The maximum likelihood method (MLM) of bottom detection allows substantial reduction in the minimum acceptable SNR in outer beams and, as a result, widening of the survey swath in deep water multi-beam echo sounder (MBES) applications. In this discussion, the arrays are installed in a Mills Cross configuration, with the transmit array oriented along-track and the receive array oriented across-track. The statistics consist of amplitude time samples and samples of phase differences between the two halves of the receive array. Transition to Cartesian coordinates allows for obtaining the inverse of the correlation matrix and, consequently, expression of the normal statistical distribution for the output of the two sub-arrays. Maximizing this distribution by depth (or by depth and surface slope) leads to finding the optimal position of bottom response in range or in time. The location of the signal peak determines the optimal estimate of range and gives bottom depth. Implementation of MLM gives the gain in SNR equal to ~3 -4 dB and up to a 15% increase in swath. The MLM approach is illustrated by data recorded in deep water in South China Sea during a trial of SeaBat 8150 MBES.