BCH vs LDPC for NAND Flash

A practical comparison of BCH and LDPC error correction for NAND, including overhead, soft information and controller complexity.

BCH and LDPC codes represent two different approaches to NAND error correction. BCH provides algebraic bounded-distance correction with predictable structure. LDPC uses sparse parity checks and iterative decoding, often gaining more from soft information and read-retry evidence.

Core idea

BCH decoders can be attractive when the required correction strength is moderate and deterministic behaviour is important. As raw error rates rise, the parity overhead and complexity needed for stronger BCH can become less attractive. LDPC can offer better coding efficiency in difficult channels but introduces iterative latency, memory traffic and a wider tail of decode effort.

Engineering relevance

The right choice depends on the NAND generation, available spare area, target UBER, controller resources, retry policy and latency budget. A fair comparison should use the same page/block size, error model and reliability target rather than comparing nominal correction counts alone.

Key points

  • BCH offers algebraic bounded-distance correction.
  • LDPC can exploit soft information and iterative refinement.
  • Overhead, latency and hardware cost should be compared at the same reliability target.
  • NAND generation and retry strategy strongly influence the system-level choice.
Public technical resource

This page explains established concepts and engineering principles. Proprietary JobMettle methods and internal ThinkQuick technology are not disclosed.