NF: We're celebrating our 50th anniversary. How long have you been in the microwave industry?
BG: My first paying job was at a research lab in Britain in 1954. At that time, I was not formally involved with microwave topics. But in the hobby years that preceded my career, dating back to 1947, I was already keenly interested in radar systems operating in the 3-cm (10-GHz) range, and avidly read whatever books I could find on this and related topics at the local library. After the war, I purchased as much ex-military radar equipment as I could afford. Using these parts, I designed and made my own oscilloscopes and TV receivers. So I suppose the cheeky answer would be, "All your 50 years!"
NF: How has the industry changed over your career?
BG: We first saw the S-shaped curve that tracked the use of bipolar junction transistors (BJTs) in place of vacuum devices starting in about 1950. That progress began to flatten off between about 1965 to 1970, as monolithic integrated circuits (ICs) gained common entry into the market. Thus, the "transistor era" was much briefer than the prior S-curve tracking-tube usage. The third S-curvetracking the ICis now at a point where further advances in electronic microtechnology are becoming extremely costly.
We can expect an equilibrium to be reached in the next 10 years, arriving at a relatively small number of "ultimate" standard processes. During this time, we can foresee an era of rapid development in nanotechnologies, using new materials (graphene being very promising) and new tokens for conveying information (most notably optical). Silicon ICs will continue to become inscrutably complex, while remaining inexpensive. In the high-frequency analog regime, such advances as silicon-on-insulator, the use of bandgap engineering (for instance, silicon germanium), the increasing use of various novel III-V materials, and "system-on-a-header" approaches to integration will become increasingly commonplace.
In the microwave arena (and many corners of the industrial market), we find vacuum devices, such as klystrons, still very much in demand. And of course, the mass-produced cavity magnetron is in every home and numerous other applications. If we define the word "microwave" to meanin today's termsthat range of frequencies from roughly 10 to 100 GHz, we find semiconductor devices to fall short in many applications and to be essentially absent in the range from 100 to 300 GHz. One well-known reason for this is that the fastest transistors can only support very small voltage swings, precluding the delivery of high microwave powers at normal impedance levels.
Other technologies, including electron tubes, are more suited to high-power microwave generation. Advanced semiconductor technologies can address receiver applications quite well. An added change we have seen is the move away from expensive, machined waveguides, using instead coaxial cables for system interconnection and coplanar lines at the circuit level. We can also expect intra-module signal coupling to be implemented using on-device antennas, rather than through package pins and across lossy substrates.
NF: Do you have any humorous stories from your career?
BG: Oh gosh...One that comes to mind goes back to my first job at the Signals Research and Development Establishment in Christchurch, England. I had designed and built an all-vacuum-tube, analog-to-digital converter (ADC) for speech signals, in rack-mounted form, as the first step in encoding them for secure, low-bandwidth transmission. I'd used many 250-V electrolytic capacitors, but was curious to check the robustness of the operation using either lower or higher supply voltages. At 350 V, the converter still workedwith a bit of a stutter. But there was an odd smell coming from the rack.
I pulled my chassis forward on its runners and peered in. Suddenly, there was a pretty loud explosion and I felt my face instantly covered with hot stuff. I staggered over to where the technician, Knobby Clarke, worked. Seeing me, he uttered in a dull monotone, "Oh. My. God," which I assumed meant that I was in for some serious disfigurement. He called the onsite medical center, to which I was rushed. When I arrived, the staff was glued to a lone TV, avidly watching a major horse race. After being transferred to a bed, I was left there to die. Eventually, when the Derby was won, a buxom nurse came to look me over. With a big smile and a little alcohol, she cleansed my face of the matrix of black capacitor stuff and the aluminum shards embedded in it.
NF: What's your biggest accomplishment?
BG: I dislike this sort of question, because it presumes but a single high moment in one's career. But no doubt the greatest glow of accomplishment came during the late 1950s, working at the Semiconductor Application Labs of Mullard Ltd. My assignment was to find new applications for the company's transistors, although supervision was vague. I did that job. But as these devices were getting faster by the month, I decided we would soon need faster oscilloscopes to observe their behavior.
When an opportunity opened up, I developed a highly innovative oscilloscope, drawing on years of experience going back to my teens. For the first time, it combined sampling (a sort of mixing) and real-time techniques in the one instrument. I designed it to look like a Tektronix 535even including interchangeable plug-ins to support high-impedance probes or 50-Ohm inputs on dual channels. It was also almost totally transistorized. In this and several other ways, it was well ahead of its time. It was my ticket to Tektronix a few years later.
NF: Can you tell us a little bit about how and why you developed the Gilbert cell?
BG: We need to clarify what that term means. During the mid-to-late 1960s at Tektronix, I was involved in the development of the "New Generation" line of oscilloscopes, which was to make extensive use of custom ICs fabricated in Tek's brand-new, in-house facility. During that period, I was allowed considerable latitude to invent new semiconductor devices and IC topologies. I made several useful discoveries. To one of them, I later (1975) proposed the formal name, "The Translinear Principle" after having spent further time in developing methodologies to apply it to analog IC design and more fully understanding the ubiquitous nature of this concept.
The earliest (1965-70) Tektronix circuits that took advantage of "TL design"in which the principal state variables are currents, while voltages are in a sense only incidental circuit variablesincluded a number of "gain cells." Amongst other useful properties, the gain of these various cells could easily and accurately be varied under current control. Analog multiplication and division are closely allied to such functionality. Thus, there also emerged two seminal cells for linear multiplication. A subset of these could be used as an RF mixer.
After presenting these ideas at the International Solid-State Circuits Conference (ISSCC) and, in 1968, publishing a JSSC paper that more fully described this mixer, Tektronix began to receive orders for itthere being yet no monolithic active mixers on the market. It was a golden opportunity that the company declined! Other TL circuits could perform signal squaring and square-rooting, as are required in measuring the RMS value of an RF signal. Their powerful nonlinear functions performed signal-processing "calculations" more accurately and more rapidly than the bolometer-based (thermal) instruments of the time, and needed no software support.
NF: Will solid-state devices ever surpass tubes in terms of RF output power?
BG: Operation at high voltages is the key. Several semiconductor materials, including gallium nitride (GaN) and extensions of that technology, exhibit much higher breakdown voltages than high-frequency silicon transistors. Since so many microwave systems operate in a 50-Ohm environment, one can argue that devices capable of a peak voltage swing of about 140 V and currents of 1 A are needed to directly drive a 50-Ohm load to 50 W of single-carrier power. That is a lot to ask of a small device. Even when operating in an efficient conduction mode, it will run very hotunless novel cooling methods are used. The fact that electron tubes have few problems when operating at high temperaturesand that large voltages can readily be handledsuggests that such devices as traveling-wave tubes and klystrons and miniaturized developments in these classes will continue to be found in high-power microwave transmitters.
But there are alternative scenarios. A familiar one is the distributed amplifier. High breakdown voltages remain important, alleviated only by impedance-matching tradeoffs. But now, the operating power can be spread over a larger area. An exciting prospect is the advent of RF powermanagement devices based on graphene, due to its very high carrier mobility, very high coefficient of thermal conduction, and other unusual properties.
NF: What advice do you have for young high-frequency engineers?
BG: First, set your sights on a career that you feel is going to bring you great joy every day of your life; nothing is more important. Second, respect the accumulated wisdom in your chosen field, but don't worship it; much of it will be incorrect. Third, never listen to the naysayers who tell you: "It can't be done" or "Who would ever need such a thing?" Such people should be politely but firmly ignored. Finally, never stop asking: "Why?" and "What if?" and "How about?" This should be your daily routine, as a young contributor to be reckoned with.