A split-spectrum processing technique for an ultrasonic flaw detection system has been developed, which improves the flaw-to-grain echo ratio in large-grained materials. The enhancement is achieved by partitioning a wide-band received spectrum to obtain frequency shifted bands, which are then processed to suppress the grain echoes with respect to the flaw echo, using a novel signal minimization algorithm. A technique for suppression of grain echoes has also been devised which takes advantage of the fact that the grain echo amplitude changes with the frequency of the incident ultrasound whereas the flaw echo amplitude does not. The combination of this technique and the new flaw detection system greatly improve the capabilities of ultrasonic evaluation of large grain materials.