The SiGe IC market in a wireless, broadband world

09/01/2000

Anthony Catalano, Technology Assessment Group Inc., Boulder, Colorado

SiGe technology is enabling silicon-based ICs to penetrate the rapidly growing market for wide bandwidth and wireless telecommunications.

In this article, we discuss projections for the emerging SiGe market; give a brief overview of SiGe technology; and identify the trends that shape the markets for SiGe telecommunications ICs, and manufacturing equipment and materials.

Present and future markets

The overall market for SiGe ICs is growing rapidly owing to the lower-cost, high-performance opportunity created at the boundary between GaAs and Si ICs. Adoption of this technology is still at an early stage, since several factors remain barriers to full market potential. These are:

1. The intellectual property is narrowly held by only a few organizations.
2. Relatively few manufacturers and especially foundries have SiGe manufacturing expertise.
3. The current lack of SiGe-based CMOS processes limits the appeal of the technology to the broader range of users. This barrier can be overcome, however, by the use of the somewhat more complex CMODFET approach.

Because sales of SiGe were only beginning in 1999, total sales were believed to be $3.7 million, but are expected to increase to $136.3 million as many new IC products reach customers. In the long term, we expect SiGe to occupy at least 36% of the overall market for wideband ICs in the wireless handset market.

As manufacturing capacity increases, sales of SiGe ICs are expected to increase rapidly, displacing both silicon and GaAs in the middle frequency/bit-rate market up to about 10Gb/sec. Within 5 years, the market for SiGe products is expected to increase to >$1 billion, as market share and the overall market grow at double-digit rates.

Technology overview

When a thin layer of SiGe alloy (typical Ge concentrations are on the order of 20-30 atomic percent) is grown on a single-crystal wafer, the SiGe adopts the atomic arrangement and spacing of the underlying silicon. If the film is kept thin, usually much less than 1mm, the compressive stress that results from cramming a larger Ge atom into the lattice improves the hole mobility. This, combined with a favorable emitter-base band offset, markedly increases the maximum operating speed of heterojunction bipolar devices (HBTs) that use SiGe for the base region. Conversely, growing a Si-layer on a "relaxed" SiGe layer (i.e., one that has been allowed to increase its atomic spacing based on increased Ge content), causes the electron mobility to increase, increasing the operating frequency of nMOS. pnp HBT devices don't benefit substantially due to less favorable band offset.

The principal device finding its way into such commercial products as low-noise amplifiers (LNAs) and power amplifiers (PAs) is the HBT. Combining SiGe HBTs with conventional CMOS for BiCMOS allows a wide array of analog and mixed-signal IC applications in the wireless and communications industry. SiGe boosts signal I/O and extends operating frequencies.

Click here to enlarge image

Table 1 compares the operating characteristics of GaAs and SiGe HBTs. The maximum frequency of operation of SiGe is close to that of GaAs. From a practical perspective, the high power added efficiency (PAE), comparable to GaAs, ensures that battery life in wireless applications is comparable to GaAs. The operating voltage is also substantially lower. These factors are important for wireless handset applications. The lower breakdown voltage of SiGe devices, however, limits certain switching and transmission applications, such as rf transmission chains in wireless systems.

The use of strained SiGe in CMOS devices increases the speed on the pMOS end by about 20%. Although this can be viewed as an alternative to linewidth reduction, since the chip area devoted to the pMOS devices can be significantly reduced, it is unclear that the advantage outweighs the added complexity and cost of fabricating the devices. Ge interferes with the formation of the high-quality oxide vital to the MOS devices. Processing requires the SiGe to be buried or capped with a thin layer of pure silicon that is sacrificed to form the oxide.

The payoff for the added processing complexity of SiGe CMOS devices comes by combining the compressed SiGe film with Si under tensile stress. Utilizing these structures in a MODFET design that incorporates a buried n- or p-type source layer increases effective mobility four-fold. Combining such CMODFET designs with SOI, low-k dielectrics, and copper suggests multi-GHz processors might be realized. The significant changes in the ordinary CMOS process, however, will likely prevent the industry from embracing the technology within the next 5 years. It is possible, however, that a less invasive type of strained Si CMOS could appear in a shorter time.

Other opportunities for SiGe lie in applications such as far-IR detectors, Si-integrated waveguides, optical switches and routers, detectors, and even light emission — all integrated onto a silicon chip. Whether any of these exciting technologies is realized depends on the eventual adoption of SiGe on a wider scale. R&D on these applications is already underway.

Applications and markets

It is useful to consider broadband applications in the frequency and bit-rate domain to gain an appreciation for where the various semiconductor IC materials fit. The frequency and bit rates for applications are determined by a combination of international industry standards and government allocation of frequency space.

Although each of the IC technologies can cover the lower-frequency range, costs generally prohibit their use. This is especially true of the III-Vs such as GaAs and InP; it is less so for SiGe as its manufacturing costs are only marginally higher than silicon alone.

The diverse range of applications in the broadband space includes local multipoint distribution systems (LMDS) that distribute information within major metropolitan centers via microwave transmission; mobile satellites; wide and local area networks (WLANs); cellular, third-generation (3G), as well as unlicensed and restricted personal communication services (U-PCS and PCS) for wireless telephony; gigabit ethernet/fiber channel; asynchronous transfer mode (ATM) switching and multiplexing for information transfer; and the synchronous optical network (SONET) protocols and systems for optical communication. Although most of these applications present an opportunity for SiGe ICs, we will concentrate on two of the largest: SONET communications ICs and wireless telephony.

A major rapidly growing market for broadband ICs is in SONET/SDH/WDM telecom systems. SONET and SDH (synchronous data hierarchy) refer to the protocol for assembling and transmitting data in an optical or equivalent electrical media. WDM refers to wavelength division multiplexing, the method of assembling several relatively closely spaced wavelengths of light within an optical fiber to carry data.

Major industry trends favor continued rapid growth. Deregulation of the telecommunications industry has resulted in competition among interexchange carriers (IXCs), local exchange carriers (LECs), and competitive exchange carriers (CLECs), each trying to build infrastructure. Competition is further fueled by residential and business customers' increasing demand for higher transmission speeds. Expend-itures for equipment by these groups as well as cable TV (CATV) service providers reached $39 billion in 1999 and will increase by about 13% to $44 billion this year.

Specific systems using wideband ICs are

1. regenerators that boost the optical fiber signal periodically along the fiber route, often by transforming the optical signal to an electrical signal;
2. terminal multiplexers that aggregate lower bit-rate signals;
3. add/drop multiplexers that separate the signal from the bulk transmission; and
4. cross-connects that distribute the digital signal to appropriate destinations.

Products in the OC-192 10Gb/sec range are becoming well established, and those based on the OC-768 40Gb/sec standard are emerging. Since the cost of installing new fiber is extremely high, the decision to carry more information via higher bit rates and wavelength division is usually straightforward. The move to higher frequencies creates a greater demand for the lower rates, too, as they must be combined through multiplexing to stuff the pipeline with the higher level of data. The growth in this wide middle range of bit rate and frequency constitutes a large market opportunity for SiGe ICs.

The move to the highest bit rates (OC-768) means increasing demand for ICs based on exotic semiconductors such as InP. InP-based devices are attractive for very high bit-rate devices because of device operating speed and because they permit an all-optical system that bypasses the requirement to transform the optical signal into its electrical counterpart. This opportunity lies about 2-5 years in the future, although the first products may appear in early 2001.

The market for all types of wideband SONET/SDH/WDM ICs was believed to be $817 million in 1999, and is expected to increase by 22% this year, to a value of almost $1 billion. SiGe products for these applications only began to enter the market in 1999. Sales in 1999 were only about $1 million, but may reach $40 million this year.

Several companies have entered the SiGe telecommunication IC industry. Applied Micro Circuits (AMC) announced several SiGe products aimed at the optical communications market. Most recently, AMC has introduced a differential crosspoint switch for WDM applications, as well as an OC-192 framer device. Other organizations, such as Conexant, Lucent, and IBM, are expected to introduce SiGe telecommunications products this year. Most major electronics manufacturers have announced either an internal effort or strategic alliance that will give them access to SiGe-based products. Table 2 on p. 62 gives a list of some of the organizations with such efforts underway. Another exciting opportunity for SiGe is presented by Amberwave, an MIT spin-out. The company is using SiGe as a mechanical platform that may allow more exotic materials such as GaAs and InP to be integrated on a silicon wafer, providing a dramatic increase in performance with the promise of reduced cost.

Wireless telephony

Major trends in wireless telephony are creating a very sizeable opportunity for SiGe sales. Service providers are experiencing heightened competition as the price for service ratchets down. The resulting customer turnover almost always results in the sale of a new telephone. Superimposed on this trend is the general increase in the number of wireless customers. To increase margins and subscriber base, service providers are bringing new services such as data transmission/internet access and e-mail to the wireless world. Although this trend is more complicated in the US compared to the rest of the world where GSM dominates, the trend from low-frequency analog cellular (~0.9Ghz) to high-frequency digital PCS (~1.9Ghz) remains strong.

Regulatory changes are also forcing an increase in the functionality of wireless telephones. Within the next several years, wireless phones will be required to have emergency locator capability (E-911) so that emergency personnel can easily identify the location of a caller. While several solutions exist, one of the more attractive is the built-in global positioning system (GPS), which uses the two frequencies of 1.57542 and 1.2276GHz. This added function provides a further opportunity for SiGe RFICs to provide the feature at a modest additional expense.

Click here to enlarge image

Although the increase in wireless handsets is robust, projections for handset sales and wireless infrastructure generally discount the possibility of wireless telephony displacing residential and business telephones. This would boost demand for handsets and infrastructure, and consequently, RFICs and related products. Already, the business component, often referred to as in-building wireless, is growing at over 50%/year. These telephones operate at frequencies similar to PCS and cellular. Handset prices are higher than consumer products, on the order of $1000, making this high-end market attractive.

As an example of robust growth, the market for wide bandwidth ICs in this sector has increased 12.2% to $2.57 billion in 1999 and is expected to grow by 28% to $3.3 billion in 2000 based only on wireless phone sales in the US and Canada. The SiGe share of this market is still quite low, as manufacturers only began rolling out products in 1999. Manufacturers who have announced wireless handset products are listed in Table 2.

Although the beauty of the SiGe process is that it is compatible with contemporary Si processing, opportunities for equipment makers lie in two areas: SiGe deposition and SiGe etch and end point detection.

Deposition of the SiGe epitaxial layer is usually accomplished by CVD, although both MBE and ion implantation are important. Among the CVD methods, the UHV-CVD method developed at IBM is attractive because it allows very low process temperatures that preserve the narrow doped features within the device and minimize dopant cross-contamination. However, the method relies on a batch rather than a single-wafer process. More conventional single-wafer CVD equipment is also attractive and manufacturers like ATMI are supplying substantial numbers of units for this purpose. Within 5 years, we estimate annual sales of about 100 single-wafer CVD units will be required to meet demand.

SiGe presents opportunities in the etch area, too. SiGe can be etched using chemistries similar to Si, but the fine features and profiles required to maintain high frequency place increased demands on processes. Novel chemistries can be an aid and end point detection plays a vital role.

Conclusion

Employing existing Si process technology, SiGe allows Si-based devices to operate at the higher frequency and bit rates normally associated with GaAs-based products. SiGe presents an opportunity not only to IC manufacturers interested in entering the telecom marketplace, but also to equipment and materials suppliers who can deliver targeted products. While we can regard SiGe as evolutionary in the 3-5-year term, new developments in the technology are certain to revolutionize silicon in the 21st century. It remains to be seen, however, how far silicon's advance will go. This will ultimately be determined by the extent to which the industry embraces SiGe.

Anthony Catalano is president of the Technology Assessment Group Inc., a consulting firm specializing in custom strategic reports. He can be reached at PO Box 19344, Boulder, CO 80308; ph 888/766-0116 or 303/442-6760, fax 303/442-4960, www.TechAGroup.com.

This article is based on the three-part report, SiGe: Silicon-based ICs for Wide Bandwidth Telecommunications, available in hard copy and CD-ROM from the Technology Assessment Group. Price: $3995.