Semiconductors Save Space Through Integration

Semiconductors have advanced from simple, discrete two- and three-terminal diodes and transistors to complex monolithic circuits with hundreds of components on a single chip.

Arguably, semiconductor technologies for microwave applications have undergone a more dramatic evolution over the last 50 years than any other facet of this industry. Just in terms of frequency coverage and output power, for example, early semiconductors struggled to produce sub-1-W output power levels at frequencies as high as 2 gHz. Five decades ago, real output power was the domain of vacuum electronic devices, such as traveling- wave tubes (tWts), and diodes were just as likely to be used as the active devices in an amplifier as were transistors.

In 1965, step-recovery diodes (d5300 series) from the sylvania subsidiary of general telephone & electronics (gte) could obtain 8 mW output power at 10 gHz by means of X25 multiplication of a 1-W input signal at 400 mHz. these diodes boasted a low junction capacitance of only 1 pF and minimum breakdown voltage ranging from 15 to 45 V, depending upon package style, with transition times of 0.7 ns or less.

In 1966, Fairchild semiconductor's model mt1050 coaxial NPN silicon planar epitaxial transistor (Fig. 1) was capable of 200 mW output power at 2 gHz and was priced at $350. it was supplied in a coaxial package with 4-pF output capacitance and 8-pF emitter- transition capacitance and offered maximum power dissipation of 2 W.

In the same timeframe, large-area step-recovery varactor diodes from motorola semiconductors were being used to generate as much as 1.8 W power at 8 gHz. the model in5153 step-recovery varactor diode sold for $28 in 100 quantity and achieved 6 W output power at 2 gHz with 50-percent efficiency.

Hewlett-Packard associates in Palo alto, ca also offered step-recovery diodes for frequency multiplication. their devices were capable of 150 mW output power at 10 gHz when fed with 2 W input power at 2 gHz.

In terms of practical applications, avantek of santa clara, ca offered their model amP-2600 low-noise transistor amplifier for a mere $2400 (Fig. 2). it had a typical noise figure of 6 db from 1.0 to 2.6 gHz, which was considered acceptable at that time since this unit was developed as a replacement for low-noise, mediumpower tWts in airborne applications. the avantek amplifier could be supplied with either 25- or 33-db small-signal gain and output power at 1-db compression of +6 to +8 dbm.

Early High-Volume IC

Such discrete devices were yet to be replaced in any major way by integrated circuits (ics), but 1966 also saw great attention paid to the first large-scale use of ICs, in a phased-array radar system. Considered novelty devices to that point, ICs were finally put to use in large numbers by Texas Instruments ( as part of the development program for the MERA X-band phasedarray radar. Designed to be the equivalent of a 60-kW vacuum-tube-powered radar, this system relied on 600 solid-state modules and antenna elements in an array to achieve comparable results in terms of radar system performance. Developed under contract to the US Air Force, the 600-element-array system made use of 100:1 pulse compression so that the power of each array module could be reduced to 1 W pulsed power at a 0.1-percent duty cycle.

Subassemblies included a 500-MHz IF preamplifier and balanced X-band diode mixer. The preamplifier (Fig. 3) was based on double-diffused planar epitaxial transistors with 1.5-GHz cutoff frequency and 3.5-dB noise figure. Across its 85-MHz operating bandwidth, the gain was 18 dB and the noise figure was 4 dB. The mixer was based on single-sided hot-carrier diodes. It provided an RF of 9.375 GHz and IF of 30 MHz, with noise figure of 9 dB.

Before the evolution of companies like Cascade Microtech and wafer probes, on-wafer testing of semiconductors was primitive; discrete devices were often evaluated in "home-brewed" test fixtures. Writing in December of 1967, the legendary Julius Lange (member of the Technical Staff of Texas Instruments in Dallas, TX) addressed the need for better device testing. Specifically, he wrote about modern tests for modern transistors, hoping to replace older transistor tests of H- or Yparameters with new tests of device scattering (S) parameters.

Lange introduced a test mount for making measurements on stripline transistors, using a mounting plate to hold the packaged transistor in place. The mount contained two low-VSWR coaxial-to-stripline adaptors to feed the 50-Ohm transmission lines which connect with the tabs on the transistor package. The contact to the input and output package leads was made by clamping the flat leads between the striplines and the upper dielectric of the test fixture.

Since it was necessary to provide matched 50-Ohm terminations to a DUT, a special slide-screw tuner was also developed, consisting of a coaxial 50-Ohm slab line with round center conductor and two slabs as outer conductor ground return, with an anodized aluminum tuning slug. The tuning slug provided a frequency range of 10 MHz to more than 9 GHz.

In terms of transistor power, Motorola Semiconductor device guru Helge Granberg reported in January 1983 of a solid state amplifier capable of 600 W output power from 2 to 30 MHz (Fig. 4). Built around four of the company's MRF150 MOSFET devices, with independently adjustable bias to the four transistors, in the range from 40 to 50 V, the solid-state amplifier operated with efficiency of 50 percent. Of course, even with that efficiency, heat dissipation was a concern for a solidstate design with this much power.

Microwave engineers have long sought some "magic" semiconductor device that would somehow defy the laws of physics and provide unparalled gain at high frequencies with very little power consumption. One such announcement came in 1980, when Fujitsu broke news about research results on a new "superspeed" semiconductor. Their claims of significantly faster electron mobility than gallium arsenide (GaAs) sent waves of skepticism through an industry that had largely considered GaAs to be the last microwave epitaxial semiconductor material that would ever be needed.

The Fujitsu researchers named their device a high-electron-mobility transistor (HEMT). It consisted of a silicon-doped AlGaAs layer sandwiched between undoped GaAs layers. At liquid-nitrogen temperatures (about -196C), the HEMT field-effect transistors (FETs) exhibited about six times the electron mobility of comparable GaAs MESFETs.

At the time, Cornell University's venerable "godfather of the transistor," Professor Lester Eastman, felt that the Japanese researchers' experiments were flawed and that the fast electron mobility was not as important as some other key device parameters, such as minimal source resistance: "What I think, is that Fujitsu has made the world's worst device with regard to resistance and negative feedback. When they played with reduced temperature, they knocked down a little of that source resistance which should have been lower by more than an order of magnitude to start with."

The HEMT device was just one of a plethora of new developments that would energize the industry over the last several decades and pave the way for many opportunities. One such example came in 1993 when AEL Defense Corp. in Lansdale, PA unveiled a line of successivedetection logarithmic amplifiers (SDLAs) based on a technology considered exotic at that timeGaAs heterojunction-bipolar- transistor (HBT) device technology.

The AEL SDLAs employed GaAs HBT chips from Rockwell International in Thousand Oaks, CA, each chip with RF amplifiers and detector circuitry. Versions were introduced to 6 GHz, with detector ranges of -65 to +5 dBm and tangential signal sensitivity (TSS) of -75 dBm.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.