Silicon photonics can deliver faster internet

Silicon-based photonic receivers simultaneously increase system density and reduce component costs for high-speed fiber-optic communications.
08 February 2013
Carl Paquet

Given the proliferation of mobile devices, operators of optical communication networks have no choice but to increase speed to support consumers' ever-increasing appetites for bandwidth-hungry applications. Video on-demand, online gaming, cloud computing, and all our favorite apps are now well anchored in our daily routines. Of course, network operators are aware that end users do not want higher monthly internet bills. Therefore, system vendors are confronted with considerable challenges: they must provide faster systems that deliver more bandwidth at low cost.

Optical networks are evolving from 10 gigabits per second (Gb/s) to 100Gb/s, with some systems making an intermediary step at 40Gb/s. On the optical transport side, the evolution to 100Gb/s is leading to coherent detection to cope with various optical fiber impairments and to be cost competitive with respect to 10G (10Gb/s) networks. As seen in the past for 10G systems, 100G coherent systems must evolve to meet the constant price pressure faced by network operators. The only two possible ways to address this situation are through reductions in operating expense or capital expense. The former implies increasing the system density, i.e., having a greater number of 100G cards in a given shelf, whereas the latter is directly addressed by improving system architecture and by reducing component costs. Our recent product development based on silicon photonics manages to meet both requirements at the same time.


Figure 1. Schematic of the silicon-based coherent receiver. PBS: Polarization beam splitter. LO: Local oscillator. TIA: Trans-impedance amplifier. MMI: Multimode interference.

When the first generation of 100G coherent systems were first introduced nearly five years ago, they disrupted the market thanks to a significant advance in digital signal processing. But these systems were expensive and big. To reduce size and cost, optical component manufacturers have recently accelerated their research and development investment in photonic integration, which entails integrating many optical functions on the same technology platform. The new technology platforms combine in one box what used to require five different boxes. Planar lightwave circuit technology, which uses silica waveguides on a silicon substrate, was the first step toward photonic integration. Today, technology platforms such as indium phosphide1 and silicon photonics2 make possible the development of new optical components that were unimaginable just five years ago. Our integrated coherent receiver based on silicon photonics is among such components.


Figure 2. Picture of the receiver assembly. A simpler configuration could further reduce the package's dimensions. PD: Photodiode. SOI: Silicon-on-insulator.

Figure 3. The silicon-based coherent receiver (bottom) is significantly smaller than a standard product (top).

Silicon photonics evolved from well-known and mature CMOS process used to build electronic components. The high refractive index contrast of silicon waveguides—which can support a bend radius as small as 3μm—written on 8-inch wafers allows for a high component density, which substantially reduces costs. While silicon photonics promises the integration of photonics and electronics,3 our work is mainly dedicated to the integration of optical functions.4

Integrated coherent receivers used for 100G coherent detection systems combine passive optical functions with high-speed photodiodes and linear trans-impedance amplifiers (TIAs). The optical signal carrying the information is mixed with a local oscillator (LO) in two different polarization stages and provides a total of eight separated electrical RF (radio frequency) outputs delayed in-phase. These output signals are sent to an external ASIC (application-specific integrated circuit), which recovers the original transmitted data stream through highly complex digital signal processing.

On the optical side, a polarization beam splitter separates the signal and a splitter separates the LO into two 90-degree hybrid mixers (see Figure 1). In our hybrid integration approach, these passive optical functions are written on a silicon photonic chip. This die is made on a wafer fabricated by the ePIXfab company using 500×220nm strip silicon waveguides. A 2D surface grating coupler acts as a polarization beam splitter for the incoming signal. The light from the LO port is coupled through a 1D surface grating coupler and is split using a 1×2 multimode interference (MMI) coupler. Both signals and LO are then mixed in two 2×4 MMI couplers. The outputs of each waveguide are connected to a 1D grating coupler, which allows the light to emerge out of the chip.

To achieve optical detection, the flip chip technique is used to package two 1×4 high-speed photodiode arrays on top of the output grating couplers. This electro-optic subassembly is mounted on a ceramic, adjacent to two dual TIAs. The heart of the coherent receiver measures only 6×8mm, which is significantly smaller than other technologies (see Figure 2). Once packaged, the silicon-based integrated receiver occupies one-third the footprint of the package defined by the optical internetworking forum (see Figure 3). A simpler RF configuration could reduce dimensions even further. This photonic integration technology allows for future coherent receiver design that could be integrated in much smaller 100G transceivers.

Silicon photonics product development for high-speed optical communications is still in its early stages with products for both line-side and short-reach applications. Optical functions can now be integrated using CMOS compatible processes, and packaging these new optical circuits will become critical in making these products cost-effective and reliable. Our research and development team is now starting to investigate how silicon photonics technology can be used to lower dimensions and cost on 100G and 400G modulators. Thanks to silicon photonics, system vendors and network operators now have promising new options to reduce operating and capital expenses. In the end, this new technology will benefit all avid internet users.


Carl Paquet
TeraXion
Quebec City, Canada

References:
1. R. Nagarajan, C. Joyner, R. Schneider Jr., J. Bostak, T. Butrie, A. Dentai, V. Dominic, Large-scale photonic integrated circuits, IEEE J. Sel. Top. Quantum Electron. 11(1), p. 50-65, 2005. doi:10.1109/JSTQE.2004.841721
2. B. Jalali, S. Fathpour, Silicon photonics, J. Lightwave Technol. 24(12), p. 4600-4615, 2006. doi:10.1109/JLT.2006.885782
3. C. Gunn, G. Masini, Closing in on photonics large-scale integration, Photonics Spectra 41(12), p. 74-79, 2007.
4. Y. Painchaud, M. Poulin, J.-F. Gagné, C. Paquet, Ultra-compact Si-photonic DQPSK demodulator, Proc. Optical Fiber Commun. Conf, p. OM3J.3, 2012.
PREMIUM CONTENT
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research