Evolutionary path to monolithic transceivers
The earliest lightwave transceivers, in the 1970s, were assembled from discrete devices and parts.* Over the past quarter-century, as active and passive IC technologies, production quantities, and applications steadily advanced, fiber-optic transceivers have evolved to hybrid configurations dominated by ICs. Compound semiconductor technology should dominate the emitter space for the foreseeable future. However, the recent announcement by Intel of its undertaking the development (and moving to production) of monolithic silicon transceivers heralds such devices as a third generation of photonic transceivers.
For this analysis, ElectroniCast defines a monolithic transceiver as a single monolithic (with the exception of emplaced emitter/detector) chip transmitter and matching receiver, packaged either together or separately, with <10% of parts/devices value as other ICs. Hybrid transceivers are defined as having over 10%-90% of photonic plus electronic circuits as ICs. Discrete-circuit-based transceivers (not addressed here) may contain up to 10% of parts/devices as ICs but not enough to meet the hybrid-transceiver definition. Over the past decade, there has been a strong trend to a higher transceiver value share held by hybrids.
The global consumption of monolithic photonic-integrated-circuit (PIC)-based plus hybrid transceivers is forecast by ElectroniCast to expand from $456 million last year to $1.54 billion in 2009 and grow further to $5.39 billion in 2014. These figures exclude discrete-circuit transceivers. The early consumption and production of these transceivers will be led by North America, with a 37% share of global consumption value in ’09, expanding to 45% by 2014, while the European share expands to 37% by ’09.
The main drivers of the evolution from discrete to hybrid to monolithic transceivers are reduction of cost and size, plus easier standardization. The trend to greater integration will reduce the assembly/test cost advantage of offshore facilities, compared to discrete-circuit-based transceivers, but offshore IC production will remain strong.
Short-reach applications (mostly <1 km) will dominate the PIC-based transceiver market. Fiber-optic short-reach interconnect within equipment (computers, servers, switches) and vehicles (automobiles, aircraft) is expanding rapidly in terms of number of links and in data rate per link. While copper interconnect vendors have made impressive advances in data rate and reach over the past decade, most major equipment vendors now consider fiber to be the preferred, more conservative solution. This viewpoint will be strengthened by the continuing rapid drop in cost per gigabit-kilometer of fiber links.
Telecommunications equipment and networks will be the leading consumer of PIC-based transceivers, with a 28% share ($426 million) in ’09 (see Table 1). Military/aerospace applications, however, although consuming smaller quantities, have substantially higher unit prices per specific photonic performance due to designing for severe environments, smaller production lots, and other factors. The military/aerospace global consumption of fiber-optic-signal links will expand at more than 30% per year to a 23% consumption share ($1.24 billion) by 2014. Most of this consumption will be captive production within major defense-system contractors.
ElectroniCast expects silicon-based 100-Mbit/sec and 1-Gbit/sec transceivers to be introduced with small shipments in 2007-08. Global consumption of monolithic 100-Mbit/sec transceivers in 2014 is forecast to be 8.4 billion units at an average unit price of $7.43. By then, the global consumption value of monolithic transceivers will accelerate past hybrid transceivers to gain a 58% consumption share ($3.12 billion) as noted in Table 2.
While current technology would enable processing 10-Gbit/sec and even 40-Gbit/sec transceivers on a single monolithic substrate, the cost for near-term market quantities would be an order of magnitude greater than that of the current approach of dividing the unit into several functional chips. Within the next decade, however, as quantity requirements expand, these higher-data-rate monolithic transceivers will achieve commercial shipments. The evolution from discrete to hybrid to monolithic circuits has always been driven mainly by cost, in photonics as in electronics, with compactness and other considerations secondary.
*J.D. Montgomery and H.F. Wolf, “Fiber Optic and Laser Communication Forecast,” Gnostic Concepts, 1976.
Jeff D. Montgomeryis chairman and founder of ElectroniCast (San Mateo, CA, www.electronicast.com), a senior life member of IEEE, and a member of SPIE and OSA.