Delay constrained linear transmission (LT) strategies
are considered for the transmission of composite Gaussian
measurements over an additive white Gaussian noise fading
channel under an average power constraint. If the channel state
information (CSI) is known by both the encoder and decoder,
the optimal LT scheme in terms of the average mean-square
error distortion is characterized under a strict delay constraint,
and a graphical interpretation of the optimal power allocation
strategy is presented. Then, for general delay constraints, two
LT strategies are proposed based on the solution to a particular
multiple measurements-parallel channels scenario. It is
shown that the distortion decreases as the delay constraint is
relaxed, and when the delay constraint is completely removed,
both strategies achieve the optimal performance under certain
matching conditions. If the CSI is known only by the decoder, the
optimal LT strategy is derived under a strict delay constraint.
The extension to general delay constraints is elusive. As a first
step towards understanding the structure of the optimal scheme
in this case, it is shown that for the multiple measurementsparallel
channels scenario, any LT scheme that uses only a oneto-one
linear mapping between measurements and channels is
suboptimal in general.