Improved Miscorrection Detection for Generalized Integrated Interleaved BCH Codes
Improved Miscorrection Detection for Generalized Integrated Interleaved BCH Codes
复制标题
DOI：
10.1109/icc45855.2022.9839067
复制
发表时间：
2022-05
期刊：
影响因子：
--
通讯作者：
Zhenshan Xie;Xinmiao Zhang
中科院分区：
文献类型：
--
作者：
Zhenshan Xie;Xinmiao Zhang
广义集成交织（GII）码可以嵌套BCH子码字以形成更强大的BCH码字。 GII 码可实现超高速解码并实现出色的纠错能力。它们是新型存储级存储器 (SCM) 的最佳候选者之一。然而，SCM 需要高码率和短码字长度。在这种情况下，GII子码字的纠错能力较小，子字的误纠会导致严重的性能下降。在之前的工作中，计算高阶嵌套综合症来检测和减轻 GII 解码中的错误。这些计算会导致较长的解码延迟，尽管它们可以通过共享其他解码步骤的硬件架构来实现。本文提出了三种通过调查导致误纠的主要错误模式来优化误纠检测的方法。第一个方案是跳过嵌套综合症检查，以检查不太可能被错误纠正的情况。为了弥补第一种方案造成的性能损失，我们的第二种方法利用 2 位扩展 BCH 码来保护每个子码字。此外，开发了第三种方案，以使用额外的奇偶校验位来保护所有子码字，同时保持码率损失可以忽略不计。还推导出公式来估计可实现的性能。应用所提出的优化，对于输入误码率为 10−3 且具有 3 个纠错子码字的示例 GII 代码，平均嵌套解码延迟减少了 43%，而性能损失和复杂性开销可以忽略不计。
The generalized integrated interleaved (GII) codes can nest BCH sub-codewords to form more powerful BCH codewords. GII codes enable hyper-speed decoding and achieve excellent error-correction capability. They are among the best candidates for the new storage class memories (SCMs). However, SCMs require high code rate and short codeword length. In this case, the GII sub-codewords have small correction capability, and miscorrections on the sub-words lead to severe performance degradation. In previous work, higher-order nested syndromes are computed to detect and mitigate miscorrections in GII decoding. These computations cause long decoding latency, even though they can be implemented by sharing the hardware architecture for other decoding steps. This paper proposes three methods to optimize the miscorrection detection by investigating dominant error patterns leading to miscorrections. The first scheme is to skip the nested syndrome checking for cases that are less likely miscorrected. To make up for the performance loss caused by the first scheme, our second approach exploits 2-bit extended BCH codes to protect each sub-codeword. In addition, the third scheme is developed to protect all sub-codewords using extra parity bits while keeping the code rate loss negligible. Formulas are also derived to estimate the achievable performance. Applying the proposed optimizations, the average nested decoding latency is reduced by 43% for an example GII code with 3-error-correcting sub-codewords at input bit error rate 10−3, while the performance loss and complexity overheads are negligible.