ALTHOUGH THE MARKET FOR OPTICAL ICs WILL SLOW THIS YEAR, IT WILL RISE FROM $851 MILLION IN 2000 TO $2.9 BILLION IN 2005. MATERIALS TECHNOLOGIES SUCH AS GaAs, SiGe, AND InP WILL SUPPORT THE MOVE TO HIGHER CHANNEL DATA RATES. GEORGE BECHTEL
The rapid growth in optical communication networks has spawned a growing and highly visible optical-communication integrated-circuits (IC) market, rivaling wireless as a new market for companies with high-speed semiconductor products. Sales of ICs operating at 2.5 to 40 Gbit/s totaled $851 million in 2000, up 42% over 1999. These ICs are needed in optical communication networks to drive laser transmitters, to amplify detected optical signals, to perform clock recovery in SONET/SDH networks, and to multiplex and demultiplex lower data rates at the electronic level.
Major semiconductor companies like Intel, previously dependent on the personal computer market, are moving into communications with a bang. Intel has invested $2 billion to buy eight networking and communication companies and plans to spend more than $1 billion in related R&D in 2001.
Two key factors will drive the optical-communication IC market in the coming years. The first is the continued increase in data rates in optical networks. Telecommunication carriers' deployment of 40-Gbit/s (OC-768) transmission will require new levels of performance from ICs. At the same time, the build-out of new, high-speed metropolitan fiber rings is increasing the number of optical connections that require high-speed ICs.
Semiconductor technologies such as silicon complementary metal-oxide semiconductor (CMOS), silicon germanium (SiGe), gallium arsenide (GaAs), and indium phosphide (InP) are now competing to meet these needs. Strategies Unlimited forecasts that the optical-communication IC market will climb to $2.9 billion in 2005 (see Fig. 1).
MARKET LOWER IN 2001
The market for telecommunications services has been growing at an accelerated pace recently. From 1998 through 2000, revenue growth averaged 11% annually compared to 8% from1994 to 1996. The growth of the Internet certainly contributed, but other changes have occurred in the marketplace. Wireless is now a significant portion of the revenue stream for some of the carriers, and others added broadband services such as cable TV to their service menus.
As revenue grows, spending increases to support new customers and services. Capital expenditure growth has averaged 24% annually over the last three years, compared to 18% annual growth for the earlier period. This increase reflects the recent backbone capacity additions (for example, new fiber and DWDM deployment) made by long-haul carriers.
Another change in the telecom market has been the rise in competitive carriers, such as Level 3, Broadwing, Williams Communications, and Qwest. Fiber deployment has been a major consumer of capital for these carriers. In the coming years, local/metro service providers are also expected to invest heavily in new equipment to address bottlenecks in access deployments. In particular, spending will rise to enable high-speed subscriber services via digital-subscriber-line technology or cable modems.
After an $82 billion investment in 2000, will the spending pace stay hot? The first quarter of 2001 has seen a rough start, with equipment suppliers forecasting reduced capital spending by their customers, the telecom carriers. With forecasts for reduced consumer and business spending in 2001, it is unlikely that the telecommunications industry will continue the investment pace of the past two years.
The impact of a slowdown in carrier spending results in a forecast for a decline of 13% in high-speed IC purchases in 2001. As component inventories are worked through in the first half of the year, orders should stabilize before the end of the year as the market resumes its growth path.
LIGHTING FIBER FUELS GROWTH
At the end of 2000, the amount of installed long-haul and metro fiber worldwide reached nearly 180 million km. The driving force behind the IC market is the growth in activated or newly lit fiber each year, not total fiber deployment. Carriers light dark fiber by deploying SONET/SDH terminals, add/drop multiplexers, and crossconnect switches, each using a number of transceiver modules that perform the optical-to-electrical (O/E) conversion for the incoming optical signals and perform the opposite conversion for the outgoing optical signals (see "Communications terminology can be confusing," p. 30). These transceiver modules (also called transponders) contain the ICs that perform transmit and receive functions. The ICs include the laser modulator driver, transimpedance amplifier, clock and data recovery, multiplexer, and demultiplexer chips (see Fig. 2).
The deployment of long-haul networks has been the major impetus for optical-communication components. The demand for backbone capacity has generated impressive improvements in device technology and communication speeds. The move to OC-192 and WDM reduced the cost of adding capacity immensely. The next move to 40-Gbit/s line rates is in its infancy. Deployment of OC-768 networks is forecast to begin in 2002, and by 2005, lit OC-768 WDM fiber is expected to account for 13% of the fiber activated.
The metro market (including interoffice) is expected to outpace the long-haul market in annual deployment growth over the next five years as well as moving to higher data rates. Transport over 10-Gigabit Ethernet networks should see trials in 2001, with considerable deployment by 2005. WDM transmission is expected to penetrate the metro/interoffice market in 2001 with the introduction of OC-48 WDM in 2001 and OC-192 WDM in 2002.
Total transceiver production is forecast to rise to 14 million units in 2005, up from 2.2 million in 2000, with the metro market growing at over 50% per year.
MOORE'S LAW AND OPTICAL ICs
In the past several years, high-speed IC technology has advanced rapidly, increasing the level of integration possible for transceivers. Moore's Law of semiconductor electronics states something to the effect that the number of transistors on a single chip doubles every 18 months. A corollary to that law is that, conversely, the number of chips per system decreases by 50% every two years (see Fig. 3). Second-generation chipsets that can reduce the chip count to four from the seven chips needed in first-generation transceivers are now available. Two analog-layer chips (laser driver and TIA) and two mixed-signal chips (serializer and deserializer) comprise the chipset (see "Transceiver chipsets move toward integration, p. 32").
Another way to reduce the number of chips is to combine electronic functions with optoelectronic functions on a single chip. Third-generation designs will see another step in integration level, bringing the chip count to possibly two. As silicon CMOS moves to smaller gates, single-chip serializers/deserializers (SERDES) are now being introduced to the market. Multirate (OC-3 to OC-48) SERDES chips are also available, eliminating the need for trans-ceiver manufacturers to stock multiple chipsets.
Optoelectronic integration has already occurred with the integrated laser modulator, incorporating an electroabsorption modulator on the same chip as an InP laser. InP is very suitable for receiver integration; products are now appearing with an InP photodiode integrated with an InP heterojunction-bipolar-transistor TIA.
At 2.5 Gbit/s, silicon technologies are likely to completely displace the more expensive GaAs ICs in the near future. Higher levels of integration may yield a two-chip transceiver. The likely higher growth in enterprise fiberoptics should drive technology to lower cost solutions, benefiting the carrier market.
Multiple chip technologies are likely at 10 Gbit/s with Si, GaAs, and the emerging SiGe technology all sharing some piece of the market. Companies such as AMCC, Nortel, and Vitesse Semiconductor are currently developing chips in multiple technologies to provide the optimum solution. The number of chips in a transceiver is forecast to fall from seven or eight to four in the next few years.
Transceivers at 40 Gbit/s may be the first significant market for InP ICs, competing with its semiconductor cousin, GaAs, particularly in the O/E conversion chips. Cost is not likely to be an issue for the next few years for 40-Gbit/s applications, but chip availability and performance will be. The possibilities of monolithic optoelectronic ICs are very real, especially for the receiver. Several laboratories have demonstrated working 40-Gbit/s transmitters and receivers using prototype SiGe ICs.
In the end, component-packaging technology at 40 Gbit/s remains a formidable development challenge. This, more than any other factor, may drive the system designers to use the very best technology in spite of the cost or yield. In this near-term quest for the best solution, all technologies should see continued investment.
George Bechtel is director of wireless market services at Strategies Unlimited, 201 San Antonio Circle, Suite 205, Mountain View, CA 94040. He can be contacted at 650-941-3438, firstname.lastname@example.org, or http://www.strategies-u.com.
Communications terminology can be confusing
The terminology used in digital communications, especially when related to digital data networks, can be confusing. The optoelectronic components, such as laser diode, optical modulator, and photodetector, perform the O/E conversion required to move data between fiber and the copper-wired world. This is frequently called the optical layer. The use of the term layer is derived from the Open System Interconnection (OSI) seven-layer model for managing communication protocols.
The physical layer devices comprise the functions that perform parallel-to-serial data conversion and electronic modulation for the transmitter, and the detection and serial-to-parallel data stream conversion on the receiver side of the transceiver.
The digital layer comprises the devices that encode or decode the data (such as SONET framers), provide error correction, and, in general, groom the incoming data into the appropriate format for individual data streams to or from user interfaces.
From a semiconductor point of view, another way to categorize the types of chips is to use the terms analog, mixed-signal, and digital (see table).
Transceiver chipsets move toward integration
The key building block in optical networks is the optical transceiver module, consisting of a number of ICs and optoelectronic components. In telecommunications, lasers at 1310 or 1550 nm are used as transmitters for the fiberoptic network. For data rates up to 10 Gbit/s, the lasers can be directly modulated by the laser driver chip. For higher data rates than 10 Gbit/s and some long-haul networks, optical modulators after the laser convert the incoming data stream from the driver into the modulated optical signal.
The multiplexer chip provides the parallel-to-serial conversion (serializer) of lower-data-rate signals into a single output data rate. For example, 16 OC-3 (622-Mbit/s) channels (parallel data streams) can be multiplexed onto an OC-192 (10-Gbit/s) optical signal (serial data stream).
For the optical receiver, photodiodes or avalanche photodiodes, a transimpedance amplifier, and an AGC amplifier are combined into a receiver module. The clock and data are recovered from the signal by the clock-data recovery chip. Similarly, the demultiplexer chip handles the serial-to-parallel (deserializer) conversion for the receiver.
Frequently, the laser driver amplifier is integrated with the laser diode to form a transmitter (TX) module. Similarly, the transimpedance amplifier is combined with a photodiode to yield an optical receiver (RX) module.