Code size constraints in b-symbol read channels: A bound analysis
Gyanendra K. Verma, Nupur Patanker, Abhay Kumar Singh
TL;DR
This work extends the study of $b$-symbol read channels to general $b\ge 2$ by deriving a suite of upper and lower bounds on $A_b(n,d,q)$, including a generalized recurrence, Johnson-type bounds, Gilbert–Varshamov-type bounds, and Elias-type bounds. It further develops a linear programming bound for $b$-symbol codes by embedding them into a larger Hamming-distance code via a simplex-based construction, linking $d_b$ to $d_H$ and enabling LP optimization with Krawtchouk polynomials. The paper also demonstrates improvements in GV-type bounds over prior results and provides constructions and examples that illustrate bound tightness and potential gaps. Collectively, these results advance the understanding of burst-error models in $b$-symbol read channels and lay groundwork for constructing codes that approach these fundamental limits in practical storage and communication systems.
Abstract
In classical coding theory, error-correcting codes are designed to protect against errors occurring at individual symbol positions in a codeword. However, in practical storage and communication systems, errors often affect multiple adjacent symbols rather than single symbols independently. To address this, symbol-pair read channels were introduced \cite{Yuval2011}, and later generalized to $b$-symbol read channels \cite{yaakobi2016} to better model such error patterns. $b$-Symbol read channels generalize symbol-pair read channels to account for clustered errors in modern storage and communication systems. By developing bounds and efficient codes, researchers improve data reliability in applications such as storage devices, wireless networks, and DNA-based storage. Given integers $q$, $n$, $d$, and $b \geq 2$, let $A_b(n,d,q)$ denote the largest possible code size for which there exists a $q$-ary code of length $n$ with minimum $b$-symbol distance at least $d$. In \cite{chen2022}, various upper and lower bounds on $A_b(n,d,q)$ are given for $b=2$. In this paper, we generalize some of these bounds to the $b$-symbol read channels for $b>2$ and present several new bounds on $A_b(n,d,q)$. In particular, we establish the linear programming bound, a recurrence relation on $A_b(n,d,q)$, the Johnson bound (even), the restricted Johnson bound, the Gilbert-Varshamov-type bound, and the Elias bound for the metric of symbols $b$, $b\geq 2$. Furthermore, we provide examples demonstrating that the Gilbert-Varshamov bound we establish offers a stronger lower bound than the one presented in \cite{Song2018}. Additionally, we introduce an alternative approach to deriving the Sphere-packing and Plotkin bounds.