Skip to content

Seismic Sources, Detectors and Noise

Seismic sources, Vibroseis correlation, air-gun bubbles, geophones, hydrophones, sampling, ghosts, coherent and random noise and arrays for CSIR NET.

  • 17 min read
  • Updated
On this page · 12 sections
  1. Prerequisites
  2. The Hook: The Fan That Hides the Oil
  3. What a Good Source Must Do
  4. Detectors
  5. Noise
  6. Memory Tricks
  7. Don’t Confuse
  8. Common Mistakes
  9. Practice
  10. Quick Revision Table
  11. Key Takeaways
  12. What’s Next

Prerequisites

The Hook: The Fan That Hides the Oil

How CSIR NET asks it. Part C (Geophysics) questions here are short and specific: which source suits which setting, what Vibroseis correlation produces, why air guns are arrayed, what a geophone measures and how its response falls below the natural frequency, Nyquist and aliasing, ghost notches, CMP fold, array response and the \(\sqrt N\) rule for stacking random noise.


What a Good Source Must Do

RequirementWhy it matters
Enough energyDeep targets, long offsets, absorbing near-surface
Broad bandwidthResolution depends on bandwidth, not only on the dominant frequency
Short, known, repeatable waveletDeconvolution and time-lapse (4D) work need a stable signature
Little source-generated noiseGround roll and air blast compete with reflections
Safe, permitted, affordableExplosives in towns or protected seas are often not allowed

Land sources

SourceHow it worksStrengthsLimits
Explosives (dynamite)Charge fired in a shot hole, ideally below the weathered layerImpulsive, broadband, high energy; good coupling below weatheringDrilling cost, permits, safety; not repeatable exactly
VibroseisTruck-mounted baseplate driven hydraulically with a controlled frequency sweep lasting several secondsControlled spectrum; usable in towns; several vibrators and sweeps can be summedNeeds correlation; harmonic distortion; poor on soft ground
Weight drop / accelerated weight dropHeavy mass dropped or driven onto a plateCheap, repeatable, quickLow energy; shallow targets
Sledgehammer, small shotgun-type sourcesHammer on a plate; small cartridge fired into the groundVery cheap; engineering refractionVery shallow only; stacking needed

Vibroseis correlation. The vibrator puts a long, low-amplitude sweep \(s(t)\) (for example a linear upsweep over a few seconds through a band of tens of hertz) into the ground. Each reflection returns a delayed, scaled copy of the sweep, and the raw record is an unreadable overlap. Cross-correlating the record with the pilot sweep compresses every copy into a short wavelet:

$$\text{correlated trace} = s(t) \star \left[s(t) * r(t)\right] = \left[s(t)\star s(t)\right] * r(t)$$

where \(\star\) is cross-correlation, \(*\) is convolution and \(r(t)\) the reflectivity. The effective wavelet is the autocorrelation of the sweep, the Klauder wavelet: zero-phase, with bandwidth equal to the sweep bandwidth. The usable correlated record is the raw record length minus the sweep length (the “listen time”).

Marine sources

SourcePrincipleUse
Air gunHigh-pressure compressed air released suddenly into waterThe standard marine source
GI (generator–injector) gunA second air pulse injected into the expanding bubble to damp its oscillationCleaner, high-resolution signature
Water gunA piston of water ejected; the cavity implodesLittle bubble pulse; precursor in the signature
SparkerElectrical discharge between electrodes vaporises waterShallow, high-resolution surveys
BoomerElectromagnetically driven plate pushes waterShallow, high-resolution
Chirp sub-bottom profilerSwept-frequency transducer with correlationVery high resolution of the upper sediments
Marine vibratorUnderwater vibrating plate with a sweepLower peak pressure alternative

Explosives are now rarely used offshore because of their effect on marine life and their poor repeatability.

The air-gun bubble. After the initial pressure pulse, the air bubble overshoots, collapses and re-expands several times. Each oscillation radiates a weaker pulse: the bubble pulse, which looks like a train of multiples. Its period follows the Rayleigh–Willis relation,

$$T_b \propto \frac{(P\,V_g)^{1/3}}{P_h^{5/6}}$$

where \(P\) is the firing pressure, \(V_g\) the chamber volume and \(P_h\) the absolute hydrostatic pressure at gun depth. Bigger guns give longer bubble periods; deeper guns give shorter ones. Remedies:

  • Tuned arrays: guns of different volumes fired together. Their primary pulses add in phase, their bubbles have different periods and add incoherently. The array improves the primary-to-bubble ratio.
  • Clusters: guns fired close together so their bubbles interact and damp.
  • Signature deconvolution in processing, using a measured or modelled far-field signature.

Ghosts

A source or receiver below the sea surface sees a second, polarity-reversed copy of every wave reflected from the surface (\(R \approx -1\)): the ghost. For depth \(d\) and water velocity \(V_w\), the ghost lags by \(2d\cos\theta/V_w\) and the summed response has notches at

$$f_n = \frac{n\,V_w}{2d\cos\theta}, \qquad n = 0, 1, 2, \ldots$$

At vertical incidence a streamer at 7.5 m has its first non-zero notch at \(1500/15 = 100\) Hz. Shallow towing preserves high frequencies but costs low frequencies (the notch at 0 Hz) and adds weather noise; deep towing does the reverse. Dual-sensor streamers and ocean-bottom recording combine pressure (hydrophone) and vertical velocity (geophone) data, whose ghosts have opposite polarity, to remove the receiver ghost. On land, a shot fired below the base of weathering produces a source ghost from that interface and from the surface.


Detectors

The moving-coil geophone

A coil suspended on springs moves relative to a permanent magnet fixed to the case. When the ground moves, the case (and magnet) moves with it, the inertial coil lags, and the relative motion induces a voltage proportional to the relative velocity (Faraday’s law). The system is a damped harmonic oscillator:

$$\ddot{y} + 2h\omega_0\dot{y} + \omega_0^2 y = -\ddot{u}$$

with \(y\) the coil displacement relative to the case, \(u\) the ground displacement, \(\omega_0 = 2\pi f_0 = \sqrt{k/m}\) and \(h\) the damping ratio.

PropertyBehaviour
OutputVoltage proportional to ground velocity above \(f_0\)
Natural frequency \(f_0\)Set by spring stiffness and coil mass; low-frequency geophones are used for refraction and deep reflection work
Below \(f_0\)Response falls off as a second-order system: about 12 dB per octave
DampingUsually set (by a shunt resistor) at a substantial fraction of critical, around 0.6–0.7, to give a flat response with little ringing near \(f_0\)
ComponentsVertical (P-wave work), horizontal (S-wave work), three-component (3C)
CouplingSpike firmly planted; poor planting causes resonance and noise

Digital MEMS accelerometers record acceleration with a flat response down to very low frequency and are used in some modern land systems.

Hydrophones

A hydrophone uses a piezoelectric element that responds to pressure, and is omnidirectional. Pairs are wired so that their responses to acceleration (from cable jerks) cancel while their pressure responses add (acceleration-cancelling hydrophones).

  • Streamers: long cables of hydrophone groups towed behind the vessel, kept at depth by depth controllers (“birds”), positioned by compasses and acoustic networks, with a tail buoy at the far end.
  • Ocean-bottom cables (OBC), nodes (OBN) and seismometers (OBS): sensors on the sea floor. 4C recording (hydrophone + three geophones) records S waves converted at reflectors (PS imaging) and allows receiver deghosting.

Recording

  • Pre-amplifier → anti-alias (high-cut) filter → A/D converter (modern systems are 24-bit) → storage, via cables, telemetry or autonomous cable-free nodes.
  • Sampling interval \(\Delta t\) is typically 1, 2 or 4 ms. The Nyquist frequency is
$$f_N = \frac{1}{2\Delta t}$$

so 2 ms gives 250 Hz and 4 ms gives 125 Hz. A frequency \(f\) above \(f_N\) is aliased to \(\lvert f - 2nf_N\rvert\), folding back into the signal band; it must be filtered out before digitising.

  • Dynamic range in decibels: \(20\log_{10}(A_{max}/A_{min})\). Each bit adds about 6 dB, so an ideal 24-bit converter spans about 144 dB.

Spread geometry and fold

TermMeaning
Group interval \(\Delta g\)Spacing of receiver groups (each a summed array)
Split spreadSource in the middle of the receiver line
End-on (off-end) spreadSource at one end; standard in marine work
Near/far offsetShortest and longest source–receiver distances
Roll-alongMoving shot and active receivers along the line for continuous coverage
CMP foldNumber of traces sharing a common midpoint

For a 2D line with \(N\) channels, group interval \(\Delta g\) and shot interval \(\Delta s\) (shots moving one way along the line):

$$\text{Fold} = \frac{N\,\Delta g}{2\,\Delta s}$$

Example: 240 channels, 25 m groups, 50 m shots: fold \(= 240\times25/(2\times50) = 60\).


Noise

Signal is whatever the survey is designed to record: in a reflection survey, the primary reflections. Everything else is noise, including perfectly good seismic waves such as refractions and multiples.

Coherent noise

Coherent noise is predictable from trace to trace, so it can be attacked by its velocity, frequency or periodicity.

NoiseRecognition on a shot recordMain remedies
Ground roll (Rayleigh waves)Low velocity, low frequency, high amplitude, dispersive fan cone through the shotGeophone arrays, low-cut filters, f-k filtering, muting, shot below weathering
Air wave (air blast)Linear event at the speed of sound in air (about 330–340 m/s), high frequencyBuried shots, filtering, muting
Direct and refracted wavesLinear first arrivalsMuting (but used for statics)
Guided waves (shallow water, near-surface layer)Dispersive, trapped energyf-k filtering, muting
MultiplesRepeat primaries at later time; lower stacking velocity; periodic in shallow waterPredictive deconvolution, stacking, Radon/f-k, wave-equation methods
Side-scattered noiseEnergy scattered from near-surface heterogeneities, often aliasedArrays, dense sampling, f-k
Power-line noiseMonochromatic 50 Hz (India) or 60 HzNotch filter, avoid lines
Swell and cable noise (marine)Very low frequency, strong on near-surface streamer sectionsLow-cut filters, tow deeper, solid streamers
Seismic interferenceAnother vessel’s shots on your recordTime-sharing, editing, denoising

Random (incoherent) noise

Wind, traffic, machinery, people walking, rain, microseisms and instrument noise vary randomly from trace to trace. Summing \(N\) traces with the same signal and independent random noise improves the signal-to-noise amplitude ratio by

$$\boxed{\sqrt{N}}$$

because signal adds as \(N\) and random noise as \(\sqrt N\). Fold 36 gives a factor of 6, about 15.6 dB. This applies to vertical stacking (repeat shots or sweeps at one location) and horizontal (CMP) stacking. It does not apply to coherent noise, which stacks like signal unless its moveout differs.

Arrays: spatial filtering in the field

Geophones in a group are summed before recording. For \(n\) equal elements in a line with spacing \(\Delta x\), a wave of apparent wavelength \(\lambda_a\) along the array gives a normalised response

$$R(\lambda_a) = \left|\frac{\sin(n\pi\Delta x/\lambda_a)}{n\sin(\pi\Delta x/\lambda_a)}\right|$$
  • Vertically travelling reflections (very long \(\lambda_a\)) give \(R \approx 1\): passed.
  • The first null is at \(\lambda_a = n\Delta x\), the array length.
  • Response repeats with full pass at \(\lambda_a = \Delta x\) (spatial aliasing of the array).
  • Design rule: make the array length at least the longest ground-roll wavelength to be suppressed, and the element spacing smaller than the shortest one.

A noise test (walkaway spread of single, closely spaced geophones) measures the velocities and wavelengths of the noise before arrays are designed.

graph LR
    N["Noise problem"] --> C{"Coherent
or random?"} C -- "Coherent" --> V{"Separable by
velocity?"} V -- "Yes" --> A["Arrays, f-k filter,
muting, Radon"] V -- "No, by periodicity" --> P["Predictive
deconvolution"] C -- "Coherent, one frequency" --> F["Notch filter
for 50 Hz"] C -- "Random" --> S["Vertical and CMP stacking
S/N up by root N"]

Memory Tricks

Don’t Confuse

ThisvsThat
Vibroseis wavelet: Klauder, zero-phase after correlationvsDynamite wavelet: impulsive, close to minimum-phase
Vertical stacking: same source–receiver position, repeated shotsvsHorizontal (CMP) stacking: different offsets, same midpoint, after NMO
Ground roll: slow, low frequency, dispersivevsAir wave: about 330–340 m/s, high frequency, non-dispersive
Coherent noise: needs velocity/periodicity toolsvsRandom noise: falls as \(1/\sqrt N\) with stacking
Source ghostvsReceiver ghost; both from the free surface (or base of weathering on land)
Bubble pulse: source effect, fixed periodvsWater-bottom multiple: path effect, period set by water depth
Temporal aliasing: \(\Delta t\) too large, \(f > 1/(2\Delta t)\)vsSpatial aliasing: trace spacing too large for the apparent wavelength

Common Mistakes


Practice

All questions below are original practice questions written in CSIR NET Part C (Geophysics) style. They are not previous-year questions.

Level 1: Foundation

  1. What is the Nyquist frequency for a 2 ms sampling interval?
  2. Which physical quantity does a hydrophone respond to?
  3. What is the phase character of the correlated Vibroseis wavelet?
Answers
  1. \(f_N = 1/(2\times0.002) =\) 250 Hz.
  2. Pressure (via a piezoelectric element).
  3. Zero-phase (the Klauder wavelet, the autocorrelation of the sweep).

Level 2: Part B Style

Q1 (practice). A streamer is towed at 7.5 m depth in water of velocity 1500 m/s. For vertically arriving waves, the first non-zero receiver-ghost notch is at:

  • (a) 50 Hz
  • (b) 100 Hz
  • (c) 200 Hz
  • (d) 125 Hz
Answer and explanation

(b). \(f_1 = V_w/(2d) = 1500/15 = 100\) Hz. The ghost delay is \(2d/V_w = 10\) ms, and a delayed, inverted copy cancels the direct arrival at frequencies whose period divides the delay: 100, 200, 300 Hz, … and 0 Hz.

Q2 (practice). Data are sampled at 4 ms without an adequate anti-alias filter. A 150 Hz noise component will appear on the digital record at:

  • (a) 150 Hz
  • (b) 125 Hz
  • (c) 100 Hz
  • (d) 25 Hz
Answer and explanation

(c). \(f_N = 125\) Hz. A frequency 25 Hz above Nyquist folds back to 25 Hz below it: \(2(125) - 150 = 100\) Hz.

Q3 (practice). A 2D line uses 240 channels at 25 m group interval with a shot every 50 m (end-on spread). The CMP fold is:

  • (a) 30
  • (b) 120
  • (c) 60
  • (d) 240
Answer and explanation

(c). Fold \(= N\Delta g/(2\Delta s) = 240\times25/(2\times50) = 60\). CMP spacing is \(\Delta g/2 = 12.5\) m.

Q4 (practice). Which statement about air-gun arrays is correct?

  • (a) Guns of identical volume are used so that bubbles add in phase.
  • (b) Guns of different volumes are combined so that primary pulses add and bubble pulses partly cancel.
  • (c) Arrays are used mainly to lower the dominant frequency.
  • (d) An array removes the source ghost completely.
Answer and explanation

(b). Tuned arrays raise the primary-to-bubble ratio: the primaries are synchronous, the bubbles have different periods and add incoherently. The ghost is a free-surface effect that arrays do not remove.

Level 3: Part C Style

Q1 (practice). A noise test shows ground roll travelling at 400 m/s with frequencies of 8–20 Hz. Reflections arrive with apparent velocity 4000 m/s and dominant frequency 30 Hz. A linear array of 12 geophones at 4 m spacing is proposed.

(i) What range of ground-roll wavelengths must be suppressed? (ii) Compute the array response for the 8 Hz and 20 Hz ground roll and for the reflection. (iii) Express each in dB and comment on the design.

Answer and worked solution
  1. \(\lambda = V/f\): 400/8 = 50 m to 400/20 = 20 m. The array length \(n\Delta x = 48\) m is close to the longest wavelength, and \(\Delta x = 4\) m is well below the shortest.
  2. \(R = \sin(n\pi\Delta x/\lambda)/[n\sin(\pi\Delta x/\lambda)]\), \(n = 12\), \(\Delta x = 4\) m.
    • 8 Hz, \(\lambda = 50\) m: \(\Delta x/\lambda = 0.08\); \(\sin(172.8°) = 0.1253\); \(12\sin(14.4°) = 2.984\); \(R = 0.042\).
    • 20 Hz, \(\lambda = 20\) m: \(\Delta x/\lambda = 0.2\); \(\sin(432°) = \sin72° = 0.951\); \(12\sin36° = 7.053\); \(R = 0.135\).
    • Reflection, \(\lambda = 4000/30 = 133.3\) m: \(\Delta x/\lambda = 0.03\); \(\sin(64.8°) = 0.905\); \(12\sin(5.4°) = 1.129\); \(R = 0.801\).
  3. \(20\log_{10}R\): ground roll −27.5 dB (8 Hz) and −17.4 dB (20 Hz); reflection −1.9 dB. The array suppresses ground roll strongly while passing reflections with little loss. The weaker rejection at 20 Hz comes from a sidelobe of the response; a slightly longer or weighted (tapered) array would flatten the reject band.

Q2 (practice). A land crew records 4 Vibroseis sweeps per vibrator point (vertically stacked) and processes the line at CMP fold 30. Consider:

(A) For random noise, the combined S/N amplitude improvement is about 11, roughly 21 dB. (B) The same factor applies to ground roll. (C) A 24-bit recorder has an ideal dynamic range of about 144 dB. (D) The vertical stack improves S/N even if the noise is perfectly repeatable from sweep to sweep.

Which are correct?

  • (a) A and C only
  • (b) A, B and C only
  • (c) C and D only
  • (d) A, C and D only
Answer and worked solution

(a) A and C only.

  • (A) \(N = 4\times30 = 120\); \(\sqrt{120} = 10.95\); \(20\log_{10}10.95 = 20.8\) dB. Correct.
  • (B) Wrong: ground roll is coherent and source-generated; it repeats with every sweep and stacks with the signal unless its moveout differs.
  • (C) \(20\log_{10}2^{24} = 24\times6.02 = 144.5\) dB. Correct.
  • (D) Wrong: repeatable noise adds coherently, exactly like signal, so vertical stacking cannot improve the ratio.

Q3 (practice). An air gun at a given depth has a bubble period of 120 ms. Using the Rayleigh–Willis scaling, estimate the period (i) if the chamber volume is increased 8 times at the same pressure and depth, and (ii) if instead the gun is lowered so that the absolute hydrostatic pressure doubles. (iii) Why does a tuned array use guns of several sizes?

Answer and worked solution

\(T_b \propto (PV_g)^{1/3}P_h^{-5/6}\).

  1. Volume × 8: period × \(8^{1/3} = 2\) → 240 ms.
  2. \(P_h\) × 2: period × \(2^{-5/6} = 0.561\) → about 67 ms.
  3. Different volumes give different bubble periods. Fired together, the first pressure peaks align and add, but the bubble oscillations fall out of step and partly cancel, giving a short, high primary-to-bubble signature that deconvolves well.

Quick Revision Table

ItemFact
Land sourcesDynamite (impulsive), Vibroseis (sweep), weight drop, hammer
VibroseisCorrelate with pilot → Klauder wavelet, zero-phase, sweep bandwidth
Marine sourcesAir gun (standard), GI gun, water gun, sparker, boomer, chirp
Bubble period\(\propto (PV_g)^{1/3}/P_h^{5/6}\); tuned arrays raise primary-to-bubble ratio
Ghost notch\(f_n = nV_w/(2d\cos\theta)\)
GeophoneMoving coil; velocity output above \(f_0\); −12 dB/octave below
HydrophonePiezoelectric, pressure, omnidirectional; acceleration-cancelling pairs
4C / dual sensorP + Z combine to remove the receiver ghost; PS imaging
Nyquist\(f_N = 1/(2\Delta t)\); 2 ms → 250 Hz; 4 ms → 125 Hz
Alias\(f\) appears at \(\lvert f - 2nf_N\rvert\)
BitsAbout 6 dB per bit; 24-bit about 144 dB
2D fold\(N\Delta g/(2\Delta s)\)
Random noiseS/N up by \(\sqrt N\) on stacking
ArrayFirst null at \(\lambda_a = n\Delta x\); length ≥ longest noise wavelength
Coherent noiseGround roll, air wave, refractions, guided waves, multiples, 50 Hz

Key Takeaways

What’s Next