Source-linked AI summary
Experimental Demonstration of Capacity Increase and Rate-Adaptation by Probabilistically Shaped 64-QAM
Fred Buchali, Georg Böcherer, Wilfried Idler, Laurent Schmalen, Patrick Schulte, Fabian Steiner
TL;DR
Coherent optical transceivers have limited rate flexibility, often requiring changes to baudrate or FEC overhead. This paper uses probabilistic shaping and a distribution matcher to adapt rate while keeping transmission settings fixed, demonstrating capacity and reach gains over 200 Gbit/s 16-QAM.
Problem
Coherent optical transceivers offer only coarse flexibility, often requiring baudrate or FEC-overhead changes, and conventional coded modulation remains below Shannon capacity.
Source: Introduction · p. 1Method
The system uses a distribution matcher to create shaped symbols and adapts their probability distribution while retaining fixed FEC overhead, modulation format, baudrate, and bandwidth.
Source: Introduction · p. 1, Rate-Adaptive Constellation Shaping · p. 2Results
15% capacity increase and 43% reach increase versus 200 Gbit/s 16-QAM were demonstrated.
Source: Conclusions · p. 3Takeaways & Limitations
Adjustable shaping selects arbitrary operating points without changing FEC overhead, constellation, or symbol rate.
Source: Introduction · p. 1, Conclusions · p. 3
Abstract
from arXiv · showhide
We implemented a flexible transmission system operating at adjustable data rate and fixed bandwidth, baudrate, constellation and overhead using probabilistic shaping. We demonstrated in a transmission experiment up to 15% capacity and 43% reach increase versus 200 Gbit/s 16-QAM.
Introduction
The paper targets coarse rate flexibility and the gap to capacity in coherent optical transceivers. It proposes probabilistic shaping for rate adaptation while keeping key transmission settings fixed.
Introduction
- Short high-SNR segments call for high spectral efficiency, whereas long low-OSNR segments may require lower-order modulation or higher FEC overhead for reliability.Source: Introduction · p. 1
- Conventional transceivers offer only coarse flexibility, often requiring baudrate or FEC-overhead changes, while conventional coded modulation remains below Shannon capacity.Gaussianlike shaping is identified as a way to overcome that capacity gap.Source: Introduction · p. 1
- Across the four illustrated PS-64-QAM distributions, greater shaping corresponds to lower entropy.The figure bars represent the probability of each modulation symbol.Source: Figure 1 · p. 1
- Probabilistic shaping and a distribution matcher enable rate flexibility and experimentally verified gains without changing FEC overhead, modulation format, baudrate, or bandwidth.The system generates non-uniform symbol sequences and is described as low complexity.Source: Introduction · p. 1
Rate-Adaptive Constellation Shaping
The system adapts transmission rate by changing the shaped symbol distribution while retaining the same FEC code. A distribution matcher, systematic FEC, and inverse matcher form the coding and modulation pipeline.
Rate-Adaptive Constellation Shaping
- The scheme assigns different probabilities to modulation symbols, unlike conventional systems that use all symbols equally.The cited scheme is proposed in [1].Source: Figure 2
- A distribution matcher converts data bits to shaped symbols, which are systematically FEC-encoded, mapped to QAM, transmitted, demodulated into LLRs, decoded, and inverse-matched.Source: Rate-Adaptive Constellation Shaping · p. 2
- The spatially coupled rate 5/6 LDPC code is chosen for its error performance and is reported to retain the same performance across constellation distributions.The authors state that other FEC schemes could in principle be used.Source: Rate-Adaptive Constellation Shaping · p. 2
- Changing the symbol distribution P changes transmission rate R while the same FEC code is retained; the experiment selects four Maxwell-Boltzmann distributions.H(P) is the entropy of P in bits.Source: Rate-Adaptive Constellation Shaping · p. 2
Experimental Setup
The transmission experiment uses a coherent optical loop with a 32-GBaud channel under test. Transmitter and receiver processing combine standard QAM modes with probabilistically shaped modes.
Experimental Setup
- The channel under test operated at 32 GBaud in a standard coherent transmission loop, with an 88-GSamples/s DAC driving a dual-polarization IQ modulator.The figure illustrates the experimental setup.Source: Experimental Setup · p. 2, Figure 3 · p. 2
- Two sets of four 32-GBaud DP-QPSK load channels with 4-nm guard bands accompanied the channel under test across three 80-km SMF spans.The spans used single-stage EDFAs with a 5-dB noise figure.Source: Experimental Setup · p. 2
- Regular QAM and PS-64-QAM P1 used blind-adaptation DSP, while shaped modes P2–P4 used data-aided processing.The receiver stored 500,000 samples for offline processing; data-aided DSP did not notably improve regular QAM or P1 performance.Source: Experimental Setup · p. 2
Results
The experiment evaluates mutual information and error-free reach for shaped and uniform constellations. Probabilistic shaping trades reach against net data rate and improves both compared with conventional systems.
Results
- Measured mutual information represents the maximum achievable rate assuming ideal FEC, while practical FECs incur a realization-dependent penalty.Source: Results · p. 2
- The tested setup fixes FEC overhead at 20% (code rate 5/6) and uses four shaped distributions.Source: Results · p. 2
- At mutual information of 4 bits per [caption text ends here], Fig. 4 reports a shaping gain of 4.3 loops, or 1032 km, for PS-64-QAM.The figure compares uniform symbols with distributions P1–P4.Source: Figure 4 · p. 3
- At 300 Gbit/s, probabilistic shaping increased reach by 25%, rising to 43% at 200 Gbit/s; the results are summarized in Table 1.Reach is defined as the maximum distance with error-free decoding, and the comparison uses the same channel bandwidth.Source: Results · p. 3, Table 1 · p. 3
Conclusions
The experiment reports capacity and reach gains over 200 Gbit/s 16-QAM and demonstrates rate adaptation through adjustable shaping while keeping key transmission settings fixed.
Conclusions
- The work reports the first optical transmission experiment of probabilistically shaped higher-order constellation schemes.Source: Conclusions · p. 3
- 15% capacity increase and 43% reach increase versus 200 Gbit/s 16-QAM were demonstrated.Source: Conclusions · p. 3
- Adjustable shaping selects arbitrary operating points without changing FEC overhead, constellation, or symbol rate.Source: Conclusions · p. 3