Gerlas van den Hoven and John Wachsman
Single-chip, semiconductor amplifiers offer cost-effective alternatives to more-complex erbium-doped fiber amplifiers in metropolitan DWDM systems. Indium phosphide-based linear optical amplifiers, in particular, can enable network designs that are dynamic, flexible, and scalable while reducing costs.
Optical networks are migrating closer to the end user. New optical functions, such as adding and dropping channels, routing, and switching, will be needed to enable processing of optical signals. Unlike in long-distance systems, processing loss will dominate fiber-span loss in metro networks.
Although the loss mechanism is different, optical amplifiers still will be needed to compensate for processing losses encountered by the optical signal. This loss, coupled with the dynamic nature and economics of metro networks, imposes new requirements on optical amplifiers.
As fiberoptic networks become more dynamic and migrate closer to the end user, optical amplifiers need to meet a different set of requirements than in traditional long-haul systems. Market drivers and corresponding amplifier trends for long-haul differ from those in metro/access networks (see Table 1).
Long-haul optical networks have historically been static, point-to-point architectures that focus on increasing capacity. This primary design goal translated into the key optical-amplifier performance parameters of high output power, high gain, low noise figure, and crosstalk immunity. The erbium-doped fiber amplifier (EDFA) has met these criteria and enabled the rapid growth in data traffic.
As optical networks move closer to the end user, there are many more nodes, with less traffic per node, dramatically altering the drivers for metropolitan network design. More nodes mean that the network must be dynamic and reconfigurable. This dynamic processing is being pushed into the optical layer. It also means that the network must be scalable, with an ability to add nodes as well as to increase capacity.Optical amplification is a key enabler of both optical processing and scalability in the metro network. Less traffic per node means that cost, not performance, becomes the primary concern. Carriers continue to demand higher port densities and better power efficiencies, which translates directly into optical components of smaller size that consume less power. The lower-cost, smaller size, decreased-power paradigm was best solved in the electronic world through integration, which will also play a key role in the future of optical networks.
In addition to these new requirements, metro/access optical amplifiers still need to meet basic performance in terms of gain, noise figure, output power, and linearity.
MANY AMPLIFIER FUNCTIONSThe traditional applications of optical amplifiers in long-haul fiberoptic networks include three distinct functions: booster amplification, in-line amplification, and preamplification (see Fig. 1). These functions are characteristic of point-to-point links and they focus on transporting the optical signal over a given distance.Key parameters for long-haul amplifiers are high output powers in both booster and in-line amplifiers to handle high channel counts in the long haul (see Table 2). Preamplifiers are characterized by low noise figure to increase the receiver sensitivity at the end of the link.
Next-generation metro and access networks will be ring or mesh formats. Similar to the long-haul amplifiers, the metro booster amplifier raises the added signals to a power level equal to that of channels already on the ring (see Fig. 2). The metro preamplifier raises the dropped signals to levels that can be easily received. The in-line metro amplifier strengthens express signals to overcome the loss encountered during transport to the next amplifier.Utilizing this design approach, every segment—a node plus the following fiber span—has "unity gain." This unity-gain segment paradigm enables the network to be modular and scalable, regardless of what functionalities are present in each node. The ring or mesh can be upgraded without detailed knowledge of the entire metro network.
METRO RING EXAMPLE
The components that enable fixed optical add/drop functionality in a metro ring will induce loss in the network. Designers can calculate typical performance characteristics of these components used in metro ring systems (see Table 3).
Without amplification, the maximum allowable loss in a 10-Gbit/s all-optical network is given by the difference between the launch power and the receiver sensitivity:
Allowable loss = Tx - Rsensitivity
= 0 - (- 16)
= 16 dB
A typical DWDM metropolitan ring to consider would be 200 km in circumference, with four intermediate fixed optical add/drop multiplexers (OADMs; see Fig. 3). Fixed optical add/drop multiplexers can add or drop a fixed wavelength at a particular location. This is the simplest form of processing in the optical layer. The total node loss (including two splices and six connectors) is given by
Lnode = 2 x Lsplice + 6 x Lconn + Lfoadm
= 2 x 0.15 + 6 x 0.25 + 2
= 3.8 dB
Assuming a launch power of 3 dBm per channel (P/ch) prior to the first span, the amount of amplification needed in the ring is given by
Gain = P/ch - 4 x Lspan - 4 x Lnode
- Rsens - link margin
= 3 - 50 - 15.2 - (-16) - 3
= 49.2 dB
This amount of gain could be implemented with two high-gain (25-dB), high-performance amplifiers. Implementation would require an amplifier with an output power per channel in excess of +12 dBm or a total output power of +22 dBm for ten channels.
If the optical amplification approach utilizes the unity-gain segment approach— one amplifier per node—the performance of each amplifier is simplified.
Each amplifier would match the loss of one node and the span preceding it. This translates into an amplifier with modest gain of 16.3 dB and the P/ch output power would be only 3 dBm or a total output power of +13 dBm for ten channels. This approach is only economically feasible with a low-cost, compact optical amplifier.
The limited flexibility of a fixed OADM will inhibit the efficiency and service velocity of the optical network. Carriers will ultimately deploy reconfigurable OADMs. These types of nodes enable any wavelength to be added or dropped at any node. This added flexibility comes with a cost: 6 to 10 dB of additional "processing" loss.It is possible to keep a similar philosophy as the previous example. In this case, two amplifiers per node would be used. The first amplifier would make up the loss of the reconfigurable node, while the second amplifier would make up the loss of the fiber span. In this example, the amplifiers would have gains in the range of 12 to 13 dB and the required output power per channel would again be limited to 3 dBm per channel.
Optical amplifiers will enable functionality and scalability in future metro networks, but are also required to upgrade even the simplest metro links. There are many 2.5-Gbit/s amplifier-less links that must be upgraded to 10 Gbit/s, but do not have enough link margin available (see "Seeking simple migration paths," p. 54).
COMPARING AMPLIFIERS
In the examples given, it is clear that the amplifier requirements for metro/access systems, in terms of performance, are different from the long-haul amplifier characteristics discussed in Table 2. Typically, gain values are much lower, and power requirements are lower since fewer channels need to be amplified.
The performance of different amplifier technologies to the target requirements of the metro networks can be compared and analyzed (see Table 4). Metro EDFAs, as expected, match well with the traditional optical-performance parameters of gain, noise figure, and output power. A mature technology, the size of metro EDFAs will remain large by comparison and do not have the cost and integration potential required by metro/access. As metro networks become more dynamic, transient suppression circuitry will need to be added, increasing complexity, size, and cost.
Erbium-doped waveguide amplifiers (EDWAs) address the size and integration issues in part. The glass waveguide is less bulky than fiber and has the potential to be further integrated with planar lightwave circuits. This technology still requires external pump lasers and sacrifices both gain and output power performance to achieve the smaller size. Transient suppression circuitry, similar to the metro EDFA, will ultimately be needed.Semiconductor optical amplifiers (SOAs), based on indium phosphide chip technology, meet the size, cost, and integration requirements, but at the expense of significant performance degradation in linearity. The linearity issues, intersymbol interference, WDM crosstalk, and switching transients will limit their application in metro networks.
The linear optical amplifier (LOA), also an indium-phosphide-based chip amplifier, addresses these issues by linearizing the gain of the amplifier.1 By combining the advantages offered by single-chip technology with the amplification performance required in metro/access applications, the LOA is well-suited to enable metro and access networks.
The LOA technology's long-term cost, size, and integration capability advantages over erbium-doped amplifier solutions will enable metro architectures to use the concept of "unity gain" segments, making the networks modular. In addition, metro and access optical networks will be dynamic and reconfigurable, adding the new requirement that optical amplifiers not generate switching transients. The single-chip LOA meets these requirements as well.
REFERENCE
- D. A. Francis et al., OFC 2001, post-deadline paper 13.
______________________________________
Seeking simple migration paths
One of the workhorses in optical interconnection is the single-channel 80-km link connecting one central office to the next (see Fig. A). The performance of this link, including critical loss elements, is governed by the power budget, which can be determined from the component specifications as presented in Table 3. The maximum span loss, Lspan, that can be supported, is given by
This formula assumes a link having eight connectors, two splices, and a power monitor in addition to the transmitter and receiver. A 3-dB link margin is included to capture dispersion penalties due to the fiber and the directly modulated transmitter laser.
For a 2.5-Gbit/s link without the amplifier, inserting numbers from the table gives a maximum fiber span loss of
Lspan = 3 - 2 - 0.4 - 0.2 - (-23) - 3
= 20.4 dB
The lower Tx power (3 dB) and reduced receiver sensitivity (6 dB) for 10-Gbit/s components means that the link cannot be upgraded without an optical amplifier. By placing an optical amplifier with 10-dB gain and an output power of 10 dBm at the transmitter, the link can be realized. The span loss that can be accommodated with the addition of the amplifier is
Lspan = Tx + G - 8 x Lconn -
2 x Lsplice - Lmon - Rxsens
- link margin
Lspan = 0 + 10 - 2 - 0.4 - 0.2 - (-16) - 3 = 20.4 dB
The performance of linear optical amplifiers (LOAs), semiconductor optical amplifiers (SOAs), and EDFA amplifiers in such a practical interconnect system can be charted (see Fig. B). Compared to the back-to-back measurement, both the EDFA and LOA transmit the signal over 80 km, with approximately 1-dB penalty, which is expected from the dispersion induced in the link by the electroabsorption-modulated signal transmitter laser. For the SOA, however, nonlinear response leads to significant distortion of the signal, inhibiting transmission.