The application of phased arrays in NDT applications is growing. State of the art ultrasonic arrays consist of many small piezo-electric elements that can be excited separately to synthesize a desired wave front. This may vary from simple plane waves to complex-shaped focusing wave fields. An implicit requirement is that the source strength (sensitivity) of all elements is equal, to prevent artifacts in the generated wave front. The same holds for the detection of ultrasonic waves. In typical commercial ultrasonic arrays, however, sensitivity variations can be significant: amplitude variations of ± 3 dB are not uncommon. Pulse-echo data can be used for calibration of element strengths. The application of pulse-echo data, however, has some limitations: its performance may deteriorate when sensing irregularly-shaped media and the application is limited to cases with identical element sensitivity in transmission and detection. For ultrasonic measurements this is not necessarily true when separate transmit and receiver arrays are used, but is also not evident when the same array is used. A new data-driven method is demonstrated that can be used to determine the frequency-dependent sensitivity of each element in a phased array in emission and detection separately. © 2011 American Institute of Physics.