Part II
Advanced optics can't do the job alone -- a new generation of electronics is necessary.
Raman amplification fits this bill perfectly because it is the highest-power, lowest-noise solution available today. To our knowledge, no one has designed OC-768 DWDM systems without it. Raman is a distributed amplification technology, which means it amplifies an optical signal over virtually the entire length of the fiber, not just a short section. Consequently, the amplification per section of fiber is lower, even though the total output power of a Raman amplifier is usually much higher than the output of a standard erbium-doped fiber amplifier (EDFA). SDL currently is the leader in Raman lasers, with a substantial time-to-market lead and the lowest-power products. JDS Uniphase and Lucent have also entered the market recently, and there are also a couple of startups targeting this area.
Yet another optical problem that's potentially a show-stopper at 40 Gbps is dispersion and distortion. At these speeds, the signal becomes distorted by polarization mode dispersion (PMD) and chromatic dispersion as it passes through the fiber. In long-haul networks, dispersion is problematic at OC-192, but it's prohibitive at OC-768. PMD is a particularly nasty problem because it's dynamic, or constantly changing with time. To correct PMD at high speeds, new subsystems need to constantly and dynamically measure and correct this problem. A few companies are specifically targeting dispersion problems. Yafo Networks has recently introduced its Polarization Mode Dispersion Compensator (PMDC), while Avanex and LaserComm products are targeted at the chromatic dispersion problem.
Breaking New Ground With High-speed Electronics
Advanced optics can't do the job alone -- a new generation of electronics is necessary. Standard silicon CMOS chips can at this time achieve OC-192 speeds for digital and some mixed-signal parts. But silicon germanium (SiGe) has been gaining traction for mixed-signal parts and gallium arsenide (GaAs) is commonly used for analog ones, such as laser and modulator drivers. For OC-768, a number of suppliers, such as AMCC and Conexant, are likely to apply SiGe to mixed-signal designs. Because of the high breakdown and swing voltages required, in addition to a high frequency limit, Conexant and Vitess plan to use other processes, such as gallium arsenide PHEMT and indium phosphide (InP), for analog parts. Our research indicates that InP will become the most prevalent OC-768 analog solution. Some vendors with military and space experience may have a leg up in this market.
Aside from individual optical and electrical components that can operate in 40-Gbps systems, circuit board and system design present unique challenges. At these high frequencies, analog electronic design becomes much more difficult and costly due to RF interference issues. however, moving from OC-192 tto OC-768 is not as big a step in this regard as moving from OC-48 to Oc-192. Consequently, experience with OC-192 should go a long way in the OC-768 market. We believe that strong manufacturing capabilities are a meaningful differentiator in the DWDM market even at OC-48 and especially OC-192, and will be even more pronounced at OC-768.
Lastly, testing high-speed systems is particularly difficult and expensive, and test equipment at a given speed usually becomes available at the same time, or soon after the systems themselves. Consequently, vendors will need to bake self-diagnostic capabilities into these systems, which is difficult.
The Ball Has Begun to Roll
Despite techncial hurdles, early lab trials have already begun. Toward the end of 2001, the first commercial deployments of Oc-768 should begin, with companies like Lucent and Alcatel pushing to be first to market and recapture mindshare lost to Nortel, which was early with OC-192. We wouldn't count out Ciena or Nortel either because of their experience with OC-192 and manufacturing capabilities, nor would we dismiss some of the newer private plays pursuing next-generation architectures that rely on advanced modulation and amplification. But there won't be a driving force until high-performance merchant semiconductors enable switch and router vendors like Cisco and Juniper, and terabit vendors Avici, Nexabit (Lucent), and Pluris, to move to 40-Gbps interfaces. If the past is any indication, this will happen about two years from initial commercial deployments, with system prices at that point falling enough to drive adoption.
>>>>>
If someone has an update on PMCS's Abrezzio's OC-192 interface development, please post.
And, don't be upset with today's drop. We gained 20 points on the week. I don't know many who can claim as much.
Cheers!
Pat |
