The frequency range of 10-100 GHz offers a gold mine of commercial and military applications. Full-blown system integration has been rare, however. Traditionally, microwave circuit designers interested in exploiting these frequencies have turned to discrete components, despite the expense of assembling and packaging them and the dearth of buyers that can afford the result.

Today this scenario is rapidly being transformed by indium phosphide heterojunction bipolar transistors (HBTs), a technology that is now yielding large-scale integrated (LSI) circuits packing 1000 to 10 000 transistors on a chip and operating at over 65 GHz. Success in integrating such fast, dense transistors has bred numerous analog, digital, and mixed-mode ICs having unprecedented performance. Moreover, such high-speed integration is not only possible but marketable, too, because consumer demand for data from the Internet, digital audio and video, and video-on-demand is on the rise.

In general, data is delivered to the consumer in two ways. Guided-wave propagation occurs along optical-fiber links and coaxial cables. Free space transmission is the territory of radio, television broadcasting, satellite links, local multipoint distribution systems, and wireless local-area networks.

For free space propagation, expensive installations are not necessarily required, but governments do regulate bandwidth and frequency use with a firm hand. As a result, new or large-bandwidth applications are forced to resort to unallocated higher carrier frequencies. For example, a new wireless local-area network may have to operate at 60 GHz. Besides possible economy, a second attraction of higher frequencies is that they use smaller antennas. Because antenna efficiency is proportional to carrier frequency, receivers of higher carrier frequencies need only compact antennas. An antenna just 10 cm in length suffices for collision avoidance automotive radar at 77 GHz, for instance.

As for guided-wave systems, cable has to be deployed, but once it is in place, bandwidth is relatively unregulated. In this kind of system (such as an optical-fiber data link), the development of high-frequency components is mainly a response to the call for higher data rates. As optical-fiber trunk lines are more fully exploited, the next generation of communications channels will come to operate at 40, 80, or 100 Gb/s, well above the current 2.5-10 Gb/s. In fact, 100-Gb/s links could be in production as soon as 2010.

These data rates raise another obstacle. The signals must be received and processed at high speed and then eventually slowed down to match silicon CMOS processing speeds of at best 1.5 Gb/s. Such a receiving circuit requires complexity at least at the LSI level.

Worth developing

The way things are tending, access to an ultrahigh-speed LSI technology will be a must for system designers. Happily, the necessary level of complexity and speed is available from HBT ICs built on indium phosphide substrates using materials with different energy bandgaps for the emitter, the base, and, sometimes, the collector. They are the only viable option today for ICs that require 30-GHz-plus frequencies and LSI complexity [Fig. 1].

[1]: Indium phosphide technology leads all others in operating frequency, but it is not expected to approach the integration scale of silicon chips. Silicon or silicon germanium (Si/SiGe) bipolar transistors and gallium arsenide (GaAs) bipolar transistors and metal semiconductor FETs (MESFETs) have intermediate combinations of speed and level of integration. The current state of the art [dark areas] can expect to improve over the next five years [light areas].

Applications that can get by with lower levels of integration but require higher frequency operation of 94 GHz and beyond, such as missile radar, are currently best served by indium phosphide high-electron-mobility transistors (HEMTs)—devices similar to depletion-mode field-effect transistors.

Indium phosphide transistors were first investigated in the early 1980s for their potential as optoelectronic communication devices. But a demonstration of an indium phosphide HBT IC was not reported until 1989, by a group at Rockwell Science Center in Thousand Oaks, Calif. Indium phosphide technology was originally funded through various U.S. government agencies interested in obtaining high-performance components for defense systems. Since then, the increased system throughput provided by these HBT ICs has encouraged a move toward commercial applications of the technology by such companies as TRW, Lucent, and HRL Laboratories in the United States and by Nortel in Canada. Research in indium phosphide monolithic microwave ICs and devices is also under way at many Japanese companies, including Hitachi, NEC, and NTT, and at many European research centers.

Currently the only obstacle to wider adoption of indium phosphide is the high cost of the 7.5-cm wafer substrate: it is about five times that of comparable gallium arsenide (GaAs).