Communication and Information Theory

Communication and Information Theory

Eshed Ram


Viterbi Faculty of Electrical Engineering, Technion

LDPC Codes with Sub-Block Access

Low-density parity-check (LDPC) codes and their low-complexity iterative decoding algorithm are a powerful method to achieve reliable communication and storage with rates that approach Shannon's theoretical limit. When used in data-storage applications, unlike in communications, re-transmissions are not possible, and any decoding failure implies data loss. To meet the stringent reliability specifications of storage devices, the error-correcting code must protect against extremely high noise levels (although most noise instances are much milder), requiring very large block lengths and complex decoding, thus degrading the latency and throughput of the device. A possible solution to this problem is codes with sub-block access that enable decoding small sub-blocks (local decoding) for fast read access while providing a high data-reliability "safety net" decoding over the large code block (global decoding). In this research, we develop an analytical framework for LDPC and spatially coupled LDPC codes with sub-block access and propose code constructions that offer sub-block decoding capabilities. In addition, we introduce a novel BP-based decoding mode which we call semi-global decoding, in which in addition to the requested sub-block, the decoder has access to some prescribed number of sub-blocks around it. We derive (and simplify) a density-evolution analysis of semi-global decoding and examine its performance over the memoryless binary erasure channel and over a channel in which variability is introduced to the channel quality.

  This research was conducted toward the Ph.D degree under the supervision of Prof. Yuval Cassuto.

Date: Thu 17 Sep 2020

Start Time: 14:30

End Time: 15:30 | Electrical Eng. Building