by Frédéric Verluise
Today most people consider optics in telecommunications the way they used to consider electronics: You have cables that you can splice, that you can connect to boards on which optical chips based on planar lightwave circuits (PLCs) perform various functions. It all seems to be just another type of electronic component, with its drawbacks (direct tapering is not easy, wiring is a bit complex, etc.) and advantages (bandwidth of the “cables,” weight, etc.). It’s easy to forget that inside the cables, the photons do not behave like electrons. In the optical world, there is still the matter of aligning prisms and lenses.Even if such activity is hidden, 90% of optics still consists of this free-space propagation and and manipulation of light, the other 10% being merely a transport matter (but what a wonderful medium). Even under the metallic cover of transponders the wavelength filtering is done in free space before the photodiodes.
Some functions can be integrated-for instance, modulating, amplifying, splitting, and in some cases wavelength slicing with arrayed waveguide gratings (AWGs). But even in these cases the coupling between media is done through free space; one still needs a lens to couple a laser into a fibre.
In addition, advanced functions cannot be achieved without the help of free-space optics, at least today, because they do not represent the huge market depicted a few years ago. Complex free-space optical architectures are implemented via what is now commonly called bulk grating technology (BGT). Two major markets require the fast-paced technological evolution BGT provides: FTTH and the deployment of new modulation and demodulation formats (DPSK, DQPSK, QAM, etc.).
FTTH represents today’s major optical telecommunications market opportunity. In Japan, the U.S., and now in Europe, more and more fibre reaches the user. Today the hardware technology is very basic. Simple Fabry-Perot (FP) lasers are exploited for slow bit rates, splitters are used, and bandwidth is shared via TDM. For people used to optical transmission, it is quite clear that this first step is under-powered, and from a market point of view the technology used is a bottleneck to future massive expansion. However, when the fibre is installed the hard part is done; bandwidth growth is a matter of equipment upgrade.
The logical evolution of those networks is to grow from TDM to WDM or even DWDM. This step is hard to accomplish for two major reasons. First, the operators don’t want to manage every user wavelength. Second, among all possible DWDM combinations no standard has really emerged. Many schemes have been tried for the first generation, based on spectrum slicing or on reflective semiconductor optical amplifiers (RSOAs), for instance. All of them exploit some feature of the DWDM multiplexer.This second issue reflects the multiplicity of possible future optical architectures. Will the classic 100 GHz spaced ITU grid be the choice? In this case, what would be the bit rate? Could it reach up to 40 Gbit/s? Will it be deployed in combination with TDM? Would it still use on/off keying (OOK, such as RZ or NRZ modulation), or will phase modulation be necessary? Or will a new wavelength spacing emerge-could it go down to 5 GHz?
The answers to these questions essentially depend on the laser technology that will make the difference in terms of costs and performance, as well as the capability of a DWDM technology to cope with the constraints imposed by the laser technology chosen.
In a way, this may look like the chicken and egg paradox. But an option exists to solve it: BGT. First of all, the main feature required for FTTH is athermal components. Since most of these networks are intended to be passive (as in a PON), no one wants to need power to thermally regulate DWDM multiplexers/demultiplexers. Traditionally, system vendors who needed athermal behavior preferred to use thin-film filters or, for high channel counts (more than eight), BGT-based multiplexers/demultiplexers. Some vendors can now provide athermal AWGs, but these are limited to the classic 100-GHz ITU grid spacing and took several years to develop.
Years are not the type of lead time appropriate for fast architecture evolution. This is why system designers exploit the flexibility of BGT more and more. Knowing that a device can be athermal below 50-GHz spacing is basically a thorn out of the pillow. Then, having a noncyclic multiplexer/demultiplexer is very important if you want to test spectral slicing to address the FP lasers.
These features only hint at the possibilities of BGT. Some years ago it seemed unlikely that advanced devices using this technology would reach commercialisation. However, BGT processes have become more reliable and the technology has demonstrated interesting behaviors that make it unique. For instance, BGT multiplexer/demultiplexer devices have demonstrated double functionality-coupling at 1,300 nm and multiplexing at 1,550 nm-and sliding cyclicity (see figure).
But FTTH DWDM is not only a matter of network engineering or performance. It is also a matter of price and volumes. Nowadays, prealignment and positioning of most of the BGT optics are automated (see photos). In the future, BGT components will be completely processed in the same manner as electronic boards, with precise pick-and-place machines, ending with an active, fast automated process to reach required performances. Even these final steps will be achieved passively. Today materials used are common glasses; tomorrow plastic elements will replace these glasses, taking a step toward moulding. It becomes easy to imagine what price could be reached by the extensive use of BGT.
The analogy between optics and electronics would lead to the equivalence of PLCs and the integrated chip-and BGT-based components and electronics boards where the optical components such as mirrors, prism, and lens would be the resistance, capacitance, and inductances of the board. In such a scenario, contrary to electronics, the “optical chip” is roughly the same size as its equivalent “optical board.”
This analogy leads to a very important point: the possibility of BGT to integrate with other technologies. For instance, one could envision MEMS or liquid-crystal cells used to build switching functions, laser chips implemented to build tunable lasers, etc.Its ease of implementation makes BGT the best technology for advanced products. One of the most active scientific fields in terms of optical transport is high-bit-rate modulation for 40 Gbit/s and beyond. For such bit rates a classical OOK approach won’t be good enough; phase modulation, so common in radio communication, begins to make sense in the domain of optical modulation. For example, differential phase-shift keying (DPSK) has experienced predeployment in the labs of major equipment vendors.
Thus, phase modulators are quite common. However, today there is only one way to demodulate the signal (i.e., to convert it from a phase modulation to an amplitude modulation) in a commercialised manner. This way is provided thanks to BGT.
An interferometer seems easy to make but requirements are quite drastic. Most system designers want an athermal device exhibiting no more than a few gigahertz drift over the temperature range. In terms of distance it means an optical piece can move less than a few nanometers over the temperature range! Few optical components vendors can achieve this kind of precision, and they all use BGT. And when tunability becomes necessary, to follow laser drifts, for instance, the integration of another technology (piezoelectrics, liquid crystal, etc.) is child’s play.
Phase modulation now seems close to being fielded, and all of the equipment vendors who aim for 40 Gbits/sec are investigating a DPSK scheme. But in the labs scientists are preparing the next step: a full amplitude-phase modulation with not only two states of each but maybe up to four. Today, creating such a modulation format is feasible thanks to lithium niobate Mach-Zehnder modulators. But demodulation is quite tricky and requires a coherent receiver, also called a hybrid 90° demodulator. PLC approaches for such demodulators were proposed with no less than six voltage control points. BGT-based products enable such a demodulator now-in a passive way, of course.
Thanks to its ease of manipulation, BGT enables the construction of custom prototypes and high-performance optical devices in a cost-effective and fast way. When a device is ready to step beyond prototyping, light processes can be set up to build a pilot line. When the time comes for massive commercialisation, manufacturing can be fully automated. Just like electronic boards, BGT can evolve from the craft-like PCB used in labs to the complex optical motherboard dedicated to mass markets.Frédéric Verluise is chief executive officer and founder of KYLIA (www.kylia.com).