Today's tunable laser vendors are striving to prove their technology still provides savings in the static networks that are likely to dominate for the next few years. The primary short-term applications are sparing and one-time provisioning (OTP).
By KEVIN AFFOLTER
Agility Communications
Three years ago, tunable lasers were attracting great interest in the market, primarily as a key enabler for the intelligent all-optical network that would provide carriers with far greater flexibility, faster provisioning, and, ultimately, much lower costs. With the current telecom slump still suffocating both carriers and vendors, the focus has shifted from a new technology-driven rush to cost savings. Technology that doesn't save carriers money just doesn't fly anymore.
Most major system vendors still believe that the long-haul network will be filled with "wavelength dynamic" systems. These networks can have new wavelengths provisioned in seconds and provide massive savings by eliminating regenerators through use of ultra-long-haul technologies. The only question--when will these systems see live network deployment?
Meanwhile, tunable laser vendors are striving to prove their technology still provides savings in the static networks that are likely to dominate for the next few years. The primary short-term applications are sparing and one-time provisioning (OTP).
Power of sparing
To demonstrate the power of the sparing application, begin with a small, simplified network (see Figure 1). The system has four nodes and each node has 32 wavelengths in each direction operating at OC-48/STM-16 (2.5-Gbit/sec) line rates. A centrally located sparing center provides for the event of a transmit/receive (T/R) circuit pack failure. If a circuit pack fails at a node, a truck-roll is needed to perform the replacement.
If the T/R uses fixed wavelength laser technology, such as distributed feedback lasers (DFBs), one spare must be held for every wavelength in the network. If a T/R retails at $10,000, the cost of the spare circuit packs alone is a whopping $320,000. Add the additional costs of space at the sparing center and the truck roll, and the figure is even greater.
If the T/R uses widely tunable laser technology, the quantity of spare T/R cards can be drastically reduced. A single spare per node affords at least the same level of security as in the fixed wavelength case.
If a tunable T/R retails at $11,000--a 10% premium--the cost of the spares is now $44,000--an 86% savings, or $276,000. The spares can now be held at the node, so the cost of space and truck rolls is dramatically reduced, or even eliminated.
If the tunable T/R is now deployed at every site as in a one-time programmable (OTP) application, there is clearly going to be a greater cost incurred, as there is a premium for tunability. The total tunable T/R cost, excluding the spares, will be $2.82 million, whereas the fixed wavelength equivalent will be $2.56 million, or $260,000 less. This is more than offset by the sparing savings. The case is even stronger at 10 Gbits/sec, since the tunable premium is significantly less than 10%.
There are many other factors to be considered that are much more difficult to quantify, but potentially even more compelling. They include:
• Maintaining a product engineering code can run in excess of $10,000 per code per year. In the previous example, the fixed wavelength approach costs $310,000 more per year.
• Forecasting, scheduling, and planning 32 wavelengths instead of one is a large overhead. Those of who have been through it once would rather not go through it again.
• Enabling a carrier to provision a service at will is priceless. With a tunable T/R, you always have the right wavelength in finished goods inventory.
• With a tunable laser, the T/R becomes future proof and all the promise of the intelligent all-optical network can be fulfilled.
Not created equal
The above discussion deals with tunable lasers in general, but not all tunable lasers are created equal. In the OC-48/STM-16 example above, there are relatively few cost effective tunable laser options.
A 2.5-Gbit/sec tunable electro-absorption modulator laser assembly (TEMLA) has been introduced for this market space. The TEMLA is based on patented sampled grating distributed Bragg reflector (SG-DBR) laser technology, but has gone a step further.
The tunable laser is monolithically integrated with a semiconductor optical amplifier (SOA) and electro-absorption modulator (EAM) to provide high power--more than 300-km dispersion limited reach on most fibers (including SMF-28) and a 40-nm tuning range (see Figure 2).
In the 10-Gbit/sec case, there are many tunable laser companies vying for a share of the spoils. These solutions currently include continuous wave (CW) tunable lasers, which are then typically spliced to a LiNbO3 modulator. Today, only two of the tunable technologies have announced qualification in accordance with Telecordia GR-468.
The most important factor for any telecom systems vendor in selecting technology is reliability. Performance differences can be accommodated to some extent, but poor reliability cannot. The SG-DBR is a solid-state laser, tuned by changing the electrical current into the laser. The assembly processes are akin to those used for a DFB and have a proven track record.
Some of the alternative technological approaches for tunable lasers, such as external cavity lasers (ECL), rely on mechanical movement to achieve tuning. This poses both reliability worries and assembly complexity when compared to the SG-DBR.
Tunable lasers have also been designed using vertical-cavity surface-emitting lasers (VCSELs) combined with an optical pump chip and a separate SOA chip. This three-chip approach results in considerable packaging complexity. The tuning, in this case, is also performed through mechanical movement of a micro-electromechanical system (MEMS) mirror.
The specifications required by long-haul 10-Gbit/sec system vendors are broadly similar. These specifications include:
• Output power must be 20mW.
• Line width must be narrow (typically less than 5 to 10 MHz).
• 25-GHz channel spacing must be supported.
• Wavelength locking accuracy must be better than +/-1.5 GHz.
Cost is the last major hurdle for the tunable laser vendors. Vendors must deliver solid-state architecture coupled with best-in-class automated assembly platform at the lowest possible cost.
