The growing sophistication of high-speed optical transmission based on 100 Gigabit-plus lightpaths and advanced ROADMs is rekindling interest in amplifier design.
Raman is a signature of the spread of 100 Gig but also the desire of being upgradable to higher bit rates
Per Hansen, II-VI
For the last decade, amplifier designers have been tasked with reducing the cost of Erbium-doped fibre amplifiers (EDFAs). "Now there is a need for new solutions that are more expensive," says Daryl Inniss, vice president and practice leader, components at market research firm, Ovum. "It is no longer just cost-cutting."
Higher output power amplifiers are needed to boost 100 Gig-plus signals that have less energy. Such amplifiers must also counter greater losses incurred by sophisticated colourless, directionless and contentionless (CDC) ROADM nodes. System vendors also require more power-efficient and compact amplifiers to maximise the chassis slots available for revenue-generating 100 Gig transponders.
Such requirements have created interest in all amplifier types, not just EDFAs but hybrid EDFA-Raman and Raman amplifiers.
"Improving the optical signal-to-noise ratio (OSNR) is of paramount consideration to enable higher capacity and reach for 100 Gig-plus lambdas," says Madhu Krishnaswamy, director, product line management at JDSU. "Raman amplification is becoming increasingly critical to delivering this OSNR improvement, largely in long haul."
Other developments include micro-amplifiers that boost single channels, and arrayed amplifiers used with ROADM nodes. These developments are also driving optical components: power-efficient, integrated pump lasers are needed for such higher-power amplifiers.
Operators' requirements span all three amplifier classes: EDFA, hybrid EDFA-Raman and all-Raman, says Anuj Malik, manager, solutions marketing at Infinera: "Some networks require a high OSNR and use hybrid amplifiers but some networks are prone to fibre cuts and hence avoid hybrid as fibre splices can cause more problems with Raman."
Raman differs from EDFA in several ways. Raman has a lower power efficiency, the optical pump power needed to pump an amplifier to achieve a certain gain and output power. This requires higher power to be launched into a Raman amplifier, raising safety issues for staff and equipment. The high launch power requires a sound connection between the Raman pump source and the fibre to avoid equipment being damaged, hence Infinera's reference to fibre splices.
Yet if Raman has a lower power efficiency, it has notable benefits when compared to an EDFA.
An EDFA performs lumped amplification, boosting the signal at distinct points in the network, every 80km commonly. Raman amplifies the signal as it travels down the fibre.
"With Raman amplification the gain is out in the fibre span, and Raman delivers a lower equivalent noise figure - a big advantage," says Per Hansen, head of product marketing, amplifier business unit at II-VI." The company II-VI acquired Oclaro's amplifier business in November 2013.
An amplifier's noise figure is a measure of performance in the network. All amplifiers introduce noise so that the input signal-to-noise ratio divided by the output signal-to-noise ratio is always greater than one. "Raman gives you a significantly better noise figure, an improvement in the range of 3 to 5dB," says Hansen.
EDFA designs continue to progress alongside the growing interest in hybrid and all-Raman. JDSU says that higher output power EDFAs, greater than 24dBm, are increasingly relevant for 96-plus channel systems that support super-channels and flexible grid ROADMs in the metro and long haul.
"Switchable-gain EDFAs to optimise the noise figure over a wider dynamic range of operation is another element enhancing overall system OSNR," says Krishnaswamy. "This is also common for metro and long haul."
Hybrid amplification combines the best characteristics of EDFA and Raman. In a hybrid, Raman is the first amplification stage where noise figure performance is most important, while the EDFA, with its power efficiency, is used as the second stage, boosting the signal to a higher level.
According to Finisar, 100 Gig uses the same receiver OSNR as 10 Gig transmissions. However, the transmission power per channel at 100 Gig is reduced, from 0 to 1dBm at 10 Gig to -2 to -3dBm at 100 Gig, due to non-linearity transmission issues. "Immediately you lose a few dBs in the OSNR," says Uri Ghera, CTO of the optical amplifier products at Finisar.
An overwhelming portion of WANs worldwide have adopted hybrid EDFA-Raman and this trend is expected to continue for the foreseeable future.
For 400 Gigabit transmission, the weaker signal sent requires the OSNR at the receiver to be 4-10dBm higher, says Ghera: "This is why you need hybrid Raman-EDFA."
Moving to a narrower channel spacing using a flexible grid also places greater demands on amplifiers. "Because of super-channels, if before we were talking about 100 channels [in the C-band], for a channel spacing of 37.5GHz it is more like 130 channels," says Ghera. "If you want the same power per channel, it means higher-output amplifiers."
The spectrum amplified by an EDFA is determined by the fibre. EDFAs amplify the 35nm-wide C-band spanning 1530 to 1565nm, and also the separate L-band at 1570 to 1605nm, if that is used. In contrast, the spectrum amplified by Raman is determined by the pump laser's wavelength. This leads to another benefit of all-Raman: far broader spectrum amplification, 100nm and wider.
Xtera is a proponent of all-Raman amplification. The system vendor has demonstrated 60nm- and even 100nm-wide spectrum amplification, broader than the C and L bands combined.
Xtera conducted trials with Verizon in 2013 using its Nu-Wave Optima platform and Raman operating over a 61nm window. The trials are detailed in Table 1.
Between 15 and 40 Terabits were sent over 4,500km and 1,500km, respectively, using several modulation schemes and super-channel arrangements. In comparison, state-of-the-art 100 Gig-plus systems achieve 16 Terabit typically across the C-band, and are being extended to 20-24 Terabit using closer-spaced channels. Using 16-QAM modulation, the reach achieved is 600km and more.
Table 1: Xtera's Verizon trial results using a 61nm spectrum and all-Raman amplification.
JDSU says hybrid amplification remains the most cost-competitive way to deliver the required OSNR and system capacity, while all-Raman can potentially increase system capacity.
Overall, it is network capacity and reach requirements that drive amplifier choice, says Krishnaswamy: "An overwhelming portion of WANs worldwide have adopted hybrid EDFA-Raman and this trend is expected to continue for the foreseeable future."
Meanwhile, the single channel micro-amp, sits alongside or is integrated within the transmitter. Operators want a transponder that meets various requirements for their reconfigurable networks. "If you look into the numbers, you want to boost the signal early on before it is attenuated," says II-VI's Hansen. "That gives you the best OSNR performance."
"This [single-channel amp] is a type that was rare in old systems," adds Finisar's Ghera. "It is also a market that is growing the fastest for us."
The micro-amp needs to be compact and low power, being alongside the power-hungry 100 Gig coherent transmitter. This is driving uncooled pump laser development and system integration.
Similar design goals apply to arrayed amplifiers that counter losses in ROADM add/ drop cards. "If you have some of the features of colourless, directionless and contentionless, you incur bigger losses in the node but you can make it up with other amps, one of these being arrayed amps," says Hansen.
Arrayed designs can have eight or more amps to support multiple-degree nodes so that achieving a power-efficient, compact design is paramount. Hence II-VI's development of an uncooled dual-chip pump laser integrated in a package. "Having four packages to pump eight amps in a small space that do not require cooling is a huge advantage," says Hansen.
The amplifier design challenges are set to continue.
One, highlighted by Infinera, is expanding amplification to the L-band to double overall capacity. JDSU highlights second-order and third-order Raman designs that use a more complex pump laser arrangement to improve system OSNR. Lowering the noise figure of EDFAs will be another continuing design goal, says JDSU.
II-VI expects further challenges in miniaturising single-channel and arrayed amplifier designs. Finisar also cites the need for more compact designs, citing putting an EDFA in an XFP package as an example.
Another challenge is producing high-power Raman amplifiers that can bridge extremely long spans, 300 to 400km. Such an amplifier must be able to read lots of physical parameters associated with the span and set the line accordingly, said Gheri.
II-VI's Hansen says the adoption of Raman and arrayed amplifiers is a good indicator of the wider deployment of next-generation network architectures. "Raman is a signature of the spread of 100 Gig but also the desire of being upgradable to higher bit rates," he says.
The article first appeared as an OFC 2014 show preview piece