This year is the 50th anniversary of Microwave Journal. As is frequently the response to such noteworthy occasions, the editors, in their unquestionable wisdom, have invited old codgers like me to create retrospectives on microwave technology. I never thought it would have happened, but the older I get, the more I tend to say things like “in my day” and to mutter more or less conventional complaints about the younger generation. In technology, we happily become the old guys that, in our youth, so thoroughly annoyed us.
This month, Microwave Journal is highlighting a 1962 article by Arthur Uhlir, one of the true pioneers of nonlinear component research. The article, while seemingly rather specialized, touches on many of the then-current considerations in the design of such components. In view of all this, it seems like a good occasion to talk a little about sources, and to show how we got where we are.
Vacuum Sources
At the time of Uhlir’s paper, there were few options for generating microwave signals. The simplest, although rarely the most desirable, was a microwave tube.1,2 Three types of tubes used for generation of microwave signals were magnetrons, reflex klystrons and backward-wave tubes. More or less conventional triode tubes could also be configured to work in the lower microwave range.
Magnetrons
Magnetrons, invented in England during the second world war, were the earliest microwave tubes. Magnetrons, more than any other technology, made radar possible, and, because of their low cost and high power capability, are still in use. Your microwave oven, for example, uses one.
Magnetrons operate by accelerating electrons and spiraling them in a magnetic field. The electrons pass apertures in a periodic structure composed of resonators, giving up energy to the electric field in the aperture (see Figure 1). The principle has been compared to creating a tone by blowing across the open top of a bottle. Like other microwave tubes, magnetrons are noisy and difficult to stabilize. As such, they are rarely used for receiver local oscillators, but for applications that require large amounts of raw power, such as radar transmitters and, of course, your microwave oven, they are ideal.
Reflex Klystrons
Reflex klystrons were common at frequencies from the lower microwave range well into the millimeter. In an amplifying klystron (see Figure 2), an electron beam first travels through a microwave cavity, where it is velocity-modulated by the electric field. The beam continues through a drift space, where the faster electrons catch up to the slower ones, forming “bunches” of electrons in the beam. When the “bunched” beam finally passes through a second cavity, it gives up some of its energy exciting the fields in that cavity, providing greater output power than input power; that power, in effect, is extracted from the beam. In a reflex klystron, only a single cavity is used. A reflector element bends the beam back into that same cavity that velocity-modulated it, effectively providing feedback necessary for oscillation. Reflex klystrons were capable of providing high power, but they were difficult to stabilize and generated substantial AM noise. In many millimeter-wave receivers, the AM noise from the klystron, downconverted by the klystron local oscillator (LO) itself, was the dominant source of noise.
Backward-wave Tubes
A backward-wave oscillator (BWO) is a variant of a travelling wave tube, or TWT. In a TWT, a wave is allowed to interact with an electron beam over a fairly long region (see Figure 3). The wave must propagate on a slow-wave structure, usually a helix, so its velocity matches that of the beam. Under the right conditions, the wave modulates the beam and the beam gives up energy to the wave, in a manner not terribly different from a klystron. The main difference, however, is that a cavity is not used, so TWTs can operate over broad bandwidths.
Backward-wave oscillators are possible because, contrary to all intuition, the beam can support a backward wave resulting from the forward-travelling wave. This creates the feedback necessary for oscillation. Furthermore, just as TWTs are broadband, BWOs can have remarkable tuning ranges. Many of us remember using BWO test sources, from the lower microwave region to the millimeter-wave, having octave-band outputs.
While BWOs are largely obsolete, TWTs are still in use. Although the preference today is for solid-state amplifiers, many spacecraft in use today still have TWTs. They are used because of their high efficiency, high power, reasonable linearity and broad bandwidth. Especially, TWTs can be made adequately reliable for space applications. It’s not an exaggeration to say that, without TWTs, broadband, high-volume satellite communication would probably not have been possible until very recently.
Although they were essential for early microwave technology, vacuum devices had limitations that encouraged system designers to abandon them as soon as adequate solid-state alternatives were available. As we have noted, tubes were notoriously difficult to stabilize. Throughout the 1960s and 70s, communication systems became progressively more sophisticated, and frequency and phase stability requirements became more severe. Although it is possible to phase-lock certain kinds of tubes (the reflex klystron is probably the easiest), it is indeed not much fun to try. Even then, the AM noise problem was severe, especially for low-noise millimeter-wave receivers used in radio astronomy and other radiometric applications. Finally, and probably most importantly, the problem of reliability motivated the move from tubes to solid-state sources. All tubes have a lifetime, usually no more than a few thousand hours of operation. Even so, and perhaps paradoxically, the tubes themselves often were reliable enough, but the necessary high-voltage power supplies were not. High voltage is inherently unreliable, especially in space applications, where rarefied gasses are easily ionized and support arcing. Another motivation was safety; as this author, who once superimposed himself across the 1800 V output of a klystron power supply can attest, high voltages are inherently dangerous.
Conventional Vacuum Tubes
Even before the 1960s, a number of vacuum tubes could generate useful signals at UHF and lower microwave frequencies.3 Most of these were triodes, tubes having a single grid. World War II military equipment often used the 955 “acorn” tube, invented in 1935.4 I have seen 955s used up to 500 MHz, but with proper cavity construction, they could possibly operate to 1 GHz. The 2C40 “lighthouse” tube’s cylindrical symmetry made it ideal for mounting in coaxial cavities. 2C40s could oscillate above 3 GHz. The 5675 “pencil” tube, also designed for use in coaxial cavities, could produce a half watt at 1.7 GHz. Finally, one of the best known tubes, and perhaps the last gasp of high-frequency receiving-tube technology, was the 6CW4 “Nuvistor” from RCA.5 It was used in many high-frequency commercial applications, primarily television tuners, and could provide low-noise oscillation and amplification to almost 1 GHz.
In the solid-state world, we focus almost entirely on resistive and capacitive parasitics. Conventional vacuum triodes, however, were limited in their operation by cathode-to-plate electron transit times. Making devices smaller and operating at higher plate voltage reduced the transit time; unfortunately, small size and high voltage were competing trade-offs and, in any case, a triode could be made only so small. It seems inevitable, then, that transistors, which have much shorter transit times, would be viewed as the next high-frequency devices, and it is no surprise that great effort was put into the development of microwave transistors. Indeed, not only were transistors improved beyond all expectations, but new, unforeseen solid-state microwave devices were created as well. It’s not an exaggeration to say that solid-state technology created a golden age of microwave electronic technology.
Solid-state Sources
Varactor Frequency Multipliers
At the time of Uhlir’s paper, practical microwave transistors didn’t exist. Rapid progress in solid-state technology was made throughout the 1960s, however, and by the early 70s we had good, low-noise bipolar transistors useful to a few gigahertz.
The earliest solid-state device useful for generating microwave energy was the varactor diode. Varactor frequency multipliers date from the late 1950s, when they were realized in both diffused-junction and point-contact form. It seems that good varactors, useful for efficient microwave frequency multiplication, were available quite early, and practical multipliers were produced well before 1960. While some aspects of the operation of reactive microwave devices were becoming clear by the mid-1950s (for example, the classic treatment of power in such devices by Manley and Rowe),6 a good understanding of the theory of such devices followed well behind practical application. For example, another paper by Uhlir at this time7 shows that the operation of even simple point-contact diodes (in which a metal contacting “whisker” actually forms a crude Schottky junction) was not well understood.
A good theoretical understanding of varactor multipliers was developed fairly quickly, however, and papers of that era show a remarkably complete theoretical basis for varactor multipliers.8,9 The paper by Uhlir, reprinted in this issue, shows a mature understanding of multiplier theory by 1962. One of my favorite books on varactor circuits, although focussed primarily on parametric amplifiers, is that of Blackwell and Kotzebue.10 It hit the streets in 1961. Not much later, Burkhardt’s classic paper on multiplier design was published;11 it became the standard for the design of such multipliers.
Early on, it was recognized that a varactor diode’s voltage was a quadratic function of charge, and, as such, could generate only second harmonics directly. Higher harmonics required the use of idlers, resonators that supported voltage components at intermediate harmonics, allowing higher harmonics to be generated through a mixing process. Although Manley and Rowe showed that 100 percent efficiency was theoretically possible in reactive multipliers, the efficiency, in practice, decreased fairly quickly with harmonic number. Uhlir’s paper takes issue with that idea, largely for theoretical reasons, and it is important to note that it was published before more complete analyses such as Burkhardt’s. Even so, it indicates quite a bit of activity in the theoretical work surrounding these components.
In the early 1960s, there existed few solid-state methods for generating microwave signals. The most common was to use a crystal oscillator, operating below 100 MHz, followed by a series of bipolar-transistor frequency multipliers. The multipliers raised the frequency to a few hundred megahertz, at which the signal was amplified and applied to a string of varactor multipliers. If the multiplier chain had more than a few varactor stages, the input power to the varactor portion often had to be a few watts. This would result in a few tens or hundreds of milliwatts at the output. To attain this modest output power, the source had to dissipate several watts of heat. Not only was this inefficient, but disposing of the heat, especially in space applications, could be a major difficulty. Apart from efficiency considerations, varactor multipliers were generally narrowband and prone to instability. Making them work well was a tricky business.
On the positive side, the use of a crystal oscillator guaranteed good stability. Transistor frequency multipliers generated little noise, and varactors, being reactive devices, generated only negligible levels of noise. As a result, such sources could be quite low-noise. This combination of low noise and stability made varactor multiplier chains extremely valuable in modern radar and communication systems.
Other Solid-state Sources
Probably the most successful solid-state microwave source is the Gunn oscillator, also known as the transferred-electron oscillator. Such oscillators make use of negative resistance that can occur in certain bulk semiconductors, all of which are III-V elements. These semiconductors have a “satellite” conduction band, in which the electron mobility is relatively low. When the electric field in the semiconductor is great enough, electrons are transferred into this band, their velocity decreases and current decreases. The resulting negative resistance can be used to create an oscillator.
The possibility of such negative-resistance oscillation was predicted in papers by Hilsum12 and by Ridley and Watkins13 before it was observed experimentally by Gunn.14 Commercial oscillators were readily available by the early 1970s, and the technology quickly matured. Indeed, microwave engineers could hardly be luckier: put in DC, get out microwaves. What could be better?
Gunn oscillators had low noise, were easy to tune electronically (so they could be stabilized in phase-locked loops), could operate at millimeter wavelengths, and had decent output power and efficiency (a couple hundred milliwatts at 30 GHz was typical). GaAs Gunns had an upper frequency limit around 100 GHz established by the time required to transfer electrons to the satellite band. Other materials, particularly indium phosphide, were not so limited, and InP oscillators above 100 GHz were regularly produced. Not cheaply, of course, but for a price, they could be obtained.
The main disadvantages of Gunn oscillators were relatively high 1/f noise, which resulted in phase noise, the need to mount the device in a microwave cavity, and like all negative-resistance devices, a tricky design and development process. As a result, Gunn oscillators required a significant amount of labor in the form of fabrication and tuning to produce them. This meant, in turn, that costs were difficult to minimize, even with significant technological advances: making a Gunn oscillator still required several hours work by both a skilled machinist and an experienced technician. This probably was the greatest cause of the Gunn’s eventual obsolescence.
Another important device was the impact-avalanche transit-time device, or IMPATT, along with its alphabet soup of cousins (TRAPATT, BARITT,…). IMPATTs are also negative-resistance devices, but their negative-resistance results from transit time through the device. The device, a kind of PIN diode, is allowed to go into avalanche breakdown at reverse-voltage peaks. Then, delays in the build-up of avalanche charge and transit through the device result in a current pulse that is out of phase with the voltage that generated it. The result is negative resistance and resulting oscillation.
Avalanche breakdown is a noisy process, so IMPATTs were unavoidably noisy. They were rarely used in receivers. On the positive side, however, they were capable of substantial power, often tens of watts at X-band. This made them serious contenders for tube replacement in various kinds of transmitting hardware. Even though the thought of using a device based on avalanche breakdown sent reliability engineers into post-traumatic shock, IMPATTs were eventually made reliable enough for space applications. They replaced a lot of TWTs in microwave systems.
What Does All This Tell Us?
Of course, it would be easy to repeat the tired idea that those who don’t learn from history are condemned forever to quote Santayana (or something like that). Even so, I find that there is often little understanding among technologists of the history of their technologies, and, especially, the phenomena that motivated their evolution. A result is research into technologies (often in academic settings, I’m afraid) that seem to push that evolutionary process backwards. The best way to prevent that malady is to understand that evolutionary process.
Microwave sources have evolved continuously from large, hot devices with a lot of mechanical parts that are often difficult to fabricate into electronic and solid-state components. The motivations have been, almost exclusively, cost (broadly defined) and reliability. Of course, performance is also a factor; however, in many cases, it has been quite acceptable to sacrifice performance for cost and reliability. A perfect example is the move from TWT to solid-state transmitters on spacecraft. The power of early solid-state amplifiers could not match that of tubes, and the efficiency was often no better. But the need to improve reliability and reduce size (for greater functionality within weight limits) was so great that solid-state amps were simply made to work: improve the receivers and antennas a bit to pick up a couple dB, accept a small reduction in link margin, and it all works.
Indeed, most tubes and many of the older semiconductor devices had a cost or reliability floor that simply could not be pushed any lower. When there appears to be no technological route around such limitations, the pressure to develop—and to adopt—new technologies becomes irresistible. That’s exactly what happened.
A continual problem in the early days of microwave solid-state sources was the lack of any concrete design procedure, to say nothing of circuit simulation capability. Producing a Gunn or IMPATT oscillator, or a varactor frequency multiplier, for that matter, was like trying to catch a very angry cat in a dark cellar. It involved many hours at a lab bench, tweaking the beast into existence. The designer, like the cat catcher, rarely emerged unscathed.
Indeed, the lack of circuit-level models for devices could be cited as a reason for the lack of systematic design methods. Such models could have helped. However, without computer circuit-simulation technology or other supporting technologies, such as accurately de-embedded microwave measurements, they would not have been terribly useful. I remember engineers in this era being openly contemptuous of computers, not entirely without reason. Using a computer in the 1960s and 70s was an exercise in slow motion that often could be outrun by some judicious lab work.
An important lesson of this experience is that the lack of a perfect, complete technology doesn’t necessarily prevent its development. A curious aspect of technological advance is evident the way that a complete technology often arrives late in the game, frequently when we are ready to move on to new ones. This emphasis on the pragmatic and a willingness to work, often in the dark, with whatever is available, has resulted in extraordinarily fast technological progress. Compare this, say, to medical research, where the mindset is invariably first to understand the theory, then hope that practical results arise from it. We technogeeks are a lot faster, and, I dare say, more successful.
References
1. R.E. Collin, Foundations for Microwave Engineering, First Edition, McGraw-Hill, New York, NY, 1966.
2. A.S. Gilmour, Principles of Travelling Wave Tubes, Artech House, Norwood, MA, 1994.
3. http://www.hem-usa.org/web-gallery.shtml.
4. http://www.antiquewireless.org/otb/acorntube.htm.
5. http://www.thevalvepage.com/valvetek/Nuvistor/nuvistor.htm.
6. J.M. Manley and H.E. Rowe, “Some General Properties of Nonlinear Elements: Part I, General Energy Relations,” Proc. IRE, Vol. 44, July, 1956, pp. 904–913.
7. A. Uhlir, “The Potential of Semiconductor Diodes in High-frequency Communications,” Proc. IRE, Vol. 46, June, 1958, pp. 1099–1115.
8. D.B. Leeson and S. Weinreb, “Frequency Multiplication with Nonlinear Capacitance: A Circuit Analysis,” Proc. IRE, Vol. 47, December, 1959, pp. 2076–2084.
9. K.M. Johnson, “Large-signal Analysis of a Parametric Harmonic Generator,” IRE Trans. Microwave Theory Tech., Vol. 8, September, 1960, pp. 525–532.
10. L.A. Blackwell and K.L. Kotzebue, Parametric Amplifiers, Prentice Hall, Englewood Cliffs, NJ, 1961.
11. C.B. Burkhardt, “Analysis of Varactor Frequency Multipliers for Arbitrary Capacitance Variations and Drive Level,” Bell System Tech. J., Vol. 44, 1965, p. 675.
12. C. Hilsum, “Transferred Electron Amplifiers and Oscillators,” Proc. IRE, Vol. 50, 1962, p. 185.
13. B.K. Ridley and T.B. Watkins, “The Possibility of Negative Resistance Effects in Semiconductors,” Proc. Phys. Soc. London, Vol. 78, 1961, p. 293.
14. J.B. Gunn, “Microwave Oscillations of Current in II-V Semiconductors,” Solid-state Communications, Vol. 1, 1963, p. 88.
Stephen Maas received his BSEE and MSEE degrees in electrical engineering from the University of Pennsylvania in 1971 and 1972, respectively, and his PhD degree in electrical engineering from UCLA in 1984. Since then, he has been involved in the research, design and development of low-noise and nonlinear microwave circuits and systems at the National Radio Astronomy Observatory (where he designed the receivers for the Very Large Array), Hughes Aircraft Co., TRW, the Aerospace Corp. and the UCLA Department of Electrical Engineering. Subsequently, he worked as an engineering consultant and founded Nonlinear Technologies Inc., a consulting company, in 1993. He is currently chief scientist of Applied Wave Research Inc. (AWR). He is the author of Microwave Mixers (Artech House, 1986 and 1992), Nonlinear Microwave Circuits (Artech House, 1988; second edition 2003), The RF and Microwave Circuit Design Cookbook (Artech House, 1998), and Noise in Linear and Nonlinear Circuits (Artech House, 2005). From 1990 until 1992 he was the editor of the IEEE Transactions on Microwave Theory and Techniques and from 1990 to 1993 was an Adcom member and publications chairman of the IEEE MTT Society. He received the Microwave Prize in 1989 for his work on distortion in diode mixers and the MTT Application Award in 2002. He is a Fellow of the IEEE.
