Power, reliability, and stability improvements extend utility of 980-nm pump lasers

Toby Strite

Stefan Mohrdiek

Berthold Schmidt

The continued phenomenal success of optical networking depends on engineers doubling the bandwidth transmittable over a single fiber strand every 9 to 12 months, thereby maintaining the pace set since the early 1990s. This pace represents roughly twice the innovation rate of the silicon industry described by the now-famous Moore`s law. The resulting erosion of the data transmission cost leaves a large pie of added value to be divided between component and system manufacturers, service providers, and the end user surfing the Internet. Erbium-doped fiber amplifiers (EDFA), which coherently amplify 1550-nm signals by converting pump laser light, are a key enabler.1 Because the process is purely optical, all signal wavelengths are amplified simultaneously, making EDFA bandwidth highly scalable at minimal incremental cost.

The bedrock of the EDFA is its pump lasers, which deliver the raw power to regenerate optical signals along the network. Doubling EDFA bandwidth once meant doubling the pump power to launch twice the wavelengths, each with adequate power to traverse the next network span. Now state-of-the-art EDFAs integrate additional complexity and functionality (with associated optical losses) to add and/or drop wavelengths, compensate dispersion, and equalize power across wavelengths, so pump laser demand has outstripped that of the overall fiberoptic market for several years running. Ongoing improvements in 980-nm pump laser power, reliability, and stability will help enable optical networking to continue to boom.

Demand for pump power is met by increasing the total number of pump lasers in the EDFA and increasing the output of each individual device. Additional lasers add to EDFA complexity and cost, however, so there is a strong desire to increase the power a single laser chip can deliver. Three power measures are relevant to 980-nm lasers: rated operating power (Pop), kink- or linear-power (Pkink), and rollover power (see Fig. 1).

The kink in the P-I curve identifies a lateral mode shift of the laser. It is a thermal effect related to the junction temperature and chip design caused by the laser waveguide favoring new modes beyond a certain operating point defined as Pkink. The effect on chip output power linearity is slight, but the kink has a dramatic effect on the fiber-coupled power since the new mode has a different beam geometry. As a result, 980-nm pump lasers are never operated near or above Pkink. Typically, a kink-margin of 10% to 20% is recommended, determined by the customer`s reliability budget, long-term laser chip degradation and the packaging technology. Both JDS Uniphase`s current generation five and future generation six 980-nm technologies are engineered to produce a median Pkink so high that in practice Pop is limited only by the customer`s reliability budget.

The rollover power is another thermal effect denoting the point in a laser`s operating curve where additional bias no longer produces additional power; that is, the temperature driven reduction of device efficiency outweighs the effect of added current. Chip rollover power is independent of laser modes and waveguiding and thus insensitive to chip-to-chip variations, making it a valuable probe of a laser technology`s underlying thermal properties. Since reliability and Pkink both depend critically on temperature, observation of rollover power is highly predictive of the Pop and median Pkink achievable in production.

Reliability improves with each generation

Any discussion of 980-nm Pop must include reliability because EDFA designers rarely pose the question, "How much power can you give me?" Rather they ask, "At 500 FIT, how much power does your technology deliver?" For each JDS Uniphase 980-nm chip technology generation, a basic reliability model is re-confirmed:

Equation 1:

FIT = A × J x × P y × exp(-Ea/kTj )

FIT, shorthand for failures-in-time, is the telecom industry standard measure for component reliability. One FIT equals a single failure per 109 device hours. More intuitively, 1000 FIT is equivalent to about 1% of a device population failing per year in the field. The key variables found to determine 980-nm laser chip reliability are drive current density (J ), operating power (P ), and junction temperature (Tj ). The relevant parameters: A, x, y, and Ea are deduced by fitting to a matrix of lifetest data in which J, P, and T j are systematically varied. The EDFA designers adopt and integrate the laser chip reliability model into their own reliability budgets (typically 1 to 2 kFIT for the whole EDFA module) to determine how hard the pump lasers can be driven.

Each new generation of 980-nm lasers operates at higher power with equal or better reliability through a combination of design and process improvements (which permit the laser to run cooler at a given power) and better screening (which lowers the pre-factor, A). On the other hand, the three key fitting parameters (x = 3, y = 1, Ea = 0.45 eV) have remained remarkably constant since JDS Uniphase produced the first commercially deployed 980-nm pump laser generation in 1993, reflecting the continuity of the underlying technology.

The current generation five 980-nm laser operates reliably at Pop = 300 mW for terrestrial telecom applications and meets more stringent undersea system reliability requirements at an industry-leading Pop >200 mW. The generation six 980-nm laser planned for introduction in 2001 will increase chip operating power by 30% without sacrificing device efficiency. New generations are achieved through continued incremental improvements of the basic technology platform based on the indium gallium aluminum arsenide (InGaAlAs) material system deposited by molecular beam epitaxy with cleaved mirrors passivated by the proprietary E2 process. Generation six design improvements also allow fiber coupling efficiencies reaching 90%. Pump laser electrical-to-optical conversion efficiency is critical since EDFA thermal loads and current requirements have grown as quickly as total bandwidth, and dealing with those issues is a significant engineering problem of its own. In undersea systems, which draw current from remote supplies that are 1000 km or more away, pump laser efficiency rivals Pop in importance. A prototype generation six pump module achieves 400 mW fiber-coupled power with 88% coupling efficiency (see Fig. 2).

Higher performance

The march of fiberoptic communication bandwidth has additional ramifications for pump lasers beyond power since higher performance systems inevitably require more tightly specified components. The EDFA bandwidth is enhanced by increasing the data rate per wavelength or increasing the number of wavelengths. Transitioning from 2.5 to 10 to 40 Gbit/s transmission reduces pulse width and spacing by 75% with each technology generation. Squeezing 16 to 32 to 80 to 160 wavelengths through the erbium amplification window requires closer spacing of the wavelengths. As a result, new systems increasingly will require more stable pump laser wavelength and power outputs. Tighter specifications have lead to the widespread introduction of fiber Bragg gratings (FBG), which lock pump laser performance over a wide operating range.

The FBG, normally written into the packaged pump laser`s fiber pigtail, performs its function by reflecting a narrow wavelength band of light back into the chip. In a proper design, the FBG feedback governs the laser output, stabilizing the output power and fixing the laser wavelength. Of particular importance is that the wavelength locking and output power stability be maintained over as wide a range of operating power and temperature as possible, since newer systems tend to drive pumps variably depending on traffic at a given time. Since it is laborious to fully test a FBG-stabilized laser over the complete possible operating range, wavelength and power specifications must be met by design to forego 100% testing. Proper stabilization has been shown over a wide operating range, demonstrating the robustness of the FBG-stabilization achievable with the generation five 980-nm pump laser (see Fig. 3).

Future improvements in optical networking bandwidth will continue to rely on the ability to transmit more information on a single wavelength, more wavelengths through a single fiber strand, and to more intelligently manage the flow of light through the network. Whether it is upgrading from 10 to 40 Gbit/s transmission, 80 to 160 wavelengths, or all-optical switching, the thirst for pump laser power will continue to grow as fast or faster as the information highway. Fortunately, pump laser designers have every indication our technology will scale to keep pace.

ACKNOWLEDGMENT

The authors wish to thank their colleagues Ulrich Pfeiffer and Sebastian Alt for their contributions to this article.

REFERENCE

1. D. Trivedi, Toby Strite and Gerlas van den Hoven, WDM Solutions, April 2000, p. 14.

And low-frequency power fluctuations are suppressed well below 0.05dB.