Components are like building blocks for RF/microwave systems. They inevitably determine the performance of the system, as well as its size and weight, and even its reliability. At one time, large systems houses would supply their own components, to be used in-house by their own system integrators. But, as this industry grew, a greater number of smaller companies were counted upon to supply the vital active and passive RF/microwave components including amplifiers, frequency-translation components, filters, and oscillators needed to make larger systems function. In fact, many of these smaller companies started as a result of their founders' experiences working in larger systems houses.
Microwave components encompass a wide range of products, from amplifiers to YIG filters, many of which are unique to the high-frequency industry. Fifty years ago, this industry used transistors, but not in power amplifiers which were exclusively energized by vacuum-tube electronics such as traveling-wave tubes (TWTs) and klystrons. An industry that saw its beginnings during World War II would need to make a transition to markets other than military electronics for survival; having a strong base of component suppliers capable of advancing performance, while also finding ways to reduce the costs of their products, would help to make those transitions.
Early component suppliers included major systems houses, since many of those companies had developed component technologies while designing and producing military electronic systems during and after the war. As an example, an advertisement run during the first year of this magazine by Raytheon Co., Special Microwave Devices Operation (Waltham, MA) displayed a waveguide circulator for use from 13.225 to 13.425 GHz (Fig. 1). It combined a pair of three-port circulators with a termination built into port 4 to achieve maximum VSWR of 1.20:1 across its 200-MHz bandwidth. The waveguide circulator boasted maximum insertion loss of 0.5 dB across the bandwidth with at least 20 dB isolation between ports, in addition to power-handling capability of 25 W average and 2 kW peak power. The circulator was 2.93 inches long and weighed 5.5 oz.
Most of the signal power in those early microwave electronic systems came from vacuum-tube electronics, and Eitel McCullough, Inc. (San Carlos, CA), also known as Eimac, was among the leading suppliers of 50 years ago. For example, the firm's model X-1100 TWT provided 5 W linear power and 10 W saturated power from 5.9 to 7.5 GHz (Fig. 2). Nominally designed for communications systems, the tube delivered 40 dB gain with a typical noise figure of 26 dB. The tube offered a number of innovations for that time, including a design that allowed ease of replacement of a defective tube in about four minutes or less. It had input and output waveguide with impedance-matching sections that would allow a user to optimize performance for either broadband or narrowband applications. The tube also featured spring-loaded screws at the cathode and collector ends to align the tube for minimum helix interception so that any tube can be used with any stack. The TWT and focusing structure were 15.25 in. long by 6 in. wide, and weighed about 7.5 lbs. It incorporated a CMR-137 waveguide flange for use with WR-137 waveguide and sold for about $1365 per tube and TWT stack. Eimac, along with a competitor that would eventually acquire it (Varian Associates), were among the leading suppliers of high-power tubes and traveling-wave-tube-amplifiers (TWTAs) during the early years of this magazine, along with top systems integrators such as Raytheon, Litton Industries, and Hughes Aircraft Co.
As solid-state designers in later years would debate the merits of different transistors for producing high output-power levels at RF/microwave frequencies, the June 1962 issue of MicroWaves featured a report on the race for high power with vacuum-electronics devices, including extended-interaction (EIA) klystrons and crossed-field amplifiers (CFAs). The state of the art in tube power at that time was about 25 MW pulsed output power at S-band frequencies and 20 to 50 kW continuous-wave (CW) output power at X-band frequencies (Fig. 3). The S-band tubes had been developed primarily for linear accelerators, while the X-band devices were for space communications and tracking, radar astronomy, and orbital scatter communications.
Raytheon Co.'s Spencer Laboratory (Burlington, MA) had invested considerable time and effort into a device called the Amplitron. The laboratory produced a developmental model of the multiple-circuit vacuum device, model QKS-976, which was predicted to deliver as much as 200 kW average power and 50 MW pulsed output power at S-band frequencies. Cooling, of course, is an issue in any type of active device with such a high power density. The Amplitron was designed to achieve improve improved cooling over existing tubes at that time through the use of several small cooling passages in each anode vane, rather than relying on a single cooling passage. Raytheon's vacuum-electronics laboratory at the time benefited from numerous gifted researchers and engineers. Among them was Bill Brown, who had been working on a variation of a CFA called the Electromagnetic Amplifying Lens (EAL). The tube achieved high axial gain, using rotating axial space-charge spokes that would amplify a signal as it traveled the length of the tube.
Vacuum-electronics innovations from the early 1960s came from sources no longer associated with such research, including General Motors Defense Systems (Santa Barbara, CA). In that June 1962 Special Report on high-power tubes, the firm reported a double-pumped parametric amplifier using two double-pumped klystrons with as much as 50 dB gain. The amplifier provided a 680-MHz bandwidth at 6 GHz with noise figure of only 1.7 dB.
Of course, any historical reflection on vacuum-electronics devices would be incomplete without a mention of the Varian brothers, Russell and Sigurd (Fig. 4), who were featured in the June 1962 issue of MicroWaves in commemoration of the then-25th anniversary of the klyston. It was on June 5, 1937 when the fundamental concept of a klystron velocity-modulating an electron beam with one cavity and extracting its energy at microwave frequencies from a second cavity occurred to Russell Varian. By August 19, he had assembled the world's first klystron, with an output frequency of 2.3 GHz.
Although they had much different personalities,Sigurd was the adventurer and airline pilot, while Russell was the scientist they believed that high-frequency radio beams could guide and detect airplanes. Because of this, they invested their time and money (about $3000) to develop a microwave oscillator. Russell had studied under William W. Hansen at Stanford University, where Hansen had experimented with resonant structures to develop high-voltage x-rays. After Professor Hansen took an interest in the Varian brothers and their project, he and David Webster, then head of the Stanford Physics Department, became unpaid workers for the Varian brothers, and their $100 budget for materials went to develop a high-power vacuum tube.
Inevitably, however, it was solid-state electronics that would claim the majority of RF/microwave applications in the years to come. As high-frequency circuit engineers refined their understandings of waveguide structures, and took advantage of improved machining capabilities, waveguide components would drop in price while improving in performance. The simplicity of interconnecting coaxial components also had its appeal, and would motivate engineers to improve the mating interfaces of coaxial connectors so that eventually they could support millimeter-wave frequencies with relative low return loss (VSWR) and low insertion loss. Improvements in connectors and their associated coaxial cables, as well as in printed-circuit-board (PCB) materials, would enable high-frequency design engineers to achieve higher frequencies with planar circuits using microstrip, stripline, and coplanar-waveguide (CPW) transmission-line technologies. This started what has been a 50-year ongoing trend to shrink active and passive RF/microwave components.
One of the earliest of passive component suppliers, and still in business today, is ARRA (Bay Shore, NY). Founded in 1957 by Harold Isaacson, it is managed today by his son Roby with very much the same attention to detail. The company is well known for one of its original product lines, direct-reading continuously variable coaxial attenuators, including extremely broadband units capable of full coverage from 2 to 18 GHz. Capable of 60 dB or more attenuation with 0.1-dB resettability of attenuation value, these variable attenuators have proven invaluable on numerous high-frequency systems where signal levels must be set with precision, including in test and measurement applications.
Passive components, such as attenuators, filters, and switches, that provided various forms of signal processing, were vital to the development of electronic-warfare (EW) and electronic-countermeasures (ECM) receivers in the decades following World War II. Manufacturers of these components sought ways to reduce loss, increase bandwidth, or improve other performance characteristics. In one case, owing to reliability issues and the need to maintain systems in the field, Philco (Lansdale, PA) developed a solid-state waveguide switch with a simple means of replacing the diode switching element (Fig. 5). Unscrewing a knurled nut on the switch's waveguide housing provided easy access to the diode mounted in the middle of the waveguide assembly, allowing for simple removable and replacement.
From a company founded by Bruno Weinschel, Weinschel Engineering (Gaitherburg, MD), also came continuously variable attenuators that were invaluable for controlling signal levels at microwave frequencies. These units relied on a dielectrically coupled thin-film resonator and precision coaxial connectors to cover bandwidths of 2.5 to 12.0 GHz and wide while handling as much as 1 kW peak (1 W average) power. The basic attenuating element was a resistive thin film deposited on a ceramic substrate. Attenuation control was capacitively coupled to the attenuating element, with a drive carriage riding on nylon runners for smooth operation and excellent resettability. A 12-GHz unit with 0 to 10 dB attenuation (Fig. 6) achieved less than 0.5 dB attenuation variation across the frequency range with less than 1.50:1 VSWR. It (model 953) sold for $195.00 in 1963.
In the early 1960s, components with octave bandwidths provided a great deal of flexibility when a target system operated over different portions of the S-band, or X-band, or other frequency range, and fledgling component suppliers sought to achieve octave bandwidths with their products. Merrimac Research and Development (Irvington, NJ), which would later become Merrimac Industries, touted their models CHT-3K and CHT-6K magic tees (Fig. 7) for use from 2 to 4 GHz and 4 to 8 GHz, respectively, for use in C-band and S-band systems. A signal injected into the component's series arm would then feed a balanced-unbalanced (balun) structure, resulting in two balanced output signals. These versatile passive components could be used as balanced mixers, equal-power splitters, frequency discriminators, phase shifters, or matched variable power dividers. They featured low VSWR of 1.50:1 across most of their octave bands, and sold for $275 each.
Five decades earlier, size was still an issue with RF/microwave systems and their components; high-power transistors were still years away and most components were single-function designs, with little thought to integrating multiple functions within a single housing. An example of the large size of power sources was the model X841D UHF radar klystron tube from Eimac, which could physically dwarf a man standing next to it (Fig. 8). The device provided 150 kW average power and 2.5-MW peak power (with 2000-s pulses) from 400 to 450 MHz and could even operate with frequency agility over a 5% bandwidth. It required 56 A peak beam current and 105 kV peak beam voltage to achieve 40% efficiency.
But some companies, such as Watkins-Johnson Co. (Palo Alto, CA), Avantek (Santa Clara, CA), and Microwave Associates (Burlington, MA), which would later become M/A-COM and then M/A-COM Technology Solutions, responded to system-level requirements for smaller components by exploring various levels of integration, typically combining several components within a single housing.
In May 1964, for example, Watkins-Johnson Co. introduced their model WJ-163 integrated front-end unit (Fig. 9) in a 19-in. rack-mount enclosure. It was a integrated preselector filter and amplifier with tunable octave-band filters based on yttrium-iron-garnet (YIG) resonators. Available in versions covering 1 to 2 GHz, 1 to 4 GHz, 4 to 8 GHz, and 8 to 12 GHz, each had a tunable passband of 30 MHz with 80 dB or more image rejection and low noise (10 dB at L/S-band frequencies and 6 dB at C/X-band frequencies). The rack-mount assembly incorporated a TWTA preceded and followed by two-stage electronically tuned YIG filters. The TWTA provided 25-dB gain and was packaged with its own regulated, solid-state power supply. The front-end unit also included a balanced mixer, two transistorized IF amplifiers, and a video detector. Priced at under $10,000, it could be used as the basis for a complete receiver.
In these early days of RF/microwave electronics, transistors had a long way to go before approaching the power levels associated with modern silicon LDMOS devices. In 1964, for example, Texas Instruments was very involved in the development of microwave transistors. Harry Cooke wrote in August 1964 about improved silicon devices from his company, such as the model TIX3016, with a cutoff frequency of 1.7 GHz at 5 mA bias current. Cooke promoted the use of this and other of TI's solid-state devices in cavity oscillator-amplifier circuits for L- and S-band applications. Unfortunately, packaging was still a serious limitation for transistors at that time, since the usable upper frequency limit for compact transistor housings (such as the TO-18 package) was only about 1.5 GHz due to parasitic inductance and high package losses.
Generation of low-noise signals has always been a priority in RF/microwave systems. For several decades, oscillators based on YIG spheres (Fig. 10, left) were in demand in many military systems and in test equipment because of their broad bandwidths, stable tuning characteristics, and low noise. Earlier manufacturers such as Watkins-Johnson and Avantek would be joined by a number of additional suppliers in the San Francisco Bay area, including YIG-TEK (Santa Clara, CA) and Omniyig, Inc. (Santa Clara, CA), and later Microsource to the north (Santa Rosa, CA). Although many of these companies brought innovation to the technology, including the use of square housings and achieving impressive continuous bandwidths, the early sources and filters were relatively large compared to more recent YIG-based products, such as TO-8 and surface-mountable YIG oscillators from Micro Lambda Wireless (Fremont, CA).
Among the early YIG oscillators, for example, was the series 7400 fundamental-frequency YIG-tuned transistor oscillators from Avantek (Fig. 10, right), which leveraged the firm's proprietary silicon bipolar transistors to cover 4 to 8 GHz. In this design, the output of the YIG oscillator drove a buffer amplifier, eliminating the need for an output isolator. The RFI-shielded source delivered +10 dBm output power into a 50-Ω load with only 2 dB variation. The YIG sphere was stabilized with temperature by means of a self-regulating heater which added to the power consumption (2 W at room temperature) and to the 22-oz. weight of the oscillator. Still, this was an early microwave source with 15 MHz/mA tuning sensitivity and -60 dBc typical spurious outputs. It was available in commercial and military versions, with the military variety selling for $1695.
By 1986, Microsource would make a dramatic change in YIG oscillator and filter packaging by introducing the MICRO YIG 1-in. cubed package for components from 0.5 to 8.0 GHz and a 1.25-in. cube for components to 24 GHz. The 1-in. oscillator operated from 2 to 8 GHz with about +15 dBm output power and -10 dBc harmonic levels, with output variations of 0.3 dB. The 1.25-in. filter operated from 8.0 to 18.6 GHz with four stages and 3-dB bandwidth of 40 MHz. It had 5-dB in-band insertion loss and 80-dB off-resonance isolation with passband VSWR of 2.0:1.
Passive component technology improved gradually and steadily over the years in terms of performance. But it really required a change in attitude on pricingas brought on by companies such as Anzac, Mini-Circuits, and even Avantek to make microwave technology accessible to commercial applications, including cellular radio. These lower costs were made possible in many cases by improved manufacturing practices to dramatically reduce the cost of many RF/microwave components. In its early years, for example, Mini-Circuits found ways to cut the costs of producing diode microwave mixers without sacrificing performance or quality. More recently (2000), the firm has applied its founding principles to create low-cost passive FET mixers such as the model HJK-21H (Fig. 11) with RF range from 1850 to 1910 MHz for wireless applications. By adopting surface-mount packaging, not only can the mixers be made smaller, but the costs associated with larger metal machined packages and coaxial connectors can be eliminated.
Around the same time, Synergy Microwave Corp. (Paterson, NJ) would introduce a new circuit approach called SYNSTRIP technology that would enable extremely small passive microwave components. It was demonstrated in two-way power splitters measuring only 0.30 x 0.25 x 0.1 in. for use in GSM and PCS cellular frequency bands. In spite of the small size (Fig. 12), the port-to-port isolation was better than 19 dB from 800 to 1000 MHz with only 0.5-dB insertion loss and less than 1.50:1 VSWR. The technology was based on the use of planar transmission structures and mixed-mode transmission lines in which there is a reduction in size without a reduction in performance or increase in insertion loss. The mixed-mode approach allows a designer to optimally distribute a circuit in different layers with different transmission-line technologies, with the meandering of lines greatly reduced.
Of course, not all technology approaches take root. A report in 1972 detailed what was offered as the ideal microwave switch for applications through 2 GHz: the mercury switch (Fig. 13). It was touted as having the ideal VSWR, infinite isolation, and minimal insertion lossin short, the ideal microwave switch, requiring almost no switching power. The switch consisted of pole pieces and contacts, armature, center barrel, and glass seals and mercury. The tiny magnetic pole piece rode on a totally enveloping film of mercury. External magnetic force was used to pull the pole piece to one contact or another, and it was held there by the surface tension of the mercury. The switch reported rise times in the nanosecond range for use at microwave frequencies.
In the 1970s, the mercury switch could be made considerably smaller than other RF/microwave switch technologies of that time, with about 0.25 dB insertion loss at 700 MHz compared to about 0.7 dB insertion loss at 12 GHz for a solid-state switch. A mercury switch measured about 0.25 in. long and 0.04 in. in diameter, with very little contact bounce, and could operate to 2 GHz. It could pass pulses with rise times faster than 150 ps with no observable distortion and only about 10 ps delay, making it attractive for many fast-pulse applications.
The importance of packaging to microwave component advancement can never be overlooked, and one of the truly groundbreaking companies in terms of high-frequency packaging was a company more typically associated with test equipment, Tektronix (Beaverton, OR). The company was one of the first to recognize the need for a true microwave package for GaAs MMIC components, introducing an eight-port package (Fig. 14) measuring 0.310 x 0.310 in. during the early stages of the "GaAs MMIC revolution" in the 1980s. As high-frequency companies discovered integration of components on GaAs and other semiconductor materials, components became smaller and less expensive but packaging was still a hurdle. This package consisted of a thin-film hybrid circuit brazed to both an etched Kovar leadframe and a tungsten/copper thermal button. It could handle more than 1 W power (+30 dBm) from DC to 18 GHz. The insertion loss between the lead frame and the bond-wire launch was less than 0.5 dB to 12 GHz and less than 1 dB to 18 GHz. Isolation between opposite leads was better than 30 dB through 18 GHz. Unfortunately, the package sold for $39 in 500 to 1000 qty. But it did pave the way for later package developments that would make IC-based components practical without sacrificing their performance.
At present, a leading area of interest for microwave components is in producing low-cost millimeter-wave devices for practical communications networks and wireless links. Ironically, millimeter-wave bands were first heralded as a growth opportunity in 1962 and perhaps the industry is still waiting.