Time‐domain beamforming has been formulated as an optimization problem with simulated annealing applied to search the large parameter space of source bearings and time series [Kupermanet al., “Optimal time‐domain beamforming with simulated annealing including application ofa prioriinformation,” J. Acoust. Soc. Am. (in press)]. Both the generality and the efficiency of this signal processing algorithm have been improved. The time‐domain beamformer was originally applied to problems involving waveforms of compact support. For more general types of signals and multiple sources, an additional constraint of minimizing the correlation between signals is required to suppress ambiguous periodic functions in the time series (each of which is common to signal pairs). The efficiency of the simulated annealing algorithm for time‐domain beamforming has been improved using the concept of latent heat release at phase transition. Simulated annealing is analogous to the cooling of a liquid to form a perfect crystal and involves artificial quantities analogous to temperature and energy. With an adaptive latent‐heat cooling schedule in which the temperature depends on the energy, the system has a greater tendency to remain in low‐energy states. The convergence rate of the optimization procedure is thus significantly increased.