Applications Widen for Silicon Photonics Paired with Coherent Transmission

By Tom Williams | Posted on January 15, 2018

After years of future promise, silicon photonics (SiPh) technology is ready for prime time — having made the transition from promise to production. With an increasing demand for more network capacity, cloud, content, and service providers want optical modules that reduce power, size, and cost. SiPh is now being used in a wide range of coherent optical interfaces, from metro and long haul to submarine data transport, to enable high-density form factors and excellent performance.

Silicon photonics can enable reduced development time, higher levels of integration, and fewer manual assembly steps than more traditional optics. The result? A powerful, easier-to-manage product that empowers cloud, content, and service providers to stay ahead of increases in network capacity demand.

Proven technology

Silicon-based photonic integrated circuits (PICs), which integrate all the high-speed optics necessary for both transmit and receive functionality, enable the density required for pluggable coherent modules. These PICs include the optical polarization-controlling functions that may require external components when integrating using Indium Phosphide (InP). By reducing the number of active alignment steps, SiPh-based products improve yield and ramp more efficiently.

While metro applications served as the primary market for SiPh’s initial service provider network implementations, SiPh is used in long-haul — even submarine — applications today. The submarine market has always required high-performance technology, even if it came at higher price points. Combining SiPh with high-performance digital signal processing (DSP) technology enables submarine-network performance equal to or better than the more expensive, discrete component-based approaches (Figure 1).

 

Applications Widen for Silicon Photonics Paired with Coherent Transmission
Figure 1. The coherent SiPh PIC reduces cost and size versus the use of more expensive, discrete components.

That said, InP technology remains important for laser functionality. But separating that laser function from the high-speed optics can produce several benefits. For example, InP generally requires thermo-electric coolers (TECs) to maintain tight temperature stability. SiPh, on the other hand, can operate over a wide temperature range with no impact on performance. The high-speed interface is optimized by putting the optics close to the DSP and moving the laser further from the DSP, where the TEC doesn’t have to work as hard to maintain constant chip temperature.

Silicon photonics provides further benefits

Using SiPh in coherent applications creates products with rich feature sets that offer high density and pluggability.

There are a number of reasons why SiPh has proven well suited for many applications, including coherent optics:

  • Yield: Beyond improving manufacturing costs, high yield reduces development time by limiting the number of variables during prototyping. When working with low-yield optics technologies, it is difficult to determine if performance limitations derive from design defects or process variation. At higher levels of integration, this uncertainty is compounded. When developing complex PICs with many integrated functions, it is imperative to know that the individual building blocks are well understood and repeatable. These attributes enable the designers to focus on optimizing the interfaces between each function.
  • Polarization Control: Coherent transmission increases the data rate via polarization multiplexing – two orthogonal polarizations are transmitted simultaneously at the same wavelength. This approach requires transmit and receive components that can manipulate the polarization state of an optical signal. When working with InP, polarization control is usually done using external components. Not only do these extra components increase material cost, they also add extra alignment steps that integration of these polarization control functions in the SiPh PIC can eliminate.
  • Thermal Operating Range: As mentioned previously, InP components are sensitive to temperature variation and must be mounted on a TEC. Since TECs have a limited control range, they fundamentally limit the operating temperature range of InP components. In addition, TECs consume significant power in cooling mode, where the thermal design is most challenging. By comparison, the optical characteristics of SiPh vary little over temperature. SiPh doesn’t require a TEC and supports a wide operating temperature range.
  • Humidity: Traditional optics degrade in high-moisture environments. For this reason, optics are packaged in vacuum-sealed gold boxes. These hermetic gold boxes contribute significantly to the cost of optical interconnects, particularly when they require high-speed interfaces. In addition, hermetic seals are historically one of the most common sources of failure for optics. Silicon is well known to be insensitive to humidity. Millions of silicon electronic components are shipped every year in non-hermetic plastic packages; moving optics to non-hermetic packaging is an important step for the industry.
  • Wafer Level Testing: In addition to having higher yields than traditional optics materials, SiPh can also be tested at the wafer level. Good die can be identified early in the process, and there is no labor wasted on material that will ultimately fail. Wafer-level test is commonplace in high-volume electronics applications, but new to the world of optics.
  • Wafer Size: Leveraging mature silicon process technologies means that much larger wafers can be made in silicon than traditional optics materials. Three-inch wafers are state of the art for InP fabs. Today’s SiPh runs on lines that accommodate 8-inch wafers or larger. These larger wafers result in an order of magnitude more die per wafer, which lowers cost.
  • Package Level Integration: As the industry continues to move toward higher data rates and lower power, the interface between the DSP and optics is quickly becoming a bottleneck. Every time a high-speed signal needs to transition across an additional interface (IC package or pluggable connector) there is loss and distortion. Compensating for this additional loss adds power dissipation, and distortion limits performance. Using SiPh enables package-level integration that can better optimize these high-speed interfaces and accelerate the realization of higher data rates at lower power.

Leveraging silicon photonics

Understanding SiPh’s benefits, how do we best use them to drive innovation? Today’s optics architecture is optimized for client interfaces in which the laser is directly modulated. This model is easily extrapolated to external modulation when the modulator technology has the same thermal and packaging limitations as the laser. Thermally sensitive components that need a TEC to maintain a constant temperature are unlikely to be integrated with a DSP chip that also dissipates power.

On the other hand, when working with SiPh, designers can optimize the high-speed interface and separate the thermally sensitive laser. For example, the laser can be placed on another part of the line card and connected to the high-speed optics through an optical fiber. This architecture enables greater thermal flexibility, a high-speed signal path with superior signal integrity, and elimination of costly hermetic packages with high-speed interfaces (see Figure 2).

siliconization of optical intercconect
Figure 2. High-speed electro-optical package integration.

SiPh and coherent are two technologies shifting the landscape of optical communications in parallel. By moving to architectures that can optimize the benefits of each, it can be possible to have the same kind of impact on access networks as we have already seen in applications from the metro core through to submarine. Using a toolbox that includes SiPh and coherent DSP technology, designers can leverage complicated modulation formats, high baud rate, and highly integrated parallel optics to optimize designs for a wide range of applications.

Ball Grid Array Packaging Technology

The transition to low-cost packaging and standard interfacing is an important next step to further the benefits of SiPh technology. As the industry moves toward 600-Gbps capacity per wavelength using higher baud rate and higher order modulation formats, traditional packaging technology can limit performance of the interface between the DSP and optics. Ball grid array (BGA) packages address this challenge by eliminating additional connectors and optical package leads, improving bandwidth and signal integrity.

Here and now

SiPh is no longer a technology of the future. Coherent modules based on highly integrated SiPh PIC technology have been deployed in applications ranging from data center interconnects to submarines. In the next phase of maturity, the industry is learning to understand how to best leverage the benefits of SiPh to achieve the pace of innovation necessary for optical networking to meet the worldwide data traffic demands that such applications as cloud computing, 5G, and the Internet of Things will drive.

Tom Williams is Senior Director of marketing at Acacia Communications. Before joining Acacia, Williams spent 14 years at Finisar Corp. (initially with Optium, which Finisar acquired in 2008), where he was director of product line management for coherent and direct detect transport products above 100 Gbps. He has also held positions at Lucent Technologies and Northrop Grumman Corp. He has an MS in electrical engineering from Johns Hopkins University and BS degrees in electrical engineering and physics from Widener University.