Noise

Key Takeaways

• Ground roll is the most problematic noise in land seismic acquisition, with energy that can be orders of magnitude larger than the primary reflection signal in the shallow part of the seismic record — ground roll (also called surface waves or Rayleigh waves) propagates along the earth's surface from the seismic source, arriving at each geophone after the direct wave and often overprinting the early primary reflections with high-amplitude, low-frequency, dispersive wave energy; the distinguishing characteristics of ground roll in the time-space (t-x) domain are its high amplitude (typically 10-100 times the primary reflection amplitude in the shallow time range), its low apparent velocity (200-800 m/s, much slower than P-wave reflections at 1,500-5,000 m/s), and its dispersive character (different frequencies travel at different velocities, causing the wave to spread out in time as it propagates); in the f-k (frequency-wavenumber) domain, ground roll occupies a distinct wedge-shaped region below the P-wave velocity trend, which allows f-k velocity filters to separate it from primary reflections without damaging the signal; surface wave noise attenuation requires the combination of source-receiver geometry design (choosing array lengths and orientations that reduce ground roll coupling at the geophones), f-k filtering, and sometimes more sophisticated surface wave modeling and subtraction methods that preserve low-frequency signal content that overlaps with the ground roll frequency range.
• Acoustic noise logging in production wells identifies fluid entry points and channeling behind casing by detecting the characteristic frequencies of turbulent flow — when reservoir fluid enters the wellbore through perforations or through channels in the cement, it flows at high velocity through the restricted openings, generating acoustic noise in the frequency range of 200-2,000 Hz; the noise amplitude is proportional to the flow velocity (and therefore the flow rate), and the frequency spectrum reflects the geometry of the restriction (high-frequency noise from tight perforations, lower-frequency noise from channels or large perforations); acoustic noise logs are run in flowing wells by lowering a sensitive hydrophone or accelerometer into the production tubing and recording the acoustic spectrum as a function of depth; the depth at which high-amplitude, high-frequency noise is detected corresponds to active fluid entry through perforations or behind-casing channels; acoustic noise logging complements production logging spinner surveys (which measure flow velocity directly) and temperature logging (which detects fluid entry through the temperature anomaly created by flowing fluid) to provide a complete picture of the vertical fluid entry profile; this information guides perforation zone isolation, zone re-stimulation, and water shut-off treatments targeted at the specific intervals identified as water-producing by the acoustic noise signature.
• Electronic noise in wireline and LWD sensors is managed through circuit design, signal averaging, and tool quality control, with different sensors having fundamentally different noise floors — resistivity tools that inject current into the formation and measure the resulting potential difference have noise floors determined by the sensitivity of the voltage measurement electronics and the signal strength that reaches the receivers after attenuation through the formation; nuclear tools (density, neutron) count discrete gamma ray or neutron events and have Poisson statistical noise that is reduced by averaging counts over a longer time window (which reduces counting statistics noise but also reduces vertical resolution by averaging over a longer depth interval at the logging speed used); NMR tools have electronic noise from the receiver coil that determines the minimum signal amplitude the tool can detect, limiting its performance in formations with very short T2 relaxation times or very low porosity where the NMR signal is inherently small; the practical consequence of these noise limitations is that tool logging speed (which determines the depth averaging interval) must be selected to achieve an acceptable SNR for the specific formation — thin beds require slow logging speeds for adequate vertical resolution, while massive beds allow faster logging with less statistical noise per depth interval.
• Drilling vibration noise in real-time MWD data must be distinguished from genuine formation variation signals — LWD tools (logging while drilling) simultaneously measure formation properties and transmit their measurements through the mud pulse telemetry channel while the drill string is rotating and the bit is impacting the formation; the combination of bit vibration, stick-slip torsional oscillation, BHA lateral whirl, and shock loads from formation transitions generates mechanical noise in the downhole tool environment that affects measurement quality; resistivity tools mounted in the drill collar above the bit are particularly susceptible to lateral and torsional vibration, which physically moves the antenna array relative to the formation between the generation and detection of the resistivity signal and blurs the measurement; gamma ray detectors are affected by statistical counting noise that compounds with measurement-time averaging in high-vibration environments where shorter depth averaging is used to reduce the effects of stand-pipe pressure fluctuations on mud pulse signal quality; separating genuine formation variation (a real resistivity change at a bed boundary) from vibration-induced noise (a resistivity spike caused by the tool momentarily tilting in the borehole) requires knowledge of the vibration history at the measurement time, provided by downhole accelerometers that record the vibration environment simultaneously with the formation measurements and allow the surface processing software to flag measurements taken during severe vibration events.
• Cultural noise in seismic surveys from roads, pipelines, power lines, and industrial facilities creates coherent noise that is more difficult to remove than random noise — random noise (from wind, thermal motion, and distant microseismic events) averages out when seismic traces are stacked (summed) because its amplitude adds incoherently; coherent noise from a nearby highway has consistent amplitude and phase relationship across neighboring geophones because the source is a localized, persistent vibration generator whose waves propagate systematically to the receivers; f-k and tau-p domain filters can suppress coherent noise whose apparent velocity is different from the primary reflections (a highway produces surface waves at 300-600 m/s while reflections appear at 1,500+ m/s), but coherent noise whose velocity is similar to the target reflection velocity (a pipeline leak generating low-frequency noise with apparent velocity in the reflection range) is difficult to remove without also removing signal; land seismic acquisition design in urban and industrial environments requires careful noise surveys before the main acquisition to characterize the cultural noise sources, their frequency content, and their apparent velocity, so that receiver layouts and filter designs can be optimized to suppress the specific noise environment at that location.