Design and Analysis of GFDM-Based Wireless Communication Systems

Authors: Wang, Yanpeng
Advisor: Fortier, Paul
Abstract: Generalized frequency division multiplexing (GFDM) is a block-processing based non-orthogonal multi-carrier modulation scheme, which is a promising candidate waveform technology for beyondfifth-generation (5G) wireless systems. The ability of GFDM to flexibly adjust the block size and the type of pulse-shaping filters makes it a suitable scheme to meet several important requirements, suchas low latency, low out-of-band (OOB) radiation and high data rates. Applying the multiple-inputmultiple-output (MIMO) technique, the massive MIMO technique, or low-density parity-check (LDPC)codes to GFDM systems can further improve the systems performance. Therefore, the investigation of such combined systems is of great theoretical and practical importance.This thesis investigates GFDM-based wireless communication systems from the following three aspects. First, we derive a union bound on the bit error rate (BER) for MIMO-GFDM systems, which is based on exact pairwise error probabilities (PEPs). The exact PEP is calculated using the moment-generating function (MGF) for maximum likelihood (ML) detectors. Both the spatial correlation between antennas and the channel estimation errors are considered in the investigated channel environment. Second, polynomial expansion-based low-complexity channel estimators and precoders are proposed for massiveMIMO-GFDM systems. Interference-free pilots are used in the minimum mean square error (MMSE) channel estimation to combat the influence of non-orthogonality between subcarriers in GFDM. The cubic computational complexity can be reduced to square order by using the polynomial expansion technique to approximate the matrix inverses in the conventional MMSE estimation and precoding. In addition, we derive performance limits in terms of the mean square error (MSE) for the propose destimators, which can be a useful tool to predict estimators performance in the high Es/N0 region. ACramér-Rao lower bound (CRLB) is derived for our system model and acts as a benchmark for the estimators. The computational complexity of the proposed channel estimators and precoders, and the impacts of the polynomial degree are also investigated. Finally, we analyze the error probability performance of LDPC coded GFDM systems. We first derive the initial log-likelihood ratio (LLR) expressions that are used in the sum-product algorithm (SPA) decoder. Then, based on the decoding threshold, we estimate the frame error rate (FER) in the low Eb/N0 region by using the observed BER to model the channel variations. In addition, a lower bound on the FER of the system is also proposed based on absorbing sets. This lower bound can act as an estimate of the FER in the highEb/N0 region if the absorbing set used is dominant and its multiplicity is known. The quantization scheme also has an important impact on the FER and BER performances. Randomly constructed and array-based LDPC codes are used to support the performance analyses. For all these three aspects,software-based simulations and calculations are carried out to obtain related numerical results, whichverify our proposed methods.
Document Type: Thèse de doctorat
Issue Date: 2021
Open Access Date: 19 July 2021
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
Description SizeFormat 
37216.pdf2.69 MBAdobe PDFThumbnail
All documents in CorpusUL are protected by Copyright Act of Canada.