Fifty years ago, RF/microwave amplifier design was more art than science. The introduction of S-parameters and the Hewlett Packard (HP) network analyzer provided practical ways to measure circuit performance, but the design phase was still complicated and frustrating “cut and try.” A small number of progressive companies and government labs, such as HP, Texas Instruments and the U.S. Army Research Center, had ongoing, in-house computer program development to support RF/microwave design, but microwave engineers elsewhere had no access to those tools.
Early computers were primarily used for business applications, and some of their operating languages, such as BASIC,1 even lacked scientific features like complex numbers and matrices. Most engineering managers were reluctant to adopt and spend money on techniques that did not start at the workbench. Circuit designers carried out approximate calculations, often with the aid of slide rules or desktop calculators, to create the first prototypes. Products were finalized through time-consuming iterations of tweaking, peaking and shielding - sometimes without reaching the original goals. At the same time, space-age demands required microwave circuits to be smaller and have higher performance. Circuit integration was the desired direction, but that required more accurate simulation.
In this article, I summarize the early-day efforts to have computational power available to design engineers, including those who worked at small companies.
FIRST-GEN CIRCUIT SIMULATORS
Figure 1 Teletype model 33ASR with telephone modem and punched-tape reader (a). Texas Instruments introduced the Silent 700 portable printer (b) in 1973, a major improvement. Sources: Rama & Musée Bolo (a) and Retro-Computing Society of Rhode Island (b) via Wikipedia.
In 1971, the University of California, Berkeley, developed a general-purpose, open-source analog circuit simulator called SPICE,2 but it lacked many of the passive components used in microwave design. For active devices, it required equivalent-circuit models (e.g., Ebers-Moll, Gummel-Poon) instead of measured S-parameters. Additionally, SPICE was only available for large computers.
Gradually, time-shared computer companies introduced affordable access to computational power for engineers,3 but circuit design programs were rare and often user-unfriendly. The most commonly available computer terminal for engineers was the noisy Teletype machine,4 with a printing capability of 10 characters per second - requiring several minutes to print a page (see Figure 1). Unlike today, when every engineer has at least one computer within reach, those Teletypes were usually shared by several designers. To make things worse, barely any of those designers knew how to type.
Early microwave circuit design programs were generally based on various two-port interconnections using S-, Y-, Z- and ABCD-matrix manipulations.5 The frequent matrix conversions required significant and expensive CPU time. My graduate degree thesis was a novel, S-parameter-based, two-port technique,6,7 and I wrote a circuit simulator program on a Santa Clara University IBM computer based on that approach. While working at Fairchild Microwave and Optoelectronics, a division of Fairchild Semiconductors, I converted the program to the General Electric time-share system’s GE-635 computers for our engineers. Later, I asked Fairchild’s management to make the new program, named SPEEDY, available worldwide to microwave circuit designers. At that point, Fairchild Microwave had the only commercial GaAs FETs in production, and SPEEDY’s large database of broadband device S-parameter data proved to be a convenient marketing tool for the company.
SPEEDY was a two-port circuit simulator, requiring engineers to describe their circuits through the interconnection of series, parallel and cascade segments. Compared to today’s schematic entry, that method sounds primitive, but until sparse-matrix techniques and faster computers became available, computational speed was crucial, and the two-port approach was speedier - hence the name. I made a special effort to simplify the program’s use by supplying a manual with examples of typical circuits.
SECOND-GEN MICROWAVE CAD
By 1972, the semiconductor industry experienced a worldwide slowdown, and I was asked to lay off some of our newly hired engineers - which I refused to do. I found work at Farinon Electric, a stable, well-managed company that produced microwave telecommunications equipment. After learning that none of their circuit designers used computers and hearing about a brand new time-sharing service called NCSS, which operated with far more powerful IBM 370 computers, I offered Farinon a deal: on my own time, I would convert my original IBM school program to the 370 system, enhance it and let all Farinon engineers use it royalty-free - if the company would pay for the related computing charges and let me have ownership of the new product. The proposed new program would also include additional circuit components for Farinon’s communications system designers, as well as optimization to find the best performance.8,9 The company agreed and for the next six months I had no free time to myself or my family.
Although I initially wrote the program in FORTRAN,10 transferring it from one time-share company to another required some effort and learning. To make computer access easier, every time-share company developed its own unique, user-friendly operating interface, like the major computer operating systems we have now. These days, we are spoiled with simple ways to store and transfer files, but in the 1970s, punched (Hollerith) cards11 were the most common way of passing a program from one computer to another. Handling and shipping those paper cards required special precautions, as you might imagine.
Figure 2 COMPACT user manual (a) and application note (b).
Figure 3 10 to 1000 MHz hybrid amplifier design optimized using COMPACT.
Figure 4 The author exhibiting at the MTT-S conference in Palo Alto in 1976.
The first version of the program contained over 1,200 lines of code, with a 45-page manual. To assist new users, I also released an application note to illustrate the coding of typical microwave circuits and the optional available output formats (see Figure 2). Offering a free lunch to my Farinon colleagues to come up with the best name, one of them suggested COMPACT: Computerized Optimization of Microwave Passive and Active CircuiTs. Everyone agreed and after I gave a short seminar, our engineers started to use the program in mid-1972. Their feedback helped immensely to clean up the initial bugs.
A few months later, a salesman from a competing time-share company, United Computing System, came by to demo their circuit simulator, called MAGIC. “We have a better program that can also find the optimum performance,” I told him and showed how COMPACT worked. Impressed, he left and returned a few days later with several managers. None of them were engineers, but the idea that COMPACT had circuit optimization made them extremely interested. “Please consider adding this program to our library on a royalty basis,” their district manager suggested. “We’ll give you free unlimited computer access and help you with the conversion to our Control Data 7600 computer, too.” I talked it over with Ed Nolan, Farinon vice president of engineering. He had no objections, but he emphasized, “Providing support for other engineers will require extra effort, and we need you here full time!” Within a few weeks, COMPACT was running on their system. When National CSS learned about the deal, they, too, offered a royalty arrangement. Soon, Cupertino-based Tymshare came along, and we installed the program on their XDX 940 computers.12 To promote the program, I submitted technical articles to trade journals and presented papers at conferences.13–16
Interestingly, not everyone endorsed circuit optimization. A well-known college professor told me, “A good engineer should always be able to find closed-form solutions instead of relying on iterative techniques.” While that was true for certain cases, optimization proved to be invaluable for active microwave circuits (see Figure 3), as well as considering the effects parasitic elements and transmission line discontinuities in passive circuits. At the same time, I always recommended using all available analytical tools to obtain the initial circuit instead of immediately jumping to optimization.
By the end of 1974, COMPACT was running on four international time-share systems without any competition. Technical support and bug fixes on the different computer systems required a tremendous effort. My wife took calls at our house, and every night as soon as I arrived home, I settled on my portable Texas Instruments computer terminal, connected it to the phone line and worked long hours. Adding new features required research and getting up early in the morning to talk with East-coast customers became part of my regular routine.
Even though I loved my work at Farinon, eventually I had to admit I was paying more attention to COMPACT and I was not being fair to the company. Still, leaving microwave engineering was too hard to consider. When I discussed the issue with my father-in-law, an oil company executive, he was not happy. “Leaving a steady job with great benefits is not a good idea! Besides, I don’t see how anyone can make a good living by selling computer programs,” he told me. Although I respected his advice, in late 1975, I announced to Farinon that I wanted to quit my job. The CEO, Bill Farinon, who was also an entrepreneur, sympathized with my situation. “We don’t want to lose you, so take a six-month leave of absence,” he told me. “We’ll maintain your health insurance coverage until then. If you still like what you’re doing, then leave us with our blessing. On the other hand, if you change your mind, pass the program support to someone else and come back.” Within a few months, I knew I had made the right decision and we parted on friendly terms. My former Farinon colleagues gave me a wonderful farewell party.
After taking a long overdue Hawaiian vacation with my family, at the beginning of 1976 I started full time program development and support out of our home. That summer, a booth at the first exhibit of the IEEE International Microwave Symposium in Palo Alto gave me the opportunity to meet with many of my customers in person (see Figure 4). Moving to a “mega house” in Los Altos Hills enabled me to have space for employees and be close to my family.
TEACHING TO MARKET CAD
Selling CAD was a challenging task because many microwave designers were not convinced of its practicality. Technical articles helped, but I found a more personable way to promote the program: short university courses.
While taking a microwave system design short-course at UCLA, I met Bob Wenzel, one of the leading filter design experts in our field. Recalling how impressed I had been with his personable, easy-going teaching style, I asked if he would team up with me to teach a microwave circuit design course. He agreed to teach two days on filter synthesis and I would follow with three days on active circuit design. In 1976, UCLA scheduled the first five-day course, titled Microwave Circuit Design. The school allowed us to use its computer lab for a two-hour, hands-on design session using COMPACT. Our first course was booked to capacity; engineers from TRW, Hughes Aerospace and other defense-oriented business groups were eager to attend. I gradually hired consultants to add new capabilities to the program, including various physical transmission line elements and discontinuity models, Monte Carlo statistical analysis, noise analysis and optimization, among others. Within a few years, other universities - the Universities of Maryland and Santa Clara in the U.S. and Cambridge and Oxford in the U.K. - asked us to teach the course. These presented excellent opportunities to introduce COMPACT to new users. In addition, we were paid well to teach.
The security considerations of defense products became a concern. Government agencies were not happy about designs performed through unsecured telephone lines. They wanted the program running on their own computers. The first company that purchased COMPACT was the Communications Research Centre of Canada, followed by several others. Installing the program on various systems often presented new challenges for us. In 1977, the president of Tokyo System Lab, a Japanese technical marketing company, offered to become our representative in Asia. We signed an agreement, they translated our user documentation and I began to teach short courses in that country, often with the support of interpreters.
THIRD-GEN MICROWAVE CAD
Early in 1979, Communications Satellite Company (COMSAT) approached me with an interesting proposal. COMSAT had been formed by the U.S. Communications Satellite Act of 1962 and had maintained a worldwide monopoly on satellite communications. In 1977, the FCC reclassified the company as a utility and began to regulate the rates charged to COMSAT’s customers. To make matters worse, the new “Open Skies” policy, which encouraged competition, ended COMSAT’s worldwide monopoly on satellite telecommunications the following year. COMSAT’s management decided to diversify and look for new business ventures, with one of their ideas automating the engineering design and manufacturing processes of high-tech companies.
