Intelligent DWDM fiber amps without dynamic GFF a smart choice

Th 121712

Everyone seems to agree that intelligent inline fiber amplifiers, able to adapt to different input-channel distributions, will be necessary to implement a wavelength-agile optical communications network. It also appears that a consensus has been reached about using dynamically adaptable gain-flattening filters (GFFs) to create such amplifiers. However, the use of a passive GFF combined with a fiber-amplifier configuration offering a constant spectral gain profile over a large dynamic range might be a much more economical, robust, and efficient way to design intelligent fiber amplifiers.

Erbium-doped fiber amplifiers (EDFAs) are the main enabler of DWDM optical communications, since they can simultaneously amplify communication channels ranging from 1530 to 1600 nm. However, there are technical considerations associated with using these EDFAs for an increasing number of optical channels over an increasing bandwidth.

One of these considerations is that the spectral gain profile of EDFAs is not uniform. Thus, some channels get more gain than others. This distortion increases as the number of concatenated fiber amplifiers over a transmission link adds up. The result is a severe restriction of available bandwidth from the transmission link. This phenomenon has been known for quite a while, as has the solution to this problem: gain-flattening filters.1 Passive GFFs inserted in the middle of a dual-stage fiber amplifier have proved to allow low noise figure amplification with good performance and spectral uniformity within 0.5 dB for the C- and L-bands.

The best-known passive GFF technologies are the tilted fiber Bragg gratings, thin-film multilayer filters, and short- and long-period fiber gratings. These filter technologies are quite sophisticated, since the spectral gain profile of EDFAs is quite complex and thus requires a complex spectral response from the flattening filter.Th 121712

Figure 1. This schematic representation of an optical crossconnect, including wavelength conversion, illustrates the complexities that will challenge future optical amplifiers.

For a point-to-point optical communications system, when no reconfiguration is anticipated, there are no more questions to be asked; the system offers good DWDM transmission performance. But if a reconfiguration involves changing the amount of total input power into the optical amplifier, as is implied by changing the number of DWDM channels, then the spectral gain profile of the EDFA changes and the passive GFF efficiency is greatly affected.

That is just the beginning of the story, since point-to-point DWDM communications systems are bound to evolve into wavelength-agile optical networks. Indeed, systems that include wavelength reuse, wavelength-selective all-optical DWDM channel switching, and eventually even all-optical packet switching will be introduced in the future (see Figure 1). That means a very large and unpredictable DWDM channel input distribution variation into EDFAs. It also means a very large and unpredictable spectral gain profile variation within the EDFAs. And ultimately, it means that passive GFFs are not efficient in such wavelength-agile systems.

The solution to such a problem would appear to be quite straightforward: a dynamically adaptable gain-flattening filter. A lot of effort has been expended in developing such active GFFs, which would be differentially activated by monitoring the total input power coming into an EDFA. Some startups have even been created around such a promising technology.

Different approaches exist toward offering dynamically adaptable gain-flattening filtering, including acousto-optic filtering. But as mentioned earlier, these filters are associated with complex spectral responses. This spectral complexity is obviously a larger challenge to implement when it is variable. Thus, spectral uniformity within 1 dB for the C- and L-bands is difficult to obtain over a 10-dB input power dynamic range incident on the actively gain-flattened EDFA. Moreover, these components being complex, they are also quite costly, potentially bulky, and certainly power hungry.

Is there another possible solution other than the "straightforward" approach? The answer is yes. If a passive GFF does not work because of the changing spectral gain profile of the EDFA, why not change the performance of the EDFA to obtain a constant gain spectrum even for variable DWDM input-channel distribution? That is quite possible to do according to fiber-amplifier theory. By keeping the ratio of pump residual to injected pump power in the amplifier cavity constant, the spectral gain profile of the EDFA is conserved notwithstanding the input-channel distribution injected in the EDFA.2Th 121713

Figure 2. The gain-locking erbium-doped fiber amplifier provides good spectral gain uniformity over the C- and L-bands.

One way to implement the constant pump residual to injected pump ratio is to use a simple WDM fiber coupler at the output of the EDFA cavity associated with a simple photodetector. Implementation of a feedback loop between this photodetector and the laser diode pump drive current controller provides a practical approach to the issue of keeping the ratio constant.

To make this configuration into the optimal solution, it has to be implemented in a dual-stage fiber amplifier where the passive GFF is inserted in mid-stage. Figure 2 shows the amplifier configuration that enables this functionality and the resultant performances.3,4 As long as the pump has a very pure spectral emission with very limited side lobes such as a grating stabilized pump, this configuration works efficiently. Spectral gain uniformity under 0.5 dB over the C- and L-bands for an input power dynamic range of 15 dB has been proven using this configuration. A 15-dB dynamic range is the equivalent of dropping 31 DWDM channels out of 32 at the input of the EDFA.

Optical amplifiers using this gain-locking EDFA configuration are considered intelligent because they can adapt to different input-signal distributions. The fact that the gain is locked on a uniform spectral profile because of the passive gain-equalizing filter only adds to the quality of the EDFA. Equalized DWDM gain per channel over 25 dB can be obtained with these gain-locking EDFAs.

The noise figure performance of the gain-locking amplifier is ensured by the dual-stage configuration. The amplifier cavity from which the residual pump power is measured does not have to be made shorter than for optimal performances since the normal residual pump power associated with an optimal erbium-doped fiber length is large enough to ensure good detection with a fine signal-to-noise ratio. Th 121714

Figure 3. The gain-locking intelligent erbium-doped fiber amplifier has demonstrated response times on the order of 100 msec, making it useful for a variety of applications.

The result is good performance at the lowest possible price using the cheapest off-the-shelf optical components and standard electronics; the fiber coupler, photodetector, and feedback loop circuit would cost less than $200. The power consumption is that of a standard electronic feedback loop, much less than what is normally associated with dynamic gain equalizers (DGEs). The use of a passive gain-flattening filter offers performances that cannot be equaled by any dynamically adjustable GFFs. The end result is a low-cost complete solution for intelligent EDFAs.

Now, before it looks too good to be true, we have to consider the possible disadvantages of this configuration. The main one, perhaps the only one, is that some pump power is wasted when a total input power under the maximum rated value is incident on the gain-locking EDFA. But this is not much of a disadvantage since instead of wasting signal power with a DGE—chopping off more of the surplus power—the pump drive current is being down-rated, which might even have a positive impact on the laser-diode pump lifetime.

Response times on the order of 100 msec have been demonstrated using this gain-locking method (see Figure 3). Such speed was obtained for sudden changes as large as 10 dB, which is equivalent to dropping simultaneously more than seven DWDM channels out of eight. In this case, the speed is limited by the electronics in the feedback loop. As for the EDFA itself, its response time is nearly instantaneous since we are talking of slight adaptations in the pumping level in the erbium-doped fiber. These speeds compare very well with what is offered by dynamically adjustable GFFs; thus, gain-locking EDFAs can be readily applied to networks that use all-optical switching at rates >100 msec for reconfiguration or wavelength-routing purposes.

But thinking ahead, we have to consider all-optical packet switching. In this case, response times on the order of a few nanoseconds, if not less, need to be implemented. That is a big challenge, but we believe it is feasible using the gain-locking approach. Again, the burden is on the feedback loop and its associated components. To obtain such high-speed performances, state-of-art electronics would become mandatory. It would also probably mean changing the packaging of the pump laser diode to ensure it can perform well at such high modulation speeds.

Such an application would present challenges, but since similar problems have been addressed in the field of optics already for transmitters and receivers, it seems these challenges can be addressed. The performance of the gain-locking, gain-flattening intelligent EDFA remains to be proven using such a high-speed feedback-loop circuitry, but there is time for such experiments—all-optical packet switching will not be required tomorrow....

The gain-locking, gain-flattening intelligent EDFA uses a passive gain-flattening filter combined with a fiber-amplifier configuration to offer a constant spectral gain profile over a large dynamic range. Such an approach might be a much more economical, robust, and efficient way to design intelligent fiber amplifiers, compared to the dynamically adjustable gain-equalizer approach.

With the gain-locking EDFA, DWDM optical communications networks offering reconfiguration and wavelength agility or routing can be readily implemented. This configuration might even allow all-optical packet switching to be implemented in the future.

Jocelyn Lauzon is director of photonics, fibers, and lasers at INO (Ste-Foy, Quebec). He can be reached at

  1. R. Wyatt, "Optical amplifiers for WDM," Optical Amplifiers and Their Applications conference technical digest, Vol. 14, July 1994, p. 110.
  2. J.C. van der Plaats, F.W. Willems, and A.M.J. Koonen, "Dynamic pump-loss controlled gain-locking system for erbium-doped fiber amplifiers in multiwavelength networks," ECOC conference publication #448, September 1997, p. 127.
  3. M. Bégin, J. Lauzon, Y. Rouleau, and Y. Mimeault, "Gain-locked dual-stage EDFA for WDM systems," Optical Amplifiers and Their Applications conference technical digest, July 2000, p. 148.
  4. M. Bégin, J. Lauzon, Y. Mimeault, Y. Rouleau, "Gain-locked dual-stage optical amplifier," U.S. patent #6,366,394, April 2002.
More in Transport