Paper Summary
Amplitudes recorded by seismic receivers are usually clipped when the signal level exceeds the
maximum signal that can be safely recorded by the instrument. Clipping leads to spectral leakage of energy from the signal into the noise band causing spectral distortion. To improve the fidelity of seismic recordings, this spectral distortion should be mitigated in the seismic processing sequence. To this end, the author has investigated Fourier transform based interpolation methods. These methods usually apply iterative reconstruction techniques such as the projection onto convex sets algorithm and can be readily repurposed for the reconstruction of clipped seismic amplitudes. In more detail, the proposed algorithm involves identification of clipped samples and reconstruction of the correct amplitudes by iteratively thresholding the spectrum. This proposed algorithm allows effective “declipping” of the seismic data. For a typical ocean-bottom seismometer record, clipping only affects the near-offset traces and corrupts only a few samples. However, carefully mitigating this clipping in ocean-bottom hydrophone data has allowed to improve the signal-to-noise ratio of the clipped samples by 14 dB. The improvement of the signal fidelity early in the seismic processing sequence benefits subsequent transform-based processing steps such as up/down deconvolution.