Ali Emre Pusane

Implementation aspects of LDPC convolutional codes

Potentially large storage requirements and long initial decoding delays are two practical issues related to the decoding of low-density parity-check (LDPC) convolutional codes using a continuous pipeline decoder architecture. In this paper, we propose several reduced complexity decoding strategies to lessen the storage requirements and the initial decoding delay without significant loss in performance. We also provide bit error rate comparisons of LDPC block and LDPC convolutional codes under equal processor (hardware) complexity and equal decoding delay assumptions. A partial syndrome encoder realization for LDPC convolutional codes is also proposed and analyzed. We construct terminated LDPC convolutional codes that are suitable for block transmission over a wide range of frame lengths. Simulation results show that, for terminated LDPC convolutional codes of sufficiently large memory, performance can be improved by increasing the density of the syndrome former matrix.

Deriving Good LDPC Convolutional Codes from LDPC Block Codes

Low-density parity-check (LDPC) convolutional codes are capable of achieving excellent performance with low encoding and decoding complexity. In this paper, we discuss several graph-cover-based methods for deriving families of time-invariant and time-varying LDPC convolutional codes from LDPC block codes and show how earlier proposed LDPC convolutional code constructions can be presented within this framework. Some of the constructed convolutional codes significantly outperform the underlying LDPC block codes. We investigate some possible reasons for this “convolutional gain,” and we also discuss the-mostly moderate-decoder cost increase that is incurred by going from LDPC block to LDPC convolutional codes.

On Deriving Good LDPC Convolutional Codes from QC LDPC Block Codes

In this paper we study the iterative decoding behavior of time-invariant and time-varying LDPC convolutional codes derived by unwrapping QC LDPC block codes. In particular, for a time-varying LDPC convolutional code, we show that the minimum pseudo-weight of the convolutional code is at least as large as the minimum pseudo-weight of the underlying QC code. We also prove that the unwrapped convolutional codes have fewer short cycles than the QC codes. These results taken together lead to improved BER performance in the low-to-moderate SNR region, where the decoding behavior is influenced by the complete pseudo-codeword spectra and by the Tanner graph cycle histogram, with the time-varying convolutional codes outperforming both the underlying QC block codes and their time-invariant convolutional counterparts.

Reduced complexity decoding strategies for LDPC convolutional codes

While low-density parity-check (LDPC) convolutional codes tend to significantly outperform LDPC block codes with the same processor complexity, large storage requirements and a long initial decoding delay are two issues related to their continuous pipeline decoding architecture [A. Jimenez Feltstrom et al., (1999)]. In this paper, we propose reduced complexity decoding strategies to lessen the storage requirements and the initial decoding delay without significant loss in performance.

Pseudocodeword Performance Analysis for LDPC Convolutional Codes

Message-passing iterative decoders for low-density parity-check (LDPC) block codes are known to be subject to decoding failures due to so-called pseudocodewords. These failures can cause the large signal-to-noise ratio (SNR) performance of message-passing iterative decoding to be worse than that predicted by the maximum-likelihood (ML) decoding union bound. In this paper, we address the pseudocodeword problem from the convolutional code perspective. In particular, we compare the performance of LDPC convolutional codes with that of their wrapped quasi-cyclic block versions and we show that the minimum pseudoweight of an LDPC convolutional code is at least as large as the minimum pseudoweight of an underlying quasi-cyclic code. This result, which parallels a well-known relationship between the minimum Hamming weight of convolutional codes and the minimum Hamming weight of their quasi-cyclic counterparts, is due to the fact that every pseudocodeword in the convolutional code induces a pseudocodeword in the block code with pseudoweight no larger than that of the convolutional codes pseudocodeword. This difference in the weight spectra leads to improved performance at low-to-moderate SNRs for the convolutional code, a conclusion supported by simulation results.

Asymptotically good LDPC convolutional codes based on protographs

LDPC convolutional codes have been shown to be capable of achieving the same capacity-approaching performance as LDPC block codes with iterative message-passing decoding. In this paper, asymptotic methods are used to calculate a lower bound on the free distance for several ensembles of asymptotically good protograph-based LDPC convolutional codes. Further, we show that the free distance to constraint length ratio of the LDPC convolutional codes exceeds the minimum distance to block length ratio of corresponding LDPC block codes.

Trapping set analysis of protograph-based LDPC convolutional codes

It has been suggested that ¿near-codewords¿ may be a significant factor affecting decoding failures of LDPC codes over the AWGN channel. A near-codeword is a sequence that satisfies almost all of the check equations. These near-codewords can be associated with so-called `trapping sets that exist in the Tanner graph of a code. In this paper, we analyse the trapping sets of protograph-based LDPC convolutional codes. LDPC convolutional codes have been shown to be capable of achieving the same capacity-approaching performance as LDPC block codes with iterative message-passing decoding. Further, it has been shown that some ensembles of LDPC convolutional codes are asymptotically good, in the sense that the average free distance grows linearly with constraint length. Here, asymptotic methods are used to calculate a lower bound on the trapping set growth rates for two ensembles of asymptotically good protograph-based LDPC convolutional codes. This can be used to predict where the error floor will occur for these codes under iterative message-passing decoding.

A Low-Cost Serial Decoder Architecture for Low-Density Parity-Check Convolutional Codes

We propose a low-cost serial decoder architecture for low-density parity-check convolutional codes (LDPC-CCs). It has been shown that LDPC-CCs can achieve comparable performance to LDPC block codes with constraint length much less than the block length. The proposed serial decoder architecture for LDPC-CCs uses a single decoding processor. Terminated data frames are sent through the processor iteratively until correctly decoded or a maximum number of iterations is reached. This architecture saves memory consumption and uses a very small number of logic elements, making it especially suitable for strong LDPC-CCs with large code memory. The proposed architecture is realized for a (2048,3,6) regular LDPC-CC on an Altera Stratix FPGA. With a maximum of 100 iterations, the design achieves up to 9-Mb/s throughput using only a very small portion of the field-programmable gate array resources.

Construction of Irregular LDPC Convolutional Codes with Fast Encoding

We propose a novel code design technique for irregular LDPC convolutional codes. The constructed codes can be encoded continuously in real time with the help of a shift-register based encoder. For moderate values of the syndrome former memory, simulation results show that the constructed codes outperform LDPC block codes with comparable hardware (processor) complexity.

A Comparison of ARA- and Protograph-Based LDPC Block and Convolutional Codes

ARA- and protograph-based LDPC codes are capable of achieving error performance similar to randomly constructed codes while enjoying several implementation advantages as a result of their structure. LDPC convolutional codes can be derived from these codes through an unwrapping process. In this paper, we review the unwrapping process as well as the pipeline decoder that allows continuous decoding of LDPC convolutional codes. Computer simulations are then used to demonstrate that the unwrapped convolutional codes achieve a convolutional gain in error performance. We conjecture that this is due to the concatenation of many constraint lengths worth of received symbols in the pipeline decoding process. The consequences of this improved performance are examined in terms of factors related to decoder implementation: processor size, memory requirements, and decoding delay (latency). Finally, given identical protograph kernels, we compare derived block and convolutional codes based on the above measures.