Who would be against the idea of applying Moore’s Law at some time in the future to optical chips in that it would revolutionize the entire cost structure of transporting signals over light? Nevertheless, only a short time ago, there was hardly a peep of public criticism when it came to the obvious shortcomings of silicon photonics for active components with the exception of Finisar being unjustly punished for telling the truth, as well as fibeReality pointing out what was being said behind closed doors at the vast majority of manufacturers. Today, excluding certain individuals in academia, who are unwilling to admit to any negative attributes of Si photonics, it has presently become almost fashionable to openly state the undeniable conclusion expressed by one vendor’s recent response to us: “The only consensus on the technology is that there is no consensus.”
In point of fact, Kaiam will actually say that the effects of Moore’s Law cannot be applied to photonics. While putting everything in silicon is good for electronics solutions, such as for transistors and modulators, the wavelength and light are too big, and so the purpose of optical integration is more about improved packaging (along with the associated cost advantages), according to the supplier.
While even the harshest critics of Si photonics will bend over backwards to say something positive, the virtually overnight transition to an almost skewering of the solution in a market that excels at hyping vaporware has been quite remarkable. It is not uncommon to hear engineering experts not only agree with Kaiam that the problem with going totally silicon is really the packaging, but also actually characterize it as a nightmare. They will also point out that Luxtera, as much as it seems a purist, admits that the CMOS process, which provides optical waveguides is incompatible with the CMOS techniques that allows for very high s and p transistors. (Therefore, Luxtera uses a hybrid flip chip up the driver chip onto its modulator waveguide chip.)
Certainly, Finisar is not holding back in expressing its full views anymore, such as that silicon photonics will not work on the receiver side. There will sure to be a pounding on any proposals that intimate that it is mature and ready for prime time, while it is currently sketchy at best.
At OFC 2015, which can always be counted on to be a hype-fest for the latest solutions, there did not appear to be a lot of vocal support for silicon photonics. While Luxtera did its best (albeit stuck in a backroom on the exhibit floor), in promoting its future prospects, it did not seem to resonate much with the audience at the show. In fact, at the rump session, somebody sarcastically congratulated the vendor for achieving volumes that equaled the amount of VCSELs produced in a relatively short period of time. (In fairness, Luxtera told us that it expects to substantially increase its manufacturing output at a quicker pace going forward.)
[written by Mark Lutkowitz]