Abstract
A derivation and discussion of a mutual information lower bound for broadband acoustic channels under a Gauss-Markov assumption is provided. This lower bound is associated with dynamic broadband acoustic response functions that obey a Gauss Markov law. Considered here is the difference of the mutual information between source and receive signals given the acoustic response and the mutual information between acoustic response and receive signal given the source signal. The derivation illuminates the features of this mutual information and permits identifying two key information loss mechanisms. The first is an added self noise term proportional to signal power that is only associated with the innovation variance of the Gauss Markov response. The second term also increases with the innovation variance but is associated with the uncertainty in the state of the channel operator at the initial time of signaling. This later variance is identified as a forward-backward channel estimation error. Platform motion effects and their impact on these two loss mechanisms are discussed and the structure of an effective signal to noise ratio (eSNR) ceiling at high SNR are identified.