The Year of the Laser
Dec 21, 2000, 12:36 PM ET
By Yusuf Haque
One solitary wavelength of light can only deliver as much information as current transmission speeds allow. Lucent (LU) appears to be the first company that will deliver 40 gigabits/second wavelengths, in the form of its Wavestar TDM 40G express. Promising to leapfrog the competition, Nortel (NT) will be the next, delivering 80 Gbps sometime next year.
Leaps and bounds in transmission speed, total capacity (via DWDM – dense wavelength division multiplexing) and routing technology (via Avici (AVCI), Charlotte's Web, among others) have yet to do much for the end user.
2001 promises to be the year when KLECS ("Killer" CLECS, such as Yipes and Cogent Communications) spread their roots in a number of cities, delivering deals like 100 mps for $1000 per month (almost 100 times faster than a T1 for the same price). So, expect the keyword for the coming annum to be connectivity – not capacity.
Interstellar Overdrive
The peer–to–peer maharishis at Napster have done substantially more than the front–end of their software lets on. Their network ideology – minus routers and minus computing hierarchy – requires connectivity to salvage wayward transmissions. Even the brief titan of the late 1900s, Microsoft (MSFT) recognized this after introducing its Windows for Workgroups, when it first permitted a relatively simple creation of P2P networks on the Windows platform.
Investor enthusiasm has been choking in a cloud of concepts that, until now, have stampeded in the direction of faster and more intelligent technologies. Technologies that ultimately have added little more than complexity to the network.
Taking for granted the static nature of computing processes, those same individuals should look over their shoulder – sneaking up from behind is a new paradigm of alternatives, heralding an evolution toward better putting the world’s computing power to work. The most exciting of these trends, and the best argument for tomorrow's connectivity, is that of peer–to–peer networks and distributed processing.
The University of California at Berkeley research project, SETI@home, is one example of a way to capitalize – albeit through donations – on idle computing power and p2p networks. Monitoring and analyzing extra–terrestrial signals is a costly and computing–intensive process – SETI shares that burden with donated idle computers that act as remote processors of data, sending it back to the central server via the Internet, once completed.
The implications for connectivity are clear – end users must have it and support networks must be in place. There are over 100 million computers connected to the Internet, stringing them together in a distributive processing architecture would yield computing power 3,333 times more powerful than IBM's ASCI White, the most powerful computer on earth.
Dense Requirements of Multiplexing
As fiber to your doorstep progressively becomes a reality – thanks to Metromedia Fiber (MFNX) for corporate, and Onsite Access for multi–dwelling units – the prospect of connectivity takes on a whole new meaning.
On the computing side, the foundations for photonic computing seem to have been laid, (U.S. patent no. 5,644,123 outlines a photonic transistor). This implies that data conveyed through beams of light will become more pervasive as the years go by, extending right up to the hardware of your personal computer – that's when photonic computing will likely put the ASCI White out to pasture for good.
The plurality of light signals in dense wavelength division multiplexing, and the demand for connectivity at all levels of the network, has given birth to a robust market for the source of light for transmission – the laser diode – and a variety of modules used to manipulate the quanta they generate.
SDL Inc. (SDLI) – expected to be married with JDS Uniphase (JDSU) sometime this month – has witnessed an almost tenfold increase in quarterly revenue since 1995. This has been based primarily on its build–out of DWDM submarine networks, and the resultant demand for its 980–nm pump laser modules.
In a simple transmission diagram, the transmitting laser lies at one end and is connected via a series of amplification modules to the light receiver at the termination point. Internal to the EDFA (erbium doped fiber amplifier) module, produced by Corning (GLW) and JDS Uniphase, is an additional set of lasers critical to the amplification process.
The number of these EDFA modules actually used in any one network, and the number of lasers to be deployed at each point, is dependent upon the length of that network and the number of channels delivered through it.
The Laser Diode
Lasers have a tremendous number of industrial applications; however, the growth of fiber–optic markets has overshadowed their non–communications applications. The lasing process, that gives birth to the high–powered beams of light, takes place inside the laser diode – for telecommunications applications, the market for laser diodes is currently almost $1.5 billion.
For transmission purposes, the market for standard 1310 nm and 1550 nm lasers – used for long distances – is fragmented between more than 12 companies. The prevalent names in this space include Furukawa, JDS Uniphase, Nortel, Corning and Lucent.
For amplification purposes, the market for laser diodes currently stands at $390 million with a target of $1.2 billion in 3 years. Amplification is currently performed through two methods. The first is through the use of EDFAs, a technology first demonstrated at Pirelli, and deployed in mass by Lucent (LU), Ciena (CIEN) and now Corning (GLW).
Overcoming one of the chief disadvantages of photons versus electrons, the three cubic inch EDFA amplify optical signals without the need for optoelectronic conversion. The second method is through the use of Raman amplification. Employing either method, or both, elicits the presence of a pump laser module that typically represents 50% of the $12,000 to $17,000 cost of an amplifier.
The onslaught of high channel counts through DWDM has necessitated higher pumping power – today's 300mW pumps in the place of 150mW pumps – and therefore a larger number of laser deployments in the system. With single–mode lasers, tuned to a fixed wavelength, the higher number of channels requires a proportionate increase in the number of lasers tuned to corresponding colors.
Nortel, Corning, Lucent and JDS Uniphase sell the lion's share of EDFAs. JDS provides the pump chips for a large proportion of the modules used in the EDFAs made by Nortel and Corning. SDL's customer base is, in this regard, a beautiful complement to JDS as it supplies to Lucent and Ciena among others. The marriage of these two carves a wedge in the pump–chip market, leaving only Corning, Furukawa and Pirelli as the competition.
The prime argument pointing to 2001 as the beginning of a significant up–trend in the market for lasers is not only the continued build–out of long–haul and submarine networks – thus far the predominant reason for investing in SDL Inc. – but also an expected breakthrough of DWDM to the metro–net level.
Due to the high cost of DWDM gear, and the time consuming SONet configuration process, metro DWDM spending has been limited. Economies are now tilting very much in favor of DWDM as being the medium for metro–fiber capacity expansion.
The figures relating to 'lit' and 'unlit' fiber could be misleading indicators of whether or not capacity demands require DWDM as a pervasive standard in metro rings. Unlit fiber, or excess capacity, isn't always in the place where capacity is required. Certain metro routes are saturated – regardless of whether Ameritech or SBC have 'lit' less than 20% of their fiber.
Implications for the local access provider have become clear: in comparison with the alternatives for boosting capacity on crowded routes, DWDM represents a significant cost advantage.
The pump laser market will unexpectedly evolve in this arena as well, in the form of 'pumplets', or low–power pump lasers to be used in tomorrow's metro amplification process. The need for amplification at the metro level comes from varying power levels in signals and the erratic environment that typifies this network segment.
Amplifiers and gain flattening filters can be used as an application to smooth these variances in signal type: bringing order and consistency to lambdas as they are prepared for drop–off to the access layer and to your doorstep. Both Corning and JDSU are developing the components for these amps – in anticipation of this very quickly adding a half–billion dollars to the existing market.
In the section above, we briefly touched upon the notion of increased channel count invoking a need for a proportionate number of fixed wavelength lasers. The year 2000 witnessed the emergence of a number of companies battling with this expensive and deharmonizing effect on networks. Solutions have emerged in the form of laser modules, tunable to a vast number of varying wavelengths within nanoseconds.
Overall, tunable lasers offer the ability to remotely provision wavelengths, deploy all–optical switching and regeneration, and provide restoration to failed optical layers. Furthermore, as opposed to fixed wavelength lasers that rely on the intelligent network to switch their signals to termination, tunables ensure that the criteria for switching are inherent in their wavelengths. A clear leader has yet to emerge in this space.
In mid–1999, MCI Worldcom (WCOM) listed tunable lasers at the top of its future all–optical technologies. At that time the categorizations for various forms of tunable laser had been determined, however, there was no clear understanding of what the networks demanded, and what would separate the successes from the failures. Ultimately, it will be ease of manufacturing and packaging that will give any one of the 5 competing technologies a significant advantage over the rest.
Currently the categories are: Distributed Feedback (DFB) lasers, produced by Lucent, JDS Uniphase, Coretek (Nortel) and Fujitsu; Vertical Cavity Surface Emitting Lasers (VCSELs), presently a toss–up between Coretek and Bandwidth9; Tunable Distributed Bragg reflectors (DBR), produced by NTT, Alcatel (ALA), Marconi, Lucent and JDS Uniphase; External cavity diode lasers (EXOs), produced by New Focus and iolon; a Grating Coupled sample reflector (GCSR) made by Altitun (ADC (ADCT)), and finally a Sampled Grating DBR made by Altitun.
The first generation of DFB lasers were tunable to two to four channels only, and were of limited use in a network besides being a back–up source in case of failure. The new generation of tunable lasers offers tunability to virtually any lambda, within nanoseconds in some cases.
The DBRs produced by Altitun and Agility offer much wider tuning bands than their DFB predecessors – in the range of 40 nanometers – yet offer the least wavelength stability. Both solutions offer wavelength stability within +/– 3 Ghz. Being able to lock wavelengths within a tiny range of their center frequency is absolutely crucial in DWDM applications.
Avanex (AVNX) currently spaces its wavelengths at 50 Ghz in its 80–channel PowerMux, however, Cao’s target is wavelengths numbering in the thousands (i.e. spacing will shrink to fractional divisions thereof). Corvis (CORV), already supplies 160 channels in its DWDM ultra long–haul transport system, however doesn’t stipulate channel spacing.
At 100 Ghz spacing, lambdas must lock within 5 Ghz of the center frequency. This drops to within a 1 Ghz lock for 25–Ghz spacing. The Altitun GCSR, which tunes between 1528 and 1565 nm, is a range perfectly suited to long–haul transmission and EDFA deployment. The product is already shipping and, as production ramps up, costs should begin to drop in the next year.
Santa Barbara–based start–up Agility, has a higher power output than Altitun, at 4 mw, and can be tuned to any of 100 wavelengths in under 10 milliseconds. These two companies are destined to compete on the bases of manufacturability and scalability. Both processes are complex, although the Agility 3040 Widely Tunable Laser involves a slightly simpler process.
The VCSEL players are not head–on competition for the DBR producers. At the moment VCSELs can only be used for short transmission distances because of the non–availability of lasers that operate at 1310 and 1550 nm bands.
VCSELs quickly consumed virtually the entire market for short–reach applications on the strength of their ability to deliver better performance at a lower cost. They emit narrow linewidths, consume less power and can be tuned simply by altering the length of the vertical cavity in which the laser is produced.
The compelling advantage offered by VCSELs for LANs will likely become equivalent for the WAN and the MAN as soon as Coretek can produce a competitive product. However, the roadblocks to success are large, and its current idea of incorporating a 1310 nm pump laser inverts the cost advantage and simplicity of the VCSEL.
Long wavelength lasers are made from indium phosphide (InP) as opposed to short wavelength devices that are based on gallium arsenide (GaAs). Manufacturing a long wavelength VCSEL involves a complex and non cost–effective epitaxial growth process that eventually yields a highly reflective mirror based on Indium phosphide.
The growth in the market for VCSELs in the short–haul space is enough to generate excitement: in 1999 it amounted to $262 million, set to grow to $3.4 billion by 2004. This progress rides on the back of the robust fiber–channel and storage–area network build–out – this fast growing intrasystem links segments – in addition to the introduction of Very Short Range (VSR) SONet. The VSR market is expected to reach $3.7 billion by 2009.
Earlier this month, George Gilder added New Focus (NUFO) to his list of telecosmic champions based solely on its external cavity tunable laser. Although somewhat reminiscent of when he added Agilent (A) to his list based solely on its photonic switching platform, New Focus and iolon may have the competitive edge in tunables now that the formerly lab–imprisoned method can be fitted onto telecom transmitter cards.
The key issue is that with 20 mW of power output, aside from the narrow–tuning DBRs from JDSU and Lucent, no other laser comes close. With the ability to tune to any transmission band it’s hard to ignore the prospect of these lasers being deployed in millions of interconnection points in tomorrow’s all–optical DWDM network.
A Brief Year in Review
During the year 2000 it became clear that the colossi of the optical transmission world were not undefeatable. Even more quickly than Lucent's growth, after being spun off in 1996, the value of the company has shriveled to levels almost comparable with Corvis on its IPO day – which at that time had no customers and no revenues.
Entrants such as Sycamore (SCMR) have followed a completely inverted trend: its business shot past a start–up period in the blink of an eye, and generated 20 million in profit for Q42000. Demonstrating that the industry's climate is very open to young entrants with superior technologies and solutions. There are no antitrust fears in this industry, efficiency is ensured by the carrier’s need for a high standard of network performance.
This was also a year of handsome rewards from the market, for virtually any company in the optical networking space. As the number of companies keeping 1999’s promises shrank, the prize converged on networkers such as Ciena (CIEN), Redback (RBAK) and Juniper (JNPR).
Next year will be different, as the recent quarter has ominously warned. Those that focus on meeting customer demands in a cost–effective way will blaze the pathway to success next year. Triumph entails the transformation of this cottage industry into one with streamlined manufacturing processes and efficient returns.
Look for acquisitive large players to slow their pace of buying technologies from young start–ups, however expect a few big mergers to take place as larger players recognize each other's talents. The IPO market, for now, has dried and the VC taps are shutting down, forcing 2001 to be a year when the nose of those who have survived this shakedown will be firmly planted to the grindstone.
Lucent’s financing commitment of $7 billion, up from $2.3 billion in 1999, is indicative that build–out will not waver. Most of the spending will be focused on DWDM in both the long–haul and at the metro level. The components and subsystems market should ramp higher as 40 Gbps systems are introduced, necessitating a move to higher–end subsystems.
The terrestrial DWDM market is set to soar, according to a recent Ryan, Hankin and Kent report, from almost $5 billion this year to $24 billion in 2004. Riding on the back of all this growth by necessity will be strong sub–markets. Narrowing it to the strongest two segments positioned to benefit: look to amplification modules as the key to healthy lambdas and tunable lasers as the jewel in the crown of connectivity.
123jump.com
*******************
Not up on this Agility company.Be nice if the projections were all credited.
Jack |
