Description

This module applies gapped predictive deconvolution to gathers that have been transformed to the tau-p (ray-parameter) domain, using time-variant design windows shaped as ellipses in the trace/time plane rather than simple flat time gates. Each elliptical boundary is defined by an apex time and a velocity, so several ellipses can be stacked to separate zones dominated by different apparent-velocity trends (for example primaries versus multiples) and a distinct decon operator is designed and applied within each zone. This gives more geologically consistent multiple attenuation than a single, offset-independent design window. You can preview and edit the elliptical windows directly as an overlay on the input gather, and optionally review before/after autocorrelation panels to judge the periodic energy removed.

Input data

Input gather

A gather that has already been transformed to the tau-p (ray-parameter) domain and carries a ray-parameter (P value) trace header. This module requires that header to build the elliptical design windows; gathers without it are rejected.

Parameters

Use L1 solver

When enabled, the decon operator is solved with a robust, iterative L1-norm solver instead of the standard least-squares (L2) solution. This makes the operator design less sensitive to isolated large-amplitude spikes or bursts of noise, at the cost of extra computation.

Default: Off. Turn on if the data contains noise bursts or spiky outliers that distort the standard least-squares design.

Number of L1 solver iterations

Maximum number of iterations the L1 solver performs while converging on the operator. More iterations can improve robustness and accuracy but increase run time.

Default: 100. Minimum 1. Only relevant when "Use L1 solver" is on.

Sparsity damping

Small stabilizing value used internally by the L1 solver to avoid numerical instability when residuals approach zero. Rarely needs adjustment.

Default: 0.0001. Minimum 1e-9. Only relevant when "Use L1 solver" is on.

Overlap taper window

Length of the smoothing taper used to blend the decon result across the boundary between neighbouring elliptical design windows, so the operator does not change abruptly as you cross from one window to the next.

Default: 0.10 ms. Increase if you see abrupt character changes at window boundaries in the output.

Predictive interval

The prediction distance (gap) of the predictive deconvolution: the operator is designed to predict, and remove, energy that repeats at this lag or longer. A short gap approaches spiking deconvolution (broadens the spectrum aggressively and can distort primaries); a longer gap targets periodic reverberations and multiples whose period is close to this value while leaving shorter-period primary character largely intact.

Default: 0.02 ms. Set it close to the dominant multiple period you want to attenuate (for example the two-way water-bottom travel time).

Noise

Prewhitening level added to the zero-lag autocorrelation before inverting for the operator, expressed as a percentage. Higher values produce a more stable but gentler decon; lower values are more aggressive but more sensitive to noise.

Default: 0.05 (5%). Increase on noisy data to keep the operator well-behaved.

Length

Total length of the predictive deconvolution operator (design window). It must be greater than the predictive interval. Longer operators can suppress multiples with more complex periodicity but require more stable statistics to design reliably.

Default: 0.2 ms. Minimum 0.

Mix trace

Number of neighbouring traces combined when estimating the autocorrelation used to design each operator. Mixing traces produces a more statistically stable operator on noisy or low-fold data, at the cost of some lateral resolution in how the operator adapts trace-to-trace.

Default: 1 trace. Minimum 1.

Velocity mute

Apparent-velocity threshold used to exclude very fast, high-apparent-velocity energy (such as direct or refracted arrivals) from the autocorrelation estimate, so it does not bias the design of the predictive operator toward non-reflection energy.

Default: 8000 m/s.

Nominal offset

Reference offset used to normalize the ray-parameter axis when building the elliptical design windows. It effectively fixes the scale of the ellipses relative to the gather's actual offset range, so choose a value representative of your maximum useful offset.

Default: 3000 m. Minimum 0.1.

Ellipse X-Y coeff

Scaling coefficient applied to the ray-parameter axis when constructing the elliptical window boundaries. It adjusts the aspect ratio (how elongated versus circular) of the ellipses in trace/time space, letting you fine-tune how the window shape follows the moveout trend in your data.

Default: 1. Minimum 0.001.

Elliptical windows

A table where each row defines one elliptical boundary that separates two decon design zones. Add one row per boundary, from shallow/steep to deep/flat; the module automatically fills the zone between consecutive boundaries (and above the first / below the last) with its own decon operator. The boundaries can also be dragged interactively as an overlay on the input gather display.

Tau0

Apex (zero-offset) time of this elliptical boundary — the two-way time at which the boundary curve crosses zero ray-parameter, analogous to picking a horizon on a stacked section.

Default: 1. Minimum 0. Set to the approximate zero-offset arrival time of the event you want this boundary to follow (for example a multiple generator horizon).

Velocity

Moveout velocity that shapes the curvature of this elliptical boundary away from its apex. Choose a value close to the stacking/NMO velocity of the event you want the boundary to track, so the design zones properly separate primaries from multiples across offset.

No fixed default; set per boundary to match the local moveout velocity trend.

Autocorrelation QC

Autocorrelation

When enabled, autocorrelation panels are computed for both the input and the deconvolved output gather, letting you visually confirm that periodic (multiple) energy has been reduced after processing.

Default: Off.

Time length

Length of the correlation window used to compute the before/after autocorrelation QC panels. A longer window reveals longer-period multiples but averages over more of the gather.

Default: 0.5 ms. Minimum 0. Only used when "Autocorrelation" is on.

Advanced

Using Balance

When enabled, the input is amplitude-balanced within a reference time window before the decon operator is designed and applied, and the original amplitude trend is restored afterward. This stabilizes the operator design in gathers with strong overall amplitude variation (for example AVO or geometric-spreading effects) that would otherwise bias a straightforward decon.

Default: Off.

Start window

Start time of the reference window used for amplitude balancing, when "Using Balance" is on.

Default: 0.01 ms. Minimum -9999.

End window

End time of the reference window used for amplitude balancing, when "Using Balance" is on.

Default: 0.05 ms. Minimum -9999.

Output data

Output gather

The tau-p gather after elliptical predictive deconvolution, with multiple/reverberatory energy attenuated within each design zone.

Autocorrelation input

QC autocorrelation of the input gather, produced when "Autocorrelation" is enabled, for comparison against the output autocorrelation.

Autocorrelation output

QC autocorrelation of the deconvolved output gather, produced when "Autocorrelation" is enabled. A reduction in periodic energy compared to the input autocorrelation indicates successful multiple attenuation.

References

Peacock, K. L. and Treitel, S., 1969, Predictive deconvolution: Theory and practice: Geophysics, 34, 155-169.

Taylor, H. L., Banks, S. C. and McCoy, J. F., 1979, Deconvolution with the l1 norm: Geophysics, 44, 39-52.