Design of Digital Communications for Strong Phase Noise Channels

Abstract

To meet the requirements of beyond 5G networks, the significant amount of unused spectrum in sub-TeraHertz frequencies is contemplated for high-rate wireless communications. Yet, the performance of sub-TeraHertz systems is severely degraded by strong oscillator phase noise. We investigate in this paper the design of digital communications robust to phase noise. This problem is addressed in three steps: the characterization of the phase noise channel, the design of the optimum receiver, and the optimization of the modulation scheme. This paper proposes a joint performance and implementation optimization. First, we address the design of the demodulation scheme for phase noise channels and propose the polar metric, a soft-decision rule for symbol detection. It is shown that performance gains are achieved for coded and uncoded systems with valuable complexity reductions of the receiver. Second, we investigate the optimization of the modulation scheme for phase noise. We demonstrate that using a constellation defined upon a lattice in the amplitude-phase domain leads to significant performance gains and a low-complexity implementation. Thereupon, we propose the Polar-QAM scheme with efficient binary labeling and demodulation. Numerical simulation results show that the proposed modulation and demodulation schemes offer valuable solutions to achieve high-rate communications on systems strongly impaired by phase noise.

Novel MIMO Technique for Wireless Terabits Systems in Sub-THz Band

“…However, all these system configurations suffer from an error floor when subjected to very high PN level (σ 2 g = 10 −1 ). In conclusion, the non-coherent SMX FSIM provides the best PN robustness among these candidates, but a PN mitigation technique and/or an APM less sensitive to PN (e.g., PAM, spiral constellations [31], Polar QAM [32], ...) are necessary at very high PN level.…”

Deep Learning-Assisted TeraHertz QPSK Detection Relying on Single-Bit Quantization

“…As a result, the distribution of each sampling point will be different from the estimated distribution obtained from the training samples. Note that the 3 Based on [36] and [37], the correlation of phase noise can be guaranteed when the frame length is smaller than ln(2)/(2π 2 T 2 s f 2 0 ), where f 0 is the corner frequency of the oscillator. Considering the oscillator of [38], the maximum frame length satisfying the correlation of phase noise may be as high as 3.5×10 6 , which is higher than NpNt +N d .…”

Scoring the Terabit/s Goal:Broadband Connectivity in 6G

“…According to [269], when the corner frequency of the oscillator is small in comparison to the system bandwidth, the phase noise can be accurately modeled by an uncorrelated Gaussian process. Using this mathematically convenient assumption, it is demonstrated in [270] that using a constellation defined upon a lattice in the amplitude-phase domain is particularly relevant for phase noise channels. More specifically, a constellation defined upon a lattice in the amplitude-phase domain is robust to PN and leads to a low-complexity implementation.…”