Scalable EDFAs simplify metro design


By J.J. Pan, Joy Jiang, Xiangdong Qiu, Kejian Guan

Elimination of gain–flattening filters and pump–laser coolers makes the narrowband EDFA practical as an amplification solution for metro applications. With fewer components than the wideband EDFA, the narrowband EDFA offers a more compact, cost–effective design.

Using a long–haul amplification system in a metro network is a little like installing a Ferrari engine in the family station wagon—more performance than you need at a prohibitive price. The "Ferrari" synonym in this case is the wideband erbium–doped fiber amplifier (EDFA)—the dominant amplification technology used in long–haul optical networks. The high price tag of high–performance wideband EDFAs is due to the use of high–power pump lasers, gain–flattening filters (GFFs), and complicated electronic controls. Wideband EDFAs are not scalable, forcing service providers to expend 100% of the total amplification cost at deployment.

In long–haul networks, this up–front pricing is acceptable because of the critical need for high output power and high gain and because the cost is distributed over many channels. In the metro world, however, service providers can rarely afford to cover the entire C–band upon deployment, primarily because they would have to wait three to five years for payback on a new system. Metro amplification solutions must provide a low initial cost of ownership while enabling carriers to implement per demand, pay–as–you–grow plans. Complicating matters, metro rings have many more add/drop nodes than long–haul networks, which contribute heavily to signal loss and necessitate amplification despite the short transmission distances. Customers frequently add and move these nodes, requiring metro networks to be more flexible and expandable than long–haul networks.

A banded DWDM structure that addresses these cost and scalability issues is not a new idea. System designers have long proposed various banded channel plans for metro networks. What is new is the appearance of low–cost narrowband amplifiers on the market. Unlike wideband EDFAs that cover the entire C–band and allow for as many as eighty 50–GHz channels, narrowband EDFAs typically amplify from one to eight channels. With fewer channels to amplify, uncooled pump lasers with lower power requirements can replace high–power lasers and GFFs, significantly reducing amplifier costs.

To minimize system costs, however, these metro EDFAs are not merely scaled–down versions of their long–haul cousins. Cascaded narrowband EDFA structures create a scalable network with respect to amplification. This allows service providers to deploy only the capacity they need initially, with the option of adding more channels over time as their revenues and usage increase.


A comparison of the power budget of narrowband and wideband EDFAs helps to understand the benefits of narrowband. For this comparison, consider an 80–km metro ring structure with 32 channels, 100–GHz spacing in the C–band, and 50% add/drop functionality. Four add/drop nodes are evenly distributed on the ring, each allowing up to four channels to be dropped and added. To obtain the minimum node–bypass loss, we will adopt a typical 4–skip–1 channel group arrangement using a wideband filter to reduce the express channel loss. We will use a typical fiber with an insertion loss of 0.25 dB/km, dispersion of 17 ps/km/nm, and a system bit rate of 10 Gbit/s (see Table 1).

This wideband configuration has eight 4–channel groups, in which ITU standards 41 through 59 define add/drop functionality (see Fig. 1, top). The base multiplexer and demultiplexer module consists of eight 4–skip–1 wideband cascaded mux/demux forming eight upgrade ports, with each connected to a four–channel 100–GHz mux/demux. The basic setup includes the base wideband modules, a wideband EDFA booster, an in–line amplifier, a preamplifier, and muxes/demuxes covering at least two bands—one for end–to–end transmission and the other for the add/drop node. A variable optical attenuator (VOA) at each add–channel before the booster EDFA balances the optical power among channels, and thereby balances the amplified output (see Table 2).

With a power penalty of 1.6 dB being the worst case, the results show an optical power margin at the receiver end of approximately 10 dB. Note that narrowband EDFAs are used at the add/drop nodes either for the purpose of balancing the add power with the power of the bypass channels or to increase receiver tolerance. Balanced power throughout the link reduces the noise figure and achieves flat signal spectrums. However, a complicated and expensive control scheme is required for transient control.

The channel plan for the narrowband configuration is the same as for the wideband approach (see Fig. 1, bottom). One narrowband EDFA covers two 4–channel groups and is connected to a 100–GHz 9–skip–1 wideband mux/demux to form upgrade ports for other bands. An in–line amplifier is not necessary in this scheme. The eight channels that are connected to each EDFA are divided into two groups, which form a base and upgrade structure. As in the wideband approach, the mux/demux is a four–channel model with a VOA installed at each add channel for power adjustment. At the add/drop nodes, two nodes use preamplifiers while the others use booster amplifiers, depending on each node's distance from the network ends (see Table 3).

Again, with a power penalty of 1.6 dB being the worst case, the results for the narrowband scheme show that the receiver's dynamic range is around 7 dB. The optical power of each channel is not balanced in the link or at the add/drop nodes. Moreover, appropriate adjustment of the unevenness among channels entering the same amplifier addresses the unequal EDFA gain profile and optimizes the output with optimum gain and a minimal noise figure. A wideband EDFA cannot make a similar adjustment because the gain across the whole C–band varies widely compared to only moderate gain ripples over a narrowband amplifier. Unbalanced input signals can degrade wideband amplifier performance.

Narrowband EDFAs eliminate the need for a GFF because channel–power balancing among the bands is not required. Instead, at the transmitter site we preset the output power from each DWDM module at –9 dBm. This eliminates both the need for monitoring other band outputs and the need to use a performance monitor.


A narrowband EDFA design for metro not only results in a scalable and flexible structure but also saves money on amplification. The wideband network configuration, in comparison, requires the following modules (assuming approximate prices): one wideband booster EDFA ($8000), one in–line EDFA ($7000), one preamplifier EDFA ($8000), and four narrowband EDFAs ($1800 each). The total amplification cost of a fully provisioned system is

$8000 + $7000 + $8000 + ($1800 × 4) = $30,200

The link does allow for a system with as few as eight channels. Four channels are dedicated to add/drop functionality, which means that at least one narrowband EDFA is required. The initial investment in amplifiers for an eight–channel system is

$8000 + $7000 + $8000 + $1800= $24,800

The cost difference between a fully provisioned system and an eight–channel system is only $5400. Although component costs may be reduced in the future, the ratio between the cost of an eight–channel system and the price of a fully provisioned system will remain about the same.

Using a banded approach, only narrowband EDFAs are used throughout the network. The total cost of amplification for a fully provisioned system is

$1800 × 12 = $21,600

However, since the amplifiers are scalable, for the same eight–channel system mentioned above the cost of amplification is only

$1800 × 5 = $9000

The difference between the two systems is $12,600, and only 43% of the full cost derives from the amplifiers. A comparison between the wideband and narrowband structure explains this significant difference in cost and scalability.

First, while narrowband EDFAs only amplify a few channels and have a relatively flat gain profile, wideband EDFAs have a wide operating wavelength range, necessitating the use of expensive, customized GFFs. Secondly, wideband EDFAs use high–power pump lasers to produce enough gain to amplify a large number of channels. Because the power emitted from the pump chip is more than 300 mW, thermoelectric coolers are needed to stabilize the output power and wavelength, which increases the cost of the EDFA.

Furthermore, the major failure mode of a pump laser is facet failure, which is related directly to high output power. Consequently, the durability of these lasers suffers. Low manufacturing yield is another concern; suppliers place yield in the single digits, which means high costs.

A narrowband EDFA requires only a single–stage pump, producing as little as 80 mW of power (see Fig. 2), eliminating the need for a thermoelectric cooler and its complicated control circuits. The lifetime of the pump chip is also extended because of its low pump output power. More important, its relaxed power requirements greatly improve the manufacturing yield, which is comparable to that of 1310/1550–nm laser–diode chips.

An eight–channel narrowband EDFA may drop seven channels at most, which translates into about a 9–dB dynamic range of transient control. A 40–channel wideband EDFA, may drop up to 39 channels and have a transient control range of more than 16 dB. Narrowband EDFAs, as a result, have far simpler control circuits, reducing the overall cost. Finally, the narrowband EDFA has a simpler structure and fewer components than the wideband EDFA, which allows a more compact design that enables integration with other optical components to reduce cost.

J.J. Pan is CEO and president, and Joy Jiang, Xiangdong Qiu, and Kejian Guan are engineers at Lightwaves2020, 1323 Great Mall Dr., Milpitas, CA 95035. J. J. Pan can be reached at

More in Transport