One optical equipment maker recently complained to FibreSystems about the high cost of tunable lasers. A familiar refrain, you may think, except that the cause is strong market demand rather than the device's complexity – an unusual turn of events for the optical components industry. "Demand has suddenly come up out of a huge bust," says Lawrence Gasman, principal analyst at market-research firm, CIR. "There isn't enough manufacturing capacity."
It's the recognition by carriers of the advantages of tunable lasers that has revived the tunable laser market in the last 18 months. Indeed, DWDM network builds at AT&T, Verizon and BT are all using widely tunables – lasers that are able to operate at any fixed ITU-T wavelength across the C- or L-band.
Tunable laser shipments are now measured annually in the tens of thousands. But despite this strong growth in demand, device makers should expect increasing competition as new entrants battle for a share of the growing revenue pie.
The big deal
Tunable lasers simplify the problem of sparing for WDM deployments. A tunable-source line card is able to cover any of the 80 wavelengths in a DWDM system should a laser fail. By contrast, a carrier using fixed lasers must keep 80 spare line cards, one for each wavelength. Bearing in mind that a fixed 10 Gbit/s DWDM transponder alone costs $3500 (€2720), this adds up to a considerable expense for equipment that is idle.
Another benefit is that tunable lasers enable carriers to provision light paths promptly. "The tunable lead time is faster – a service can be brought up rapidly," explains Steve Robinson, vice-president of product management at Canadian equipment maker Meriton Networks. That translates to a much faster time-to-revenue.
"Most tunable lasers that we see deployed are only [used] for one wavelength and it stays there," explains Adam Price, director of high-end optics for US optical components manufacturer Bookham. "The tunable laser can be used to find the [most suitable] wavelength to use in terms of what offers the best system link budget."
Equipment reuse also becomes possible. One wavelength can be taken out of service and another deployed using the same tunable laser. "It is not dynamic wavelength provisioning but flexible provisioning," explains Robinson, although he notes that dynamic configurable networks remain a firm fixture on carriers' roadmaps.
Carriers also consider the deployment of tunables as an investment that future-proofs their networks. This provides a built-in competitive edge as they move towards full mesh-based optical networks where light paths can be set up and changed quickly and easily. Industry analysts emphasize that such networks are a way off, but that the deployment of tunable lasers and reconfigurable optical add–drop multiplexers (ROADMs) is a key prerequisite.
"Tunable lasers and ROADMs form part of a bigger trend – what some people like to call agile networks," explains Gasman. However, he thinks that the strong tunable laser sales have nothing to do with ROADMs. "ROADMs do not have to rely on tunable lasers, sparing is the driver."
In contrast, US equipment vendor Cisco Systems does make the link between tunable lasers and ROADM deployments, regarding them as key to advancing the wider adoption of WDM. "The tunable laser and ROADM solve the two remaining barriers to make WDM easy to use," explains Jeff Maddox, senior manager, product line management for Cisco's DWDM products.
According to Maddox, WDM networks comprise three key elements: amplifiers, filters and lasers. Tunables are replacing fixed sources while ROADMs are replacing fixed filters. Such flexible building blocks enable any wavelengths to be used in any combination. "You don't need 100 optical PhDs [for network planning and operation]; with tunable lasers, ROADMs and software, you can use DWDM in a far simpler way," he says. Such networking capabilities also bring operational savings for carriers.
Market growth
Both long-haul and metro DWDM networks are driving tunable laser growth. "Long haul is undeniably the growth part and has more channels," comments CIR's Gasman. "But metro is now growing faster than long haul. Metro may have fewer channels but it has a lot more nodes."
Industry analyst Ovum-RHK says that 29,000 tunable lasers were shipped in 2005 and it expects sales of 60,000 in 2006 – representing a second year of significant growth for the component makers. It also predicts a 64% compound annual growth rate through 2010. US-based tunable-laser start-up Santur says that in 2005 it shipped fives times the number of lasers that it did in 2004. "We expect to do the same again this year," says Gurinder Parhar, Santur's vice-president of business development and marketing.
New network builds are a significant reason for this growth. "Carriers in Europe and North America, like BT, are saying 'tunability is what I need in the network'. That is why there is demand, much to the surprise of the industry," says Hendrik van der Meer, product marketing manager for the optical platform division at Intel, US.
Another major factor is the ongoing transition from 2.5 to 10 Gbit/s wavelengths – the "currency of choice" for tunable lasers, as Canadian equipment vendor Nortel Networks puts it. "Two years ago, 75% of wavelengths on the line side were at 2.5 Gbit/s," explains Meriton's Robinson. "Now that has switched, with some 60–75% at 10 Gbit/s." Video-on-demand, linked to carriers' general upgrading of their access networks, is also a driver for metro growth.
As for prices, a tunable laser currently costs twice as much as a fixed laser. CIR says that the price of a 10 Gbit/s fixed laser is between $700 and $900, while a widely tunable laser costs between $1300 and $1600. The price premium of a tunable is more attractive when the laser is sold as part of an optical transponder: fixed and tunable 10 Gbit/s transponder prices are $3500 and $4500, respectively.
This difference in the price premiums between selling a discrete tunable laser and a tunable transponder shows the success of equipment vendors in beating down transponder prices. It also shows the higher prices that tunable laser makers can charge while so few players are shipping the product.
According to CIR, the biggest suppliers of tunable lasers are Santur, Intel and JDSU, which late last year acquired tunable laser start-up Agility Communications. Other players include Syntune of Sweden and Bookham, both of which have products in qualification, plus Pirelli Broadband Solutions of Italy and Paxera of the US.
Manufacturers of tunable lasers can be divided into two camps. The likes of Santur and Syntune sell their devices – butterfly packaged lasers and assemblies – to transponder makers and systems vendors. Other tunable laser makers, like Intel, JDSU and Bookham, primarily use their tunable lasers within their own transponders. As such, they sell butterfly packages, assemblies and transponders to systems vendors.
The systems vendors can also be split into two camps: those that buy transponders and those that also buy discretes for their own designs, although the trend is definitely shifting towards transponders. "Cisco is a transponder-focused firm," explains Maddox, stressing that it does not do "discrete designs". Nortel, conversely, buys both transponders and discrete lasers. The discretes are used for proprietary line-card designs where extended DWDM system performance like optical reach needs to be delivered.
Standard approach
The shift towards buying transponders explains why equipment makers welcome the development of the Optical Internetworking Forum's multisource agreement for an integratable tunable laser assembly (ITLA). The merit of the ITLA is simple enough: module and equipment vendors can now buy a standards-based assembly that fits within a 300-pin large-form-factor (LFF) transponder, rather than having to buy a laser and develop the drive electronics themselves. Moreover, being a standard, vendors can choose their ITLAs from several suppliers.
The ITLA has yet to reduce tunable laser prices, notes Kevin Green, vice-president of marketing and sales at Syntune (which is now sampling its ITLA), "but it should". "It will allow vendors to design-in three or four guys and play the price game," he says. An ITLA is currently 20% more expensive than a butterfly packaged tunable laser.
However, not all tunable laser makers are keen on ITLAs. "We focus most of our development at the transponder level," explains Kevin Affolter, director of JDSU's tunable lasers business unit. "Some 75% or more [of systems vendors] are shifting to transponders and don't want ITLAs," says Affolter, though he admits that JDSU may yet develop one.
"The ITLA is very good news, with more choice and a cost commensurate with increased competition," says Michel Belanger, Nortel's senior technical advisor for next-generation optical networks. However, he thinks that the technologies employed in an ITLA will be best applied to standard applications. "The ITLA will not be sufficient for some applications, which may call for discrete components coupled with more sophisticated control electronics design," he says.
The advent of the ITLA and equipment makers' increasing adoption of transponders shifts the design effort from the equipment maker back to the component vendors. The focus on transponders, coupled with the fact that tunable lasers have been in production for several years, means that a device's actual tuning mechanism is now much less of a concern for equipment makers.
Two laser-tuning methods are commonly used: an external-cavity design and an integrated indium-phosphide (InP) distributed Bragg reflector (DBR) design. Santur uses a third approach, comprising an array of temperature-tuned distributed feedback (DFB) lasers.
Pirelli and Intel both use external-cavity lasers. Here, the cavity length is extended outside the laser chip, enabling a highly selective grating to be used. The downside of this scheme is that external-cavity designs can result in relatively larger devices.
Pirelli's laser features a silicon-based nanostructured tunable mirror with a filtering function that is able to tune over the whole C-band by changing the applied driving voltage. Thus, the laser doesn't require actuators or MEMS to change its operating frequency. "It has few pieces. We have a standard gain chip inside and there is more than one supplier for this," explains Alex Schiffini, product manager for photonic components and modules at Pirelli Broadband Solutions. "A current controls the laser chip and a voltage controls the tunable mirror."
Intel's external-cavity laser uses two thermally tuned etalon filters for controlling the wavelength and wavelength locking. By leveraging the Vernier effect, the filters can be thermally adjusted to a desired channel anywhere within the C- or L-band. This design also boasts no moving parts, ramping its reliability and lifetime performance. Both Intel and Pirelli have a cost reduction roadmap for their devices and stress the ease of device control.
JDSU, Syntune and Bookham use integrated InP devices based on DBR lasers. A tunable DBR comprises an active gain region and a passive region that contains the Bragg grating. The attractions of this design are its relatively small size, lack of moving parts, a fast switching speed, and the ability to integrate a modulator. Syntune, for example, claims that its laser has a tuning time of less than 50 ns.
Santur's laser is based on an array of 12 DFB lasers. The output wavelength is controlled using a MEMS-based tilting mirror to select the most appropriate DFB laser. Fine-tuning of the wavelength is achieved by changing the temperature of the laser chip.
"We always look at the [tuning] technology used," explains Meriton's Robinson. "If the laser is mechanical it undergoes robust testing. If its mechanism is based on changing the electro-optic characteristics, we use humidity tests to test the chip is reliable."
However, Vladimir Kozlov, the founder of transceiver market-research company Lightcounting, believes that the tuning mechanism – and implied reliability – is no longer a consideration for vendors. "The fact that the laser makers are still in business after five years proves they have resolved reliability issues," he says. Nortel's Belanger agrees. "We are employing all the different types [in our systems]."
The tunable laser metrics that do concern systems vendors include the device's output power, switching speed and its cost. Meriton imposes an additional requirement, selecting suppliers that can provide 2.5 and 10 Gbit/s tunable lasers, since both are used for its metro regional 6400 optical transport platform. "We future-proof by choosing devices with high output power and fast switching speeds, though 50 ms is not a key requirement at present," Robinson explains.
Nortel downplays the importance of device switching speed, pointing out that it is the control software at the optical layer that dictates network performance. "It [the device's switching speed] alone does nothing for your customer," explains Belanger. This type of control software is used to optimize the performance of the optical system under all operational conditions.
Among other tasks, the software sets the optical power level of the elements (such as transmitters and amplifiers) such that any operation can be performed without perturbing the previously installed wavelengths. It also constantly reviews the system performance, detecting any deviations that will require compensation. "I'm completely convinced such a control system is of primary importance," comments Belanger, adding that Nortel has developed proprietary software for these functions.
Finally, laser size is another important factor. The more compact the laser, the easier it is to develop smaller tunable transponders and even tunable pluggable transceivers.
What next?
Tunable laser makers must expect a highly competitive market to emerge as companies chase overall modest revenues despite the forecasted growth. With demand tight, firms have to think about adding manufacturing capacity, but with costs of around $10 m, this must be considered carefully.
Lightcounting's Kozlov notes that a tunable laser start-up selling 10,000 units – equivalent to 17% market share in 2006 – will earn $13 m. "Can you run a profitable firm with 10 or 20 people on $13 m and not be pressured by VCs [venture capitalists] to go public?" he asks. Moreover, analysts predict that more optical component makers will enter the market, especially from Japan. Firms to watch include NEC, Mitsubishi Electric and Furukawa of Japan, as well as Opnext and CyOptics of the US. The result would be some dozen tunable laser players.
Syntune's Green agrees that a firm selling 10k or 20k lasers is not likely to survive and predicts that some of the smaller start-ups will be bought or combine in some way. It may even be the case that some of the five new entrants will drive some of that change. "We believe that a smaller set of players [than the dozen] will compete," he says.
"This is not a highly profitable market; there are no high gross margins," explains Kozlov. "It is impossible to forecast disruption, but honestly the signs are not there." The implication is that the tunable laser market, despite the expected growth, will struggle to support all of the players.
Meanwhile, laser makers are improving their offerings by developing more integrated products and driving costs down. Syntune is developing a monolithic device that combines a tunable laser, modulator and semiconductor optical amplifier, and expects to sample it later this year. Most existing 10 Gbit/s transponders use a separate lithium niobate modulator. "This is an expensive and a bigger modulator, but it offers great performance," explains Green.
Bookham has already developed a hybrid tunable laser and modulator, although both are InP devices. It is also developing a monolithic version for use in a tunable XFP transceiver design. JDSU is also working on an integrated laser and modulator product. "We are working to integrate a 10 gig laser with a Mach–Zehnder modulator," says Affolter. "It is one of the beauties [of an InP design]."
Smaller sources will help the shift to smaller-form-factor (SFF) modules and will ultimately lead to a tunable XFP transceiver. Fixed-wavelength laser DWDM XFP modules already exist. In January, Fujitsu of Japan unveiled a 10 Gbit/s DWDM transceiver application with a reach of up to 80 km in an XFP form factor, while at the OFC conference in March, Opnext, Mitsubishi Electric and Avanex, US, announced the formation of a multisource agreement to support DWDM at up to 80 km using the extended size XFP-E form factor, which is the width of two XFPs.
Green believes that the next product to hit the market will be a 300-pin SFF tunable transponder. Existing line-card designs are able to hold a single 300-pin LFF transponder. Using 300-pin SFFs, two to four modules can be hosted, delivering a 4 × 10 Gbit/s capacity line card.
Compact pluggables such as XFP-E and XFP will follow. Syntune expects to have an XFP design in two to three years' time. And Bookham's Price says that it will have a tunable XFP as early as 2007. In contrast, Intel thinks that 300-pin modules will remain the form factor of choice until 2010 and stresses the considerable design challenges of developing a tunable XFP. "There are power, heat and space issues," explains Intel's van der Meer. "It will be a long time before we see [tunable] XFPs shipped in any volume."
Cisco stresses that DWDM will benefit router and switch platforms just as it has its ONS 15454 transport platform. And XFPs are widely accepted on such platforms. As such, XFPs will benefit Cisco's overall IP-over-DWDM strategy. In turn, carriers like BT expect the capacity of router nodes in the core of their networks to grow to 100 Tbit/s. Such capacities will inevitably force the density on DWDM line cards to rise, enabling XFPs to play an important role.
The different philosophies held by Nortel and Cisco regarding discrete tunable lasers are evident in their views on XFPs. Nortel sees the demand for tunable and pluggable modules as distinct. "You can always take advantage of a pluggable's higher density and it promises to take $1000 out of the cost," says Belanger, "but a pluggable has many compromises."
That said, as a systems vendor, Belanger is keen to know when such a cost reduction will be passed onto him.