|
<< Click to Display Table of Contents >> Navigation: Filters > DC bias removal |
DC bias removal eliminates low-frequency baseline drift and constant offset components from seismic traces. A DC bias is a non-zero mean value that shifts the entire trace amplitude up or down, which can corrupt amplitude-dependent processing steps such as deconvolution, spectral analysis, and stacking. This module removes that bias by fitting a trend to each trace and subtracting it, leaving only the true seismic signal.
DC bias commonly arises from instrument electronics, near-surface conditions, or data format conversion artefacts. This module is typically applied early in a processing flow, before spectral balancing, deconvolution, or any operation that assumes zero-mean traces. Three correction modes are available: a simple constant-level removal, a linear ramp removal, and a higher-order polynomial trend removal for more complex baseline shapes.
Connect the seismic data source to be corrected. This can be any standard seismic data connector in the processing flow, such as a SEG-Y file handle, an in-memory gather item, or the output of a preceding processing module. The module processes all traces present in the connected data.
The input seismic gather to be processed. Each trace in the gather is corrected independently. The gather can be in any domain (shot, CMP, offset, etc.) and the sample interval is read automatically from the gather header — no manual configuration is required.
Selects the shape of the trend that is estimated and removed from each trace. Three options are available:
Constant (default) — removes a uniform amplitude offset from the trace. The mean level is estimated using a half-period sine weighting function applied across the full trace length. This weighting gives less influence to samples at the very start and end of the trace, producing a smooth, edge-stable result. Use this option when the baseline shift is approximately flat across the entire trace.
Linear — fits and removes a straight-line (first-order polynomial) trend from each trace using singular value decomposition (SVD) least-squares fitting. This is appropriate when the baseline drifts linearly over time, for example due to instrument warm-up effects or acquisition-related ramp artefacts.
Non linear — fits and removes a higher-order polynomial trend whose degree is set by the Polynomial Degree parameter. The polynomial is estimated per trace via SVD least-squares fitting. Use this option only when the baseline drift has a clearly curved or oscillatory shape that cannot be captured by a constant or linear model. Caution: high polynomial degrees can inadvertently fit and remove genuine low-frequency seismic energy — verify results on a sample of traces before applying to the full dataset.
Controls the order of the polynomial used to model the baseline trend when DC bias type is set to Non linear. The default value is 5, and the minimum value is 1. A degree of 1 is equivalent to a linear fit; degree 2 is quadratic; degree 5 allows a moderately complex curved trend. This parameter is ignored when DC bias type is set to Constant or Linear. As a practical guide, start with low degrees (2–3) and increase only if the residual drift is clearly non-linear. Avoid degrees above 7–8 on typical seismic records, as the polynomial may oscillate and remove valid long-period reflections.