ROY RUBENSTEIN THE FUTURE
digital signal The future of coherent
With SP-QPSK, only three of the four constellation points are used for each symbol. A third fewer constellation points mean less data is transported but the benefit of SP-QPSK is extended reach due to the greater Euclidean distance between the symbol points, created by carefully mapping the sequence of symbols. This results in 2.5dB of extra gain compared to PM-QPSK, for a reach beyond 5,000km. And using the PSE-2’s 45Gbaud/s symbol rate, the fewer constellation points of SP- QPSK can be compensated for to match 100Gbit/s PM-QPSK at 33Gbaud/s. Ciena, meanwhile, first introduced multi- dimensional coding with its WaveLogic 3 Extreme DSP-ASIC and has extended its use with the WaveLogic Ai. WHAT NEXT? The latest coherent devices from the three vendors are implemented using a 28nm CMOS process. The next-generation devices will likely use a 16nm CMOS process followed by a 7nm process node. A 16nm DSP-ASIC will support 65Gbaud while 80-100Gbaud rates will be possible at 7nm, says Infinera’s Mahajan. Using these high baud rates will improve overall transmission capacity while the move to each new CMOS generation will improve the DSP-ASIC’s power efficiency. But the optical modulator alongside the DSP-ASIC will also need to work at 65Gbaud while the processor’s A/D and D/A converters must operate at 130 gigasamples-per-second, twice the symbol rate. “These are engineering problems that will be solved in the next two years,” says Eisenach. Ciena’s Roberts expects to be designing coherent DSP-ASICs for some time yet.
processing
Has the industry reached a tipping point of peak DSP improvement or is a complete rethink of performance improvement required? Roy Rubenstein reviews recent developments and trends.
ROY RUBENSTEIN
S everal leading optical transport vendors unveiled their latest coherent digital signal processors last year. The coherent DSP-ASICs from Infinera, Nokia and Ciena all use sophisticated techniques to achieve new levels of optical transmission performance over metro and long-distance networks. The goal of the designs of coherent DSPs is to maximise the traffic capacity available on a given optical route. But given the effectiveness of the latest generation of coherent DSPs, is there any scope for further improvement? “It is getting harder and harder,” admits Kim Roberts, vice president, WaveLogic science at Ciena. “Unlike 10 years ago, nowadays there seem to be no factors of 10 available for improvement.” CAPACITY-REACH The latest set of coherent chips offer a choice of modulation schemes and symbol rates while employing ingenious coding schemes. Previous DSP-ASICs offered a basic set of modulation formats, such as polarisation multiplexing, binary phase-shift keying (PM-BPSK), quadrature phase-shift keying (PM-QPSK) and 16-ary quadrature amplitude modulation (PM-16QAM). The latest designs of coherent chips add PM- 8QAM and even PM-64QAM as well as specialist coding schemes. As for symbol rates, Nokia’s PSE-2 supports 33Gbaud/s or 45Gbaud/s while Ciena’s WaveLogic Ai supports 35Gbaud/s or 56Gbaud/s. Using the higher baud rate, the same number of bits can be sent using a lower modulation scheme, which extends the system’s reach. For example, Nokia’s PSE-2 supports two 200Gbit/s formats using either 16-QAM or 8-QAM. With 16-QAM the reach is 1000km but rises to 2000km with 8-QAM. Alternatively, using 45Gbaud and 16-QAM, 250 gigabits can be sent per wavelength over 900km. Ciena uses the WaveLogic Ai’s 35Gbaud/s symbol rate for existing 50GHz fixed- grid networks for 100Gbit/s to 250Gbit/s
data rates in increments of 50Gbit/s. At 56Gbuad, 400Gbit/s single-wavelength transmission is possible across a flexible grid network. At 56Gbaud, the WaveLogic Ai delivers 100Gbit/s to 400Gbit/s optical channels in 50-gigabit increments. CODING SCHEMES The vendors are also using coding schemes to enhance transmission capabilities. “Vendors use a lot of different terms essentially for the same thing: applying some type of coding to symbols or to symbols across polarisations,” says Randy Eisenach, senior product marketing manager, optical networks at Nokia. “It [the coding scheme] increases the effective Euclidean distance between the constellation points and therefore improves overall performance.” Infinera’s FlexCoherent DSP-ASIC, part
Coherent DSP-ASIC
28nm CMOS
16nm FinFET
7nm FinFET
Baud rate - Gbaud/s
33-45
60-80
80-100
Gbit/ lambda
200-400
400-600
800
Capacity C-band/ Terabits Reach 2000-4000km Capacity C-band/ Terabits Data centre interconnect 150-300km
20
24
28
35 (150km)
Mid 30s
~40
30-32 (300km)
Power W/Gig (relative)
x
x/2.5
~x/6
Comparing capacity capabilities. Source: Infinera.
He identifies several areas such as FEC and countering non-linear effects, each an area where a decibel of gain can be reclaimed. But design engineers will need to use more mathematics and encapsulate the algorithms in more logic gates. The good news, says Roberts, is that Moore’s law continues to provide CMOS gates. So yet further capacity and bandwidth improvements look possible – even inevitable – only the systems developers will certainly have to think out of the box.
of its Infinite Capacity Engine, which also includes its 1.2 terabits photonic integrated circuit, uses matrix-enhanced PM-BPSK, a form of averaging that adds a decibel of gain. “Any innovation that adds gain to a link, the margin that you give to operators, is always welcome,” says Pravin Mahajan, Infinera’s director of product and corporate marketing. Nokia’s PSE-2 uses coding for its set-partition QPSK (SP- QPSK). Standard PM-QPSK uses amplitude and phase modulation, resulting in a 4-point constellation.
www.opticalconnectionsnews.com
23
ISSUE 8 | Q1 2017
Made with FlippingBook - Online magazine maker