Process based cost modeling of emerging optoelectronic interconnects : implications for material platform choice

Abstract

Continuously increasing demand for processing power, storage capacity, and I/O capacity in personal computing, data network, and display interface suggests that optical interconnects may soon supplant copper not only for long distance telecommunication but also for short reach connection needs. In the search for a standard, the current debate in the optoelectronic industry is focused on the technical and economic challenges of the next generation interconnect. Technological advances over the past few years have given new strength to a silicon-technology platform for optoelectronics. The possibility of extending a mature and high-yield Si CMOS manufacturing platform of the electronic industry into the optical domain is an area of intensive interest. Introducing new photonic materials and processes into the mature electronic industry involves a convergence of knowledge between the optoelectronics and semiconductor IC manufacturers. To address some of the technical, market, and organizational uncertainties with the Si platform, this research explores the economic viability and operational hurdles of manufacturing a 1310 nm, 100G Ethernet LAN transceiver. This analysis is carried out using the process-based cost modeling method. Four transceiver designs ranging from the most discrete to a high level of integration are considered on both InP and Si platforms. On the macro-level, this research also explores possible electronic-photonic convergence across industries through a multi-organization, exploratory roadmapping effort. Results have shown 1) integration provides a cost advantage within each material platform.

(cont.) This economic competitiveness is due to cost savings associated with the elimination of discrete components and assembly steps; 2) a total cost comparison across material platforms indicates at low volume (less than 1.1 million annual units), the InP material platform is preferred, while at high volume (greater than 3 million annual units) the Si material platform is preferred. Furthermore, this study maps out the production cost at each technology and volume projection, and then compares this cost with price expectation to determine the viability of the transceiver market in the datacom and computing industry. Results indicate that annual production volumes must be in the tens of millions unit range to provide the minimum economies of scale necessary for designs to meet the trigger price. These results highlight that standards and a set of common language are essential to enable converging technology markets.