An algorithm for separation of signals according to their coherency is proposed. The algorithm, called Superstack, is used on common depth point data after normal moveout corrections have been applied. The algorithm can be regarded as an iterative stacking procedure. After each stack, input values are changed depending on the consistency of the output of the previous stack. The Superstack algorithm is able to provide better separation of signals showing a different degree of horizontal consistency (coherency) than the normal horizontal stack. All noise signals in a CDP‐gather may be assumed to show less horizontal consistency than the primary reflections. On this basis the algorithm will give an improved primary/multiple and an improved primary/noise ratio. The Superstack is very general, should not be especially expensive in terms of computer time, and can easily be implemented in routine processing. Copyright © 1979, Wiley Blackwell. All rights reserved
