Issue



Demand for advanced communication devices drives the compound semi market*


07/01/2002







In the late 1990s, the focus of the compound semiconductor IC industry changed from small-scale demand for niche applications to high-volume demand for advanced communications systems.

The increasing level of data, voice, and multimedia transmission through wired and wireless communications systems, for example, has resulted in the need for higher operating frequencies and more bandwidth. The growth of digital cellular phones, PCS products, pagers/text messengers, GPS navigation products, satellite communications, and wireless networks has forced the wireless communications industry to move to higher, less congested frequency bands.

Meanwhile, the growth of traditional voice traffic and higher levels of data traffic arising from facsimile communications, computer networking, cable and high-definition TV, and on-line and Internet services are driving the telecommunications and data communications industries to boost transmission rates and increase capacity. As a result of these increased demands, compound semiconductor technology has emerged as an effective alternative or complement to silicon-based solutions in many advanced communications applications.


Figure 1. Compound semiconductor market forecast.
Click here to enlarge image

Figure 1 shows that the market for compound semiconductor ICs grew tremendously in 2000, but like the rest of the IC industry, suffered in 2001. A rebound is expected to start in 2002 and gain momentum in 2003 and 2004.

Compound semi materials
Compound semiconductor materials like gallium arsenide (GaAs), indium phosphide (InP), and silicon germanium (SiGe) have intrinsic electrical properties that give them performance advantages over pure silicon, such as higher frequency operation, improved signal reception, better signal processing in congested bands, and greater power efficiency. They can also be bandgap engineered, which allows them to be optimized for speed or power in a given application.

SiGe is a compound in a different sense than III-V compounds like GaAs and InP. While Si and Ge are both column IV elements, SiGe devices are not built on SiGe wafers. Instead, SiGe ICs begin with a standard silicon substrate; a small amount of germanium is then added to the transistor structure, extending silicon performance.


Figure 2. a) 2001 compound semiconductor market share ($2.1 billion); and b) 2006 compound semiconductor market share forecast ($7.2 billion).
Click here to enlarge image

Compound semiconductors are still largely based on GaAs. SiGe is expected to gain a significant amount of market share by 2006, however (Fig. 2). Although it is difficult to make direct comparisons between SiGe technology and other competing technologies like GaAs, SiGe stands out in certain areas because it is compatible with conventional silicon processes.

First, SiGe devices can be processed on existing silicon wafer fabrication lines with only slight modifications. That means that SiGe chips can benefit from the economies of scale that today's high-volume 200mm fabs can provide, thereby lowering per-die costs. Second, the advantages of CMOS technology in low-power, high-density digital signal processing can be merged with the high speed of the SiGe heterojunction bipolar transistor (HBT) in a BiCMOS process. In wireless communications, for example, an SiGe BiCMOS process allows for the RF and baseband circuitry to be integrated on the same chip, bypassing the IF stage where the radio signal is converted to a signal that can be processed by the baseband DSP. This direct conversion or zero-IF design technique is expected to be a major driver behind the increased use of SiGe technology.

InP has recently gained interest because it has the potential to extend IC performance beyond the capability of other semiconducting materials. Its high electron mobility allows it to attain cutoff frequencies as high as 250GHz, enabling IC design that supports 40Gbits/sec fiber communications applications, with the potential for 80Gbits/sec or higher operation. InP also requires a lower supply voltage than GaAs (the base-emitter voltage is 0.7V for InP HBTs vs.1.4V for GaAs HBTs), which translates into devices that consume less power. InP HBTs also exhibit better noise and linearity performance than GaAs HBTs.

Major suppliers
Several of the world's leading compound semiconductor suppliers are listed in the table. After holding the top spot for several years, Vitesse lost this place in 2001. Besides being hurt by weakness in the network equipment industry, the company converted its lower-speed components (i.e., ≤2.5Gbits/sec) from GaAs to CMOS, as well as added CMOS-based network processors and switch fabric devices to its product line. It is now broadening its process portfolio to meet a range of performance and/or integration levels.

RF Micro Devices (RFMD) claimed the top position in 2001. The company's revenues declined during the year, but not to the extent that other companies' revenues dropped. The cell phone industry's switch from GaAs MESFET to GaAs HBT technology for power amplifiers in handsets has been appreciated by RFMD, which has promoted the technology since licensing it from inventor TRW in 1996. The company has also worked to diversify its business beyond cell phones and into other areas like wireless networks and digital set-top boxes.

TriQuint's sales in 2001 dropped relatively little. The company's products and services, probably the most diverse of the leading compound IC suppliers, span RF and millimeter-wave frequencies and employ analog, digital, and mixed-signal circuit designs. Its products are used in wireless communications, telecommunications, data communications, and aerospace systems. TriQuint not only offers standard and customer-specific IC products, but also foundry services.

IBM has been producing SiGe ICs for its own internal use for several years. It has also been the biggest driving force behind SiGe technology commercialization. Since opening its SiGe manufacturing capability to the outside world in 1998, the company has signed numerous foundry customers, including Alcatel Microelectronics, AMCC, Intersil, Leica Geosystems, National Semiconductor, Nortel, Conexant, Qualcomm, RF Micro Devices, Sierra Monolithics, and Tektronix.

A pure-play compound semiconductor IC foundry industry has recently emerged. Trying to do for the GaAs and InP businesses what UMC and TSMC have done for the silicon world, several companies are setting up shop to focus on manufacturing GaAs and/or InP ICs for other suppliers. Not surprisingly, most of these companies are building their fabs in Taiwan.

Future technology trends
Motorola, in collaboration with IQE of the UK, has demonstrated the growth of GaAs (or other III-V compounds) on silicon substrates with high yields. This could lead to significant cost savings in GaAs device manufacturing because of the economic benefits of silicon wafers vs. smaller, more expensive compound wafers. In addition, the technology opens up a variety of integration possibilities, such as combining silicon circuitry on the same chip as optoelectronic components.

The key to the new technology is the use of a thin layer of strontium titanate (strontium titanium oxide, SrTiO3, or STO) between the silicon and GaAs layer. The STO acts as a compliant buffer layer, reducing mechanical strain and the thermal mismatch between the silicon and GaAs (Fig. 3). Motorola has formed a subsidiary called Thoughtbeam Inc. to develop, sell, and license the new technology.


Figure 3. Motorola's GaAs-on-silicon technology. Source: Motorola.
Growing GaAs or other III-V compounds on silicon substrates could lead to significant cost savings in GaAs device manufacturing.
Click here to enlarge image

There is also momentum gathering behind packaging GaAs/InP RFICs alongside passive components in small multichip modules. The modular approach shrinks area requirements, reduces electrical parasitics, and minimizes the required RF competence of the electronic system manufacturer (e.g., cell phone maker).

RF Micro Devices shifted a significant portion of its power amplifier business to modules in 2001. One of the main motivations behind TriQuint's merger with Sawtek in 2001 was to combine TriQuint's GaAs ICs and Sawtek's SAW filters into RF modules for wireless applications.

Click here to enlarge image

The first chips for OC-768 (40Gbits/sec) optical networks are expected to ship in 2002, and InP has been chosen by many companies as the material for the devices. Vitesse, Anadigics, Alpha, startup Inphi, and TRW-spinoff Velocium, for example, are all designing InP ICs for OC-768. Meanwhile, AMCC and Infineon are staying with SiGe instead of switching to InP.

Several companies have released details on new SiGe process capabilities for foundry services. All of the new services are BiCMOS processes targeted at mixed-signal devices in wireless and wired communications applications.

The companies are also offering multiproject wafer programs to help lower the cost of SiGe prototypes and small-volume products by fabricating more than one chip design on each wafer.

Finally, IBM says it has built an HBT with a switching frequency of 210GHz using a 0.18μm SiGe process. Although it was once thought that silicon-based transistors were incapable of exceeding 200GHz, IBM believes the technology will be used to build communications chips with clock rates of 100GHz in 2003.

*This article is adapted from the 2002 edition of The McClean Report, "Section 14: Technology Trends," by Bill McClean, Brian Matas, and Trevor Yancey. For more information on the report, contact IC Insights Inc., 13901 N. 73rd St., Ste. 205, Scottsdale, AZ 85260; ph 480/348-1133, fax 480/348-9745, e-mail [email protected], www.icinsights.com.