members visiting our place, either for business reasons, privately, or both. Of course, during each one of my few US visits, I went to Maynard to see "how the guys were doing" over at DEC.

Special life-long friendship grew out of the numerous visits with John Kirk and, particularly, with John Clarke. Together with his wife Marcia, he visited the old continent many times over the years.


Afterword

On why I took up my pen

  • A friend of mine - we studied together at the ETHZ - motivated me taking advantage of this very special IEEE Global History Network, First-Hand History forum to make a contribution of my own. This aroused my curiosity, and after he had given me a practical introduction of how to proceed on this generous platform, I was on-board.
  • Then I have to admit, that I got to like the idea that someday, some of my eight still young grandchildren might pick up a copy and be surprised to find a professional horizon of their grandfather somehow broader than gardener with a knack for vintage motorbikes. And how impressed they would be, finding out what their admired grandmother - as a young woman with two babies to take care of - had accomplished, having supported her husband in every respect with grace, never ceasing optimism and an incredible energy.
  • A history buff myself, I then began to perceive, that I may have some obligation to contribute to this very special slice of electronics history. Withholding my recollection, omitting publishing technical information not otherwise available, thereby undervaluing the efforts of all my colleagues and contributors in the struggle for success, would have to be most likely graded as neglect.
  • Evidently this story testifies the coincidence of several seasoned electronic technologies culminating in a single piece of an iconic industrial design which deserves status by its own merits. TTL technology applied in small-scale (SSI) and medium-scale (MSI) integration together with core memory technology rendering prospering life for an already seasoned, perhaps most elegant ever invented, minimalist work of logic in the new form of the PDP-8/E with its OMNIBUS-based versatility and simplicity.
  • Then, there are all these enthusiastic collectors of PDP-8 machines with related equipment, software and documents, contributing their means and time for the preservation of most precious historic material. If this story could bring them any insight they deserve or even be of some help on their undertaking, it would be of greatest satisfaction to me.
  • After all this is a tale and an appreciation for a bunch of wonderful, proud, focused people, serving in a remarkable company environment with hot marketers, clever engineers, motivated sales people, hardworking production and helpful service personnel, struggling and giving their best for their common goal of rapidly getting loads of new machines out of the door.
  • Finally I hope readers will have fun while riding this OMNIBUS. For me it was a call for an encore after an exciting play.


References

Big help from people who did not know they did help ….

  1. John G. Linvill and James F. Gibbons: Transistors and Active Circuits, McGraw-Hill, 1961
  2. Dr. Walter Guggenbühl, Dr. Ing. Max J. O. Strutt and dipl. Ing. Willy Wunderlin: Halbleiterbauelemente, Birkhäuser Verlag, 1962
  3. H. Melchior: Analytische Darstellung von Tunneldiodenkennlinien und die Berechnung der Verzerrungen und Mischvorgänge in Tunneldiodenstufen, Scientia Electrica Vol. 9, Prof. Dr. Max Strutt, Institut für höhere Elektrotechnik der ETH Zürich, 1963, pp. 93-108
  4. Frederick E. Terman: Radio Engineering, McGraw-Hill, 1947
  5. Reference Data for Radio Engineers, Howard W. Sams & Co. Inc., 1968
  6. Patent US3659201 A: "Apparatus for measuring the Muzzle Velocity of a Projectile", filed on Aug. 5, 1970, published Apr. 25, 1972, priority date: Aug. 8, 1969, inventor: Remo Josef Vogelsang. Also published as: CH515506 A, SE358473 B, NL7011763 A, NL159502 B, JPS4935708 B1, GB1284480 A, FR2065676 A1, DE2038733 A1, DE2038733 B2, BE754626 A1
  7. Tom Gardner: Oral History of Grant Saviers, Computer History Museum, recorded May 17, 2011, CHM Reference number: X6162.2011
  8. John Voelker: The PDP-8, IEEE Spectrum, 1988, Vol.25, No.11, pp. 86-92
  9. DEC: PDP-8/E & PDP-8/M Small Computer Handbook, Digital Equipment Corporation, 1971.
  10. DEC: PDP-8/E Maintenance Manual, Volume 1, DEC-8E-HR1B-D, Digital Equipment Corporation, 1971
  11. DEC: PDP-8/E PDP8/F PDP8/M Internal Bus Options Maintenance Manual, Volume 2, DEC-8E-HR2C-D, Digital Equipment Corporation, 1972, 1973
  12. DEC: PDP-8/E PDP8/F PDP8/M External Bus Options Maintenance Manual, Volume 3, DEC-8E-HR3, Digital Equipment Corporation, 1972, 1973
  13. Ambros P. Speiser: Digitale Rechenanlagen, Springer Verlag, 1967
  14. DEC Public Relations Department: "on line" monthly publication August 1970, p. 2
  15. Donald E. Murphy Stephen A. Kallis Jr.: Introduction to Data Communication, DEC, 1968
  16. John E. McNamara: Technical Aspects of Data Communication, DEC, 1978, p. 138
  17. H. Egli, W. K. Huber, H. Selhofer and R. Vogelsang: A New Digitally Controlled, Computer Compatible Quadrupole Mass Spectrometer, Balzers AG, pp. 413-419


Salt and Pepper

Tales, some serious some funny !

  1. OMNI-Smile.png T_h_e___S_t_i_n_g_______________________ This happy ending made me forget the preceding week's turbulences. For the promotion, the sales team got support by Chuck Conley for software and by me for hardware. Of course was I excited, that I got chosen for the two-week West Coast show. But there was a catch. Just two days before the scheduled transport of all the show's requisites we had found the workbench of prototype #1 empty when coming to work. As it later turned out - after #1 had been spotted someplace in Texas - one hyper-motivated salesman simply could not resist "borrowing" it. This certainly enabled him convincing his customers, that his promises were more than vaporware. Now I found myself stuck with the untested prototype #2. Some of the modules required wired bug fixes. Luckily, together with Jim Manzari, Branch Manager Field Service Palo Alto and a PDP-8 production veteran possessing the indispensable skills to handle the programmers console with virtuosity, we were a perfect debugging team. The morning when the promotion started - after our first twilight night shift - we had the lights flickering on the console. By the end of the three-day promotion, we had the thing running, ready for WESCON. Once I got confident that we would win this battle and seeing a relaxing weekend coming up, I called my wife Stephanie to join me in San Francisco for the weekend and possibly the week thereafter. We both enjoyed exploring this great area for the first time. Stephanie appreciated this change of scene as a dear gift. She had really deserved a break for her enormous contribution supporting me professionally and doing whatever she could to facilitate a happy family life over all these years.
  2. OMNI-Smile.png O_h_m_'_s___L_a_w_______________________ My older brother and I fixed the roof of an abandoned henhouse and whitewashed it to make it a "cottage" for our sisters. Inside we even built up a wooden wall as separation of "kitchen" and "living room". Then I installed an electric light and an electric doorbell. Now our sisters were the happy owners of the greatest attraction in the whole neighborhood. But resupply of flashlight batteries hurt our budget enormously. How come the doorbell on our house does not need them, I was wondering? My investigations led to a hardly accessible, little black box with taps marked 3/5/8 volts, located in a store room. All I had to do was to connect to this inexhaustible power source. It took me weeks to get the necessary insulated wires and put up a line from the front of our house across to the back and via a pipe to a shack, from there finally to the "cottage". What a disappointment ! It did not work after I had hooked up the installation. Hopping from line-splice to line-splice, checking with a bulb, I was caught by surprise that the filament of the bulb got dimmer and dimmer the farther away from the little box I was checking. Well, at least I discovered - not yet Ohm's law - but the transformer. This whole undertaking did not run smoothly and ended rather panicky. When placing back the metal cover of the transformer after I had properly connected my wires - standing wobbly on a ladder, needing one hand to keep my balance - I accidentally provoked a short circuit to the mains, forcing the fuse to break the line which was feeding most of the light bulbs of the house. Because of my mother's anxious behavior whenever a fuse happened to break, I was aware, that the electrical installation of our house might have been in a somewhat shaky condition and probably prone to cause fire. Under these circumstances I could not help but flee the site and keep waiting from a save distance in the fields, until I could see in the evening that the lights in the house got turned on.
  3. OMNI-Smile.png 1_-_C_y_l_i_n_d_e_r___4_-_S_t_r_o_k_e___ During winter when I was ten, my father asked me to "clean" the machine (short for cleaning, painting, greasing or in other words do full service and restoration). The machine was a concrete mixer in combination with a winch (windlass) for a hoist. It was the most valuable, most needed equipment of his business then. I had spent many hours fiddling with the engine. I had found out the trick of positioning the two feet in diameter open-running fly-wheel into position that enabled also a small creature like me to hand start this fat hunk. This engine was idling so slowly, it was easy to distinguish all strokes of a full cycle by just listening to the "thump"-"thump" of the exhaust and the carburetor's sucking noise. Simultaneously observing the movements of valve rods and rocker levers - with open covers - one could perfectly understand the function of a 4-stroke engine not even knowing the name of the principle. The job I had just fetched was a dream coming true. Now it was my machine. Coming back from work, my father saw that I had taken the whole thing apart, including its gasoline engine - my object of interest. After he had noticed that I had neatly put down all pieces in logical order for reassembling, he asked: 'Are you able to put everything together again?' 'Yes' I replied; this answer was good enough for him. There was a funny twist to this story. Assembling the engine, I could rely on marks at the valve gear for proper valve timing. For the ignition I was forced to find the spark's timing by trial and error. Later at a construction site, I noticed two burly type workers haggling about which one of them should start the engine, because of its tremendous backstroke. I intervened using my trick. They could not believe the smooth starting. I was afraid fixing this problem during construction season and thereby possibly provoke a breakdown. The following winter I shifted the ignition by one notch.
  4. OMNI-Smile.png B_i_g___S_h_o_t_s______________________ The company had its own "Berufsschule", a professional school for their 600 apprentices. Besides regular teachers, some company professionals also acted as instructors. Book-keeping - at a level sufficient for a craftsman's workshop - was taught by the company's chief controller. Small wonder, everyone was able to pass with best grades. Draftsmen were instructed by the company's chief of standardization. As a side benefit for the company, standardization got partially updated by the nomadic apprentices through infiltration.
    _ OMNI-Smile.png C_o_a_c_h_i_n_g_______________________ Group instructors were conducting the first-term basic courses. They were aided by assistant instructors (apprentices) for support of individuals in the groups. Twice I was awarded this position for half a year. It was a coaching task with a clear goal. You had to get an apprentice improving his individual skills to the level feasible for himself, not necessarily a standard level. You were also supposed to help him form a balance of self-criticism and self-esteem. At the time he decided to show his piece of work to the instructor for approval, he had to be confident it would pass. In case a novice really ran into a mess, you had to sort out what was still ok, helping him in clearing up the debris and then get him back on track. Then he needed a bump start und a clap on the back for his self-esteem. This task required you being demanding but still accepted as one of them. I learned how to upset people, and how not to. I acquired an instinct for assessing individual's potentials and limits, physical and emotional.
    _ OMNI-Smile.png C_o_u_n_t_e_r_f_e_i_t_e_r_s_____________ Some guy (an apprentice) someday discovered waste pieces from stamping sheet-metal, matching size and weight of a regular Swiss currency coin. This particular coin fitted the coin slots of snack machines being posted at all railroad stations. As word spread - what had started inconspicuously, built up to sort of an avalanche. Soon, all the snack machines of a whole region were empty, first because of the looting, then because they did not get replenished any longer by its owner-distributor. Of course he noticed the continued collection of this kind of fake money. So the whole racket blew up. However, there were too many culprits to catch 'em ! It was one of those spicy incidents, helping overcome low spirits at the often boring commuting by train (in my case three hours a day).
  5. OMNI-Smile.png T_h_e___O_v_e_r_k_i_l_l__________________ One of the interviews deserves a special note. My second meeting was arranged by a renowned company doing business on peripheral equipment. From two o'clock until a quarter to five I was kept waiting. My expectations were low at that time. Then finally, reading my resume and browsing through my technical documentation, interest flared up on the personnel manager's face. After a breezy phone call - it was closing time - I was escorted to a group leader in engineering. He came to the point immediately. He pointed out some problems they had encountered in an almost finished project, mainly concerning the interaction of electronics and mechanics. Deadlines were endangered. After questioning me in detail about some of the graver issues, he got convinced that I was the competent engineer to fill this specific void in their team. And I thought he was right. He called two other group leaders who had related problems to solve, to join for the interview. After about an hour, it was me who asked questions. Fully immersed in the matter, we all got carried away somehow, forgetting about this being an interview. They decided among themselves to whom I would have to report, in case. Returning to the personnel department at half-past eight that evening, the personnel manager expected a deal. And he expected it right then. I was presented a formal offer with a salary beyond anything I had expected. Now I had a problem because I felt I needed ample time for reflection. For technical discussions my English was quite adequate. However, to express my hesitation politely and gracefully, was well beyond my linguistic skills. I could read his face, that my words got interpreted as brutal bargaining to press for higher salary. He got really upset but nevertheless increased the bid one notch higher. We kept doors open when terminating the meeting, although with mixed feelings on both sides.
  6. OMNI-Smile.png R_a_n_d_o_m_ _E_r_r_o_r_s_____________ The author wrote at the end of 2009 after he accidentally discovered Grant Saviers' mail address: Hello Grant, You might remember, it has been almost exactly 40 years ago at DEC, when you asked me to look into the time-random, location-sensitive problem we ran into just before shipping the first batch of those wobbling, apple-pie sized, fixed-head disk drives. I can still quite closely recall how we figured out the puzzle of the effect of a combination of electro-statically charged read/write heads together with the discharging trigger caused by the stray-field of the hysteresis motor. Application of some conducting glue for attaching the heads was the simple solution. Since then, the continuous evolution of disk drives was driven to an unbelievable scale. I am aware, that you contributed a great deal to that surprising development (on and on, the death of "electro-mechanical" storage devices was predicted...) ….. Grant's immediate response (Jan. 1, 2010): Hi Remo, yes I remember those days and the quite mysterious pulses. This project was my first work on disk drives and quite an introduction to the novel problems that every generation since has encountered. It is quite amazing to see the evolution, a terabyte powered from your USB port. Once an engineer, always an engineer. ……
  7. OMNI-Smile.png P_D_P_-_8___D_o_c_u_m_e_n_t_s_________ Here an assortment of useful documents for more details on the PDP-8/E:
    • The encyclopedic PDP-8 article comprises the whole family of PDP-8 machines and also provides useful links to related subjects.
    • The pdp8/e & pdp8/m small computer handbook describes operation, installation, maintenance, interfacing, programming and theory of the basic system and all regular options.
    • The pdp8/e maintenance manual volume 1 includes installation procedures, theory of operation and maintenance procedures of the basic system.
    • The pdp8/e pdp8/f pdp8/m internal bus options maintenance manual volume 2 includes installation procedures, theory of operation and maintenance procedures of non-peripheral options.
    • The pdp8/e maintenance manual volume 3 includes installation procedures, theory of operation and maintenance procedures of peripheral options.
    • Available access to: PDP-8/E manuals, PDP-8/E engineering drawings, engineering drawings of OMNIBUS Devices.
  8. OMNI-Smile.png D_o_n___W_h_i_t_e_______________________ Don was a man of the highest level of professionalism, very decent, accessible and not lacking humor. Besides, he was the living PDP-8 encyclopedia and inherent communications center of our engineering. Asked for help, he would give advice or review a design part - not by go/no go judgments or offering his own solution - but by explaining why the status quo of the object under consideration would affect or limit functionality or quality and what had probably been overlooked to fulfill expectations. Responsibility was kept where it belonged. Since we met first - during my job interview - we both felt mutual respect and goodwill. We also discovered an affinity of how things should get done. When either of us handed out a design of something to somebody, it may have had still errors in it or some details were marked pending, but it was complete to the minutest detail. Hence technical discussions between us were very rare. Don, visiting my box-office calling me "Mister Swiss" and doing some small-talk, meant something like: I am informed, I approve it and I like it. And what? Well the pending thing, such as the timing circuit proposal I had put on his desk the day before. In case I needed some private advice regarding governmental issues or similar things, I was glad I could consult him.
  9. OMNI-Smile.png B_a_r_g_a_i_n___M_i_n_i_n_g_____________ Chasing for expedient parts, I came across a 16-diode array in a 14-pin DIL package, fitting the needs of the stack board and also a 4-transistor array in a 14-pin DIL package which could replace bulky TO-5-size items for xy-drive circuits. The electrical data would match perfectly, but their more than ten-fold list prices beyond our cost target were a killer argument. Not for me. Why not tap some military potential. The parts were manufactured according to military specifications, including second sourcing. With my trump card of a yearly demand of some 200'000 devices, and the vendors possibly stuck with a set of rarely used tools, equipment, procedures, the potential for a win-win situation was likely. For deadline reasons I had decided to make layouts fitting these preferred parts, hedging my bet with feasible discrete substitute devices. After meetings with vendors to sort out specifications adapting to our non-military requirements, it turned out that I had guessed right. The fact that several vendors were part of the game helped too.
  10. OMNI-Smile.png B_l_a_s_t___a_n_d___S_h_o_c_k____________ or "He who will not hear must feel." For the sake of shaping team spirit, a fitting prank was ok in our group. A newcomer - let's call him Nobody - apparently obstinate to unwritten rules of the team, was repeatedly using his neighbor's test machine, unsolicited, for his own needs at his leisure. This could be a pain in the neck, losing for instance a test routine, tediously pieced together with great effort. As a precaution, the affected guy tossed the machine key into his top drawer whenever leaving his workbench. To no avail. Later on, when Nobody drew the draw: A bang, a smoke cloud, a toxic stench and a big man with trembling hands, pale in his face like a ghost ! Then silence. Then roaring laughter by a dozen devils around Nobodys crash field. One of the guys, conspiratorially informing the crowd, had placed a selenium photoresistor he had dug out from a surplus parts bin into the top drawer, directly connecting it to a fitting power supply. Of course, as a good engineer he had secretly tested the dark/bright explosion trick in advance.
  11. OMNI-Smile.png Q_u_a_r_t_z___C_r_y_s_t_a_l______________ Commercial-type quartz crystals at that time were over-specified, over-priced, over-sized, bulky, clumsy pieces, unfitting at least. They were mostly used in rather exotic equipment, such as frequency or calibration standards. I preferred quartz crystals mounted in a HC-49/U type case as used in military or similar applications. However, their list price was out of proportion for our application. It was Friday afternoon when I first met the representative of this rather small company I had spotted as a possible supplier at purchasing department's Bill Burns office. I showed this representative my specification which actually was a list of drastically lowered requirements, compared to their datasheet. With mouth agape he registered, that I even preferred a low resonance quality factor (lowering requirements of the oscillator circuit for safe startup), when everybody else was demanding the contrary. Then the surprise: "a demand of 10'000 pieces a year". I made clear, that the part was of interest to us in the price range of commonly available simple delay lines. Early Monday morning the sales representative showed up, presenting his formal offer and already delivering fifteen samples. By their own initiative they had spot-welded a "third leg" for perfect, simple mounting. I was fascinated. I checked the pieces the same day, and Bill closed the deal by sending them his first order.
  12. OMNI-Smile.png 1_1_0_3___A_n_t_e_ _P_o_r_t_a_s__________ In spring 1971 I had Bruce Tarpley design and build a 4k by 12-bit words single-slot semiconductor DRAM memory based on Intel's 1103 memory chip. The inclusion of the necessary refreshing posed no problems with the available capabilities of the OMNIBUS. For the time being, we deferred issues like the handling of the operating gap at refresh time and the inherent volatility with its consequences regarding systems and programs. In a short time we had a prototype board running. However, it would not sustain our base memory-diagnostic programs. We informed the Intel people and lent them a board together with the information necessary to have it connected to their off-the-shelf memory exercisers. A few months later, they came to see us, bringing the board back with replaced chips. The diagnostics would then run and we loaded our FOCAL plot program OMNI-Smile.pngplot which did also run ostensibly. Together we left the lab for a coffee break and everybody was in a relaxed mood. Coming back, hitting the return key to restart another plot, the machine stalled. Myles and Stanley from Intel frowned. We stopped the machine for a few minutes, letting cool down the memory, reloaded the programs, and when we restarted, it would run again. A peculiarity of this plot program provides an indispensable memory test. Waiting at the end of the plot for restart, the program actually beats the same few memory cells permanently while waiting in a tiny loop for somebody to hit the key. This rises the temperature of these cells dramatically, forcing a flawed circuit to fail. Apparently Intel's memory expertise was still limited. We let the Intel people experiment at their discretion, offering support by one of our technician. I do not know, whether they had gained any insights or not. And they never asked any questions.
  13. OMNI-Smile.png U_n_d_i_s_g_u_i_s_i_n_g__________________ Whenever a single core of a fully manufactured stack (by the way "stacks" had become "planar") was detected out of specification at the supplier's final tests, he had to replace it. Three wires had to be cut, the core replaced and six splices soldered. Five to ten such corrections per stack were considered normal. With a forecast of 1000 stacks per month and more to come, an intolerable waste of resources and an inherent, disguised quality problem. This was worrisome. I changed covers previously made of printed circuit board material to transparent material. Inspecting all incoming batches of the various vendors, I sorted out the sample with the lowest number of splices and then placed it at Bill Burns desk in purchasing. He sure knew how to go about playing off one vendor against the other. It did wonders. It did not take too long, and all stacks got supplied without splice repair.
  14. OMNI-Smile.png G_a_l_e_-_W_a_r_n_i_n_g_________________ Designing the OMNIBUS, expansion beyond two OMNIBUS assemblies (that means only one box) had not been considered. Rather late in the game, the requirement for a System Expander Box was added. Doubling the number of slots by using a System Expander Box, inclusive its necessary cable connections, causes rise times for open-collector signal lines to increase by a factor of three to four, rendering it non-conforming to the applied timing scheme. This was a most aggravating situation. Conditional adapting of cycle time for two box systems would dangerously dilute the OMNIBUS concept. Mentally trying all sorts of swerves, I finally discovered a rescuing solution. An elegant scheme was to speed-up the relevant, clamped open-collector signal line rise times, by providing higher voltages to its pull-up resistors during the impulse times of the TP2, TP4, INT STROBE signals. Unfortunately, the Bus Loads M8320 had been considered a low risk item, allowing earliest possible high manufacturing rates. Therefore, loads of them had to get supplemented with wired fixes. Alarming was the fact, that I got to know about an Expansion Box first, when I noticed one of them on the workbench of a fellow worker, who was unsuccessfully trying to make the assembled system work. Following, there was one other incident with wrecking potential, carrying the scent of recklessness and irresponsibility. Besides lacking any noticeable benefit, the intended intervention would have disrupted the anyway fragile production process. Thus, I extinguished the tinder before it reached the powder keg. Then I began preparing for stormy weather.
  15. OMNI-Smile.png D_e_-_E_s_c_a_l_a_t_i_o_n________________ Since the project's start there was persistent bickering by production - but never a clear decision - regarding producibility and quality of the new board size of 10½ x 8½ inches. DEC at this time was able to crank out printed circuit boards at very low cost, done with well depreciated equipment which was still adequate for the their 2½ x 5 and 5 x 5 inch size Flip-Chip modules. There may have been good reasons for not updating production with state-of-the-art equipment. But lacking proper communication of what could, and what could not be done was more than annoying to engineering. Ridiculing themselves by coming up with first batches of boards warped like pancakes, they lost the too-big-size argument for good. Later someday, out of the blue sky, a person from production together with someone from quality control showed up, telling me that production would be halted for precautionary reasons because the number of feed-through holes per board exceeded the sustainable limit for guaranteed quality. Their faces told me plenty; not taking their concerns very seriously at their terms, would endanger the project. This was to be considered inappropriate but delicate. Well, it was their ill luck, hitting the wrong guy this time. Four years before, I was actively involved when transition from riveted-hole to feed-through holes took place. I knew the state of the art. I asked them for some time for reflection. We had just released the designs of the basic machine boards for rework. So I informed the model shop about production wavering quality of feed-through holes and instructed them - highlighted by some statistic hocus-pocus (It came in handy that I had attended, besides English, Statistics at Northeastern the year before.) - to always place two holes instead of one. Hence the twin feed-through holes on all the PDP-8/E boards ! This was a last flare up, that production dared intimidating engineering in any manner. Afterwards, whenever they addressed me regarding a problem, there was a real problem I had to take care of.
  16. OMNI-Smile.png B_l_i_n_d_f_o_l_d_e_d_____________________ or "Don't count your chickens before they're hatched." This looked like peanuts ! We already had a running prototype of the MR8-E Read-Only Memory, a device which had to be factory-wired according to the information content specified by the user (a predecessor of the semiconductor mask memory). Now, one of the guys doing the routine tests on a module, which had only required slight rework for production release, could not get the thing going. Called a core rope memory, it consists of a matrix of 12 transformers, one per bit. Programming is done by weaving "word line wires", one per address, inside or outside of the ferrite transformer cores to signify 'zero' or 'one' respectively. When an address is selected, current flows through that particular line and induces a voltage in the sense windings of those cores that have a line winding passing them. The memory would not run however, because the sensed voltages were attenuated to an unusable, intolerably low level. The many hours we spent, too many hours - feeling blindfolded - just re-testing, swapping parts, and comparing measurements until we finally found the reason. The legs of the mounted, U-shaped ferrite cores were leading through perforations of the printed circuit boards. On the bad version these perforations were plated through, forming a pair of single-turn secondary coils, absorbing almost all secondary energy available.


Copyright © 2013 by Remo Vogelsang