From electronic to photonic integrated circuits

Without dispute, one of the greatest technological innovations of the twentieth century was the development of the integrated circuit (IC). From its roots at Fairchild Semiconductor and Texas Instruments where pioneers Robert Noyce, Gordon Moore and Jack Kilby first integrated transistors into silicon and germanium, the technology showed immediate promise over the discrete, single-function electronic devices that were available at the time, and today makes possible devices which could not have been imagined in the days of single-transistor devices.

The impact of ICs stems from the ability to monolithically integrate ever more transistors and other electronic components into a single device. This spawned the now famous "Moore's Law" (see figure 1) that has since predictably delivered exponentially greater processing power, lower cost per device, improved reliability, reduced space and power requirements, and enabled countless new devices for a wide range of applications.

Figure 1: Moore's Law predicts a doubling of transistor density every 18 months. This has been practically validated and has led to significant decreases in costs per transistor.

One can ask the question "what role do electronic ICs play in optical networks, since at their core such networks manage photons, and not electrons?" The answer lies in the value and functionality that is provided by electronic ICs in accessing and managing the services and applications delivered over a reliable transport network.

This includes monitoring data transmission performance, tracking service level agreements, switching different data streams into larger transmission facilities, enabling rapid in-service network reconfiguration and flexible service add/drop, and providing robust fault detection and protection to prevent service outages.

Thus while purely optical technologies such as Wavelength Division Multiplexing (WDM) and optical amplifiers enable capacity scalability and extended optical transmission between nodes, nearly all other value-added service functionality is implemented using a combination of electronic ICs and system software (see Figure 2).

For example, electronic ICs used within SONET/SDH or data switches provide feature-rich grooming and reconfiguration of digital data streams carried by optical networks. Depending on the functionality implemented in the chosen electronic IC, it can either switch packets, cells, or TDM bit streams at a cost of a few tens to hundreds of dollars.

Figure 2: Electronic ICs cost-effectively enable a wide range of data processing and performance improvement capabilities.

Electronic ICs are also used to monitor network performance, for example by measuring performance monitoring (PM) bytes within either SONET/SDH or digital wrapper overhead. Combined with functions implemented in system software, these capabilities are used to provide robust and rapid network protection and service restoration.

Figure 3: Digital signal processing such as FEC and EDC implemented in electronic ICs can improve optical system performance at much lower cost than all-optical technologies.

Electronic signal processing is also increasingly being used to improve optical transmission performance. For example, high-gain Forward Error Correction (FEC), Electronic Dispersion Compensation (EDC), and optical modulation techniques are used to recover degraded bits, mitigate the degradation of optical signals due to chromatic or polarization mode dispersion, and increase system reach.

Electronics provides these benefits at a cost of tens to hundreds of dollars per IC; essentially the cost of processed silicon real-estate (see figure 3). In contrast, "all optical" technologies are more expensive, more complex, or deliver less functionality. And therein lies the fundamental advantage of electronics over optics; the extremely low cost by which feature-rich and value-added processing can be implemented.

So if electronic ICs are so cost-effective, why has there been so much effort put into the development of "all-optical" networks that seek to minimize Optical-to-Electrical-to-Optical (OEO) conversions? In such an "all-optical" network, electronic processing is relegated purely to the edges of the network, while service manipulation within the core is done in the photonic domain. Since this limits the ability to implement value-added electronic processing, one can logically question the desirability of such an approach.

The problem has not been the cost of electronic ICs, but rather the conversion cost of transferring data from the optical domain into the electronic domain (see figure 4). Thus one can think of OEO conversion as imposing a "tax" on the use of electronic ICs. With today's OEO cost structure, the tax is so high as to create a disincentive to the wide-spread use of electronics in optical networks, despite the cost-effective functionality and performance benefits that this would bring.

Figure 4: The cost of optical components required to implement an OEO conversion are significant compared to the cost of electronic ICs used to manipulation the data in the electronic domain.

This reason OEO conversions are expensive has been the need for many discrete, single-function optical components required for each OEO conversion, including lasers, modulators, wavelength lockers, detectors, attenuators, WDM multiplexers and de-multiplexers. In a typical optical transport system, each OEO conversion may require up to half a dozen optoelectronic or optical components, and a fully deployed 40-wavelength WDM terminal node may therefore use upwards of 120 or more components interconnected by 260 or more fiber couplings (see figure 5)!

As the number of OEO conversions in the network scales with capacity and across many network elements, the total cost incurred from this multitude of components increases to the point of making frequent OEO conversion architecturally unsustainable and economically unviable. In addition, since each component is separately fabricated, packaged, burned-in and tested, the opportunity for manufacturing efficiencies is significantly less than has been possible for electronic ICs.

Figure 5: A typical conventional DWDM terminal requires many discrete, single function optical components—which therefore incurs high cost for ubiquitous OEO conversion.

It was into this existing industry paradigm that Infinera sought to address the fundamental root cause of the problem, expensive OEO conversion, and in response developed the industry's first large-scale monolithic Photonic Integrated Circuit.

Back to Top >>

© 2008 Infinera Corporation.