Deconvolution
SEARCH
Either gap or spiking deconvolution operators may be computed from autocorrelation functions. The operator is designed using Wiener-Levinson recursion. Zero-phase operators can also be applied if required. This can be applied to either prestack or poststack data.
Design time-windows are specified within which the autocorrelation function is computed and can be spatially varying (e.g. subline-crossline-offset or based upon a horizon or water depth). Multiple windows can also be specified for time-varying deconvolution, with the deconvolution parameters able to vary between the different windows. Application time-windows can be similarly specified and do not need to match the design time-windows.
Autocorrelations may be averaged across traces before computing the deconvolution operator. The averaging can be across an entire group or part of a group of traces (shot record, for example), or across specified ranges of traces (three filters per shot, for example) or across a running average of user specified length (different filter for each trace computed from averaged autocorrelations). All options are available for prestack data, but only the running average method is available poststack.
.jpg)
Ideal seismic response would be a single sharp reflection for each sub-surface rock layer boundary. Actual seismic response is less than ideal because our output pulse is not perfectly sharp and changes its shape while passing through the Earth.Deconvolution 'deconvolves' our output pulse from the seismic response and converts it into a cleaner, sharper, less confusing pulse.
Q compensation
Q compensation corrects for the attenuation of high frequencies by the earth. The compensation applied may be either for both amplitude and phase corrections, or for either phase or amplitude only. The compensation applied may be limited to a maximum frequency. For areas with variable water depths, the start time for the Q compensation may be specified to follow the water bottom reflection time and handles the change in travel path within the water-layer as a function of offset. If no Q values are available, Q estimation can be performed using spectral ratio analysis.
Spectral whitening
Spectral whitening techniques level or broaden the bandwidth of seismic data. In addition to deconvolution methods outlined above, other methods include frequency dependent balancing and frequency-dependent AGC.
Tau-p deconvolution
In the x-t domain, reverberation energy on shot records is not periodic at non-zero offset because of the effects of move out. To combat this, prestack deconvolution may be applied in the Tau-p domain where, for a 1.5D earth, reverberation is periodic at all p values. Data are forward transformed to the Tau-p domain with a sufficient p range to provide an accurate inverse transform. Gap deconvolution, as described above, is applied in this domain before the inverse transform is performed. The gap may vary as a function of p according to a supplied constant velocity value. Typically minimal trace averaging will be performed, since the deconvolution operator will need to vary as a function of p. Tau-p deconvolution can be applied in either shot or common receiver domain. This approach is routinely used for multiple attenuation on shallow marine data.
