Even though most systems have yet to fill the C-band (1520 nm to 1570 nm) and L-band (1560 nm to 1610), researchers are providing yet another transmission option by moving into the S-band (1450 to 1510 nm) with thulium-doped fiber amplifiers (TDFAs), as well as single-point Raman amplifiers (see oemagazine, March 2001, page 54). The technology is advancing with researchers at the Optical Fiber Communication (OFC) Conference (Anaheim, CA; 1822 March), reporting TDFA conversion efficiencies as high as 48% and materials advances that may render the technology more practical.
Thulium-doped devices achieve gain from the 3H43F4 transition of thulium ions (Tm3+). Because the lifetime of the upper level is significantly shorter than that of the lower level (1.7 ms vs. 11 ms), it is difficult to form the population inversion necessary for gain. Researchers have solved this problem using various forms of upconversion pumping in which Tm3+ ions are first excited to the 3F4 level by pumping at 1200 nm or 1550 to 1650 nm, then subsequently pumping to the 3H4 level by excited state absorption using 1050 nm or 1380 to 1410 nm wavelengths. Upconversion pumping also can be performed by a single pump laser operating at 1050 nm, though it is somewhat less efficient.
The most common host for TDFAs is fluorozirconate (ZBLAN) fiber, which has a glass transition temperature about 700° C less than that of silica fiber, a disparity that prevents the use of fusion splicing to connect the material with conventional transmission fiber. In addition, the material is hygroscopic and forms microcrystallites over time.
To provide an alternative, Brian Cole and collaborators at the Naval Research Laboratory (Washington, D.C.) successfully doped silica fiber with Tm3+, using 3.7 m of the fiber to demonstrate a TDFA (paper #TuQ3). The group pumped the fiber with 750 mW of 1047 nm output from an Nd:YLF laser, and up to 1100 mW at 1410 nm from a fiber Raman laser, using both single- and dual-wavelength pump schemes. For the dual-wavelength approach, they achieved small-signal gains as high as 12 dB with a maximum pump-to-signal conversion efficiency of 12.5% and a 28% slope efficiency. Output coupling losses of about 0.7 dB suggested that the actual efficiencies may be twice those values.
Standard TDFAs produce gain from 1450 to 1480 nm. Researchers also are focusing on shifting the gain band of TDFAs to between 1480 and 1510 nm. The key to gain-shifting is maintaining a low population inversion between the 3H4 and 3F4 levels, and it can be accomplished by either increasing the Tm3+ doping levels or by using a dual-wavelength pumping scheme.
In the case of increased doping levels, the energy from a Tm3+ ion excited to the 3H4 level transfers its energy to a ground state Tm3+ ion by cross relaxation, which lowers the population inversion between 3H4 and 3F4. Shin-ichi Aozasa and colleagues at NTT Photonics Laboratories (Ibaraki, Japan) pumped a ZBLAN fiber doped to 6000 ppm with 500 mW of output from diode laser pump sources around 1400 nm to obtain a power conversion efficiency of 42% (paper #PD1). The double-pass system consisted of a 3 m- first stage and a 5-m second stage.
Meanwhile, Fabien Roy and colleagues at Alcatel Research and Innovation (Marcoussis, France) used a dual- wavelength fiber Raman laser (1240 nm, 1400 nm) with Tm3+-doped ZBLAN to achieve 48% power conversion efficiency (paper #PD2). The 1240-nm pump wavelength populates the 3H5 level, which populates the 3F4 level via non-radiative decay. The subsequent absorption of a 1400 nm photon bumps the Tm3+ to the 3H4 level. The group used a 15-m fiber doped to 2000 ppm in a single-stage configuration.
The various projects sound promising, though not immediately practical. That's all right, according to Jeff Livas, CTO for the transport division of Ciena Corp. (Linthicum, MD), since it's unlikely that the S-band will be required any time soon. "You can up the spectral efficiency in the C-band quite a bit before you have to think about any other band," he observes.
There may be other technical problems with the S-band. Dispersion-shifted fiber (DSF) features a zero-dispersion wavelength λ0 centered at 1550 nm. With the advent of dense wavelength division multiplexing, DSF created problems because dispersion properties varied as a function of wavelength across the C-band. As a result, today's nonzero dispersion shifted fiber (NZDSF) features a λ0 shifted to just outside the C-band.
Unfortunately, that can place it right in the S-band. "To the extent that DSF isn't good for the C-band, NZDSF isn't good for the S-band," Livas says. "It doesn't mean you can't use it, it just means it's trickier." It's likely that researchers will find a way around these difficulties, but for now, they have their work cut out for them.