Phase Noise Compensation for CFBMC–OQAM Systems Under Imperfect Channel Estimation

Abstract

Among many multi-carrier systems, circular filter-bank multi-carrier offset quadrature amplitude modulation (CFBMC-OQAM) is one of promising candidates for future wireless communications. This paper studies the impact of phase noise and its compensation for CFBMC-OQAM under imperfect channel estimation, which has not been done before. In the presence of phase noise, a two-stage phase noise compensation algorithm is proposed. In the first stage, the channel frequency response and phase noise are estimated based on the transmission of a preamble. Such a preamble is designed to minimize the channel mean squared error. In the second stage, the estimated channel obtained from the first stage together with pilot symbols are used to compensate for the phase noise and detect the transmitted signal. Simulation results obtained under practical scenarios show that the proposed algorithm effectively estimates the channel frequency response and compensates for the phase noise. The proposed algorithm is also shown to outperform an existing algorithm that performs iterative phase noise compensation when phase noise impact is high. INDEX TERMS FBMC, OFDM, CFBMC-OQAM, phase noise, preamble design, channel estimation.

Compensation of Phase Noise and IQ Imbalance in Multi-Carrier Systems

“…To estimate the channel impulse response, the interdependency between the channel h and the PN P 1 in (21) has to be resolved. Similar to [14], [20]- [22], the interdependency can be decoupled by first obtaining an estimate of the channel which is a function of P 1 . The estimated channel is then substituted into the optimization problem which minimizes the MSE of PN estimate.…”

“…It is pointed out that, to improve the quality of the estimated channel, the minimum mean squared error (MMSE) method is used in this paper rather than the least square (LS) method as in [14], [20], [21]. Furthermore, the proposed algorithm estimates the channel impulse response rather than the channel frequency response as proposed in [22]. The reason is that the length of a channel impulse response is usually much shorter than the length of its corresponding channel frequency response, and as such, estimating the channel impulse response usually requires less This work is licensed under a Creative Commons Attribution 4.0 License.…”

“…It can be seen that the most complicated operations in computing Λ related to matrix inversion and multiplication are reduced from N -dimensional matrices to only N/2-dimensional matrices. The matrix inversion in Λ can be further simplified if the preamble is designed such as A H A = I N/2 [22] and the length of the channel impulse response is much shorter than the half length of the preamble s 1 . Specifically, Λ can be rewritten as…”

“…Another disadvantage of iterative algorithms is their long latency. In [14], [20]- [22], the authors propose non-iterative algorithms to compensate for the PN impact. Specifically, in [14], the PN and channel are first jointly estimated by solving a constrained quadratic optimization problem, but under overly simplified constraints.…”

“…Specifically, in [14], the PN and channel are first jointly estimated by solving a constrained quadratic optimization problem, but under overly simplified constraints. The optimization problem in [14] is revisited in [20]- [22] with more realistic constraints. Compared to [20], the proposed algorithms in [21], [22] have lower complexity.…”