The Non-Binary Low Density Parity Check (NB-LDPC) codes constitutes an interesting category of error correction codes, and are well known to outperform their binary counterparts. However, their non-binary nature makes their decoding process of higher complexity. This PhD thesis aims at proposing new decoding algorithms for NB-LDPC codes that will be shaping the resultant hardware architectures expected to be of low complexity and high throughput rate. The first contribution of this thesis is to reduce the complexity of the Check Node (CN) by minimizing the number of messages being processed. This is done thanks to a pre-sorting process that sorts the messages intending to enter the CN based on their reliability values, where the less likely messages will be omitted and consequently their dedicated hardware part will be simply removed. This reliability-based sorting enabling the processing of only the highly reliable messages induces a high reduction of the hardware complexity of the NB-LDPC decoder. Clearly, this hardware reduction must come at no significant performance degradation. A new Hybrid architectural CN model (H-CN) combining two state-of-the-art algorithms - Forward-Backward CN (FB-CN) and Syndrome Based CN (SB-CN) - has been proposed. This hybrid model permits to effectively exploit the advantages of pre-sorting. This thesis proposes also new methods to perform the Variable Node (VN) processing in the context of pre-sorting-based architecture. Different examples of implementation of NB-LDPC codes defined over GF(64) and GF(256) are presented. For decoder to run faster, it must become parallel. From this perspective, we have proposed a new efficient parallel decoder architecture for a 5/6 rate NB-LDPC code defined over GF(64). This architecture is characterized by its fully parallel CN architecture receiving all the input messages in only one clock cycle. The proposed new methodology of parallel implementation of NB-LDPC decoders constitutes a new vein in the hardware conception of ultra-high throughput rate decoders. Finally, since the NB-LDPC decoders requires the implementation of a sorting function to extract P minimum values among a list of size Ns, a chapter is dedicated to this problematic where an original architecture called First-Then-Second-Extrema-Selection (FTSES) has been proposed.
The Non-Binary Low Density Parity Check (NB-LDPC) codes constitutes an interesting category of error correction codes, and are well known to outperform their binary counterparts. However, their non-binary nature makes their decoding process of higher complexity. This PhD thesis aims at proposing new decoding algorithms for NB-LDPC codes that will be shaping the resultant hardware architectures expected to be of low complexity and high throughput rate. The first contribution of this thesis is to reduce the complexity of the Check Node (CN) by minimizing the number of messages being processed. This is done thanks to a pre-sorting process that sorts the messages intending to enter the CN based on their reliability values, where the less likely messages will be omitted and consequently their dedicated hardware part will be simply removed. This reliability-based sorting enabling the processing of only the highly reliable messages induces a high reduction of the hardware complexity of the NB-LDPC decoder. Clearly, this hardware reduction must come at no significant performance degradation. A new Hybrid architectural CN model (H-CN) combining two state-of-the-art algorithms - Forward-Backward CN (FB-CN) and Syndrome Based CN (SB-CN) - has been proposed. This hybrid model permits to effectively exploit the advantages of pre-sorting. This thesis proposes also new methods to perform the Variable Node (VN) processing in the context of pre-sorting-based architecture. Different examples of implementation of NB-LDPC codes defined over GF(64) and GF(256) are presented. For decoder to run faster, it must become parallel. From this perspective, we have proposed a new efficient parallel decoder architecture for a 5/6 rate NB-LDPC code defined over GF(64). This architecture is characterized by its fully parallel CN architecture receiving all the input messages in only one clock cycle. The proposed new methodology of parallel implementation of NB-LDPC decoders constitutes a new vein in the hardware conception of ultra-high throughput rate decoders. Finally, since the NB-LDPC decoders requires the implementation of a sorting function to extract P minimum values among a list of size Ns, a chapter is dedicated to this problematic where an original architecture called First-Then-Second-Extrema-Selection (FTSES) has been proposed.
This paper proposes a unified framework to describe the check node architectures of non-binary low-density parity-check (NB-LDPC) decoders. Forward-backward, syndrome-based, and pre-sorting approaches are first described. Then, they are hybridized in an effective way to reduce the amount of computation required to perform a check node. This paper is specially impacting check nodes of high degrees (or high coding rates). Results of 28-nm ASIC post-synthesis for a check node of degree 12 (i.e., a code rate of 5/6 with a degree of variable equal to 2) are provided for NB-LDPC over GF(64) and GF(256). While simulations show almost no performance loss, the new proposed hybrid implementation check node increases the hardware and the power efficiency by a factor of six compared with the classical forward-backward architecture. This leads to the first ever reported implementation of a degree 12 check node over GF(256), and these preliminary results open the road to high decoding throughput, high rate, and high-order Galois field NB-LDPC decoder with a reasonable hardware complexity.
Conception du décodeur NB-LDPC à débit ultra-élevé Les codes correcteurs d’erreurs Non-Binaires Low Density Parity Check (NB-LDPC) sont connus pour avoir de meilleure performance que les codes LDPC binaires. Toutefois, la complexité de décodage des codes non-binaires est bien supérieure à celle des codes binaires. L’objectif de cette thèse est de proposer de nouveaux algorithmes et de nouvelles architectures matérielles de code NB-LDPC pour le décodage des NBLDPC. La première contribution de cette thèse consiste à réduire la complexité du nœud de parité en triant en amont ses messages d’entrées. Ce tri initial permet de rendre certains états très improbables et le matériel requis pour les traiter peut tout simplement être supprimé. Cette suppression se traduit directement par une réduction de la complexité du décodeur NB-LDPC, et ce, sans affecter significativement les performances de décodage. Un modèle d’architecture, appelée "architecture hybride" qui combine deux algorithmes de l’état de l’art ("l’Extended Min Sum" et le "Syndrome Based") a été proposé afin d’exploiter au maximum le pré-tri. La thèse propose aussi de nouvelles méthodes pour traiter les nœuds de variable dans le contexte d’une architecture pré-tri. Différents exemples d’implémentations sont donnés pour des codes NB-LDPC sur GF(64) et GF(256). En particulier, une architecture très efficace de décodeur pour un code de rendement 5/6 sur GF(64) est présentée. Cette architecture se caractérise par une architecture de check node nœud de parité entièrement parallèle. Enfin, une problématique récurrente dans les architectures NB-LDPC, qui est la recherche des P minimums parmi une liste de taille Ns, est abordée. La thèse propose une architecture originale appelée first-then-second minimum pour une implantation efficace de cette tâche.
Non-binary low-density parity-check (NB-LDPC) codes can achieve better error-correcting performance than binary LDPC codes when the code length is moderate at the cost of higher decoding complexity. The high complexity is mainly caused by the complicated computations in the check node processing and the large memory requirement. In this paper, a novel check node processing scheme and corresponding VLSI architectures are proposed for the Min-max NB-LDPC decoding algorithm. The proposed scheme first sorts out a limited number of the most reliable variable-to-check (v-to-c) messages, then the check-to-variable (c-to-v) messages to all connected variable nodes are derived independently from the sorted messages without noticeable performance loss. Compared to the previous iterative forward-backward check node processing, the proposed scheme not only significantly reduced the computation complexity, but eliminated the memory required for storing the intermediate messages generated from the forward and backward processes. Inspired by this novel c-to-v message computation method, we propose to store the most reliable v-to-c messages as “compressed” c-to-v messages. The c-to-v messages will be recovered from the compressed format when needed. Accordingly, the memory requirement of the overall decoder can be substantially reduced. Compared to the previous Min-max decoder architecture, the proposed design for a (837, 726) code overGF(25) can achieve the same throughput with only 46% of the area.
When the code length is moderate, non-binary low-density parity-check (NB-LDPC) codes can achieve better error-correcting performance than their binary counterparts at the cost of higher decoding complexity. The check node processing is a major bottleneck of NB-LDPC decoding. Its computation complexity can be reduced by an iterative forward-backward scheme. Nevertheless, large memory is required in this scheme to store intermediate results. In this paper, a novel check node processing scheme and corresponding VLSI architectures are proposed for the Min-max NB-LDPC decoding. The proposed scheme first sorts out a limited number of the most reliable input messages. Then the forward and backward processes are carried out using only these messages. In addition, an innovative approach is developed to derive the output messages from only a small number of intermediate results without noticeable performance loss. Accordingly, both the memory requirement and computation complexity are greatly reduced. For an example (837, 726) NB-LDPC code over GF(25), our proposed check node processing architecture can achieve at least 34% area reduction and about the same throughput compared to previous efforts based on the original forward-backward scheme.
Non-binary codes are highly efficient error-correcting codes for short message sizes. They are also naturally adapted to coded modulations, and therefore, can be used to approach channel capacity. However, non-binary decoders built on sets of high cardinalities (greater than or equal to 64) are still marginally used due to their high decoding complexity. This thesis studies and proposes new algorithms for simplifying the processing of the parity nodes for two types of non-binary error- correcting codes: Non-Binary Low-Density Parity Check (NB-LDPC) and Non-Binary Polar Codes (NB-PC). For NB-LDPC codes, a new decoding technique called the Best-Request-Default (BRD) algorithm has been proposed. In the context of truncated message exchange, it becomes interesting to relax the extrinsic information principle and make messages exchanged over a connection between a variable node and a parity node interdependent. More precisely, part of the parity node's processing will be dedicated to responding to the "requested" symbols by the variable node (the algorithm's "request" term). The BRD algorithm makes it possible to significantly reduce the size of the messages exchanged, and therefore the complexity of the decoder, without any significant degradation in performance. The decoding of high-cardinality NB-PC codes is still a largely unexplored field, and the state-of- the-art consists of adapting the Extended Min-Sum algorithm developed for NB-LDPC codes to NB-PC codes. We propose two new optimizations: firstly, a reduction in the complexity of parity node processing through asymmetrical processing of its inputs based on a relative reliability criterion (Asymmetrical Extended Min-Sum algorithm), and secondly, an ad-hoc simplification of each parity constraint according to the polarization of the messages arriving at it (Polarization- Aware Polar Decoder). The combination of these two optimizations reduces complexity by a factor of 2 compared with the state-of-the-art, again without any noticeable degradation in decoding performance.
Binary Low-Density Parity-Check (LDPC) codes are a type of error correction code known to exhibit excellent error-correcting capabilities, and have increasingly been applied as the forward error correction solution in a multitude of systems and standards, such as wireless communications, wireline communications, and data storage systems. In the pursuit of codes with even higher coding gain, non-binary LDPC (NB-LDPC) codes defined over a Galois field of order q have risen as a strong replacement candidate. For codes defined with similar rate and length, NB-LDPC codes exhibit a significant coding gain improvement relative to that of their binary counterparts.Unfortunately, NB-LDPC codes are currently limited from practical application by the immense complexity of their decoding algorithms, because the improved error-rate performance of higher field orders comes at the cost of increasing decoding algorithm complexity. Currently available ASIC implementation solutions for NB-LDPC code decoders are simultaneously low in throughput and power-hungry, leading to a low energy efficiency.We propose several techniques at the algorithm level as well as hardware architecture level in an attempt to bring NB-LDPC codes closer to practical deployment. On the algorithm side, we propose several algorithmic modifications and analyze the corresponding hardware cost alleviation as well as impact on coding gain. We also study the quantization scheme for NB-LDPC decoders, again in the context of both the hardware and coding gain impacts, and we propose a technique that enables a good tradeoff in this space. On the hardware side, we develop a FPGA-based NB-LDPC decoder platform for architecture prototyping as well as hardware acceleration of code evaluation via error rate simulations. We also discuss the architectural techniques and innovations corresponding to our proposed algorithm for optimization of the implementation. Finally, a proof-of-concept ASIC chip is realized that integrates many of the proposed techniques. We are able to achieve a 3.7x improvement in the information throughput and 23.8x improvement in the energy efficiency over prior state-of-the-art, without sacrificing the strong error correcting capabilities of the NB-LDPC code.