Optical amplification is migrating from undersea and long haul roles to the power-hungry metro networks. But as network needs exceed the capabilities of EDFAs, Raman amplification is looking increasingly attractive.
By Bertrand Clesca Corvis
Optical amplification has enabled the widespread deployment of wavelength division multiplexing (WDM) optical systems, because it allows a single device to simultaneously amplify all channels - whatever the bit rate or protocol.
Although demonstrated by 1962, the erbium doped fibre amplifier (EDFA) enjoyed a research and development renaissance in the late 1980s. Simultaneously, other options for optical amplification were also investigated by this time, namely Brillouin and Raman amplification.
By 1993, EDFAs were introduced in commercial products because they offered the best performance/cost trade-off for the network requirements at that time.
In the late 1990s, the constantly increasing demand for optical network reach and capacity began to test the limits of EDFA technology and provided the impetus to reconsider the performance/cost trade-off of Raman and other amplification technologies.
EDFAs contain a few tens of metres of special fibre doped with erbium (Fig. 1).
A 1480nm or 980nm pump laser excites the erbium ions in these few metres of fibre where optical amplification takes place. EDFAs boost signal wavelengths in the 1530-1560nm range (the conventional or "C" band). EDFAs can also be designed to amplify signals in the 1560-1610nm range (the long or "L" band). Within each amplification band, numerous optical channels at 10Gbit/s can be packed with channel spacing as close as 25GHz (0.2nm) in advanced product designs.
EDFAs are typically used to provide between 17 and 27dB (or, in linear scale, between 50 and 500) of gain, with a total output power around +20dBm (100mW).
EDFAs have performed fairly well in single-channel and traditional point-to-point WDM systems for nearly a decade. However, as performance and capacity pressures increased, EDFA systems showed some inherent limitations, such as:
- Gain variation effect: An EDFA's gain spectrum is not uniform. Unfortunately, there is nothing that can be easily adjusted in an EDFA to correct for this, so it is common to use external gain-flattening filters. To make matters worse, the gain shape changes as the total optical power of the signal channels changes at the input of the amplifier. This is not good news if channels need to be added to or dropped from the system, nor is it compatible with using static gain flattening filters.
Fibre ageing is another factor causing modification in the total optical power at the input of the EDFA.
The attenuation of a typical fibre span of 100km can double (3dB increase) through the lifetime of the system (say, 15 years).
- Limited gain bandwidth spectrum: EDFAs cannot effectively amplify signals above or below the "C" or "L" bands. New doped fibres for additional amplification bands (the short "S" band, below 1520nm) have always been either prohibitively expensive or have other operational inadequacies that limit their applicability.
Another option is to keep a single EDFA optical amplifier and to expand at the most the "C" band in a continuous way in order to offer a kind of super "C" band ranging, for instance, from 1520-1580nm. However, such a design is not expected to become available within the next few years.
- Gain transients: In addition to non-uniform gain, the output power of an EDFA can oscillate wildly if the power at its input is changed too rapidly. In fact the unintentional (in the event of a fibre cut) or intentional (optical switching) dropping of a few channels can disrupt all the other channels sharing the same fibre.
- Noise performance: Optical amplification is obtained at the expense of wideband optical noise that is added to the optical channels. Furthermore this optical noise accumulates as the signals pass through more optical amplifiers. In order to meet the output end Optical Signal-to-Noise Ratio (OSNR) requirement for high capacity and very long reach, it would be necessary to increase the power of optical channels. However, this approach would result in non-linear effects within the optical transmission fibre that severely corrupt the optical pulses (Fig. 2).
EDFAs cannot keep up
Although EDFAs were the right choice 10 years ago, this option cannot meet all of today's requirements. For metro applications, more cost-effective designs need to be developed in order to significantly decrease the price of metro WDM products. For core applications, increasing capacity and connection lengths need alternative option with higher performance and flexibility, as offered by Raman amplification.
Raman amplification operates along many kilometres of the actual transmission fibre, rather than just within a short length of special fibre at discrete locations. This technique capitalises on a non-linear interaction between photons and molecules in the glass itself, with no special doping of the fibre needed.
Specifically, this amplification scheme harnesses a non-linear effect called Raman scattering and exploits it for a productive use. This is carried out by launching a high-power pump laser beam with a shorter wavelength than the optical channels into the transmission fibre that serves to amplify the channels, via stimulated Raman scattering.
Raman amplifiers show significant advantages over EDFAs in high-performance systems. These advantages become evident at higher capacities and more advanced WDM systems, especially with all-optical networks:
- Gain variation insensitivity: Gain from Raman amplification is dynamically controllable and relatively insensitive to fluctuations and changes in the signals being amplified. The Raman gain spectrum, though non-uniform, is different from the EDFA's in one important respect. The shape remains constant with changes in signal power and pump power, the amplitude of the gain varies in proportion to the amount of pump power, and the wavelength range at which the gain spectrum is centred follows the wavelength of the pump laser exactly. Consequently, by carefully adjusting the pump power and wavelength, any desired spectral shape can be obtained - and varied dynamically.
- Unlimited gain bandwidth spectrum: Unlike EDFAs, which boost wavelengths in the 1530-1610 nm range, Raman amplifiers can increase the signal strength of any wavelength by pumping at 13THz above the desired frequency (or about 100nm less than the desired wavelength in the 1550nm region). This means that Raman amplification can provide amplification at both the "C" and "L" bands - and potentially into the "S" band of much shorter wavelengths.
- Instant adjustment to drastic changes in power level: Raman amplifiers tolerate abrupt changes in the number of wavelengths. If a fibre is cut or a significant amount of traffic is added or dropped, Raman gain is relatively insensitive to these fluctuations in power level. Raman amplifiers are the most efficient solution for carriers who want "point-and-click" provisioning of new capacity.
- Best noise performance: Raman amplification delivers significant long-haul performance improvements. When used on typical network spans, compared to EDFAs, Raman amplifiers enable transmission over much longer distances without optical-electrical regeneration. The key is Raman's superior optical signal-to-noise ratio, which is the result of distributed amplification.
Although EDFAs have spurred a tremendous increase in optical system performance and cost effectiveness, that technology is now imposing a distinct upper limit on the capacity x distance metric that needs to be overcome in order to address continuously increasing traffic capacity demands and connection distances. Raman amplification, and the bandwidth-engineering options it affords, is proving itself now to be the most manageable and deployable solution for high-capacity optical core networks.
Product Marketing Manager
Bertrand Clesca works for Corvis in Paris, France.