Few industries show as much promise for transforming the world. The optical industry holds the potential of giving everyone access to information everywhere, all the time. The race to light the world with optical technology should, in the end, help light human consciousness.

Special Features

Editor's Note: Bright promise For every optical company that has stumbled, there are dozens more that are promising to make optical switches smarter, lasers easier to tune, chromatic dispersion more manageable, Dense Wavelength Division Multiplexing ever more dense.

Photon profits
Remember back when any optical player that had a new twist or small variation on a switch, router, chip or marketing scheme could get funded, go public and watch its stock soar? Not today

Pushing the limits
Now that 10-Gbps traffic is the ascendant norm, and 40 Gbps is next year's battleground, physicists are pushing the limits of just how fast information can travel across optical networks and still emerge intelligible on the other end. Optical network companies are teaming physicists with engineers to tap more bandwidth, sharpen fuzzy or dim signals and correct errors on the fly.

See the light bounce
Replacing the current generation of switches with true optical ones not only eliminates the bottleneck caused by converting the signal, but creates a network that is more manageable, has much higher capacity, is faster to provision and is capable of supporting all kinds of new services, boosters say. It's not a question of whether to move in the direction of purely optical fabric, but how - and how quickly.

Test of time
Under pressure from carriers, manufacturers need to quickly move concepts from prototype to mass production. Carriers have their own set of pressures from customers eager to deploy complex enhanced services, with high reliability, at faster speeds, in smaller boxes.

Chips under glass
Using technology as old as World War II and as young as last week's breakthrough, chipmakers are transforming optical networks with devices that slice, dice and relight optical fibre and all its traffic.

Editor's Note: Bright promise

By Rebecca Cantwell

I remember my first visit to an optical laboratory, at the University of Colorado about a dozen years ago. Watching light beams dance across the darkened room as scientists explained how information would soon travel the globe on channels of light, I could hardly envision such a transformation.

But optical technology is, after all, based on the speed of light.

A couple of years ago at a micro-electromechanical systems (MEMS) research center at Bell Labs in New Jersey, a research scientist showed me through powerful microscopes how tiny mirrors channel waves of light. That same day, I watched a pure optical system in another Bell Labs facility beam bits of information from a hilltop clothed in fall colours straight into the lab.

These technologies still seem magical to me, but they are marching into our daily lives. Experts predict photons will be to this century what electrons were to the last. And with reality come some cold business lessons.

Some of the optical companies that sprouted on a dream and shot up with showers of easy money from investors have already withered. Venerable companies such as Lucent Technologies are paying the price for not moving quickly enough, Bell Labs notwithstanding.

But for every optical company that has stumbled, there are dozens more that are promising to make optical switches smarter, lasers easier to tune, chromatic dispersion more manageable, Dense Wavelength Division Multiplexing ever more dense.

Beyond the technological promise, the maturing optical industry is learning to focus on the need to move quickly from the laboratory to mass production, so that customers can count on their products.

Optical companies are also targeting their efforts to offer service providers a range of options in the products they sell. As they get ever closer to the end user, these products and technologies are solving the thorny problems of metro-area systems and their multiple demands.

But underneath the challenges of both technology and business lies a deeper importance.

Few industries show as much promise for transforming the world. The optical industry holds the potential of giving everyone access to information everywhere, all the time. The race to light the world with optical technology should, in the end, help light human consciousness.

Photon profits

Remember easy money?

Remember back when any optical player that had a new twist or small variation on a switch, router, chip or marketing scheme could get funded, go public and watch its stock soar?

Not today. Not with tech stocks taking a beating, carriers slowing their build-out plans and venture capitalists tightening their hold on money.

The optical market is overcrowded with players. Companies late to the market, those with shoddy business plans and those with untested technology will be the first to succumb. But even those with good ideas are in trouble if they placed their bets on wrong solutions, says Scott Clavenna, a principal analyst at optical research firm Pioneer Consulting.

Be a little late developing 10-gigabit-per-second optical switches, and see your stock fall 85 percent, as Lucent Technologies' shares did. Deal with too many customers that can't pay you, such as many of the struggling competitive local exchange carriers, and you're deep in the red. Have a great product but no way to ramp up manufacturing volume, and you're ripe for being acquired.

"Everything can't win, even if it is good," Clavenna says. "Carriers can't be asked to pick from 20 vendors for a single product. That's just too many."

A company might have a good technology and a workable system, but carriers still won't buy it if they've moved on to something new.

Ciena found that out when it made a run at the metro market. Ciena acquired Omnia Communications and its Asynchronous Transfer Mode-based platform for US$429 million in stock in 1999. Omnia had first-rate engineers, but Ciena ultimately abandoned the project because the industry was turning away from ATM.

Ciena was smart enough to do two things: It put Omnia's engineers to work on projects that were more marketable, and it jumped right back into the metro market late last year when it acquired Cyras and its K2 platform, pegged to the venerable Synchronous Optical Network (SONET) technology.

"The first thing we learned was that when you find yourself in a hole, stop digging," Ciena Chief Executive Patrick Nettles says. The Omnia engineering team went on to design Ciena's hot Core Director optical switch. "The second thing we learned was that when you're looking at an acquisition, you should focus on the customer contacts and traction that the company has," he says.

Ciena slipped only once, so it is counted among the winners as 2001 moves into its second quarter. Last month, the company sparked a spike in the stock market when it announced revenue of $352 million for the quarter, up 130 percent from the same quarter a year earlier. Nettles projected that Ciena's 2001 revenue would be 95 percent to 105 percent higher than last year's.

Nettles believes that optical technology stands at the crossroads this year. "It's a datacentric world, and data behaves in a different way than voice," he says. With fat pipes and fast switches speeding bandwidth at 10 Gbps, the Time Division Multiplexing solutions of SONET and ATM are increasingly irrelevant, he says.

The newer, faster Internet Protocol (IP) network will cost less to build and will have operational efficiencies. "The sea change will fundamentally differentiate the carriers that use that solution from those that don't," Nettles says.

The North American optical transport market is expected to double by 2004 to $45 billion per year, according to telecom research firm RHK. By then, 60 percent of vendor equipment revenue will be in IP systems, while the shares for the older frame relay and ATM offerings will shrink, analysts predict.

Lasting legacy

Lucent and Nortel Networks have long been the leaders in supplying equipment for telecommunications networks. But both had commitments to older technologies for networks built for voice conversations, and both were late with high-speed equipment for datacentric next-generation networks.

Nortel's mascot could be Buck, Jack London's wonder dog of the Canadian north: huge yet fast, smart, nimble and cutthroat when it has to be.

A network equipment company born of the breakup of Canada's staid old national telephone company, Nortel owns an astounding 53 percent of the overcrowded and burgeoning Dense Wavelength Division Multiplexing equipment market. DWDM divides light into dozens of channels - each capable of carrying voice, video and audio traffic - and is expected to be the highest-growth area in the red-hot optical market.

What could go wrong? How about last month's forecast by Nortel's CEO John Roth that the company's sales would grow just 15 percent this year, an estimate that sent its stock plunging 33 percent?

Nortel is an agile giant by most standards, but in optical networking it seems there is always another company with a newer architecture, a more sensitive finger on the pulse of the market.

Nortel is "perfectly positioned to deliver old solutions," Nettles says. "They have their whole business built around SONET and ATM. They talk a lot about new systems with new-generation products, but as far as we can see, they're not delivering much of their touted open-system DWDM systems. They have infrastructure and sales, but they're missing product."

But expert observers say Nortel is still plenty nimble. "No incumbent vendors are going to stay on top by continuously leveraging simply what they have," says Grier Hansen, optical analyst at Current Analysis. "Nortel is a very good example of incumbent vendors that have seen the space explode and know there are a lot of compelling technologies out there from start-ups, and have taken the necessary steps to compete with them."

Service providers will particularly feel the tightening money supply, Nettles says. "There's going to be a separation of the haves and have-nots - those that have capital and can move ahead with deployments vs. those who will have to take a conservative wait-and-see approach," he says. Level 3 Communications, McLeodUSA, Qwest Communications International, Sprint and others that anticipated the tightening and prefunded their build-out plans will enjoy a widening advantage over those without money, he says.

Deploying the hottest new technology will bring advantages to carriers. Last month, Corvis and Williams Communications sent a signal a record-breaking 4,000 miles without regeneration. Williams, one of the new-generation carriers not tarrying in the long-distance voice market, has committed to purchasing $300 million worth of Corvis' optical products.

Corvis' all-optical products can slash network costs in half because the signals never have to be regenerated, says Shyam Jha, vice president for marketing communications at Corvis. Still, it's tough competing with the legacies of Lucent and Nortel, he says. "They have the advantage of incumbency," Jha says. "We're the new kids on the block, and loyalty counts for something. We have to convince them that this is a better technology."

Ten years ago, when all routing was done by mainframes, Cisco Systems had a tough time convincing companies that they could save up to 90 percent by buying something called a router, Jha notes.

"What Corvis has going for them that no one else has, is they have a switch that works," Hansen says. "Their competitors have lambda routers in trials." Corvis took some heat last quarter for relying on only a few key customers. "But they have one in a network now passing traffic," he says. "That solidifies the validity of their solution."

Hansen applauds the newer companies such as Corvis, Sycamore Networks and Tellium on their all-optical products. "Certainly, it's harder to get in now in the all-optical market," he says.

Says Corvis' Jha: "Different people see the market evolution differently. The bandwidth barons - Williams, Qwest and Broadwing - are the first to build true optical networks, while the legacy long-distance companies - Sprint, WorldCom and AT&T - are slower. The game is changing rapidly. The newcomers such as Broadwing are selling bandwidth to the granddaddies."

Lucent is everyone's favorite whipping boy, but after overhauling product development and assembly at its optical plant, it is vowing to be the first to market - targeting later this year - with 40-Gbps switches. It also will bring to market this year a lambda router capable of switching up to 1,024 signal channels.

Lucent needed a shock to its complacency, but will come back, says Roger Wery, an analyst at Pittiglio Rabin Todd & McGrath. "They still have some amazing technology and will produce some great products," he says.

Lucent must trim its complex layers, so the spin-off of its microelectronics division, Agere Systems, and the reported interest in selling its fiber-optics plant are probably good moves. Layoffs, the spin-off, the possible sale and the recent US$4.5 billion in bank loans will net Lucent plenty of money to buy cutting-edge technology and vault it into the lead when carriers go on their next spending spree, experts predict.

Nortel, too, will bounce back, Wery says. A good way to start would be to recognise that competitor Cisco has about one-third the employees, with total revenue in the same ballpark. "This may be a trigger to be more nimble and get rid of some of their past and their overhead, to trim down their cost structure and become a sleek athlete," he says.

Nortel may have slipped a bit, but it hasn't been complacent. Last month, the company announced a suite of products that can make Web surfing easier, and help advertisers target banner ads to individual Internet users. A week later, Nortel purchased a Swiss subsidiary of rival JDS Uniphase, paying $2.3 billion for a company that makes 40 percent of the world's 980-nanometer pump lasers. It's those pump lasers that are needed for each channel of a DWDM system.

Huge pipes are not enough

The fastest switches powering the biggest pipes won't necessarily win the game. If size and speed is the only differentiator, everyone will fight to be the discount king, and wavelength will become a commodity in a long price war.

Carriers are looking to vendors that can combine hardware switches with intelligent software so they can offer different classes of Internet services to their customers.

"If you really want to offer carrier-class services over a data network, you need a way to guarantee the performance of the packets," Light Reading's Clavenna says. "The only way is with new forms of switches and routers, and equipment at the edge" of the network.

Celox Networks in January came out with a switch that can serve 6 million subscribers at a time and deliver 16 levels of service, from firewall protection to voice-over-IP to bandwidth on demand. It allows service providers to offer the basic $29.95-per-month service, but then easily ramp up to about $60 per customer with the lure of the gold-plated options.

The services-creation field is crowded with new products from CoSine Communications, Lucent's Spring Tide Networks, Nortel's Shasta line, Quarry Technologies and Redback Networks, among others. They're all stuck in second gear, waiting for a killer application to come along so carriers can fill their bandwidth and start clamoring for faster switches that can deliver even more traffic.

Equipment vendors would do well to remind themselves how pragmatic and bottom-line oriented their carrier customers are - especially now that financing is tight.

Global NAPs, an eastern seaboard carrier, is buying voice packet switches from Convergent Networks because it needs to make money now, and can do so with equipment geared to ATM protocols. Frank Gangi, president of Global NAPs, is convinced that ATM has a lot more life left in it - and some advantages over IP. "IP doesn't have quality of service inherent in its protocol, but ATM was written from the outset with QOS in mind," Gangi says.

The fact that IP works best with the cheap-to-install, popular and newly speedy Ethernet technology doesn't daunt Gangi. "For every argument I can make about how great ATM is, they can make an equally passionate argument about IP," he says. "But quality of service is just not there with IP. We couldn't migrate to any solution, no matter how cheap or fast or sexy, if your phone call is going to sound any different than it sounds on the traditional circuit network. If you've ever used IP telephony, it drops and garbles too much."

With equipment from Convergent, Marconi and Sycamore, Global NAPs' network "isn't a spider-web mass of 50 lines in 50 places," Gangi says. "It's one big, fat pipe into an ATM cloud. It's much more efficient and you get real savings. Plus, it's about a tenth the cost of traditional stuff. And it offers new services that Class 4 switches didn't dream of."

ONI Systems' Hugh Martin doesn't agree about ATM's future, but that's what makes horse races. He's betting on IP "for one simple reason: Never bet against Ethernet," he says.

Ethernet, a protocol most commonly used with IP, is cheap and simple to install. The rap has been that it's slow. But with optical fiber, Ethernet has moved up from 10-megabit-per-second speeds to 100-Mbps to gigabit-per-second speeds. "And we're less than nine months away from 10-gig [Gbps] Ethernet," Martin says. "It's just relentless. It's a technology that never loses. And Ethernet is IP. Over time, ATM is going to go away."

The dozen companies making intelligent devices for what they believe will be an IP future are trying to put a lie to the claim that IP and Ethernet don't allow service providers to offer different classes of service.

Playing well with others

As optical fibre moves closer to the customer, the challenges and the competition escalate. Equipment vendors that can form partnerships with component or fiber makers are at an advantage.

Quantum Bridge Communications makes optical equipment for the edge, where businesses are connected to the network core. Last month, it announced a partnership with Corning whereby they become each other's preferred vendors. "We help our customers get access to Corning fiber, and Corning helps its customers who want optical to reach their business to get access to our equipment," says Jeff Gwynne, vice president of marketing at Quantum Bridge. "We look at partnerships to provide fuller solutions."

Quantum Bridge also allies with companies such as Marconi and Nortel so it can focus on innovation and let others worry about providing interfaces with different technologies and protocols, Gwynne says.

Analysts point to another reason why they think the optical market will soar - or at least outperform the entire high-tech sector. Salomon Smith Barney expects the shared-storage business to grow to $40 billion per year by 2003. That's increasingly a metro-area business, because the storage centres need to be close to large businesses. And with the bandwidth those centres demand, the only conduit that can deliver it is, of course, optic fibre.

The optical space isn't as hot as it once was, but there will be continued exploding demand for bandwidth, "and that's not going away," says John Kane, CEO of Telseon, which lights fibre in the metro area and sells it to service providers. "The cheapest and easiest way to deliver that is over optical networks."

Telseon last month won $175 million in financing from investors sold on its business plan. Investors were just as sold on Yipes Communications, which plays in the same space, but sells services directly to businesses. Investors gave Yipes $200 million last month. Both companies completed fibre-optic rings around 20 U.S. cities by the end of last year.

"Money is sitting on the sidelines until they understand where the bottom of the market is," Kane says. "Money will then start to flow back into the market, albeit cautiously."

To survive as an equipment vendor today, "don't get involved in religious wars," Gwynne suggests.

He urges vendors to provide solutions for big incumbent carriers, new competitors, cable television and those that run various protocols and technologies. "You can't say, 'I'm going to be an IP bigot, or an ATM bigot,'" he says. "You have to provide equipment that guarantees their investment in any of that. Listen to your quality customers. Most of what they want is going to be widely commercially applicable. Give them total solutions and love and care. That's what's going to separate winners from losers."

And one more thing: "Run like hell," he says. "Speed is life. We have a saying around here: 'The guys that are quicker and most agile will win the race.'"

Pushing the limits

Back when 10-Gigabit-per-second and 40-Gbps speeds were just a glimmer in an engineer's eye, no one worried much about bits bumping into one another on wavelengths of light.

But now that 10-Gbps traffic is the ascendant norm, and 40 Gbps is next year's battleground, physicists are pushing the limits of just how fast information can travel across optical networks and still emerge intelligible on the other end. Optical network companies are teaming physicists with engineers to tap more bandwidth, sharpen fuzzy or dim signals and correct errors on the fly.

Among network equipment companies, customer service and brand-name loyalty may separate the winners from the losers, analysts say. But among components players, cutting-edge technology will still make the biggest difference as carriers attempt to meet the public's apparently insatiable appetite for bandwidth.

Delicate balance
The first order of business is dealing with chromatic dispersion - the tendency of a light beam to spread as it travels through fibre. In an optical network that uses Dense Wavelength Division Multiplexing (DWDM) to segregate traffic onto dozens of different-colored wavelengths, chromatic dispersion degrades the optical signal, requiring costly signal regeneration.

But there's a delicate balance to be struck - different wavelengths of light travel at different velocities, and without enough dispersion, the signals aren't separated and wavelengths will interact with each other. If there's too much dispersion, the signal can't be detected.

Carriers have turned to single-mode fiber from Lucent Technologies and Corning to minimise signal dispersion. But the fibre best at minimising dispersion can't keep up with the new high speeds. Each time the speed is quadrupled - from, say 2.5 Gbps to 10 Gbps - the dispersion tolerance drops 16-fold.

"Up until 10 gigabits [per second], everything worked pretty well," says Mark Barratt, vice president of business development at LaserComm. "At these speeds, the fact that each wavelength is a bit different has created a problem."

Long-haul and ultra-long-haul systems trying to move traffic at 10 Gbps or faster need extremely precise tools to manage the dispersion. At high speeds, merely a change in the ground temperature where the fiber is buried can increase dispersion. "It's to the point where we think it will take a device that is actively changing the dispersion in concert with the DWDM system's fluctuations," says Mark Stubbe, vice president of marketing at LaserComm. "We refer to it as dynamic variability."

LaserComm has developed a single box that solves dispersion problems on long-distance fiber networks. Its 10-inch by 10-inch by 1.5-inch box includes a spool of fibre that reverses the dispersion effect, sharpening the signal of each channel on an 80-channel fibre.

"Without the device, which goes into DWDM subsystems of optical amplifiers, the bands would have to be corrected one at a time," Stubbe says.

LaserComm is testing the Hi-Mode Dispersion Management Device with several DWDM system developers, but hasn't announced any paying customers yet. The company expects to announce a box later this year that can correct dispersion over all channels of a 40-Gbps system.

LaserComm's next stop is the longer-wavelength L-band. As bandwidth demand soars, carriers hope to double today's output by tapping 160 channels on the L-band. In February, LaserComm announced a device to manage chromatic dispersion in the L-band.

Avanex offers another way to compensate for chromatic dispersion - a device with two tilted parallel mirrors. Off-the-shelf micropositioners tune the mirrors to bounce light back and forth in a way that slows or speeds the traffic flow through the dozens of wavelengths. It's another way of assuring that the wavelengths arrive simultaneously so the signal isn't lost in the noise.

Avanex's advantage is its wider tunability and smaller size, says Giovanni Barbarossa, senior director of research and development. It doesn't use a special high-mode fibre, or any fibre at all, so "you don't need to spool the fibre or be constrained by the bending radius of fibre."

Avanex says it has shipped its product to all the major manufacturers of equipment for high-speed optical networks.

Reamplification redux

Traditionally, carriers have used Erbium-Doped Fibre Amplifiers to revive weakened signals. EDFAs excite erbium atoms that release energy to the light beam as it passes through the amplifier. They work well in the C-band where wavelengths are between 1,530 nanometers and 1,563 nm in length. Erbium also can work on the L-band, at 1,570 nm to 1,610 nm.

But erbium reamplification is only good for about 80 km and it doesn't work on the S-band, where wavelengths are shorter, between 1,485 nm and 1,520 nm. The inability to reamplify S-band wavelengths makes today's optical network operate at about half its potential efficiency.

When email was the primary data traffic, the C-band was plenty. But now that consumers want video-on-demand, and bosses want videoconferencing on Fridays, carriers want access to the S-band.

Raman amplification attacks both problems and is at the center of a hot new battleground. In Raman amplification, the fibre itself is used as the gain medium, so the signal weakens much less over a long distance. The launch power of the transmission signal doesn't have to be as high, so fewer amplifiers and less power are needed to carry the information cross-country. Raman amplification is needed about every 100 km.

Xtera Communications is refining Raman amplification so it can awaken the S-band of light waves. "We can increase bandwidth by 50 percent" without adding more fibre or switches, says Chief Executive Jon Bayless, the former chairman at Ciena.

In February, Xtera announced that it had amplified 16 channels in the S-band, using discrete Raman technology to transfer energy from the pump laser to the light traveling through the amplifier.

No one has announced customers for the S-band solution yet, but Bayless suspects Xtera isn't alone in developing the technology. Fibremaker Corning and several equipment vendors also are reported to be developing S-band Raman amplification gear.

Short-haul solutions

Not all the physics challenges have to do with more wavelengths traveling greater distances. ONI Systems CEO Hugh Martin says his company is more interested in moving traffic down the street, across town and from suburb to suburb.

On the long-haul network, information can travel from Los Angeles to Dallas without being touched by an optical switch. But the metro area is crowded with switches and couplers. It has its own regeneration and reamplification rules, and as speeds begin to zoom, the solutions get more complex.

"One minute, the traffic has to move down the street to the next building; the next, most of the traffic has to get 20 miles away," Martin says. On a 20-mile trek across a metro area, the signal passes through dozens of switches, each robbing the pulse of some of its accuracy.

ONI Systems builds a platform that incorporates devices from companies like Avanex, LaserComm and Xtera with its own Dynamic Power Control technology.

ONI's variable-gain amplifiers monitor the optical power of every wavelength in the metro area and change how much amplification exists in the system at any moment.

If the blue wavelength is on a milk run to the office down the street, the signal won't degrade, even if it's not on full power. But if blue suddenly is switched to a crosstown route, signal clarity becomes a factor and ONI's intelligent switch turns up the juice.

Most networks use a fixed system, adjusting, say, the blue wavelength to travel a 20-mile route, never longer than that, never shorter than that. ONI's dynamic power control can tweak the power and correct the dispersion, so that any wavelength can travel any distance and deliver intelligible data, voice or video.

"In real time, we adjust the power level of every colour," Martin says.

Ciena and Nortel Networks reportedly are working on similar systems, but ONI so far is the only company with announced customers - Qwest Communications International and Williams Communications.

"We either make the signal much brighter at the input, or have amplifiers along the way to juice the wavelength," Martin says. "At 10 gigs the signal-to-noise ratio gets important, and at 40 gigs it gets even worse."

See the light bounce

By Lisa Greim Everitt, Special To Interactive Week

Light bouncing off bubbles, tilting off tiny mirrors, refracting through crystals and shifting across holograms.

Cross a seventh-grade science fair with the inventions of Rube Goldberg, add a dash of Dr. Seuss and - behold! - the state of the art in all-optical switches.

Replacing the current generation of switches with true optical ones not only eliminates the bottleneck caused by converting the signal, but creates a network that is more manageable, has much higher capacity, is faster to provision and is capable of supporting all kinds of new services, boosters say. It's not a question of whether to move in the direction of purely optical fabric, but how - and how quickly.

In the rush for bandwidth, optical device purveyors must weigh reliability, speed and throughput against cost and yield. "All engineering choices are trade-offs," says Gary Austin, general manager of Lucent Technologies' optical switching division. "The fundamental value proposition is not because it's neat technology or optics for optics' sake, but because it's really the most economical way to build switches."

Lucent has placed its bet on micro-electromechanical systems (MEMS) - micro-electromechanical arrays of tiny mirrors made of silicon that power its WaveStar LambdaRouter. The silicon plates flex on springs, creating motion without friction, Austin says.

As the number of ports increases, the number of cross points in a wave-guide system quickly becomes difficult to manage. A 256-port switch requires more than 65,000 cross points, Austin notes, and a version of the WaveStar LambdaRouter with 1,024 ports is due to ship by midyear. Lucent began testing the WaveStar LambdaRouter last summer on GlobalCrossing's next-generation network; tests with four other carriers are ongoing.

Corvis is one step closer than Lucent to full-fledged deployment of a photonic switch. So says Shyam Jha, Corvis' vice president of marketing, who notes that Broadwing, Qwest Communications International and Williams Communications all have Corvis equipment on the job. Broadwing expects to have revenue-generating traffic by the end of March, he says.

Corvis declined to discuss the details of its optical switch, but analysts suspect it's similar to Lucent's MEMS technology.

Meanwhile, Agilent Technologies took the library of thermal ink-jet patents it inherited from its parent company, Hewlett-Packard, and, working with Alcatel, translated them into a photonic switching technology. The Agilent switch uses bubbles in the index-matching fluid to reflect light, a solution that Agilent says is simple, proven and inexpensive to manufacture.

Jennifer Pigg, executive vice president at The Yankee Group, says MEMS shows promise because the mirrors can adjust microscopically in three dimensions. "With any luck, you can go diagonally," Pigg says, increasing the number of possible gates and enabling faster provisioning and fault recovery by creating a fully redundant switch fabric with far fewer ports. "It really does look like the future," she says. "It saves a huge amount of money."

The Yankee Group predicts MEMS technology will account for 65 percent of the all-optical switch market. Bubble technology such as Agilent's will capture 20 percent of the market, with the remaining 15 percent divided among other technologies. The firm predicts the all-optical switch market will grow from a standing start in 2000 to more than US$2 billion by 2005.

"The name of the game is to help carriers reduce the cost of providing bandwidth," Lucent's Austin says.

While The Yankee Group predicts the lion's share of the all-optical market will go to MEMS and bubble technologies, there's still room for alternative approaches. Start-up companies with interesting angles on the all-optical switch problem have been popping up like mushrooms. Among them are:

Light Management Group, a company that in December demonstrated its fibre-optical line switch. The switch is small, does not have any mechanical parts and uses an "acousto-optical deflector set" to randomly connect the channels from the array to a single fiber line.

Chorum Technologies, maker of all-optical processors and subsystems, uses liquid crystals instead of hard contacts to manipulate optical signals in its PolarWave family of products. Chorum recently acquired liquid-crystal display specialist Polytronix to help build out its line of filters, switches and processors.

Trellis Photonics has developed a solid-state technology that relies on wavelength-specific holograms written in potassium lithium tantalate niobate crystals. By building crystals into a matrix of rows and columns, any wavelength can be routed to any fibre.

The Yankee Group's Pigg also sees telecommunications potential in research at Harvard University, where scientists used chilled gases to slow, then stop, the passage of light. Two independent research teams managed light by beaming it through a chamber of gas, first slowing the light, then holding it at a standstill. The researchers revived the beam with a second flash of light, duplicating the shape, intensity and other properties of the original beam.

"We've been able to hold it there and just let it go, and what comes out is the same as what we sent in," Ronald Walsworth, one of the researchers at the Harvard-Smithsonian Center for Astrophysics, told The New York Times.

Test of time

By Lisa Greim Everitt, Special To Interactive Week

As equipment manufacturers and service providers rush to build next-generation networks, new approaches to optical tests and measurement are gaining ground.

Under pressure from carriers, manufacturers need to quickly move concepts from prototype to mass production. Carriers have their own set of pressures from customers eager to deploy complex enhanced services, with high reliability, at faster speeds, in smaller boxes.

Service providers must be able to verify the performance of individual components - by bench-testing, then field-testing variables like chromatic dispersion, polarisation, attenuation and the geometry of the fibre itself. Finally, the network needs to be monitored, and troubleshooting must be done as a regular part of its operation, checking bit-error rates, service levels, fault isolation, provisioning and power-balancing.

That's a broad range of requirements. What a field tech needs to do the job on-site is entirely different from what is needed in the lab, where test instruments the size of small refrigerators are not uncommon.

"The primary challenge is the speed at which the measurement needs to be made," says Jennifer Pigg, who follows optical networking at The Yankee Group in Boston. In the all-optic network, plugging in electronic test equipment provides proof of Heisenberg's theorem - the act of monitoring can degrade the network's performance.

As dynamic provisioning and other intelligent services move from pipe dream to reality, analysts say, expect test functions to migrate from the lab into the network, integrated into transport equipment.

Lucent Technologies was among the first to explore this market, teaming with Digital Lightwave, provider of optical network testing gear.

"They realised that in order to speed up deployment, they need to put the test and diagnostic functions into the transport equipment," says Digital Lightwave engineer Doyle Mills. In January, JDS Uniphase took a minority stake in Avantas Networks, a network testing equipment company.

Last fall, Digital Lightwave added Internet Protocol to the list of technologies analyzed by its portable Network Information Computers and Network Access Agents. Carriers such as 360Networks, Level 3 Communications and Livingston UK have purchased Digital Lightwave testers for installation, maintenance and remote performance monitoring.

Sniffer Technologies, a unit of Network Associates, uses Digital Lightwave's Dense Wavelength Division Multiplexing (DWDM) analyser in its network management products, allowing technicians to pick out one circuit from the hundreds within an optical fiber and run diagnostics on its physical and applications layers.

Among component manufacturers, the name of the test-and-measurement game is to figure out ways to get a brilliant idea from prototyping to mass production as quickly as possible.

Claiming that test and measurement accounts for 60 percent of the cost of optical manufacturing, Canadian test kit maker Exfo Electro-Optical Engineering is working on a system that automatically tests DWDM components, eliminating the need for manual testing. Exfo, which bought Burleigh Instruments last year for its fibre-alignment expertise, has said it will spend $20 million to triple its manufacturing capacity.

Another test equipment vendor, Newport, took a similar tack when it bought Kensington Laboratories in February. Newport plans to add Kensington's robotics and motion-control expertise to its line of test and assembly equipment for semiconductor and optical device manufacturing. "This is particularly important as system speed and material handling features are added to our test and assembly tools," says Newport Chairman and Chief Executive Robert Deuster.

Plenty of test-and-measurement companies are claiming "firsts" in the fast-moving market. GN Nettest, a division of Denmark's GN Great Nordic Group, in January released a test suite for Multiprotocol Label Switching. GN Nettest claims its product is "the first complete MPLS testing solution," covering functional, conformance, emulation and interoperability testing.

Valiant Networks says it pioneered a vendor-neutral approach with its third-party independent testing lab. Valiant's 6,000-square-foot "UltraNOC" provides network design, interoperability testing and other services for service providers and equipment manufacturers. Anritsu Co. of Richardson, Texas, says it is the first to support OC-192 (10-gigabit-per-second) speeds in its testing gear. Luciol Instruments, a Swiss start-up, says it is the first to develop exquisitely sensitive photon-counting technology to measure both glass and polymer optical fiber.

Agilent Technologies says it is the first to focus on testing the optical control plane - the protocols and hardware connecting different switches and routers. For carriers eager to move onto all-optical networks and offer bandwidth on demand, the control plane holds the key - but each switch vendor uses its own protocol, from Sycamore Networks' Broadleaf to Tellium's StarNet to Ciena's Optical Signal and Routing Protocol. Open optical networking seems a long way off.

While vendors, carriers and the Optical Internetworking Forum fight among themselves over standardisation, Agilent's optical control plane analysis software sidesteps the problem of competing protocols. It converts and displays everything in hexidecimal code, carrying new protocols on a small bandwidth-control channel, rather than taking up a whole wavelength.

While deployment of all-optical networking will be gradual, Agilent expects plenty of business from optical equipment developers eager to see how well their equipment meets market requirements.

Chips under glass

By Bill Scanlon

Using technology as old as World War II and as young as last week's breakthrough, chipmakers are transforming optical networks with devices that slice, dice and relight optical fibre and all its traffic.

The companies are not short on astonishing claims: Gemfire has a device that it says can pump erbium atoms onto eight channels at a time, slicing regeneration costs in half. Vitesse Semiconductor says it has a single-chip optical device that cuts the need for regeneration by decoding incoming data loaded with errors and restoring it to what the sender had in mind. Zolo Technologies says it can separate 200 channels in a far simpler way than ever before.

While the optical switch wars make headlines, components engineers seem to be having more fun, designing chips the size of playing cards, postage stamps or wafers that will simplify life for service providers and their business and residential customers.

The chips - optical or semiconductor - go into the equipment that systems makers are selling to carriers. But someday soon, they'll be the systems themselves, replacing routers, cross-connects and switches.

"The components guys drive the tech engine that feeds the evolution of the network," says Ed Rodriguez, head of KPMG's Electronics Industry Group. "We're almost seeing systems-on-a-chip on the network side, just as Intel did in microprocessors. It's very realistic that they'll go down that same path with high-horsepower routers and switches."

Of course, it's sometimes difficult to sort out what looks great on the wafer table from what can be produced in volume and work in a real, live, lit network. And it's tough distinguishing between all-optical and semi-electronic solutions - or, for that matter, figuring out why the carrier or end user should care.

"There's a lot of fiction out there," says Simon Keeton, optical manager at Vitesse's forward-error-correction unit. "All-optical networking is a buzzword right now. But you'd be hard-pressed to find anyone delivering true all-optical solutions, because of the physical limitations."

Some 30 percent to 60 percent of the costs of optical networks are in regenerating the signal about every 50 miles. The greater the distance signals can travel without needing regeneration, the faster carriers will adopt optical networks for all traffic.

Vitesse's forward-error correction leans on technology developed by World War II code breakers to clean up signals. It's a stand-alone hardware device that sits close to the fibre, using parallel buses to slow the information down. The data is channeled into a single-bit stream running at the speed of the network, perhaps 10 gigabits per second, and handed off to a multiplexer. After the traffic is converted into electric signals and demultiplexed, it goes into Vitesse's single-chip algorithm device for decoding and error correction.

All-optical devices are transparent and don't care about technologies or protocols. But, made with bubbles or mirrors, they're hard to manufacture in volume, are prone to failure and are relatively dumb devices.

Optical technology so far can't process information, monitor performance or determine whether a stream of traffic needs top priority - such as a phone call - or is merely coach-class service - such as email.

"You can do more intelligent switching in the electronic universe," Keeton says. "We're going to see people look more toward optical switches and at new ways of monitoring that signal. But we're never going to get the same level of monitoring that you can get with an electronic switch."

Optical Switch is building optical switching components for systems that require just a handful of channels per fibre because that's where 95 percent of the demand is today, says Chief Executive Gary B. Nabhan. In three years, when $15 billion worth of optical components are expected to be sold, 80 percent of orders will still be for systems with 32 channels or fewer, he predicts.

"There's a tendency in the investment world to think the transition to an all-optical network will occur overnight," Nabhan says. "I don't think so. I think it will be phased in over five or 10 years."

Optical Switch has invented and patented a tool that automates setup of a chip laboratory's optical table, where the components are manufactured, Nabhan says. "It used to take our Ph.D.s all day to set up a table and produce one or two wafers," he says. "Now that we've computerized and automated the process, we can process multiple wafers per minute."

And producing products quickly may become a key differentiator.

"In this industry there's an awful lot of hype, people saying, 'I can do this many channels,' " says Zolo Chief Technology Officer Andrew Sappey. "The question you really need to ask is not what 20 Ph.D.s are doing in a lab, but 'what is it you can deliver in two weeks?' "

Despite the current slowdown in the systems business, there remains a severe shortage of some of the components that make those systems work.

"The ability to manufacture and get these things out is a huge deciding point for systems vendors on whose components they are going to use," says Grier Hansen, an analyst at Current Analysis. "They're going to pick a component they can actually get."

Copyright © 2009 CBS Interactive, a CBS Company. All Rights Reserved.
ZDNET is a registered service mark of CBS Interactive. ZDNET Logo is a service mark of CBS Interactive.