Constant Envelope Signaling in MIMO Channels

TL;DR

This paper investigates the capacity and optimal signaling strategies for MIMO channels under a constant envelope constraint, revealing the structure of capacity-achieving distributions and degrees of freedom.

Contribution

It derives the capacity-achieving distribution for 2x2 MIMO channels with constant envelope signals and characterizes the degrees of freedom for full-rank channels.

Findings

Capacity-achieving distribution has finite mass points on the circle.

Degrees of freedom for n x n channels is n-1 with uniform hypersphere distribution.

Power allocation schemes differ from conventional methods as channel condition number varies.

Abstract

The capacity of the point-to-point vector Gaussian channel under the peak power constraint is not known in general. This paper considers a simpler scenario in which the input signal vector is forced to have a constant envelope (or norm). The capacity-achieving distribution for the non-identity $2\times 2$ MIMO channel when the input vector lies on a circle in $\mathbb{R}^2$ is obtained and is shown to have a finite number of mass points on the circle. Subsequently, it is shown that the degrees of freedom (DoF) of a full-rank $n$ by $n$ channel with constant envelope signaling is $n-1$ and it can be achieved by a uniform distribution over the surface of the hypersphere whose radius is defined by the constant envelope. Finally, for the 2 by 2 channel, the power allocation scheme of the constant envelope signaling is compared with that of the conventional case, in which the constraint is on the average transmitted power. It is observed that when the condition number of the channel is close to one, both schemes have a similar trend while this is not the case as the condition number grows.