Abstract
Generalized mutual information (GMI) has become a key metric for bit-interleaved coded modulation (BICM) system design and performance analysis. As residual phase noise (RPN) normally exists after imperfect phase estimation, the mostly used mismatched Gaussian receiver is suboptimal for GMI analysis in phase noise. This letter thus analyzes the GMI of BICM systems using 8-ary quadrature-amplitude-modulations (QAM) in the presence of both RPN and additive white Gaussian noise (AWGN). We will use the maximum-likelihood receiver derived in our earlier work to calculate the GMI of Star-8QAM, Circular-8QAM, Rect-8QAM and 8PSK. The explicit symbol and bit log-likelihood ratios are specifically expressed in amplitude-phase form with RPN considered. Numerical results are given in details to show the GMI comparison in phase noise and the GMI loss compared to the pure AWGN case. It is shown that Star-8QAM is much more tolerant to large RPN. Moreover, the ratio parameters of Rect-8QAM and Star-8QAM are optimized to maximize the GMI as the RPN variance increases. We also optimize the bit mapping for non-Gray 8QAM.
Highlights
Thanks to coherent detection, receiver digital signal processing, soft-decision forward error correction (SD-FEC) and other techniques in optical communications, advanced modulation formats have received increasing attention to maximize the fiber capacity [1]
Optical fiber communication systems usually rely on bit-interleaved coded modulation (BICM) channel coding schemes without iterative demapping, and generalized mutual information (GMI) has emerged as a practical tool for system design [7], [8]
The BICM mutual information, i.e., the GMI, is a more appropriate metric for performance comparison as bit mapping and interleaving are taken into account
Summary
Receiver digital signal processing, soft-decision forward error correction (SD-FEC) and other techniques in optical communications, advanced modulation formats have received increasing attention to maximize the fiber capacity [1]. Even though RPN is considered in [16], the GMI analysis therein is still based on using a mismatched channel model, i.e., a mismatched Gaussian receiver This Gaussian receiver is not the right criterion for performance analysis in phase noise, as shown in our earlier work [17]. The optimum receiver with RPN considered is given as [17, eq(7)] based on the received amplitude and phase information in polar coordinates Using this maximum-likelihood (ML) receiver in [17], accurate GMI analysis with respect to both AWGN and RPN is imperative for BICM systems. We will analyze the BICM mutual information of different 8QAM constellations (Star-, Circular-, Rect- and 8PSK) in phase noise. Due to the uniform input, the bit-wise GMI I(Bi, r) is expressed in terms of the bit LLR as [9, eq(18)]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
