Marine seismic data acquisition has conventionally been carried out with source arrays positioned at a single depth. The resulting source-ghost notch is often taken as the limit for the bandwidth of the data. In broadband seismic data acquisition, multilevel sources (MLSs) are now being used to fill in these notches. Here we present a modeling study in which variable source-depth acquisition (VSDA) is compared with conventional single-level source and MLS acquisition. Our results indicate that an overall (not all over) improvement of signal-to-noise ratio is obtained using VSDA compared with the other source strategies. Due to the nature of VSDA, having different far-field signatures for different source depths, the amplitude variation with offset behavior of the processed VSDA data is not ideal. In stacked images, these variations are mostly seen in very shallow reflectors ().