+44 (0)24 7671 8970
More publications     •     Advertise with us     •     Contact us
 
Technical Insight

Optical networking components gear up for 40 Gbit/s and beyond (Cover Story)

The next generation of optical networks will place heavy demands on components. Ira Deyhimy of Vitesse Semiconductor examines how InP is meeting the challenge.
Today s optical networking infrastructure handles the bulk of communication traffic in the developed world. Voice, data, wireline, or wireless traffic ultimately traverses the optical network at some point. At the heart of the optical network infrastructure is the "optical subsystem", which converts the aggregated digitized traffic between the optical and electronic domains. Certain common functions appear in the optical subsystem, regardless of transmission protocol or whether the signaling is between continents or within equipment cabinets. These generic functions include the optical-to-electrical conversion, the deserialization or serialization of the digitized data, and protocol-specific framing and de-framing of the data used to encapsulate or strip the data from the overhead associated with the protocol being used. As an example, in the case of Ethernet, this function is called the media access controller (MAC). The demand for greater volume of traffic motivates higher serial transmission data rates, and in turn, gives rise to technology alternatives. OC-48 (2.5 Gbit/s) has traditionally been served by GaAs MESFET and silicon bipolar technologies, and more recently by CMOS. Higher standard transmission rates, OC-192 (10 Gbit/s) and OC-768 (40 Gbit/s) are being addressed by several technology alternatives, such as GaAs, SiGe, and InP. GaAs MESFETs Historically, Vitesse Semiconductor has applied high-speed digital GaAs technology to develop multiplexers and demultiplexers for the physical layer of OC-48 optical networks. The company s MESFET-based process runs on 6 inch GaAs and requires only 16 mask layers. The gate lengths are 0.25 m and the MESFETs exhibit ft values of 80 GHz. However, with advancements in a wide range of process technologies, the OC-192 physical layer components now going into production are not only being implemented in GaAs, but also in SiGe and even in 0.13 m CMOS (see ). The competition amongst various materials and technologies is fierce in the pursuit of 10 Gbit/s applications. The stringent performance requirements for the physical layer components needed in next generation optical networks, operating under the OC-768 standard (40 Gbit/s), will eliminate many of these technologies as potential solutions. InP increases the speed Though they are sufficiently fast for 10 Gbit/s operation, AlGaAs/GaAs HBTs with an ft of 60 GHz and SiGe HBTs with an ft of 90 GHz will not be fast enough to support the bandwidth and error correction requirements for 40 Gbit/s operation. The only current alternative is an InP-based HBT, where a 1.0 m device exhibits an ft of 160 GHz, a value high enough to make these devices suitable for 40 Gbit/s applications. Another advantage of InP HBTs is their lower supply voltage requirement. The emitter-base turn-on voltage is only 0.7 V, compared to 1.4 V for GaAs HBTs. This will allow InP HBTs to operate at lower voltages, and therefore consume less power. The lowering of operating voltages becomes extremely important as overall module operating voltages are pushed down to 3.3 V. In addition, InP HBTs, where both the emitter and collector are implemented in InP, exhibit breakdown voltages in excess of 10 V, far higher than SiGe devices. This makes InP HBTs compatible with the high breakdown voltage requirements of laser drivers and modulators. These considerations indicate that InP-based technology will be the best and perhaps only technology capable of supporting 40 Gbit/s systems. Modules as components Even as standards evolve and system requirements expand, most optical networks share many commonalities. All these systems require an optoelectronic layer that converts signals from optical to electrical or electrical to optical, to move data in and out. Next comes the physical layer as described above, followed by the framing layer that frames or deframes according to the particular protocol being used. After framing the data is processed, then the switching fabric takes the stream of data and switches it to a different channel and passes it out again. Currently, within each of these layers, is a myriad of components. Typically, OEMs incorporate these components into modules that in turn are sold to those manufacturers implementing optical networks. However, within the near future, driven by market needs, cost reduction and performance improvements, the "components" will no longer be the elements within the modules, but the modules themselves. Greater integration In order to maintain high growth rates in the optical networking arena, it is no longer sufficient to simply supply to OEMs devices such as MUX/DEMUXs. Such devices are used by the OEMs to fabricate physical layer modules that go into the line cards of an optical networking module. Large OEMs such as Lucent and Nortel have in the past designed and fabricated custom layouts for the production of physical layer modules. This effort results in minimal added value to the overall optical network. In future, these OEMs may focus more on overall systems integration from the module level, and the development of network engineering software. While ever faster devices will be required (as in the case of MUX and DEMUX), many of the functions need to be reduced to compact assemblies, or in some cases even single chips. In the near future, the components being supplied by manufacturers such as Vitesse will actually be the entire module a single black box that contains all the electronic and optical components together, forming an optical transponder module (see ). InP-based modules Currently, systems are migrating from OC-48 (2.5 Gbit/s) to OC-192 (10 Gbit/s) and in the next few years to OC-768 (40 Gbit/s). As this happens, the fundamental component within the physical layer will no longer be represented by such elements as a MUX or clock and data recovery (CDR), but rather by the transponder module. This implies that the current approach of choosing the best technology or material for a MUX or CDR will need to be replaced by an approach using a single technology capable of supporting a wide range of functions, including lasers, electronics and detectors. Fortunately, many of these optical components, in particular the lasers and detectors, are already being implemented in InP-based materials. Currently, InP is the only material that can also support the electronic portion of a 40 Gbit/s system, therefore making it ideally suited to large levels of on chip integration containing a wide range of different InP-based device types. Jumping the processing hurdles However, the process technology to fully implement a transponder module using InP-based materials does not yet exist. While excellent detectors and lasers are currently being fabricated in this material system, only a relatively small level of integration has been demonstrated with InP-based transistors. In an environment of volume manufacturing in which a high level of transistor integration must be coupled into a single InP chip with optical components, the process development will be a great challenge. But in the long run, Vitesse believes that this is the only way to succeed. Module suppliers that are only assemblers of sub-components will not be able to compete on performance or cost in the face of competition from those manufacturers that view the module as the fundamental component. Challenges for assembly Beside the material processing challenges are the mechanical aspects of assembly and alignment of the sub-components of the module. Currently, many modules are assembled by hand, and require in situ calibration and adjustments. While this approach may work for production runs limited to a few hundred modules, future requirements will dictate that tens of thousands of modules be fabricated. Manual assembly will need to be replaced by automated systems, a challenge every bit as great as the implementation of a robust, manufacturable electronic/optical integrated InP-based process. The roadmap toward a fully integrated transponder will therefore evolve in step-wise fashion. Initially just the detector and amplifier would be fabricated in a single chip. This would be followed by the modulator and the driver, then perhaps a component that consists of the modulator, driver, and the waveguide amplifier. The vision is of a transceiver module with a high-speed serial connection to the outside world where MUX/DEMUX, CDR and related functions are carried out, and an optical module with detectors, lasers and modulators. Eventually this would lead to a fully integrated transponder module. Optical modulation A variety of optical modulation techniques are available, the choice of which depends on factors such as the distance the optical beam transverses between electronic conversions, and the data rate. The simplest modulation technique is direct modulation of the current through the laser diode. This technique is typically used for relatively short distances and low data rates. Other devices used for higher data rates and/or longer distances include electro-absorption (EA) modulators and lithium niobate (LiNbO3) modulators. EA modulators utilize the effect of an electrically induced shift in the absorption band edge of a semiconductor material. LiNbO3 modulators utilize an electrically induced phase-shift in the optical beam, which is generally applied in a Mach-Zehnder interferometer configuration. LiNbO 3 vs InP modulators Currently the highest performance at high data rates is achieved using LiNbO3 modulators. These modulators have the disadvantage of being bulky and requiring large voltage swings (78 V) which are increasingly difficult to produce at higher data rates (e.g. 40 Gbit/s). Among alternative approaches for modulation at high data rates, InP is emerging as a strong candidate. The development of either semiconductor-based EA modulators, or LiNbO3 modulators operating in a differential mode by taking advantage of the different transport characteristics of these devices when fabricated in alternative crystallographic directions, are two possible solutions. However, the key challenge in the implementation of InP-based devices for these applications lies not in the individual performance capabilities of the devices, but in the ability to integrate sufficient numbers of transistors, along with optical components, in a cost effective manner. The successful development of such a process will open the way to a low-cost, high throughput, module manufacturing capability that will support 40 Gbit/s optical networking systems. Conclusion InP is a very attractive candidate for developing components for the "optical subsystem" at 40 Gbit/s and higher. This material has the potential of data rates well beyond 40 Gbit/s, and is applicable to the electro-optical conversion devices (lasers, detectors and modulators), as well as the electronic serialization/deserialization components.
×
Search the news archive

To close this popup you can press escape or click the close icon.
×
  • 1st January 1970
  • 1st January 1970
  • 1st January 1970
  • 1st January 1970
  • View all news 22645 more articles
Logo
×
Register - Step 1

You may choose to subscribe to the Compound Semiconductor Magazine, the Compound Semiconductor Newsletter, or both. You may also request additional information if required, before submitting your application.


Please subscribe me to:

 

You chose the industry type of "Other"

Please enter the industry that you work in:
Please enter the industry that you work in: