14 June 2011
CMOS technology enables higher data rates over poorer quality optical fibre
Pressure is growing on the companies which supply telecommunications operators (telcos) to develop equipment which carries more data at lower overall cost. In a way, it's the same story heard in every other area of the electronics market, but in this instance, it's on a much larger scale.
In particular, attention is focusing on fibre optic communications, used primarily for long haul telecomms links, but finding increasing application over shorter distances.
One of the reasons behind the pressure is the exponential growth in the amount of data being handled. Telcos can, in theory, deal with this by moving to higher data rates; and the moves are being made from 10 to 40 and, in the near future, 100Gbit/s. But faster data rates don't necessarily solve the problem, because if the signals are passed over old fibre, they degrade rapidy as data rates increase.
One approach under active consideration is coherent networking and communications specialist PMC-Sierra is developing SoCs for the application. Jay Bennett is manager of product marketing for the company's communications products division. "Coherent networking technology was actively investigated in the late 1980s and early 1990s in order to improve reach," he said. "However, the advent of wavelength division multiplexing and erbium doped fibre amplifiers met the need for immediate capacity improvement and reach extension for 10G networks. These innovations delayed the requirement for coherent technology until now."
Coherent technology improves the quality of signals transmitted by increasing their spectral efficiency. According to Bennett, the dsp technology that enables this can also eliminate a range of equipment deployed along the fibre link. Currently, optical impairment compensation equipment needs to be used at 80km spacings.
In Bennett's opinion, POLO 40G will enable equipment developers to take coherent networking away from niche deployments, where it has only been used on the most important routes, to something that can be deployed broadly. "POLO enables the two or three slots needed by non coherent equipment to be collapsed to a single slot. This is a most challenging area, but we have been able to demonstrate a 25% increase in optical reach."
View Figure 1
Bennett says there is a 'lot of buzz' around 40G. "Carriers can see clear value in coherent networking because it eliminates a whole class of optical impairment equipment. With non coherent networks, you have to put a lot of equipment in the ground. Coherent transmission means no capital or operational costs."
He believes non coherent 40G transmission will be phased out quickly. "In long haul, capacity is king," he claimed. "And it brings more options for shorter range metro networks."
But what are the problems which POLO 40G and coherent networking solve? "As the transmission rate increases to 40G," Bennett explained, "it requires the symbol rate to increase. With non coherent transmission, this results in more optical impairment – chromatic dispersion, for example. This makes it challenging to recover the original signal at the receiving end. Because of this, transmission reach can be reduced by a factor of 16.
"This is compensated for through the use of en route equipment, but this brings other problems; such as non linear effects on the transmitted signal."
Bennett dived deeper into the technology. "If you send data at 10G, you rely on amplitude. If you do this at 40G, there is significant signal degradation. Coherent transmission increases the amount of information transmitted per symbol so, instead of relying solely on amplitude, you can also include phase information."
The overall result is that, by adopting a coherent networking strategy, optical impairment is moved into the electrical domain, which means the power of Moore's Law can be brought to bear.
View Figure 2
The key to the development of POLO 40G, said Bennett, is the availability of 40nm cmos technology. "It enables low power solutions that can be air cooled and it's 16W power consumption is groundbreaking." He says PMC is the first merchant silicon supplier of such technology, which enables existing networks to be upgraded while controlling costs.
Removing the impairment equipment also improves latency by about 20%, said Bennett; something operators are looking to offer the financial community, but which is also applicable in data recovery applications.
But perhaps the most interesting application of coherent networking is to allow poor quality fibre to carry higher data rates. "When 2.5G mobile communications systems were being rolled out, a lot of G.652 fibre was laid. This was found to have a high degree of polarisation mode distortion, making it a poor conductor at 40Gbit/s," he noted. "By moving impairment compensation into the electrical domain, operators can now light low quality fibre and use it for higher transmission rates."
The performance of 40G coherent networking is enabled by the modulation scheme. Using dual polarisation quadrature phase shift keying (dp-qpsk), data can be transmitted at 11Gbit/s, rather than the 44Gbit/s required by differential phase shift keying (dpsk). While dp-qpsk can operate with cheaper rf optics than dpsk, it does require complex digital signal processing at the receive end.
When the 40G signal is received by POLO, the X and Y polarisations and their I and Q phase components are processed by an a/d converter and passed to the dsp engine, which comprises seven blocks.
The quadrature error filter block provides signal error detection and correction functions, correcting for impairments introduced in the receiver's optical and analogue processing. The frequency offset removal block performs a frequency rotation on data samples from the quadrature error filter, bringing the signal to baseband, after which the chromatic dispersion (CD) compensation block removes the bulk of CD from the X and Y polarisation channels.
The a/d converter runs at twice the incoming symbol rate in order to resolve the data stream and to satisfy Nyquist. Each data channel is processed by an independent FIR filter. The data resampler uses information from the symbol acquisition and tracking components in order to output synchronous data samples.
This is followed by polarisation mode dispersion/polarisation dependent loss (pmd/pdl) compensation, in which adaptive equalisation is used to compensate for cross polarisation interference, supplemental chromatic dispersion and adjacent symbol interference.
The symbol timing acquisition and tracking block uses complex symbol radius matching in order to output accurate acquisition and tracking signals. It also uses a symbol lock indicator for the other components in the dsp. Meanwhile, carrier frequency acquisition is achieved using an FFT with appropriate averaging and peak frequency component detection.
Finally, the carrier phase recovery algorithm uses a block phase estimator and a phase rotation function to remove residual frequency and phase errors.
A further processing step after the dsp engine is forward error correction using PMC's Swizzle technology. This delivers a performance, says the company, which approaches the theoretical maximum defined by the Shannon Limit. "We've done a lot of work on Swizzle," Bennett maintained, "and POLO is the first product to integrate the technology."
POLO 40G will sample later in 2011for lead customers, packaged in a 480ball 23 x 23mm fcbga.