The optical component industry is changing from one in which existing InP and discrete designs are replicated in SiPho to a new approach under which the unique benefits of SiPho are better exploited. Several announcements made recently at OFC 2016 underscore that companies are beginning to migrate to these key applications for silicon photonics.
Recall Cisco’s CPAK – essentially a clone of the CFP2 in performance. Despite being in SiPho it offered little differentiation. Other vendor efforts by Kaiam and Luxtera centered around 100G PSM-4, exploiting the benefits of parallel architectures in silicon but pushing the solution cost from the optics to the fiber interconnect. These approaches offered no compelling opportunity for SiPho as a disruptive technology.
This situation is now changing. There are now examples of products that simply couldn’t be built in other technologies, though many of these have yet to reach production (i.e. large volume). Several examples were present at OFC 2016:
InPhi ColorZ
InPhi ColorZ was the blockbuster announcement of the show and it made waves as the industry and investors imagined how the market is re-aligning in it’s wake. The merits and applicability of this solution for the broad market can be debated (and will be reviewed in a later article), but the value of SiPho in making this work is real.
InPhi announced a partnership with Microsoft to build a low cost / low power 80km or “ZR” solution. This 2x50G PAM-4 approach uses a SiPho TOSA/ROSA with a PAM-4 DSP bundled into a QSFP28. It is non-tunable, and breaks the 3.5W limit of QSFP, but it is a remarkable piece of integration. It will be shipping in volume into Microsoft networks by 4Q16.
This product has a very material impact on demand for standalone DCI boxes that are optimized for shorter reaches such as the Infinera CX. The technology is very difficult to deploy, though, which will prevent ubiquity.
With the announcement, Inphi is now just one step from becoming a full-blown optical component vendor. While the company is working with an outside assembler to make and test the modules, the core components are all built by Inphi in collaboration with an undisclosed fab partner for the PIC. There is concern at Inphi that the company may be perceived as competing with its own customers, and this was one of many reasons that precluded Inphi from selling modules. That said, Inphi’s decision is surprising, given the success that Acacia has experienced in designing key components and compelling users to purchase them as a subsystem with higher prices though lower margins.
Acacia CFP2-ACO
Speaking of Acacia, the company announced a CFP2-ACO based on SiPho which leverages existing designs currently used in its products. Production is scheduled for 2H16, and it is a class-II module that most of the market (except Ciena) is eager to welcome.
On the surface, this new introduction appears to be yet another effort to clone existing solutions. However, there is a competitive advantage in using SiPho for shorter metro reaches when it is paired internally with Acacia’s semiconductor optical amplifier (SOA). Power consumption is the key figure here – the company claims less than 12W – but will not disclose exactly how much less.
Assuming this device has a significant power advantage over InP-based solutions for certain applications, it will enable the next Holy Grail of Coherent – a CFP2 DCO. This is a coherent module with the DSP inside, permitting pluggable coherent in traditional switch and router applications. This is the product the Chinese market desperately wants, as do router vendors such as Juniper and Arista and Webcos like Microsoft. With 80km performance and higher density like CFP4, this could be a great product and would complement the Inphi solution.
MACOM L-PIC
Similar to Inphi, MACOM is one step away from becoming a module vendor after announcing a CWDM-4 silicon PIC. The company has developed the means to combine its facet laser technology with silicon photonics to potentially create a higher yield CWDM4 transmitter solution. To this the company then matches its own single chip driver electronics and retimer to create nearly a complete solution (absent the ROSA).
Summary
In addition to these noteworthy announcements at OFC, there was great interest in the recent progress made by Aurrion as it pursues the growth of III/V materials directly onto silicon. Kaiam also announced a new CWDM4 design based on SiPho with a MEMS-based alignment approach. Remember, it isn’t likely that all of these solutions will live to see production, but it is clear that silicon photonics is evolving from being used to build knockoffs to a higher purpose as it better leverages the benefits of its technology against those of InP.