Seismometry: Seismometers, Response and Recording
CSIR NET geophysics: inertial seismometer equation, damping, response curves, velocity transducers, broadband and strong-motion sensors, digitisers and noise.
- 21 min
On this page · 16 sections
- Prerequisites
- The Hook: The Mass That Stays Still
- The Inertial Seismometer Equation
- Frequency Response
- From Mass Motion to Voltage: Transducers
- Digitising the Record
- Instrument Response and Correction
- Seismic Noise
- Networks and Arrays
- Memory Tricks
- Don’t Confuse
- Common Mistakes
- Practice
- Quick Revision Table
- Key Takeaways
- What’s Next
Prerequisites
- Earthquakes: Causes, Measurement and Seismicity — the Part B picture of a seismometer as a mass on a spring, and the magnitude scales that depend on it.
- Seismology Basics and Earth’s Internal Structure — the waves the instrument records.
- Sampling Theorem, FFT, Z-Transform and Convolution — Nyquist frequency, aliasing and convolution, all used when a seismogram is digitised and corrected.
- BSc mechanics: the damped, driven harmonic oscillator (see Physical Sciences).
The Hook: The Mass That Stays Still
The Inertial Seismometer Equation
Let \( u(t) \) be the ground (frame) displacement and \( z(t) \) the displacement of the mass relative to the frame. A mass \( m \) on a spring of stiffness \( k \) with viscous damping \( c \) obeys Newton’s law in the inertial frame:
$$m(\ddot z + \ddot u) = -kz - c\dot z$$Dividing by \( m \) and writing \( \omega_0^2 = k/m \) and \( 2h\omega_0 = c/m \):
$$\boxed{\ddot z + 2h\omega_0\dot z + \omega_0^2 z = -\ddot u}$$- \( \omega_0 = 2\pi/T_0 \) is the natural (angular) frequency; \( T_0 = 2\pi\sqrt{m/k} \) is the natural period.
- \( h \) is the damping constant (fraction of critical damping). \( h = 1 \) is critical; \( h < 1 \) under-damped (the mass rings); \( h > 1 \) over-damped.
- The ground enters only through its acceleration \( \ddot u \). The instrument is fundamentally an accelerometer whose output we then interpret.
For a pendulum seismometer the same equation holds with \( \omega_0^2 = g/l \) for a simple pendulum (or the corresponding expression for a compound or garden-gate pendulum); long natural periods are obtained by geometry (inclined hinge axis, astatic springs) rather than by a very soft spring.
Free oscillation and the logarithmic decrement
With no ground motion and \( h < 1 \), the mass oscillates with a decaying amplitude at the damped period \( T_d = T_0/\sqrt{1 - h^2} \). The ratio of successive peaks on the same side gives the logarithmic decrement:
$$\delta = \ln\frac{A_n}{A_{n+1}} = \frac{2\pi h}{\sqrt{1 - h^2}} \quad\Longrightarrow\quad \boxed{h = \frac{\delta}{\sqrt{4\pi^2 + \delta^2}}}$$This is how damping is measured in a calibration (release test): displace the mass, let go, read two successive peaks. For small damping, \( h \approx \delta/2\pi \).
Frequency Response
For harmonic ground motion \( u = U e^{i\omega t} \), try \( z = Z e^{i\omega t} \). Then \( \ddot u = -\omega^2 U \) and
$$\frac{Z}{U} = \frac{\omega^2}{\omega_0^2 - \omega^2 + 2ih\omega_0\omega}$$The displacement amplitude response is
$$\boxed{\left|\frac{Z}{U}\right| = \frac{\omega^2}{\sqrt{(\omega_0^2 - \omega^2)^2 + 4h^2\omega_0^2\omega^2}}}$$and the phase lag is \( \phi = \tan^{-1}\left[\dfrac{2h\omega_0\omega}{\omega_0^2 - \omega^2}\right] \) (the sign of the overall response depends on the convention for \( z \); only magnitudes and the limiting cases matter for the exam).
The three limiting cases
| Ground frequency | Approximate equation | Mass motion \( z \) is proportional to | Instrument behaves as |
|---|---|---|---|
| \( \omega \gg \omega_0 \) (periods much shorter than \( T_0 \)) | \( \ddot z \approx -\ddot u \) | \( -u \), ground displacement (gain 1) | displacement meter (long-period pendulum seen from high frequency) |
| \( \omega \approx \omega_0 \) | damping term balances \( \ddot u \) | amplitude \( U/(2h) \), lagging by 90° | resonance: gain \( 1/(2h) \) |
| \( \omega \ll \omega_0 \) (periods much longer than \( T_0 \)) | \( \omega_0^2 z \approx -\ddot u \) | \( -\ddot u/\omega_0^2 \), ground acceleration | accelerometer |
| Heavy damping, \( h \gg 1 \), middle band | \( 2h\omega_0\dot z \approx -\ddot u \) | \( -\dot u/(2h\omega_0) \), ground velocity | velocity meter (rarely used this way) |
Two consequences Part C likes:
- To record ground displacement over a band, make \( T_0 \) longer than the longest period of interest. That is why classical long-period instruments had long natural periods.
- To record acceleration of strong shaking, make \( T_0 \) very short (natural frequencies of tens to hundreds of hertz). That is the design of strong-motion accelerometers.
Choosing the damping
- \( h \ll 1 \): a large resonance peak; the record is dominated by ringing at \( T_0 \).
- \( h = 1/\sqrt{2} \approx 0.707 \): the displacement response above \( \omega_0 \) is maximally flat (no peak); at \( \omega = \omega_0 \) the gain is \( 1/(2h) = 0.707 \), the half-power (−3 dB) point.
- \( h = 1 \): critical damping; no overshoot in the release test.
Most seismometers are therefore damped to about 0.7 of critical. The Wood–Anderson torsion seismometer used to define \( M_L \) had a natural period of 0.8 s and damping of about 0.8 (its nominal static magnification is discussed in the note below).
From Mass Motion to Voltage: Transducers
Electromagnetic (moving-coil) velocity transducer
A coil attached to the mass moves in the field of a magnet fixed to the frame (or vice versa). Faraday’s law gives a voltage proportional to the relative velocity \( \dot z \):
$$\boxed{V = G\,\dot z}$$where \( G \) is the generator constant (V per m/s). Above the natural frequency \( \dot z \approx -\dot u \), so the output is proportional to ground velocity. This is the principle of the geophone used in exploration (natural frequencies of a few hertz to tens of hertz) and of classical short-period seismometers.
Damping is supplied electrically: current flowing through a shunt resistor across the coil produces a force opposing the motion. Lower shunt resistance → more current → more damping. Changing the shunt changes \( h \) without touching the mechanics.
| Sensor output | In the flat band it is proportional to | Response to ground displacement falls off |
|---|---|---|
| Mass displacement (optical lever, capacitive) | ground displacement (above \( \omega_0 \)) | as \( \omega^2 \) below \( \omega_0 \) |
| Moving-coil voltage | ground velocity (above \( \omega_0 \)) | as \( \omega^2 \) in velocity below \( \omega_0 \) |
| Accelerometer | ground acceleration (below \( \omega_0 \)) | flat in acceleration up to near \( \omega_0 \) |
Force-feedback (broadband) seismometers
Modern broadband seismometers do not let the mass move freely. A capacitive displacement sensor detects the tiny motion of the mass, and a feedback coil applies the force needed to hold the mass nearly stationary. The feedback current is the signal. Advantages:
- the response is set by the electronics, not the spring, so it is stable and linear;
- a flat response to ground velocity over a very wide band, typically from about 0.01 s (100 Hz) out to periods of hundreds of seconds;
- a large dynamic range: one instrument records microseisms and the body waves of distant great earthquakes.
Broadband instruments record three components (vertical Z, north–south N, east–west E). Horizontal components are tilt-sensitive, which makes long-period horizontal noise worse than vertical, so they need careful installation (vaults, boreholes, thermal insulation).
Strong-motion accelerometers
Near a large earthquake a sensitive broadband instrument clips (saturates). Strong-motion accelerometers have very short natural periods and low gain, stay on scale for accelerations of the order of \( g \) and more, and give peak ground acceleration (PGA) and the response spectra used in engineering. Modern instruments use force-balance or MEMS (micro-electro-mechanical) sensors.
Other sensors you should recognise
| Instrument | Measures | Note |
|---|---|---|
| Strainmeter (extensometer, borehole dilatometer) | change in distance between two points: strain, the spatial derivative of displacement | sensitive at very long periods, including tides and slow slip |
| Tiltmeter | rotation of the ground surface | long-period deformation; volcano monitoring |
| Gravimeter (superconducting) | gravity changes | Earth tides, free oscillations at the longest periods |
| Ocean-bottom seismometer (OBS) | three-component motion plus a hydrophone (pressure) | marine seismology; tilt and current noise on horizontals |
| Rotational sensors, fibre-optic sensing (DAS) | ground rotation; strain along a fibre | newer techniques |
Digitising the Record
Sampling and the Nyquist frequency
The analogue signal is sampled at interval \( \Delta t \). The highest frequency that can be represented is the Nyquist frequency
$$f_N = \frac{1}{2\Delta t}$$Energy above \( f_N \) is aliased into lower frequencies, so an anti-alias low-pass filter is applied before (or, in oversampling converters, during) digitisation. A station sampling at 100 samples per second has \( f_N = 50 \) Hz; long-period channels are often decimated to 1 sample per second or less. See Sampling Theorem, FFT, Z-Transform and Convolution for aliasing in detail.
Dynamic range
Dynamic range is the ratio of the largest to the smallest signal the system can record, in decibels:
$$\boxed{DR = 20\log_{10}\frac{A_{max}}{A_{min}}\ \text{dB}}$$An ideal \( n \)-bit converter has \( 2^n \) levels, so each bit adds \( 20\log_{10}2 \approx 6.02 \) dB. A 16-bit converter gives about 96 dB; a 24-bit converter gives about 144 dB nominally (about 138 dB if one bit is counted as the sign). In practice the noise of the electronics reduces the effective number of bits. Older analogue (paper or film) recorders managed only about 40 dB, which is why a single instrument could never record both a small local event and a large teleseism on scale.
Timing
Every travel-time study depends on absolute time. Modern stations take time from GNSS (GPS) receivers to well under a millisecond. Clock errors show up as station-wide travel-time residuals (see Travel-Time Residuals and Tomography).
Instrument Response and Correction
A seismogram \( s(t) \) is the ground motion \( u(t) \) convolved with the instrument impulse response \( i(t) \) (plus noise):
$$s(t) = u(t) \ast i(t) \quad\Longleftrightarrow\quad S(\omega) = U(\omega)\,I(\omega)$$The response is published as poles and zeros of the transfer function, a gain (sensitivity) and a normalisation. Removing it (instrument correction) is a deconvolution: divide by \( I(\omega) \) in the frequency domain, with a band-pass taper so that noise is not blown up where \( |I(\omega)| \) is small. Then the record can be converted to displacement, velocity or acceleration, or convolved with a standard response (for example a simulated Wood–Anderson instrument to measure \( M_L \)). The deconvolution idea is the same one used in Filters, Deconvolution and Fractal Analysis.
For a moving-coil sensor, the transfer function from ground velocity to output voltage is \( G\,\omega^2/(\omega_0^2 - \omega^2 + 2ih\omega_0\omega) \) (up to sign): two zeros at the origin and two poles set by \( \omega_0 \) and \( h \). Referred to ground displacement it gains one more factor of \( i\omega \), so three zeros at the origin. The statement to remember: velocity response flat above \( \omega_0 \), falling as \( \omega^2 \) below it.
Magnification of an analogue instrument is the ratio of trace amplitude to ground amplitude at a given period. The static magnification of an optical-lever torsion seismometer is the trace amplitude per unit mass displacement at long period.
Seismic Noise
Seismometers record motion all the time. The background is seismic noise, and it sets the smallest detectable earthquake.
| Noise | Typical period | Cause |
|---|---|---|
| Cultural noise | under about 1 s (above about 1 Hz) | traffic, machinery, people; strong daily and weekly cycle |
| Wind and local effects | short periods | wind on trees, buildings and the sensor housing |
| Secondary (double-frequency) microseisms | about 5–10 s, the strongest noise peak | interaction of opposing ocean waves creating pressure fluctuations at twice the ocean-wave frequency |
| Primary microseisms | about 10–20 s, same as ocean swell | ocean waves breaking and shoaling on coasts |
| Long-period noise | tens to hundreds of seconds | atmospheric pressure, temperature, tilt (worse on horizontals) |
Ways to beat noise: sites on hard rock away from towns and coasts, deep vaults and boreholes (cultural noise decays with depth), thermal shielding, and arrays. The worldwide reference curves of minimum noise (the new low- and high-noise models) are used to judge station quality.
Noise is also a signal: cross-correlating long noise records at two stations recovers the surface-wave Green’s function between them, the basis of ambient-noise tomography (see Travel-Time Residuals and Tomography).
Networks and Arrays
- Global networks of standardised, calibrated broadband stations made modern global seismology possible: uniform instruments mean comparable amplitudes and travel times. The World-Wide Standardized Seismograph Network (WWSSN), built in the 1960s with matched short-period and long-period instruments, is the historical example; its records helped map global seismicity and focal mechanisms that supported plate tectonics.
- National networks in India are operated by the National Center for Seismology (under the Ministry of Earth Sciences), which reports Indian earthquakes.
- Seismic arrays are clusters of closely spaced instruments. Delaying and summing traces along a trial direction (beamforming, delay-and-sum) improves signal-to-noise ratio by about \( \sqrt{N} \) for \( N \) sensors with incoherent noise, and measures the slowness (inverse apparent velocity) and back-azimuth of an arriving wave. The Gauribidanur array in Karnataka is India’s well-known example; arrays were developed largely for detecting and identifying underground nuclear explosions.
For a plane wave crossing an array, the apparent velocity along the surface is
$$v_{app} = \frac{v}{\sin i}$$where \( i \) is the angle of incidence from the vertical. A steeply arriving teleseismic P wave has a large apparent velocity; slowness \( p = \sin i/v = 1/v_{app} \) is the ray parameter (see Seismology Basics).
Memory Tricks
Don’t Confuse
| This | vs | That |
|---|---|---|
| Natural period \( T_0 = 2\pi\sqrt{m/k} \) | Damped period \( T_d = T_0/\sqrt{1 - h^2} \), slightly longer | |
| Displacement meter: ground periods shorter than \( T_0 \) | Accelerometer: ground periods longer than \( T_0 \) | |
| Moving-coil output: proportional to relative velocity \( \dot z \) | Capacitive sensor output: proportional to relative displacement \( z \) | |
| Broadband: force feedback, flat velocity over a wide band, very sensitive | Strong-motion: short \( T_0 \), low gain, stays on scale near the source | |
| Primary microseisms: same period as ocean waves | Secondary microseisms: half the period (double frequency), stronger | |
| Dynamic range: ratio of largest to smallest recordable signal | Bandwidth: range of frequencies recorded | |
| Nyquist frequency \( 1/(2\Delta t) \) | Sampling frequency \( 1/\Delta t \) | |
| Magnification: trace/ground amplitude at a given period | Generator constant: volts per m/s of mass velocity |
Common Mistakes
Practice
All questions below are original practice questions written for this page. They are not from any CSIR NET paper.
Level 1: Foundation
- Write the equation of motion of an inertial seismometer in terms of \( h \), \( \omega_0 \) and ground displacement \( u \).
- For ground periods much shorter than the natural period, what does the mass displacement record?
- A station samples at 40 samples per second. What is its Nyquist frequency?
Answers
- \( \ddot z + 2h\omega_0\dot z + \omega_0^2 z = -\ddot u \).
- Ground displacement (with unit gain and opposite sign).
- 20 Hz.
Level 2: Part B Style
Practice Q1. A seismometer mass of 0.5 kg hangs on a spring of stiffness 19.74 N/m. Its natural period is about:
(a) 0.5 s (b) 1.0 s (c) 2.0 s (d) 4.0 s
Answer: (b) 1.0 s
\( T_0 = 2\pi\sqrt{m/k} = 2\pi\sqrt{0.5/19.74} = 2\pi\sqrt{0.02533} = 2\pi\times 0.1592 = 1.000 \) s.
Practice Q2. In a release test, successive peaks on the same side are 8.0 mm and 2.0 mm. The damping constant \( h \) is about:
(a) 0.11 (b) 0.22 (c) 0.40 (d) 0.70
Answer: (b) 0.22
\( \delta = \ln 4 = 1.386 \). \( h = 1.386/\sqrt{39.48 + 1.92} = 1.386/6.434 = 0.215 \).
The small-damping shortcut \( \delta/2\pi = 0.221 \) is close; (a) halves it, and (d) is the design value, not the measured one.
Practice Q3. A moving-coil seismometer with natural period 1 s and generator constant 200 V/(m/s) records a ground motion of period 0.2 s and velocity amplitude 5 µm/s. Assuming the instrument is in its flat velocity band, the output voltage amplitude is:
(a) 0.2 mV (b) 1.0 mV (c) 5.0 mV (d) 40 mV
Answer: (b) 1.0 mV
At 0.2 s the ground period is well below \( T_0 = 1 \) s, so \( |\dot z| \approx |\dot u| \). \( V = G|\dot u| = 200\times 5\times 10^{-6} = 1.0\times 10^{-3} \) V.
Practice Q4. Which statement about seismic noise is correct?
(a) The strongest natural noise peak is at periods of 0.1–0.5 s and is caused by wind. (b) Secondary microseisms have about half the period of the ocean waves that generate them. (c) Cultural noise increases with depth in a borehole. (d) Long-period noise is usually lower on horizontal components than on the vertical.
Answer: (b)
Secondary (double-frequency) microseisms are generated at twice the ocean-wave frequency, i.e. half the period, and form the strongest peak (about 5–10 s). Cultural noise decreases with depth; horizontals are noisier at long periods because of tilt.
Level 3: Part C Style
Practice Q5. A seismometer has \( T_0 = 1 \) s and \( h = 1/\sqrt{2} \). Which statements are correct?
- Its displacement gain at a ground period of 0.5 s is about 0.97.
- Its displacement gain at a ground period of 2 s is about 0.24.
- At a ground period of 10 s its mass displacement is approximately proportional to ground acceleration.
- Its damped period in a release test is exactly 1 s.
(a) 1 and 2 only (b) 1, 2 and 3 only (c) 3 and 4 only (d) 1, 2, 3 and 4
Answer: (b) 1, 2 and 3 only
With \( x = T_0/T \) and \( 4h^2 = 2 \):
- \( T = 0.5 \) s, \( x = 2 \): \( 4/\sqrt{9 + 8} = 0.970 \). True.
- \( T = 2 \) s, \( x = 0.5 \): \( 0.25/\sqrt{0.5625 + 0.5} = 0.243 \). True.
- \( T = 10 \) s is far below \( \omega_0 \), so \( z \approx -\ddot u/\omega_0^2 \): accelerometer behaviour. True.
- \( T_d = T_0/\sqrt{1 - 0.5} = 1.414 \) s, not 1 s. False.
Practice Q6. A broadband station digitises with a 24-bit converter. The largest signal on scale corresponds to a ground velocity of 1.3 cm/s. Ignoring electronic noise and taking the full \( 2^{24} \) levels:
(i) What is the nominal dynamic range in dB? (ii) What ground velocity corresponds to one count (least significant bit), if the full scale spans −1.3 to +1.3 cm/s? (iii) If the electronics add noise equal to 16 counts rms, by how many dB is the usable range reduced?
Answer and worked solution
(i) \( 20\log_{10}2^{24} = 24\times 6.02 = 144.5 \) dB.
(ii) Full span = 2.6 cm/s over \( 2^{24} = 16{,}777{,}216 \) counts: \( 2.6\times 10^{-2}/1.678\times 10^{7} = 1.55\times 10^{-9} \) m/s per count, about 1.5 nm/s.
(iii) Noise of 16 counts raises the floor by a factor of 16: \( 20\log_{10}16 = 24.1 \) dB. Usable range ≈ 144.5 − 24.1 ≈ 120 dB, i.e. effectively 20 bits.
Practice Q7. A small-aperture array records a P wave that crosses two sensors 3.0 km apart along the propagation azimuth with a delay of 0.25 s. The P velocity beneath the array is 6.0 km/s. Which statements are correct?
- The apparent velocity is 12 km/s.
- The angle of incidence from the vertical is 30°.
- The slowness is about 0.083 s/km.
- A wave arriving vertically would show zero delay between the sensors.
(a) 1 and 4 only (b) 1, 2 and 4 only (c) 1, 2, 3 and 4 (d) 2 and 3 only
Answer: (c) 1, 2, 3 and 4
- \( v_{app} = 3.0/0.25 = 12 \) km/s. True.
- \( \sin i = v/v_{app} = 6/12 = 0.5 \) → \( i = 30° \). True.
- \( p = 1/v_{app} = 0.0833 \) s/km. True.
- Vertical incidence: \( v_{app} \to \infty \), zero delay. True.
Quick Revision Table
| Idea | Fact or formula |
|---|---|
| Seismometer equation | \( \ddot z + 2h\omega_0\dot z + \omega_0^2 z = -\ddot u \) |
| Natural period | \( T_0 = 2\pi\sqrt{m/k} \) |
| Damping from decrement | \( \delta = \ln(A_n/A_{n+1}) \), \( h = \delta/\sqrt{4\pi^2 + \delta^2} \) |
| Damped period | \( T_d = T_0/\sqrt{1 - h^2} \) |
| Displacement response | \( x^2/\sqrt{(1 - x^2)^2 + 4h^2x^2} \), \( x = \omega/\omega_0 \) |
| Limits | \( \omega \gg \omega_0 \): displacement; \( \omega \ll \omega_0 \): acceleration; at \( \omega_0 \): gain \( 1/2h \) |
| Best damping | about 0.7 (maximally flat) |
| Moving coil | \( V = G\dot z \); velocity above \( \omega_0 \); shunt resistance controls damping |
| Broadband | force feedback, flat velocity about 0.01 s to hundreds of s |
| Strong motion | short \( T_0 \), low gain, PGA |
| Nyquist | \( f_N = 1/(2\Delta t) \); anti-alias filter |
| Dynamic range | \( 20\log_{10}(A_{max}/A_{min}) \); 6.02 dB per bit |
| Response removal | deconvolution with poles and zeros, band-limited |
| Microseisms | secondary 5–10 s (double frequency), primary 10–20 s |
| Array | delay-and-sum; \( v_{app} = v/\sin i \); SNR gain about \( \sqrt N \) |
Key Takeaways
What’s Next
- Earthquake Sources, Moment Tensor and Focal Mechanisms: what the instrument is recording, and how the source is inferred from it.
- Travel-Time Residuals and Tomography: timing and picking accuracy feed straight into location and imaging.
- Filters, Deconvolution and Fractal Analysis: the signal-processing tools behind response removal.
- Seismology: Quick Revision: the whole chapter on one page.