|
<< Click to Display Table of Contents >> Navigation: General > Wavelet conversion |
The Wavelet conversion module designs and applies a shaping filter that transforms one wavelet into another. Given an input (original) wavelet and a target (desired) wavelet, the module computes a Wiener-Levinson least-squares shaping filter using cross-correlation and the Toeplitz autocorrelation of the original wavelet, then applies the filter by convolution. The result is a wavelet that best approximates the desired shape in the least-squares sense, as well as the shaping filter itself which can be used to apply the same conversion to seismic data.
Wavelet conversion is used in well-to-seismic tie workflows to match an extracted seismic wavelet to a desired zero-phase Ricker wavelet, or to match a synthetic seismogram wavelet to the phase of the real seismic data. The module can also convert between minimum-phase and zero-phase wavelets as part of a phase correction workflow.
The original wavelet to be converted — the wavelet that the shaping filter will act on. Connect the wavelet gather that represents the current wavelet present in the data (for example, a wavelet extracted from seismic data using the Wavelet detection module).
The desired target wavelet — the shape the output wavelet should match as closely as possible. Connect the wavelet gather representing the desired output shape (for example, a synthetic Ricker wavelet at the correct dominant frequency, generated using the Wavelet generator module).
The pre-whitening (noise) level applied during the Wiener-Levinson filter design, expressed as a percentage of the peak autocorrelation value. Default: a small percentage value. Pre-whitening stabilises the filter solution by preventing division by very small spectral values — this is particularly important when the input wavelet has spectral notches or a limited bandwidth. Increasing the noise level produces a smoother, more stable filter at the cost of a less exact wavelet conversion. A value of 0.1–1% is typically appropriate; use higher values (5–10%) if the input wavelet is noisy or band-limited.
The length of the Wiener shaping filter in seconds (the filter is then converted to samples internally using the wavelet sample interval). A longer filter can capture more complex phase and amplitude corrections between the original and desired wavelet, but may also fit noise in the input wavelet. The filter length should be at least as long as the dominant period of the wavelet. For typical seismic wavelets with dominant frequencies of 30–60 Hz, a filter length of 0.1–0.2 s is usually sufficient.