As complexities in OEICs increase, prices fall and the market rises
Integrating optical and electrical functions on one device presents many challenges and potential payoffs. The market for OEICs is expected to grow from $360 million in 2000 to $680 million in 2001, and to $2.6 billion by 2005.
Jeff D. Montgomery and Stephen Montgomery
Development of optoelectronic integrated circuits (OEICs) has proceeded worldwide during the past 18 years and high-volume production has been achieved in lower performance, light-emitting diode-based OEICs. Until recently, however, the production quantity potential of high-performance OEICs was not big enough to generate a profitable return on the large investment costs of phasing OEICs into manufacture. The surge of lightwave communication toward high-volume interconnection to homes, desktops, and small businesses now provides the necessary payoff potential. Major, worldwide component producers, such as Alcatel, Agere Systems, Nortel Networks, NEC, Agilent Technologies, and Tyco are tooling up for production of OEICs. They are joined by numerous other established companies and startups.
ElectroniCast conducted a two-level analysis of global consumption of OEICs, and of the devices and parts used in their fabrication (see Fig.1). This study focused on high-performance OEICs, classified as being designed for operation at 155 Mbit/s (OC-3) and higher data rates, plus those combining optical emitters or detectors with electronic devices and/or optical waveguides on a single substrate.
These OEICs may be fabricated as hybrid integrated circuits (ICs) by mounting die plus other devices and parts on a supporting substrate, or by processing all elements monolithically in a semiconductor substrate. Hybrid IC substrates typically incorporate optical waveguides.
The trend toward increased integration of optical, photonic, and electronic devices is driven almost entirely by the expectation of achieving lower-cost component production, which brings a competitive market advantage. For a prospective vendor (or captive producer), however, the unit cost of the OEIC must include the amortized engineering, tooling, capital equipment, market introduction, and other costs of bringing the OEIC to commercial production. These front-end costs are quite substantial, ranging from several hundred thousands to tens of millions of dollars.
Therefore, the decision of whether, and when, to undertake OEIC development requires an analysis and forecast of several factors. First, the total available (merchant) market (or captive demand) for the particular function (quantity, average price, value) must be calculated year-by-year over the next several years. The fact that average prices will drop rapidly, inversely proportional to the quantity increase, must be considered. The potential market share to be captured must also be forecasted, knowing that competitors may develop similar products. Finally, synergism with other products must be realized.
PERFORMANCE NOT A DRIVER
Through the mid to late 1980s, major investment took place, especially in Japan, in the development of high-data-rate OEIC transmitters and receivers aimed at improved performance. At that time, the focus was on 1.0 to 2.5 Gbit/s components. The primary premise was that the shorter internal interconnect distances achieved through integration would permit much better performance, compared to conventional assembly of discrete devices. However, researchers following the discrete device path were able to continue their advances through 2.5 Gbit/s (and, more recently, 10 Gbit/s transitioned into commercial production, and 40 Gbit/s now into play). Because high transmission rates have been achieved through conventional discrete designs, the argument for higher performance through OEICs has less support now.
In some applications, unusual circumstances will support much higher prices for OEICs versus discrete device-based components. In military and aerospace applications, for example, (especially in retrofit/upgrade programs) the goal may be to achieve a 10- to 50-fold increase in performance in the same drastically limited volume (and weight) previously occupied by discrete device-based components. Ability to withstand unusually high shock, vibration, radiation, and other environmental extremes are additional factors favoring OEICs. In these situations, a 10-times higher price, or even 100-times, can be justified. For this reason, military-aerospace will be a leading customer for new OEICs.
Transceivers of 155 and 622 Mbit/s are leading candidates for integration, starting with the photodiode-plus-transimpedence amplifier, expanding to encompass additional electronic die, on both the receiver and the transmit side. By 2010, the global consumption of these transceivers will reach about 45 million per year (fiber to the home, plus fiber to business, plus other), and most of these will be integrated to varying degrees.
TOWARD GREATER COMPLEXITY
Most consumption of OEICs consists of relatively simple circuits. As volume increases and high yield of simpler circuits is achieved, consumption will trend to increasingly complex OEICs (see Fig. 2). This movement will moderate the downward plunge of average prices of overall OEIC functional categories.
Optoelectronic integrated circuits are usually defined as combining optical and electronic functions on a single substrate (either hybrid, monolithic, or a combination of both). Included within some hybrid OEICs, however, is the optical integrated circuit (OIC), which combines multiple passive optical components and/or active (such as laser diodes and/or photodiodes) optical components on a substrate, or in a fused, discrete component integrated assembly. A pertinent example of a high-potential OIC is the chip incorporating wavelength stabilizers plus electro/optic (E/O) modulators, used with the multiwavelength laser-diode emitter array plus combiner, aimed at dense wavelength-division multiplexing (DWDM) applications (see Fig. 3)
A strong trend, especially over the past eight years, has been toward greater integration of OICs formed by the fused fiber coupling of a number of passive and active optical elements. Leading examples are the passive optical gain block and active optical gain block of optical fiber amplifiers (see Fig. 4).
As with OEICs, several advantages are to be gained by the integration of optical components, assuming adequate production quantities. The argument for increased reliability is that all assembly and interconnect functions are accomplished under a single quality control procedure, in a relatively mature process. This reduces assembly and packaging costs. The resulting package is more compact, which is crucial to holding down implementation costs. Such OICs bring a competitive advantage to vendors with a wide selection of component parts. They also permit trade-offs of performance specifications between OEIC elements.
TELECOM OEICs WILL DOMINATE
Early North American consumption of OEICs will be led by telecommunications equipment and telecom customer connection component consumption. Over the next decade, as the emphasis in fiberoptic networks shifts to the metropolitan/access networks and fiber is connected to tens of millions of homes, the use of high-production quantity, moderate performance, price-sensitive transceivers, and other components will rise dramatically. Private data networks will increase their typical data rates and the use of single-mode fiber, while the telecom industry will shift to shorter distance, moderate data rates, and cost-sensitive links. These trends mean the two applications will become increasingly similar, driving increases in OEIC demand from both sides. The telecommunications equipment consumption of OEICs represented approximately 90%—or $270 million—of the global total in 2000.
Market totals for OEICs are growing rapidly, from $360 million in 2000 to a projected $680 million in 2001. Astronomical growth will continue, with the market expected to reach $2.6 billion by 2005.
The market for transceivers and other optoelectronic components is price-elastic. Falling prices will be more than offset by growth in quantities shipped. Component producers, therefore, are focusing on ways to drastically reduce production costs. Analysis by several developers has shown that the biggest payoff in cost reduction will come from automation of module assembly.
The factory cost of OEICs is projected to include 25% to 75% of the cost outlay in the alignment, positioning, anchoring, testing, and package closure required. Generally, the higher the performance requirements, the higher the packaging costs. Most of the packaging costs, in turn, consist of the cost of human labor. Automation can greatly reduce labor cost.
Automation, or semiautomation, however, does not come without its own cost. The establishment of an automated facility that can produce several hundred thousand hybrid OEICs per year is estimated to be in the $5 million to $15 million range, depending on the complexity and precision of the components to be produced. Photolithographic wafer processing, precision pick-and-place equipment, laser welding, reflow soldering, automated testing, and a fully equipped development laboratory represent major investments that must be amortized in product prices within a few years.
Beyond cost reduction, assembly automation provides other advantages. Quality control can become more effective with the elimination of human inconsistency. Wide swings in output volume can be more easily accommodated. On the negative side, however, some relaxation of device dimensional tolerances—with related modest reduction in component performance—is typically required.
Since the cost of an automated assembly facility is not much different for the highest data rate, most expensive OEICs (such as 2.5 and 10 Gbit/s) than for moderate data rate (155 and 622 Mbit/s), automation will be extended to these highest priced units.
MINIATURIZATION IN DEMAND
The expected 20- to 30-fold expansion of global network throughput capacity over the next decade will require central office, access node, and subscriber equipment with greatly increased capabilities. Yet installation space is already crowded and adding space is very expensive. As a result there is strong pressure from the end customers to get much greater throughput without increasing the size (hopefully, decreasing the size) of switches, transport terminals, and other equipment. Part of this adaptation will be achieved by using fewer but faster components. There is also considerable pressure to reduce component size through integration and miniaturized packages.
The competitive environment is different for each OEIC and device or part category, and also is different in the various world regions. Many companies specialize in selected products in this field.
Transmitters, receivers, and transceivers start with diodes. For the short-to-medium distance LAN and access networks that will dominate OEIC usage over the next decade, these must be low-cost devices. They are then combined with electronic circuits (ICs) and passive parts to become a complete transmitter, receiver, or transceiver.
For premises (private network) equipment, where performance requirements are often less stringent (and cost pressures are even more severe), equipment producers are more amenable to buying complete transmitters and receivers. This market encompasses optoelectronics for transmission up through OC-24 (1.244 Gbit/s), for moderate distances, evolving to OC-48 near term and OC-192 by 2003. Datacom transceivers, however, will remain mainly discrete circuitry (individually packaged IC die on printed wiring board) over the next decade.
Along with the newness of OEICs, important efforts are under way to carve out a new business/product segment. Historically, communication equipment companies bought or produced performance-specified diodes and connected them to electronics on other equipment boards to form the complete transmit or receive function.
FOR THE (FAR) FUTURE
It is technically feasible to integrate laser diodes and/or photodiodes, electronic integrated circuits, and optical elements such as WDM filters, isolators, and external modulators on a III-V substrate. This technology has been used operationally in multimillion-dollar military/aerospace photonic subsystems. Looking ahead to future high-volume/low-cost requirements, especially fiber-to-the-home, several companies are investing heavily in this type of OEIC development.
Jeff D. Montgomery is the founder and chairman and Stephen Montgomery is the president of ElectroniCast Corp., San Mateo, CA. They can be reached at 650/343-1398 or e-mail: email@example.com.