Dispersion–management devices optimize amplifier design


By Uri Levy and David Menashe

The noise figure of line amplifiers has a major impact on system performance. High–order mode fiber–based dispersion–management devices offer system designers a simple and cost–effective path to improving the noise figure without major changes to system design.

With the accelerated demand for low–cost broadband service, optimizing the EDFA subsystem becomes key to driving down the cost per bit per kilometer. A major challenge in optically amplified systems is the accumulation of amplified spontaneous emission (ASE), which leads to a degradation in the optical signal–to–noise ratio (OSNR) along the link.

Standard erbium–doped fiber amplifier (EDFA) systems deploy three types of amplifiers in the link—a preamplifier, a line amplifier,; and a power amplifier—that respectively compensate loss at the launching site, at each span site, and at the receiving site. As link lengths have increased, with more spans and thus more line amplifiers, the ASE noise generated at span sites has become a major limitation for system designers.

We present a high–level design for a cost–effective, optimized line–amplifier device based on advanced component technology at the amplification site. The potential improvement in OSNR performance should significantly contribute to maximizing performance and minimizing costs in advanced DWDM systems.


In modern terrestrial transmission systems it is necessary to perform advanced optical–signal conditioning functions at each line amplification site. The most important such function for high bit–rate systems—10 Gbit/s and beyond—is dispersion management. Other functions typically include channel power sensing and equalization. In addition, add/drop operations for individual WDM channels may be performed.

Each of the optical modules that perform these functions introduces loss, resulting in the well–known system–design issue of whether to place the modules before or after the line amplifier. To minimize system noise, the modules should be placed after the amplifier, whereas to maximize pump power utilization¸ they should be placed before the amplifier.

A technologically mature solution is to use a two–stage amplifier, with mid–stage access for the optical conditioning modules. This design calls for a low–noise preamplifier to compensate the loss of the preceding span, followed by a high–power booster amplifier to compensate the loss of the mid–stage modules and launch the desired power into the next span (see Fig. 1).

The system noise introduced at the amplification site is characterized by the total noise figure (NF) of the two–stage amplifier. The NF is defined as the decrease of the SNR at the output of the amplifier, assuming the SNR at the input is shot–noise–limited. The total NF is a function of the NF of each individual stage, as well as the gain of the preamplifier stage and total loss of the mid–stage modules. The higher the gain of the preamplifier stage, and the lower the loss of the mid–stage modules, the lower the total NF will be. Recent advances in pump–laser technology have led to relatively low–cost, high–output–power pumps for the preamplifier stage, enabling the gain of this stage to be increased beyond that found in currently installed systems.

However, the modules used for dispersion management in the mid–stage remain a major limiting factor for system designers. Current dispersion–compensating modules (DCMs) are based on single–mode dispersion–compensating fiber (DCF). These specially designed negative–dispersion fibers typically have a small effective area, on the order of 20 μm2, and thus suffer from stronger nonlinear effects when high signal power is launched into them. This small effective area limits the allowed output power (and therefore gain) of the preamplifier stage.

Another serious limitation is the high loss associated with DCMs—typically 10 dB for modules compensating single–mode fiber (SMF) spans. The power evolution of a single channel in current line amplifier designs, as dictated by these limitations, is shown by the curve labeled "DCM" in Fig. 1. Reducing these limitations results in a significant improvement in the total NF of line amplifiers.

In recent years, a new type of advanced dispersion–management device (DMD) has been introduced to the optics industry.1, 2 This device is still based on a negative–dispersion fiber, but instead of the signal propagating in the fundamental mode, in the specially designed fiber in these devices the signal propagates in a higher–order mode. The fiber is thus called a high–order mode fiber (HOM–F). In operation, mode transformers in the DMD convert the signal from the basic mode of the transmission fiber to a select high–order mode, and back again. The negative dispersion of the HOM–F reverses the pulse–spreading effects of chromatic dispersion created when light propagates through transmission fiber.

One advantage of using high–order mode technology to manage dispersion is the characteristic low loss of the DMDs. Low loss is achieved by incorporating in the DMD a shorter length of HOM–F with much higher negative dispersion per km. Unlike conventional single–mode DCF, achieving high negative dispersion per km does not come at the expense of a decreased effective area.

Plotting module loss versus the compensated span dispersion for a DMD and a conventional DCM is quite revealing (see Fig. 2). At small span dispersion, the DCM exhibits lower loss, due to the added fixed loss of the mode transformers in the DMD. However, at about 400 ps/nm (25 km of SMF at 1550 nm), the situation reverses and the advantage of the DMD becomes apparent as the span dispersion grows. At 1700 ps/nm (100–km SMF), the loss difference between a DCM and HOM–based DMD is approximately 6 dB.

Without exception, the magnitude of all nonlinear effects (NLEs) is proportional to the ratio of the effective length (Leff) and the effective area (Aeff) of the fiber. Therefore, to minimize the noise or pulse distortion that result from nonlinearities requires short effective length and large effective area.

Low sensitivity to NLEs → small Leff / Aeff

Short effective length and large effective area imply low sensitivity to NLE generation,; which is precisely the case with HOM–F–based DMDs. The use of a high–dispersion fiber implies a short effective length, and propagating in a high–order mode results in a large mode effective area. Compared to regular DCF, the HOM–F effective length is up to four times shorter and the HOM effective area is up to four times larger, so the sensitivity of the HOM–F to NLE generation is up to 4 × 4 = 16 = 12 dB times smaller.

For example, it is interesting to compare the sensitivity of a DCF–based DCM and a HOM–F–based DMD to one type of nonlinear effect—stimulated Brillouin scattering (SBS). The DCM exhibits an SBS threshold at approximately 7 dBm, but the DMD is not affected by SBS until at least 20 dBm—an advantage of 13 dB or more (see Fig. 3). Other tests of monitoring system Q versus input power show at least a 10–dB advantage of the HOM–F–based DMD over the DCF–based DCM.

The high power tolerance and low loss of HOM–F–based DMDs open up new possibilities in the design of line amplifiers, as illustrated by the curve labeled "DMD" in Fig. 1. Compared to the "DCM" case, the output power of the preamplifier stage has been increased by 6 dB,; well within the margin allowed by the improved power tolerance of the DMD. Furthermore, the low loss of the DMD means less gain is needed from the booster amplifier. These factors combine to produce a significant decrease in the total NF of the amplifier.


The total NF of a line amplifier is mainly determined by NFpre, the noise figure of the preamplifier stage. However, the noise figure of the booster amplifier NFboost, usually larger than NFpre, also impacts the total NF. Mathematically, this is written as:

Here all quantities are linear (not dB)—Gpre refers to the gain of the preamplifier, and Tmid refers to the transmittance of all mid–stage modules. The equation illustrates what was stated previously: the larger Gpre, and the larger Tmid(the smaller the mid–stage loss), the smaller the contribution of NFboost, and the smaller the total NF.

The following specific design example illustrates the previous equation, and demonstrates the NF improvement that can be expected using an advanced, high–power, low–loss DMD. The details of the example are summarized in the table, with the free design variables being the loss of the dispersion–management device and the output power of the preamplifier. The design assumes a span loss of 25 dB.

The NF of the amplifier can be plotted as a function of the total output power of the preamplifier and the loss of the DMD (see Fig. 4). The results clearly illustrate the NF advantage of combining high preamplifier output power with low loss of the DMD.

Note that the assumption that NFpreand NFboost are fixed at 4 dB is not strictly valid. Typically, increasing Gpre and Tmidwill lead to decreased NFpreand increased NFboost. However, these changes are consistent with the general trend shown in Fig. 4, and in some cases can make this trend more pronounced.

For a conventional DCM based on single–mode DCF, with a loss of 10 dB and an input power limited to –4 dBm per channel (+14 dBm total power), the total NF will equal 5.5 dB. If instead a DMD with a loss of 6 dB that tolerates an input power of +2 dBm per channel without any additional nonlinear penalty is used, then the total NF will decrease to 4.2 dB. This constitutes a significant improvement in the overall OSNR performance of the system, allowing a greater degree of design freedom.


The reduction in NF of the line amplifiers may be directly translated into increased system reach. Assuming a system limited only by OSNR, the relation between system reach and line–amplifier NF is easily plotted (see Fig. 5). Based on the preceding example, replacing a conventional DCM with a DMD allows an increase of approximately 40% in system reach.

Furthermore, additional routing functionality, such as add/drop or optical crossconnects, can be added to the system. In fact, the lower loss of advanced DMDs directly allows designers to increase the mid–stage loss allocated to routing modules.


In reality, modern transmission systems are limited not only by OSNR performance, but also by nonlinear effects in the transmission fiber. Thus, although the improvement in OSNR described previously is significant, system designers still face the limitation of nonlinear effects. Recently, much effort has gone into developing new transmission fibers that reduce nonlinear effects.3, 4, 5 These fibers typically have a large effective area—larger than 100 Μm2—that allows more power to be launched into the fiber with fewer nonlinear effects.

Once these new transmission fibers are deployed, system designers will need to increase the power level throughout the system to take full advantage of these fibers. Clearly, power levels within the amplifier will also need to be increased, resulting in the need for higher power tolerance in amplifier components and dispersion–management modules.

Uri Levy is vice president of optical networking and David Menashe is a research physicist at LaserComm, 2600 Technology Dr., Suite 900, Plano, TX 75074. Uri Levy can be reached at ulevy@lasercomm–inc.com.


1. A. H. Gnauck et al., Electron. Lett. 36, 1946 (2000).
2. S. Ramachandran et al., IEEE Photon. Tech. Lett. 6, 632 (2001).
3. K. Tanaka et al., ECOC 2001, Mo.F.3.6, Amsterdam 2001.
4. M. Vasilev et al., OFC 2002, THQ4, Anaheim 2002.
5. H. Sugahara et al., OFC 2002, FC6, Anaheim 2002.

More in Home