Silicon Photonics: Widespread Usage on Active Components Doubtful

Avatar photoPublished on
blog 1486467496 5

While the amount of industry chatter about silicon photonics rose substantially in 2014, partially because of certain market research firms looking to sell more reports, the technological hurdles, especially with insertion loss, make extensive use of active, combined components, such as for modulators and photodetectors, unlikely in the foreseeable future. Although there have definitely been numerous demonstrations of integrating with these CMOS fabrication processes, like with receivers, both the performance and cost are superior with standard methods.

Certainly for passive devices, silicon lends itself quite well to integration, including for AWGs. Instead of using micro optics, combining waveguides with attenuators is being accomplished with Si. With 100G receivers, the delay lines are being integrated with the phase combiners with either silicon or Indium Phosphide.

Given the increased excitement level, there may be the mistaken impression that the idea behind Si photonics is fairly new. About 45 years ago, a prominent engineer at Bell Laboratories envisioned employing the same production techniques for optics as was used for integrated circuits. Moreover, Intel, which definitely has a vested interest in moving Si, has been working on silicon photonics for over a dozen years.

Speaking of suppliers, with the exception of a few notable large companies with extensive R&D resources, there is only a relatively small number of other vendors substantially touting silicon photonics. Yet even Huawei, which has seemingly never hesitated to develop and promote every conceivable variant in the optics space, especially when it comes to leading-edge technology, has hardly spoken publicly about the concept – other than to mention an acquisition in the space.

Unless pressed for a response, the vast majority of optical equipment vendors both on the component and system sides of their houses have remained silent, including Infinera, which is still the only full-system firm to fully integrate all of its functionality on a single chip. These realizations alone indicate that the deployment of silicon photonics for active solutions in a big way is hardly around the corner.

The only major qualification is Cisco Systems, which will continue to sell its CPAK transceivers, especially to its “captured” base of customers. However, this exception will not change the overwhelming economic and technical challenges faced by the silicon photonics market in general.

[written by Mark Lutkowitz]



  1. You need to make a clearer distinction between “Long Haul” (>10km) vs “Short Haul” (<1km) applications for Silicon Photonics. Silicon Photonics has only ever really been promising for Short Haul applications which is where Luxtera has delivered 100K's of transceivers. In such short haul applications, the cost benefits of integration are not adversely impacted by any compromises compared to discrete technologies performance.

    Your examples of Infinera and 100G delay line + phase combiners are examples which pertain to Long-Haul.

    If your assessment is that active Silicon Photonics will not have widespread use in the Long-Haul, I'd agree.

    But I would disagree if you were making the same conclusion about short-haul applications.

  2. Hi Roger,

    I have known you a long time and respect your opinion very much. However, assuming that you are correct at less than a kilometer, you are kind of supporting my point in that it is very short haul. We are pretty much excluding metro and even some access applications.

    At least publicly, the proponents of silicon photonics always make the lower cost argument, but never appear to explain the math. Would you be willing to take a crack at that challenge?



Comments are closed.