Abstract
Granular gain of low-dimensional lattices based on binary linear codes is estimated using a quantization algorithm which is equivalently a soft-decision decoding of the underlying code. It is shown that substantial portion of the ultimate granular gain is achieved even in limited dimensions.