We consider the problem of transmitting a Gaussian source over a 
block-fading  Gaussian channel. We analyze several source/channel models 
in terms of distortion exponent, that is, how fast the end-to-end 
average distortion decays to zero as the SNR increases. As opposed to 
most of the literature, we consider SNR-asymmetric channels, correlated 
fading and source side information. In particular, we focus on 
Successive Refinement (SR) and on Multiple Description Coding (MDC) with 
more than two descriptions. Our contributions include: 1) showing that 
determining the distortion exponent reduces to solving a linear 
programming problem; 2) proposing a general framework to study the 
distortion exponent for MDC with any number of descriptions; 3) 
quantifying analytically the impact of source side information and/or of 
fading correlation in terms of offset with respect to the case of no 
side-information and/or uncorrelated fading case; 4) studying the 
distortion exponent  for interference channels.