Achieving 1/2 log (1 + SNR) on the AWGN channel with lattice encoding and decoding

Uri Erez*, Ram Zamir

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review


We address an open question, regarding whether a lattice code with lattice decoding (as opposed to maximum-likelihood (ML) decoding) can achieve the additive white Gaussian noise (AWGN) channel capacity. We first demonstrate how minimum mean-square error (MMSE) scaling along with dithering (lattice randomization) techniques can transform the power-constrained AWGN channel into a modulo-lattice additive noise channel, whose effective noise is reduced by a factor of √1+SNR/SNR. For the resulting channel, a uniform input maximizes mutual information, which in the limit of large lattice dimension becomes 1/2 log(1 + SNR), i.e., the full capacity of the original power constrained AWGN channel. We then show that capacity may also be achieved using nested lattice codes, the coarse lattice serving for shaping via the modulo-lattice transformation, the fine lattice for channel coding. We show that such pairs exist for any desired nesting ratio, i.e., for any signal-to-noise ratio (SNR). Furthermore, for the modulo-lattice additive noise channel lattice decoding is optimal. Finally, we show that the error exponent of the proposed scheme is lower bounded by the Poltyrev exponent.

Original languageEnglish
Pages (from-to)2293-2314
Number of pages22
JournalIEEE Transactions on Information Theory
Issue number10
StatePublished - Oct 2004


Dive into the research topics of 'Achieving 1/2 log (1 + SNR) on the AWGN channel with lattice encoding and decoding'. Together they form a unique fingerprint.

Cite this