|
<< Click to Display Table of Contents >> Navigation: Filters > Impedance Post PSDM processing |
Impedance Post PSDM processing performs acoustic impedance inversion on pre-stack depth-migrated (PSDM) seismic gathers. It recovers the subsurface acoustic impedance model directly in the depth domain by solving a linear inverse problem: the observed seismic reflectivity is modeled as the convolution of the time derivative of the wavelet with the logarithm of impedance. The result is a depth-domain impedance section or gather that can be used for reservoir characterization and lithological interpretation.
The inversion is regularized using Tikhonov (first-difference) constraints and an initial impedance model to stabilize the solution in low-frequency and low signal-to-noise zones. The module supports two solver strategies: a Split Bregman L1-norm solver (more robust to noise and sparse features) and a Least-Squares Conjugate Gradient solver (faster for large problems). Optional post-processing filters — including a Laplacian spatial filter, a Laguerre-Gauss FK filter, and a rho (power-spectrum) filter — can be applied to the output to enhance structural continuity or adjust the spectral balance of the result.
Use this module after PSDM migration when you want to convert depth-domain seismic amplitudes into an absolute or relative acoustic impedance volume. An extracted wavelet and an initial background impedance model (for example, from well logs or tomographic velocity) are required as inputs.
The seismic dataset container providing access to the input gathers. Connect the depth-domain PSDM gather dataset that you want to invert for acoustic impedance.
Handle to the SEG-Y file associated with the input dataset. This is used internally to read trace data in the correct format.
Trace header information for the input gathers. Headers carry geometry coordinates, bin assignments, and elevation/datum values used for topography cropping.
The primary seismic gather to be inverted. This should be a depth-domain PSDM gather containing reflection amplitudes that will be used as the right-hand side of the inversion system. Each trace in the gather is processed independently to recover the impedance log at the corresponding lateral position.
Reference to the 2D stack line geometry associated with the input data. Used to define the spatial extent and ordering of traces along the processing line.
Crooked line geometry descriptor for 2D surveys with non-straight acquisition lines. Provides the lateral geometry needed for correct spatial positioning of output traces.
The 3D bin grid defining the spatial layout of CMP bins. Required for 3D datasets to map traces to their correct surface locations.
Sorted and indexed trace header table used to retrieve gathers efficiently during processing. This index is built automatically from the input dataset sort order.
The seismic wavelet used to build the forward-modeling operator. The module takes the time derivative of this wavelet, applies a Hanning taper over the wavelet time window, and uses the result as the convolution kernel in the inversion. This wavelet should match the dominant wavelet present in the input gathers — use a wavelet extracted from the actual seismic data for best results. A single-trace gather is expected, where the trace contains the wavelet samples.
A background acoustic impedance model provided as a gather with the same trace layout as the input seismic data. The logarithm of this model is incorporated into the inversion system as a low-frequency constraint, anchoring the absolute impedance level and preventing drift in zones with weak reflectivity. This model is typically derived from well-log impedance values interpolated or extrapolated across the survey area, or from a tomographic velocity model scaled by density.
The length of the wavelet time window used to extract the derivative operator from the input wavelet, in seconds. Default: 0.1 s (100 ms). Only the first samples of the wavelet up to this window length are used for the convolution operator, and a Hanning taper is applied to reduce end effects. Set this value to approximately the duration of the dominant wavelet in the data. Using a window that is too short will truncate the wavelet and introduce ringing; using one that is too long will include noise or side lobes.
A global amplitude scaling factor applied to both the input seismic data and the wavelet operator before building the inversion system. Default: 1.0. Increasing this value amplifies the weight of the seismic data term relative to the regularization and model constraint terms, which can improve data fidelity when the seismic amplitudes are very small. Decrease it if the inversion is unstable or produces artefacts due to noise amplification.
The weight of the Tikhonov (first-difference) regularization term in the inversion system. Default: 0.01. This term penalizes large vertical variations in the recovered log-impedance and ensures a smooth, stable solution in zones of weak or absent reflectivity. Higher values produce smoother, more model-conformant results; lower values produce results that follow the seismic data more closely but may be noisier. Tune this parameter in conjunction with Amp Factor to balance data fit against solution smoothness.
Selects the numerical solver used to solve the impedance inversion system. Default: enabled (Split Bregman). When enabled, the Split Bregman L1-norm solver is used — this solver is robust to outliers and noise, and promotes sparse solutions, which is often desirable for impedance inversion in areas with distinct layer boundaries. When disabled, a Least-Squares Conjugate Gradient (LSCG) solver is used instead, which minimizes an L2-norm and may converge faster for large, well-conditioned datasets. Try both options if the result quality is unsatisfactory.
When enabled, samples above the surface topography (between the datum level and the actual surface elevation) are zeroed out in the output. Default: enabled. This prevents physically meaningless impedance values from being output for the above-ground portion of depth traces when a floating datum or non-zero elevation is used. Leave this enabled for all surveys with topographic relief.
When enabled, applies a Laplacian spatial filter to the output impedance data. Default: enabled. The Laplacian operator enhances lateral contrasts and sharpens structural features by emphasizing second spatial derivatives, similar to an edge-enhancement filter. This can improve the visibility of faults and layer boundaries in the impedance section. Disable this filter if the output appears over-sharpened or if you prefer a smoother, more geologically interpretable result.
When enabled, applies a Laguerre-Gauss filter in the frequency-wavenumber (FK) domain to the output data. Default: disabled. This filter computes a two-dimensional FK transform, applies a Laguerre-Gauss-shaped weighting that emphasizes energy at specific spatial frequencies, and transforms back. It can be used to attenuate spatially aliased noise or to reshape the 2D spatial spectrum of the impedance output. Enabling this option reveals the Laguerre weight parameter below.
Controls the width of the Gaussian envelope in the Laguerre-Gauss FK filter. Default: 1.0. This parameter is only active when the Laguerre gauss filter is enabled. Larger values broaden the filter and pass a wider range of spatial frequencies; smaller values narrow the filter and concentrate the pass-band. Adjust this parameter to balance spatial resolution against noise in the FK domain.
When enabled, replaces the output gather values with the second vertical difference (Laplacian in depth) of the seismic input, rather than the inverted impedance. Default: disabled. This option is useful for diagnostic or QC purposes: the second difference highlights rapid amplitude changes in depth and can reveal reflector positions and multiple events. It is not intended for production impedance output.
When enabled, applies a rho (frequency-power) filter to the output data before writing. Default: enabled. The rho filter multiplies the amplitude spectrum by f raised to the power specified in FF power, which reshapes the spectral balance of the output. This is commonly applied after impedance inversion to compensate for the spectral coloring introduced by the wavelet convolution operator. Enabling this option exposes the FF power, Shift phase angle, and Normalize amplitudes parameters.
The exponent n applied to frequency in the rho filter (amplitude spectrum multiplied by f^n). Default: 1.0. This parameter is only active when Use ff filter is enabled. A value of 1.0 applies a first-order spectral whitening boost to high frequencies. Increase the value to apply stronger high-frequency enhancement; reduce it (toward 0) to apply less boost. Negative values would attenuate high frequencies. This parameter is visible only when Use ff filter is enabled.
A constant phase rotation angle (in degrees) applied within the rho filter. Default: 0 degrees. This parameter is only active when Use ff filter is enabled. Rotating the phase can help align the impedance output with expected well-log polarity conventions. A value of 90 degrees converts between zero-phase and minimum-phase character, while 180 degrees inverts polarity. This parameter is visible only when Use ff filter is enabled.
When enabled, normalizes the trace amplitudes after applying the rho filter to preserve the overall amplitude level. Default: enabled. Without normalization, the rho filter can significantly amplify or attenuate absolute amplitudes depending on the power exponent. Normalization ensures the output amplitude scale remains consistent with the input. Disable this only if you specifically need the unnormalized filtered amplitudes. This parameter is visible only when Use ff filter is enabled.
When enabled, the module automatically connects its input and output data items to matching datasets in the processing flow without requiring manual wiring. Useful for standard pipeline configurations where dataset names follow conventions.
Controls behavior when NaN or infinity values are detected in the input traces. Fix replaces bad values with zero before processing. Notify reports bad values to the log without modifying data. Continue silently ignores bad values. Choose Fix to ensure the inversion system is not contaminated by invalid samples, which would cause numerical failures in the solver.
When enabled, computes the difference between the output and input gathers and writes it to the Gather of difference output. This is useful for QC: the difference section shows what the module changed in the data, helping you assess whether the inversion result is physically meaningful.
Selects whether the processing runs on the CPU or GPU. GPU execution can significantly accelerate the inversion when large numbers of gathers must be processed. Requires a compatible GPU and appropriate driver installation.
Configures distributed (cluster) execution options. When enabled, processing can be split across multiple compute nodes to accelerate throughput for large 3D datasets. Refer to the cluster configuration guide for node setup.
The minimum number of gathers sent to each compute node or thread in a single processing chunk. Larger bulk sizes reduce inter-process communication overhead but increase memory usage per node. Tune this value based on available memory and the size of individual gathers.
When enabled in distributed mode, caps the number of CPU threads used per cluster node. This prevents the module from consuming all cores on a shared node and allows other jobs to run concurrently.
An optional text suffix appended to the distributed job name. Use this to distinguish multiple simultaneous runs of the same module in the cluster job queue.
When enabled, allows manual specification of CPU core affinity for the processing threads. Enabling this option reveals the Affinity parameter for fine-grained control over which CPU cores are used.
Specifies the CPU core or set of cores to which processing threads are bound. This parameter is only active when Set custom affinity is enabled. Use this on multi-socket servers to keep threads on a specific NUMA node and avoid cross-socket memory latency, which can improve performance for the sparse linear algebra operations used by the solver.
The number of CPU threads used for parallel gather processing. Each thread processes an independent gather, so this value can be set up to the number of physical CPU cores available. Higher values increase throughput on multi-core workstations. Setting this too high relative to available cores may cause context-switching overhead and reduce performance.
When enabled, bypasses this module and passes input data through to the output unchanged. Use this to temporarily disable the impedance inversion step during workflow testing without removing the module from the processing sequence.
The output dataset container carrying the inverted impedance gathers in the time domain. Connect this to the next module in the processing flow or to a storage node for saving results.
Handle to the SEG-Y output file for writing the inverted impedance traces. Connect this to a SEG-Y writer node if you want to export the results directly to disk in SEG-Y format.
Propagated trace headers for the output gathers. Geometry and sorting information from the input is preserved and passed through to the output, ensuring correct positioning of the inverted traces.
The primary output gather containing the inverted acoustic impedance traces. Each output trace corresponds to the recovered log-impedance profile at the lateral position of the corresponding input trace, expressed on the same depth grid as the input. This output is the main product of the module and should be connected to subsequent interpretation or storage steps.
The 2D stack line geometry descriptor propagated to the output. This is carried through unchanged from the input and is required by downstream modules that need line geometry information.
The crooked line geometry descriptor propagated to the output for non-straight 2D survey lines. Passed through unchanged from the input.
The 3D bin grid definition propagated to the output. This is required by downstream 3D visualization and storage modules to correctly position the impedance volume in the survey coordinate system.
The sorted trace header index propagated to the output. This index preserves the gather sort order from the input and allows downstream modules to efficiently retrieve traces by bin or offset.
When Calculate difference is enabled in Settings, this output carries the residual gather representing the difference between the output and input gathers. Inspecting this output helps quantify how much the inversion changed the data relative to the original seismic input and can be used to detect processing artefacts.
An additional output gather explicitly tagged as depth-domain, carrying the same inverted impedance data as the primary output gather but with the domain attribute set to Depth. This output can be connected to depth-domain visualization or storage nodes that require explicit domain labeling to display the data on a depth axis rather than a time axis.