% mncNcorrns.tex
\documentstyle[11pt,llncslike,symbols]{article}
\begin{document}
\pagestyle{empty}
\thispagestyle{empty}
\bibliographystyle{plain}
\title{Errata for the paper:\\
 Good Error-Correcting 
	Codes \\ based on Very Sparse Matrices}
\author{David J.C. MacKay\\
	Cavendish Laboratory,
	Cambridge, CB3 0HE. \\
	United Kindom.
	\verb+mackay@mrao.cam.ac.uk+ 
}
\date{\today}
\maketitle
\thispagestyle{empty}
\newcommand{\qoo}[1]{q^{(#1)}_{00}}

% \section{Corrections for the paper \protect\cite{mncN}}

\noindent
(a) In the statement of theorem 1 \cite[page 403]{mncN},
\begin{quote}
 a desired block error probability $\epsilon$
\end{quote}
should be replaced by
\begin{quote}
 any desired block error probability $\epsilon$.
\end{quote}

\medskip

\noindent
(b) 
In the proof of theorem 2 (page 406),
% \cite[page 406]{mncN},
 the sentence 
\begin{quote}

 For large $t$ it is evident (c.f.\ Fig.\ 1) that
 $\left( \lambda H_2^{e} ( w/L ) + \frac{1}{M} \log \qoo{wt} \right)$
 attains its largest value at $w=w^*$.
\end{quote}
should read
\begin{quote}
 For large $t$ it is evident (c.f.\ Fig.\ 1) that,
 if it is positive for any $w\leq w^*$,
 the term
 $\left( \lambda H_2^{e} ( w/L ) + \frac{1}{M} \log \qoo{wt} \right)$
 attains its largest value at $w=w^*$.
\end{quote}

\medskip

\noindent
(c)
 It should have been noted that theorems 3 and 4 (page 403) were
  originally
 proved by Gallager \cite{Gallager63} using a   stronger
 ensemble of codes.
\medskip

\noindent
(d) A factor of $2^{-M}$ was omitted
from the argument of the logarithm  in equation (43).

\medskip

 I thank Kamil Zigangirov for drawing attention to these errors.

\bibliography{/home/mackay/bibs/bibs}
\end{document}