AGC Time Constant
The AGC time constant, more precisely called the AGC gate length or AGC window length, is the duration of the sliding time window over which root-mean-square (RMS) amplitude energy is computed during automatic gain control (AGC) normalisation of seismic trace data. It is the single most consequential parameter in any AGC application because it determines the temporal resolution of the gain function: a short gate length produces a rapidly varying gain that equalises amplitude at small time scales, aggressively suppressing genuine amplitude anomalies caused by lithology and fluid content; a long gate length produces a slowly varying gain that preserves more of the gross amplitude trend while still correcting for the broad energy decay caused by geometric spreading and intrinsic attenuation. In seismic processing practice, gate lengths are specified in milliseconds and typically range from 100 ms (short, aggressive normalisation) to 2,000 ms (long, conservative normalisation). The choice of gate length is rarely innocent: short gates applied to datasets intended for amplitude-versus-offset (AVO) analysis or direct hydrocarbon indicator (DHI) work destroy the lateral amplitude continuity that defines a bright spot or a dim spot, while long gates applied to noisy datasets may fail to suppress the amplitude variations caused by coherent and incoherent noise, leaving the processed section difficult to interpret structurally. For any quantitative amplitude work including AVO, DHI, 4D monitoring, or acoustic impedance inversion, the only acceptable alternative to the AGC gate-length debate is to eliminate AGC entirely and apply true relative amplitude (TRA) processing, which corrects only for physical spreading and attenuation without normalising out formation-controlled amplitude variations.
Key Takeaways
- Short AGC gate lengths (100 to 200 ms) effectively destroy lateral amplitude continuity and should never be used when amplitude information is needed for interpretation: A 150 ms gate at 50 Hz dominant frequency spans approximately 7 to 8 reflection events, meaning the RMS normalisation is computed from just 7 to 8 individual reflectors. Any amplitude anomaly narrower in time than about 75 ms (half the gate) will be partially equalised away, and anomalies narrower than 30 ms (common for thin-bed DHIs in the Cardium or Glauconitic at 40 to 50 Hz) will be completely suppressed. On legacy 1990s WCSB seismic processed with 100 to 150 ms gates, Cardium flat spots and Mannville bright spots were often invisible on the standard processed section even when the original field tapes confirmed genuine amplitude anomalies on TRA-reprocessed versions. Short-gate AGC is acceptable for structural interpretation displays where the goal is maximising reflection coherence across the full two-way travel time range, but must be explicitly flagged in the processing report as incompatible with amplitude-based interpretation.
- Long AGC gate lengths (500 ms to 2,000 ms) preserve more true amplitude variation but still fail to meet the requirements of AVO or 4D seismic analysis: A 1,000 ms gate at 40 Hz spans approximately 40 reflection events, so the normalisation is computed from a broad window that includes contributions from many different reflectors across a range of depths. Gross amplitude trends caused by changing lithology over hundreds of metres will still be partially suppressed, but short-wavelength amplitude anomalies corresponding to individual hydrocarbon-saturated beds (which occupy 20 to 60 ms of two-way time in WCSB clastic plays) will be less affected. Long-gate AGC is sometimes used as a compromise display parameter on sections where both structural and semi-quantitative amplitude interpretation are needed, but AVO gradient maps and near-minus-far amplitude difference extractions must still be performed on TRA-processed data without any gate normalisation, even a long-gate one, because any spatially varying gain corrupts the subtle near-to-far amplitude ratio that defines AVO behaviour.
- The AGC gate length must be documented in the SEG-Y processing report and traceable in the trace-header scalar fields for all seismic data submitted in WCSB regulatory and data-sharing workflows: SEG-Y Revision 2 (2017) standardises bytes 169 to 170 of the binary trace header as the amplitude scalar applied to each sample, allowing the gain values to be recorded alongside the trace data. Processing contractors on major WCSB programmes (Shell Canada Montney 3D, ConocoPhillips Duvernay 3D) are contractually required to log the AGC gate length, target RMS value, and application stage (pre-stack vs post-stack) in the processing sequence documentation. This traceability is required by the CSEG data-exchange standards and by the AER's requirements for seismic data archival under the AER Manual 007 (Petroleum Industry Report). Without gate-length documentation, downstream users cannot determine whether amplitude anomalies on the section are genuine or artefacts of the normalisation, making the dataset unsuitable for Crown land technical evaluation by the AER or NEB.
- AGC gate length selection interacts critically with the noise model, and an inappropriate choice can either fail to suppress noise or suppress signal along with noise: If coherent noise (multiples, ground roll, guided waves) has higher amplitude than the primary reflections in a given time window, a short-gate AGC will scale up the primaries and scale down the noise in equal proportion, partially suppressing the noise but at the cost of introducing gain stripes wherever the noise amplitude is spatially variable. Conversely, if the noise is spectrally similar to the primary reflections, no AGC gate length will separate them, and the normalisation will simply equalise both together. In marine seismic processing in the Gulf of Mexico and offshore eastern Canada (Jeanne d'Arc Basin, Grand Banks), a 500 ms gate is the most common compromise for standard structural display, with separate noise-attenuation workflows applied before AGC to remove water-bottom multiples that would otherwise drive the short-time gain and suppress primaries. The Jeanne d'Arc Basin projects of Equinor, ExxonMobil, and Chevron Canada all document this two-step approach in their published processing reports from Seabird/CGG programmes.
- 4D seismic requires that AGC gate parameters, if used at all, be held absolutely identical across baseline and monitor surveys, and time-lapse amplitude differences must be computed on the raw amplitude versions before any AGC is applied: The amplitude change between a baseline survey and a monitor survey in a producing reservoir is typically 2 to 8% of the absolute amplitude, a signal far smaller than the gain variations introduced by any AGC with a gate length below 2,000 ms. If different gate lengths were applied to the baseline (e.g., 500 ms) and the monitor (e.g., 750 ms), the difference map will show apparent amplitude changes that are entirely attributable to the gain mismatch and have no relationship to fluid substitution. In Alberta's Weyburn-Midale CO2 enhanced-recovery 4D programme, which monitored CO2 flood fronts over multiple vintages from 2000 to 2014, all gain functions were fixed at identical parameters across all vintages and applied only after the baseline-monitor cross-equalization step, with the difference maps computed from the pre-gain versions to preserve the small but economically critical fluid-front amplitude signal.
The Gate Length Selection Problem in Practice
In WCSB seismic processing, gate length selection is often treated as a display-quality decision rather than a scientific one, particularly when processing contracts are structured around deliverable products (migrated stack in SEG-Y format) rather than interpretation objectives. A processor optimising for visual quality on a paper or PDF display will choose 150 to 250 ms gates, which produce clean, coherent-looking sections with good reflection continuity across the full two-way time range. A processor optimising for amplitude integrity will choose the longest possible gate (1,500 to 2,000 ms) or eliminate AGC entirely in favour of TRA, accepting a section that looks noisier at depth in exchange for preserving the amplitude signal at the reservoir interval.
The problem is that most seismic data in the WCSB was acquired and processed for structural mapping, not amplitude analysis, and the processing contractors were instructed to optimise for display quality. When an operator later decides to run an AVO or DHI study on the same data, the short-gate AGC is already embedded in the product and the field tapes may or may not be available for reprocessing. In Alberta, the AER requires that raw field tapes be submitted to the Crown seismic data library within 12 months of acquisition, making TRA reprocessing possible on most datasets back to the late 1980s. For older vintages processed before mandatory archival requirements, the AGC-processed stack is often the only surviving product, and amplitude interpretation must be treated as qualitative at best.
Relationship Between Gate Length and Processing Frequency
The effective temporal resolution of an AGC normalisation is not just the gate length in milliseconds but the ratio of gate length to the dominant period of the seismic wavelet. A 200 ms gate at 25 Hz dominant frequency spans 5 full wavelet cycles (adequate to estimate RMS from multiple independent reflection events), while a 200 ms gate at 100 Hz spans 20 cycles (very aggressive equalisation). This means that high-frequency data (100+ Hz) requires proportionally longer gates to achieve the same degree of amplitude preservation as lower-frequency data at the same gate length in milliseconds. In Montney 3D seismic programmes that achieve 80 to 120 Hz dominant frequency through near-surface static correction and careful source design, processing contractors sometimes apply 300 to 400 ms gates on the rationale that these are "long" relative to typical WCSB practice, but the actual normalisation is equivalent to a 100 ms gate on 25 Hz data and destroys most of the Montney amplitude information.
Fast Facts
The AGC time constant concept originates from analogue electronics, where the RC time constant of the automatic gain control circuit in an amplifier determined how rapidly the gain could track amplitude variations in the input signal; the digital seismic processing equivalent (gate length in ms) inherited the term "time constant" from this analogue heritage. SEG-Y Revision 2 (2017) standardises bytes 169 to 170 for amplitude scalars, enabling gate length traceability in modern processing workflows. The Canadian Society of Exploration Geophysicists (CSEG) published formal guidelines on AGC gate length selection in its Model-Based Interpretation handbook (2019), recommending a minimum gate of 1,000 ms for any dataset used as input to acoustic impedance inversion and prohibiting gates below 500 ms on any data intended for AVO analysis. The dominant processing platforms used by Calgary contractors, SLB Omega and CGG Geovation, both expose gate length as an explicit required parameter in the AGC module, and the parameter value is written into the SEG-Y processing-history header at the time of application. In the Scotian Basin offshore Nova Scotia, CGG processed the 2018 Deep Panuke 3D reprocessing with a 1,500 ms gate specifically to preserve amplitude information for DHI mapping by Shell and ExxonMobil exploration teams, and the project was cited in the CSEG Geoconvention 2019 proceedings as a best-practice example of gate length selection driven by interpretation objective rather than display quality.