Electronics Guide

What Makes a Sensor Distributed

Fiber sensing systems fall into three families, and the difference is where the information is encoded. A point sensor places a transducer at one location and uses fiber only to carry light to and from it; a Fabry-Perot pressure cell at a fiber end belongs here, and the fiber is a wire that happens to be made of glass. A quasi-distributed sensor writes many discrete transducers into one fiber and separates their responses by wavelength, by time of flight, or by both. Fiber Bragg grating arrays are the dominant example, and one fiber routinely carries tens to hundreds of gratings, each with a known location and a calibrated sensitivity. The fiber between gratings still measures nothing.

A fully distributed sensor has no transducers. It exploits the fact that any real glass fiber scatters a small fraction of the guided light back toward the source at every point, and that the properties of that backscattered light depend on the local temperature, strain, and vibration of the glass. The instrument resolves the fiber into contiguous resolution cells and reports one value per cell, so the channel count is set by the interrogator and the fiber length rather than by anything that was installed.

Three properties follow. Coverage is complete, which is why distributed sensing dominates applications where the location of the event is the unknown. The cable is passive, carrying no power, no electronics, and no metal if the design does not require it, which suits explosive atmospheres and high-voltage environments. And the sensing element is cheap while the instrument is expensive, which inverts the usual economics: adding a thousand channels costs a reel of cable, while improving the measurement means replacing the interrogator.

The instrument is the interrogator and the fiber the sensing cable. The interval over which one reported value is averaged is the spatial resolution or, in acoustic systems, the gauge length. The interval between reported values is the sampling interval or channel spacing, normally finer than the spatial resolution and not to be confused with it. Specifying a system by channel spacing alone is a common procurement error.

The Three Scattering Mechanisms

Light in silica is scattered by three distinct processes, each returning a different part of the spectrum. Scattered light from a single launched wavelength shows a strong central line at the launch frequency, a pair of weak lines a few tens of gigahertz to either side, and a pair of much weaker and broader bands displaced by roughly thirteen terahertz. These are the Rayleigh, Brillouin, and Raman returns, and distributed sensing is built on all three.

Rayleigh Scattering

Rayleigh scattering is elastic: the scattered photon has the same frequency as the incident photon. It arises from microscopic inhomogeneity in the refractive index of the glass, frozen in when the fiber cooled through the glass transition during drawing. These density and composition fluctuations are far smaller than a wavelength and permanently fixed in position, so the scattering pattern of a given piece of fiber is a stable, essentially random signature unique to that fiber.

Rayleigh scattering is the dominant intrinsic loss mechanism in silica fiber, and its strength varies as the inverse fourth power of wavelength, which is why long-haul links moved to 1550 nanometers, where attenuation is close to 0.2 decibels per kilometer. Scattering is nearly isotropic, so only the fraction falling within the numerical aperture of the guided mode in reverse returns to the instrument, a recapture fraction of order one part in a thousand.

Being elastic, Rayleigh scattering carries no temperature or strain information in its frequency. What it carries is phase. The returning field is the coherent sum of contributions from all scattering centers within one pulse length, and the optical path lengths between those centers change when the fiber is stretched or heated. Rayleigh-based systems therefore measure a change in fiber length, either as a change in the interference pattern of the return or as a shift in the wavelength at which the local signature repeats. This is the basis of distributed acoustic sensing and of high-resolution strain sensing.

Brillouin Scattering

Brillouin scattering is inelastic and arises from the interaction of light with acoustic phonons, the thermally excited sound waves traveling in the glass. A moving acoustic wave imposes a Doppler shift, so the backscattered Brillouin lines are displaced from the launch frequency by an amount set by the acoustic velocity in the core and by the optical wavelength. For standard single-mode silica fiber near 1550 nanometers that shift lies close to eleven gigahertz, and reported values across common fiber types and telecommunications wavelengths span roughly nine to thirteen gigahertz.

The acoustic velocity depends on the density and elastic moduli of the glass, and both change with temperature and applied strain, so the Brillouin shift is an almost perfectly linear function of each. Widely reported coefficients for standard single-mode fiber at 1550 nanometers are approximately 0.05 megahertz per microstrain and approximately one megahertz per kelvin. Measuring a frequency rather than an intensity is the great advantage of the approach: frequency is immune to connector loss, bend loss, source power drift, and detector gain drift, all of which corrupt an intensity measurement.

The Brillouin gain spectrum in silica has a natural linewidth of roughly thirty megahertz, corresponding to an acoustic phonon lifetime near ten nanoseconds. That lifetime is not a detail; as shown below, it sets a hard floor of about one meter on the spatial resolution achievable by conventional Brillouin time-domain methods.

Raman Scattering

Raman scattering is also inelastic, but it involves molecular vibration rather than acoustic waves, and the frequency shift is correspondingly much larger. In silica the Raman return occupies a broad band centered near thirteen terahertz from the launch frequency, a wavelength offset of roughly one hundred nanometers at 1550 nanometers. The down-shifted band is the Stokes component and the up-shifted band the anti-Stokes component.

The anti-Stokes process requires the molecule to already be vibrationally excited, and the population of those states follows Boltzmann statistics. The anti-Stokes intensity is therefore strongly temperature dependent while the Stokes intensity is only weakly so; reported relative sensitivities for typical silica glass at room temperature are about 0.83 percent per kelvin for the anti-Stokes band and about 0.096 percent per kelvin for the Stokes band. Their ratio depends on absolute temperature and largely cancels the common-mode effects of launch power, splice loss, and detector responsivity.

The decisive property of Raman sensing is what it does not respond to. Vibrational populations are essentially insensitive to the modest strains a fiber sees in service, so a Raman system reports temperature unambiguously, while a Brillouin system must always disentangle two unknowns from one frequency. The price is a very weak return and a measurement that is fundamentally an intensity ratio rather than a frequency, with all the calibration burden that implies.

Optical Time-Domain Reflectometry

Nearly every distributed architecture inherits its ranging principle from optical time-domain reflectometry, the technique developed to locate faults in telecommunications cable. The instrument launches a short optical pulse and records the returning backscatter against time; light returning at time t was scattered from a position equal to the group velocity times t divided by two. With a group index near 1.47, light travels roughly 204 meters per microsecond in silica, so each microsecond of delay corresponds to about 102 meters of fiber.

Spatial resolution follows directly from pulse duration. A pulse occupies a length in the fiber equal to the group velocity times its duration, and because it travels out and back, contributions from a region half that long arrive together. Ten nanoseconds gives about one meter of resolution, one hundred nanoseconds about ten meters, a common default in acoustic systems, and one microsecond about one hundred meters.

The temptation is obvious and so is the obstacle. Halving the pulse duration halves the resolution cell, but it also halves the pulse energy, and backscattered power scales with pulse energy, so the change costs three decibels of signal-to-noise everywhere along the fiber. Raising peak power to compensate fails above a threshold, because the pulse then drives nonlinear processes: modulation instability broadens and depletes it, self-phase modulation distorts its spectrum, and stimulated Brillouin scattering can divert a large fraction of its energy into a backward wave unrelated to the measurement. Every time-domain system therefore works against a peak-power ceiling the fiber itself imposes.

The remaining lever is averaging. Noise in successive traces is largely uncorrelated while the signal is not, so averaging N traces improves signal-to-noise by the square root of N: ten decibels requires one hundred traces and twenty decibels requires ten thousand. Because traces cannot be launched faster than the round trip allows, averaging converts directly into acquisition time, which is why static temperature systems can afford to average for minutes and dynamic acoustic systems cannot afford to average at all.

A more elegant route is pulse coding. Rather than one pulse, the instrument launches a sequence drawn from a code with good autocorrelation properties, such as a Simplex or cyclic code, and recovers the single-pulse response by correlation, raising total launched energy without raising peak power. Reported Brillouin results using a 511-bit cyclic code achieved one-meter spatial resolution with temperature and strain accuracies of about 2.2 degrees Celsius and 44 microstrain at 50 kilometers.

Phase-Sensitive OTDR and Distributed Acoustic Sensing

A conventional OTDR uses a source with a short coherence length, so the many scattering contributions within one pulse add in intensity and the trace is smooth. Phase-sensitive OTDR does the opposite, using a narrow-linewidth laser whose coherence length exceeds the pulse length so that contributions add as fields rather than powers. The result looks like noise but is not: it is a stable interference pattern, a one-dimensional speckle, determined by the frozen-in positions of the scattering centers in that particular fiber. Because the pattern is fixed by the glass, it changes only when the optical path lengths between those centers change, and stretching a section by even a nanometer alters the relative phases there. Comparing successive traces therefore reveals dynamic strain along the entire fiber, at every point, at the trace repetition rate. This is distributed acoustic sensing.

Coherent Detection and Phase Demodulation

Early systems simply watched the speckle intensity fluctuate. That detects that something happened and roughly where, but the relationship between intensity change and strain is nonlinear, non-monotonic, and different at every point, so the waveform cannot be recovered. Modern systems use coherent detection: part of the source laser is split off as a local oscillator and mixed with the return on a balanced receiver, producing a beat signal from which amplitude and optical phase are separately demodulated in the digital domain.

The optical phase is the useful quantity because it is linear in fiber elongation, equal to first order to the elongation multiplied by the propagation constant and by a photoelastic correction factor of roughly 0.78 for silica, which accounts for the fact that straining the glass changes its refractive index as well as its length. With calibrated phase demodulation a DAS system is a strain or strain-rate instrument rather than a mere event detector, and its output can be compared directly with geophones and strain gauges.

Gauge Length

The phase difference between two positions separated by a distance L reports the average axial strain over that interval, and L is the gauge length. Choosing it is a genuine engineering decision. A short gauge length localizes the measurement well but subtracts two nearly equal phases, so the strain signal is small relative to demodulation noise. A long one gives a strong signal but averages over the interval, attenuating features shorter than the gauge and introducing a notch response at wavelengths that fit an integer number of times into it. Seismic acquisition matches the gauge length to the shortest wavelength of interest; intrusion detection matches it to the expected extent of the disturbance. It approximates the pulse length in simple systems, but where sampling is finer than the pulse it can be chosen independently in processing.

Fading and Its Mitigation

The speckle nature of coherent Rayleigh backscatter creates the characteristic failure mode of phase-sensitive systems. At some positions the contributions within a resolution cell interfere destructively, leaving almost no returned field. Sensitivity there collapses and phase demodulation becomes ill-conditioned, and the literature describes the consequences plainly as varying sensitivity, random dead zones, and false disturbance signals. Polarization fading, which occurs when the polarization state of the return is orthogonal to the local oscillator, produces the same symptoms.

Because the interference pattern depends on optical frequency, the standard remedy is frequency diversity: interrogate with several probe frequencies and combine the results, since a fade at one frequency rarely coincides with a fade at another. Multi-frequency and frequency-swept, or chirped, pulses achieve this within a single interrogation, and chirped pulses additionally permit pulse compression, with reported results of thirty-centimeter resolution from a two-microsecond pulse swept over 420 megahertz. Polarization diversity in the receiver addresses the polarization term. A newer hardware route is engineered fiber carrying weak, deliberately written scattering centers at regular intervals, which raises the return by tens of decibels and replaces random speckle with a deterministic pattern, at the cost of a fiber that is no longer ordinary cable.

Bandwidth and Range

A new pulse cannot be launched until the previous one has cleared the fiber, or returns will overlap ambiguously. The maximum repetition rate is therefore the reciprocal of the round-trip time, and the highest measurable frequency is half of that. For a fifty-kilometer fiber the round trip takes about 490 microseconds, the repetition rate is just over two kilohertz, and the Nyquist limit is about one kilohertz; a two-kilometer fiber supports tens of kilohertz. Range and bandwidth are locked together by the speed of light, and the only escapes are to multiplex several optical frequencies within one round-trip window or to divide a long route among several interrogators.

Range itself is set by attenuation, and at 0.4 decibels per kilometer round trip, forty to fifty kilometers is a typical limit for an unamplified system. Distributed Raman amplification, remotely pumped erbium-doped sections, and hybrid schemes extend this considerably; the literature reports phase-sensitive systems reaching 175 kilometers with distributed amplification and direct detection, and 108 kilometers using time-gated frequency-domain interrogation with Raman amplification. Amplification is not free, since it adds spontaneous emission noise and its own nonlinear constraints.

Data Rates

DAS is a signal-processing problem as much as an optical one. A fifty-kilometer fiber sampled every five meters, giving ten thousand channels at two kilohertz with four bytes per sample, produces eighty megabytes per second, or roughly 6.9 terabytes per day, from one instrument. Raw phase is normally filtered in the frequency-wavenumber domain, where propagating disturbances appear as lines whose slope is their apparent velocity, so a train at thirty meters per second separates cleanly from a seismic wave at two kilometers per second. Edge processing that reduces raw phase to detected events before transmission is now standard in pipeline and perimeter installations, and supervised classifiers trained on site-specific data materially reduce false alarm rates. Such models transfer imperfectly between sites, so commissioning includes staging known events at known positions.

Optical Frequency-Domain Reflectometry

Where the time domain trades pulse energy against resolution, the frequency domain avoids pulses altogether. In optical frequency-domain reflectometry the source is a continuous-wave laser swept linearly in optical frequency. Light returning from a scatterer at distance z was launched earlier, when the laser frequency was different, so mixing it with the current output produces a beat note whose frequency is proportional to z. A Fourier transform of the recorded beat signal converts the frequency axis into a distance axis in one operation.

Spatial resolution is set by the total optical frequency excursion of the sweep rather than by any pulse duration, and a sweep of several terahertz corresponds to resolution in the tens of micrometers. Range is limited instead by the coherence length of the laser and by detection bandwidth, since the beat frequency grows with distance. The result is the mirror image of the time-domain trade: extraordinary resolution over short fibers, typically tens of meters, and no access to the tens of kilometers time-domain systems handle routinely.

For strain and temperature measurement, OFDR is normally combined with Rayleigh spectral correlation. The frozen-in scattering signature of a fiber section acts as a random but repeatable grating; recording it once as a reference and again later reveals a spectral shift proportional to the change in local strain or temperature, in direct analogy with the wavelength shift of a fiber Bragg grating but available continuously. Luna, for example, states gauge pitches of 0.65 millimeters over a twenty-meter sensor and 2.6 millimeters over a hundred-meter sensor, with a strain range of plus or minus fifteen thousand microstrain. Those are vendor specifications rather than independently verified figures, but they indicate the regime the technique occupies: composite coupons, aerospace structures, medical device shape sensing, and battery thermal mapping, where the object is meters long and the features of interest are millimeters wide.

Brillouin Interrogation Architectures

Brillouin systems measure the frequency shift of the backscattered light, which requires resolving a spectrum, not merely a power, at every point along the fiber. Two families of instrument do this, and the difference between them decides the cable topology of the entire installation.

Spontaneous Brillouin: BOTDR

Brillouin optical time-domain reflectometry launches a single pulse and analyzes the spontaneous Brillouin backscatter it generates, using coherent detection against a local oscillator to resolve the shift. Its practical virtue is that it needs only one end of the fiber, so a break still leaves everything up to it measured and the break itself located, which for a buried pipeline or an offshore umbilical is a decisive advantage. The price is signal strength: spontaneous Brillouin backscatter is far weaker than Rayleigh backscatter, so BOTDR needs heavy averaging and generally accepts coarser spatial resolution. A demonstrated BOTDR sensor length of 57 kilometers came with a spatial resolution of twenty meters.

Stimulated Brillouin: BOTDA

Brillouin optical time-domain analysis launches a pulsed pump from one end and a continuous-wave probe from the other, with a controllable frequency offset between them. When that offset matches the local Brillouin shift, the waves couple through the acoustic wave and energy transfers from pump to probe. Sweeping the offset and recording the amplified probe against time maps the Brillouin gain spectrum at every position, whose peak gives the local shift.

Stimulated interaction is orders of magnitude stronger than spontaneous scattering, so BOTDA delivers better spatial resolution, better frequency precision, or shorter measurement times than BOTDR, and in favorable conditions all three. Reported results include ranges near one hundred kilometers at sub-meter to two-meter resolution with temperature resolution of about one to 1.5 degrees Celsius, and 150 kilometers at two-meter resolution using inline amplification.

The structural cost is that BOTDA needs access to both ends of the fiber. In practice this means a loop, out and back along the same route, which doubles the cable and halves the addressable route length, and it means a single break disables the measurement entirely rather than degrading it. A further limitation is pump depletion, in which the pump is drained by amplifying the probe near the launch end and can no longer produce accurate gain further along.

The Phonon Lifetime Limit

Building the acoustic wave that mediates the Brillouin interaction takes time. If the pump pulse is much shorter than the roughly ten-nanosecond phonon lifetime, the acoustic wave never reaches steady state, the measured gain spectrum broadens toward the reciprocal of the pulse duration, and the frequency peak becomes correspondingly harder to locate. Since ten nanoseconds corresponds to about one meter of fiber, conventional BOTDA saturates near one meter of spatial resolution, and pushing below it costs frequency precision directly.

Three approaches break the limit. Differential pulse-width pair techniques measure with two pulses of slightly different duration, both longer than the phonon lifetime, and subtract the results, recovering centimeter-scale resolution from long-pulse measurements. Pre-excitation and dark-pulse schemes keep the acoustic wave in steady state and modulate it briefly. Correlation-domain methods, notably Brillouin optical correlation-domain analysis, modulate both pump and probe so they correlate strongly at only one position, which is then scanned electronically; reported resolutions reach one centimeter and, in specialized demonstrations, three millimeters, though the range is confined to meters or a few hundred meters.

Dynamic Brillouin Measurement

A conventional BOTDA measurement is slow for two reasons: the probe frequency must be swept across the gain spectrum in tens of steps, and each step must be averaged. Acquisition times run from seconds to minutes, which is fine for a structure that creeps and useless for one that vibrates. Slope-assisted BOTDA removes the sweep by parking the pump-probe offset on the steep flank of the gain spectrum, where amplitude varies almost linearly with frequency shift; the measurement becomes a single intensity reading, with reported dynamic strain sampled at two hundred hertz over thirty meters of fiber at three-meter resolution. The trade is dynamic range, since the linear flank spans only a fraction of the thirty-megahertz gain bandwidth. The absolute ceiling remains the round-trip time: for a hundred-meter fiber that is about one microsecond, implying a maximum acquisition rate near one megahertz with no averaging or sweeping at all.

Raman Distributed Temperature Sensing

Raman DTS is the oldest commercially mature distributed technique and remains the default when temperature is the only quantity of interest. A pulsed laser launches into the fiber, a filter assembly separates the anti-Stokes and Stokes bands from the far stronger Rayleigh return, and two detectors record their intensities against time; their ratio, corrected for calibration, yields absolute temperature at every point.

Published performance for commercial systems is a spatial resolution near one meter with accuracy within about one degree Celsius and temperature resolution near 0.01 degrees Celsius, over distances that can exceed thirty kilometers. Standard cables operate to 85 degrees Celsius and specialized designs to 700 degrees Celsius. As always these figures are not simultaneously achievable; the quoted resolution applies at the near end and degrades with distance as the anti-Stokes return weakens.

Differential Attenuation and Its Correction

The central calibration problem in Raman DTS is that the Stokes and anti-Stokes bands are separated by roughly two hundred nanometers and therefore do not experience identical loss. Any wavelength-dependent loss along the fiber biases the ratio and appears as a spurious temperature gradient, and it is not static: bending, connector degradation, and hydrogen ingress all change it over the life of an installation, which the literature identifies as a critical issue for stability.

The standard remedy is a double-ended, or loop, configuration in which the instrument interrogates the same fiber alternately from both ends and combines the traces. Taking the geometric mean of the forward and backward backscatter ratios cancels the differential loss term to first order, because a loss appearing early in one direction appears late in the other. This is why DTS cables are so often installed as loops even though the technique needs only one end. Single-ended operation remains common where a loop is impractical, and it then depends on a differential-attenuation coefficient determined at calibration and assumed constant.

Fiber and Wavelength Choices

Raman DTS has historically used multimode fiber, typically graded-index 50/125, because the larger core captures substantially more backscattered light. The penalty is higher attenuation and modal dispersion, which limits practical range to roughly ten kilometers. Single-mode fiber extends the range at the cost of a weaker return, and it is the choice for long pipeline and submarine work and wherever the same fiber must also serve a Brillouin or acoustic system. Source wavelength is a similar compromise: 1064 nanometers offers a strong Raman cross section and inexpensive high-power sources but higher attenuation, while 1550 nanometers offers the lowest loss, telecommunications components, and coexistence with data traffic on the same cable.

The Performance Trade Space

The single most useful thing to understand about distributed sensing is that four quantities are coupled and none can be improved without spending one of the others. Vendors quote each separately, and a specification listing a best case for each describes four different operating points, not one.

Spatial resolution is the length of fiber over which one value is averaged; in time-domain systems it is set by pulse duration, and shortening the pulse reduces returned energy in proportion. Sensing range is limited by the optical budget, since attenuation costs about 0.4 decibels per kilometer round trip at 1550 nanometers. Acquisition time is the price of averaging, which improves signal-to-noise only as the square root of the number of traces, so a factor of ten in measurement resolution costs a factor of one hundred in time. Measurement resolution, in degrees or microstrain, is set by signal-to-noise at the point in question; in Brillouin systems it is the precision with which the peak of a thirty-megahertz-wide gain spectrum can be located, which demands a correspondingly high signal-to-noise ratio.

The couplings work as follows. Improving spatial resolution by a factor of two costs three decibels of signal, repayable by four times the averaging, which quadruples acquisition time, or by accepting worse measurement resolution, or by shortening the range. Doubling the range costs several decibels at the far end, with the same three options, and halving acquisition time costs three decibels. There is no free parameter anywhere in the loop, because the quantity being divided up is the energy the fiber returns, and that is capped by the nonlinear threshold of the glass and by attenuation.

Two engineering moves genuinely enlarge the space rather than trading within it. Pulse coding raises total launched energy without raising peak power and buys several decibels outright, and optical amplification restores signal at the far end, which is what makes hundred-kilometer measurements possible. Everything else on a datasheet is a choice of operating point. The practical consequence for procurement is to ask for the four numbers together, and to ask whether the quoted spatial resolution is the ten to ninety percent transition width of the response to a step change, which is the meaningful definition, or the sampling interval, which is not. A system reporting values every 0.25 meters may well have a spatial resolution of two meters, and the difference decides whether a one-meter hot spot is seen at its true temperature.

Strain and Temperature Cross-Sensitivity

Brillouin sensing measures one frequency and must infer two quantities from it. With coefficients near 0.05 megahertz per microstrain and one megahertz per kelvin, a one-kelvin change produces the same shift as about twenty microstrain, so an uncorrected ten-kelvin seasonal swing masquerades as two hundred microstrain, roughly forty megapascals of stress in steel. No amount of instrument precision fixes this; it is a physical degeneracy that the installation must break. Four approaches are used in practice.

Mechanical decoupling. A loose-tube cable holds the fiber inside an oversized gel-filled or dry tube with deliberate excess fiber length, so the fiber floats free and follows only the temperature of its surroundings, measuring temperature and nothing else. A tight-buffered or strain-coupled cable is bonded to its jacket and to the structure and follows both.

Dual-fiber cables. The most common commercial solution places two fibers in one cable, one loose and one strain-coupled, and subtracts the temperature-only reading to recover strain. It works because the two fibers are separated by millimeters and share the same thermal environment, which is exactly the condition a spatially separated reference cannot guarantee.

Combined Brillouin and Raman measurement. Because the Raman anti-Stokes to Stokes ratio responds to temperature and not to strain, a single fiber interrogated by both techniques yields two independent equations in two unknowns. Hybrid instruments that share one sensing fiber between a Brillouin and a Raman channel are commercially available, and this is often the cleanest solution when only one fiber can be installed.

Specialty fiber with two Brillouin observables. Fibers supporting several acoustic or optical modes exhibit multiple Brillouin peaks whose strain and temperature coefficients differ, so the two effects separate within one fiber. Few-mode, photonic-crystal, and polarization-maintaining fibers have all been used this way. The method is elegant and the fiber is not ordinary, which limits it to installations that can justify a non-standard cable.

Rayleigh-based OFDR strain sensing has the same degeneracy and the same remedies. Raman DTS is the one technique with no cross-sensitivity to resolve, which is a large part of its continuing commercial appeal.

Cable Design and Mechanical Coupling

A distributed strain system measures the strain of the fiber. Whether that equals the strain of the structure is a question about cable design and installation, not about optics, and it is where field results most often disappoint. The literature on Brillouin sensing cables states the requirement plainly: a strain cable must be fixed to the structure with tight mechanical coupling between sheath and fiber, whereas a temperature cable must have loose coupling and controlled fiber over-length precisely to prevent strain transfer.

Strain transfer through the intervening layers is described by shear-lag analysis. Strain applied at a jacket surface reaches the fiber core only after passing through the adhesive, the jacket, and the primary coating, each of which shears. The average transfer rate over a bonded length follows a hyperbolic expression in a shear-lag parameter set by the moduli and geometry of those layers, and reported transfer efficiencies commonly fall between sixty and ninety-five percent. A stiff adhesive such as an epoxy near 3.5 gigapascals transfers strain over a short distance, while a soft acrylic near 200 megapascals spreads it over a much longer one and blurs localized features.

That blurring sets a resolution limit independent of the interrogator. A crack in concrete opens over a millimeter, but the strain it induces in a bonded fiber spreads over a length set by the shear-lag parameter, giving a peak whose full width at half maximum is twice the natural logarithm of two divided by that parameter. Cracks closer than this width merge into one peak, and a centimeter-resolution interrogator does not help if the cable smears the strain field over half a meter.

Cable design also sets how much strain can be measured before something breaks. A bare silica fiber can be strained to roughly three percent before failure, and reported designs using elastomeric jackets that transfer stress nonlinearly have measured eight to nine percent by allowing controlled slip. Compression is a separate problem, because a fiber bonded only at its ends buckles rather than shortening with the structure, so continuous bonding is necessary if compressive strain must be measured.

For acoustic sensing the coupling question is different but no less decisive. A DAS cable responds primarily to axial strain, so its sensitivity to a wave arriving broadside is much lower than to one arriving along the cable, and this directional response shapes every seismic and intrusion application. Whether the cable contacts the ground directly, is cemented behind casing, hangs in a fluid-filled borehole, or lies loose in a duct changes the coupling by tens of decibels. Borehole seismic work therefore prefers cemented installations to tubing-conveyed ones, and in perimeter security the burial depth and soil compaction set the achievable detection range.

Environment sets the remaining constraints. Hydrogen liberated in downhole and subsea environments diffuses into silica and raises attenuation, particularly at longer wavelengths, which is why carbon-coated hermetic fibers and pure-silica-core designs are specified for wells. Ionizing radiation darkens conventional germanium-doped fiber, so nuclear installations use radiation-hardened fiber. Coating sets the temperature ceiling: standard acrylate reaches roughly 85 degrees Celsius, polyimide around 300 degrees Celsius, and metal-coated fibers higher still, which is the basis of the cables used in steam-assisted wells and furnaces.

Optical Budget and Installation Practice

Distributed systems are far less tolerant of a poor optical path than telecommunications links are, because a link only needs enough power to close, whereas a sensing system spends every decibel on measurement resolution. Connectors are the usual culprit. A mated pair loses a few tenths of a decibel and, more importantly, produces a discrete Fresnel reflection that can saturate the receiver and blind the instrument for the following meters, exactly as a reflective event creates a dead zone in a conventional OTDR trace. Fusion splices are preferred everywhere in the sensing path, and where connectors are unavoidable, angled-polished types suppress back reflection. A launch fiber of several hundred meters keeps the near-end dead zone and the receiver recovery transient away from the measurement. Bend loss deserves particular attention because it is wavelength dependent and corrupts Raman DTS ratios directly; a tight coil left in a splice enclosure during installation can appear years later as an apparent temperature offset downstream.

Calibration practice differs by technique. Raman DTS needs at least one reference section at a known temperature, normally a coil inside the instrument in a controlled bath. Brillouin systems need a reference Brillouin frequency for the specific fiber, since the shift varies by tens of megahertz between fiber types and even between batches. Rayleigh OFDR strain measurement is inherently relative and requires a reference scan of the undisturbed fiber; if that reference is lost, absolute strain is lost with it.

Existing telecommunications infrastructure introduces its own considerations. Dark fiber in a duct is often usable for acoustic sensing, but its route is known only approximately, its ground coupling is uncontrolled, and it may contain splices, slack loops in handholes, and sections of different fiber type. Mapping fiber distance to geographic position, usually called tap testing, is a necessary field exercise, since slack coils occupy fiber length but no ground distance. On a live cable the sensing wavelength must be allocated in the wavelength plan, and inline optical amplifiers block backscatter entirely because they are unidirectional.

Applications

Wells and Reservoirs

Downhole temperature profiling was the first large commercial market for DTS and remains a major one. A fiber run behind casing or in a control line reports the temperature profile of the whole well continuously, revealing which zones are producing, where water or gas is entering, and how injected fluid is distributing. Because flowing fluid carries a thermal signature, a temperature profile is effectively a flow profile read through a thermal model of the wellbore. Adding DAS on the same or a parallel fiber extends this to flow-induced noise, sand production, perforation-cluster efficiency during hydraulic fracturing, and vertical seismic profiling with the well as the receiver array; the associated instrumentation is treated in the article on downhole and well logging electronics. Geothermal wells present the same problem hotter, which is where metal-coated fiber and hydrogen-resistant designs earn their cost.

Pipelines

A fiber laid alongside a pipeline detects leaks by two independent mechanisms. Escaping gas cools by Joule-Thomson expansion and escaping liquid changes the local thermal environment, both of which a DTS system sees as a temperature anomaly at a known position. Simultaneously, DAS detects the acoustic signature of a leak, and of third-party interference such as excavation, often at ranges of tens of meters from the cable. The same fiber tracks pigs and detects ground movement. The operational value lies less in detection than in localization: an alarm naming a position to within a few meters along a hundred-kilometer line changes the response from a search to a repair.

Power Cables and Thermal Rating

The current-carrying capacity of a buried or submarine power cable is limited by conductor temperature, and the steady-state calculations of the IEC 60287 series necessarily assume worst-case soil thermal resistivity and ambient conditions along the whole route. In reality the limit is set by a few hot spots, typically where the cable passes under a road, through a duct bank, or near another circuit. A DTS fiber in the sheath or in an integrated optical ground wire measures the true profile and feeds a real-time thermal rating calculation that infers conductor temperature from the measured surface temperature and the loading history. Operators routinely find that the true constraint is a handful of meters out of many kilometers, and that the circuit can carry more than its nameplate rating for most of the year. The same fiber gives early warning of joint overheating.

Perimeter, Rail, and Traffic

A buried DAS fiber along a fence line, a border, or a facility boundary detects footsteps, vehicles, and digging, and classifies them by their spectral and temporal signature; the economics are compelling for long perimeters because the marginal cost of another kilometer is the cable. On railways the same technique tracks every train as a moving acoustic source, giving train location without trackside equipment, and detects rockfalls, broken rails, cable theft, and trespass, often on fiber already installed along the right of way for signaling. In urban settings, fiber in existing telecommunications ducts has been used to observe traffic and to characterize shallow ground structure from the resulting noise.

Civil Structures and Geotechnics

Distributed strain sensing suits structures whose critical location is not known in advance. A Brillouin fiber cast into a concrete segment, bonded along a bridge girder, or wrapped around a tunnel lining reports the strain profile of the whole element and reveals cracking wherever it appears. Piles, diaphragm walls, and anchors are instrumented the same way, and the resulting load-transfer profile is far more informative than the handful of point gauges it replaces. Dams and levees use DTS to detect seepage, since infiltrating water changes the local thermal regime, and unstable slopes use distributed strain to detect movement along a whole hillside.

Seismology on Telecommunications Fiber

The most striking recent development is the use of existing telecommunications cable as a dense seismic array. Ajo-Franklin and colleagues reported in Scientific Reports in 2019 that dark fiber in a conventional conduit could be used for near-surface characterization and for detecting broadband seismic events, establishing that unused strands of ordinary installed cable are usable seismic instruments. Submarine cables extend the idea to the ocean floor, where conventional instrumentation is scarce: work on the MARS cable in Monterey Bay analyzed a twenty-kilometer section of a fifty-one-kilometer armored cable with a ten-meter gauge length, stacked to twenty-meter channels for roughly a thousand channels, and recorded a magnitude 3.4 earthquake in March 2018 along with ocean surface gravity waves and interface waves.

The appeal is channel count and geography. A single fifty-kilometer cable at ten-meter spacing provides five thousand recording points along routes that cross oceans, mountains, and cities where no seismometers exist. The limitations are equally clear: the measurement is single-component axial strain rather than three-component ground motion, the coupling is uncontrolled and varies along the route, the instrument response requires careful calibration, and the geometry is whatever the cable installers chose decades ago.

Standards and Specification

Standardization has followed commercial adoption. The IEC 61757 series covers fibre optic sensors, with a generic specification and detail specifications per measurement type. IEC 61757-1-1:2020 addresses strain measurement using fiber Bragg gratings, the quasi-distributed counterpart to the techniques described here; IEC 61757-1-2:2023 specifies distributed strain measurement; IEC 61757-1-4:2025 covers absolute strain measurement based on spectral correlation analysis of Rayleigh backscattering signatures in single-mode fiber, that is, the OFDR method above; and IEC 61757-2-2:2016 specifies distributed temperature sensing. These documents matter chiefly because they define measurement terms consistently, which is what makes competing instruments comparable.

A defensible specification names these together rather than separately: sensing range; spatial resolution, defined as the ten to ninety percent width of the response to a step change and stated at the far end of the range; sampling interval, stated separately; measurement resolution or repeatability, distinguished from accuracy, at a stated time and position; measurement time; and the fiber type, cable construction, and connector plan assumed. Acoustic systems add gauge length, frequency response, and whether the output is calibrated strain, strain rate, or an uncalibrated intensity.

Acceptance testing is worth specifying too. Immersing a known length of installed cable in a stirred bath and confirming both the reported temperature and the reported position validates the whole chain, including the distance mapping that field splices and slack coils disturb.

Conclusion

Distributed fiber optic sensing rests on one idea that separates it from every other optical sensing technique: the fiber has no transducers in it. The measurement comes from scattering that occurs everywhere along the glass, so coverage is continuous by construction, and the same cable that senses carries the signal home. Rayleigh scattering, being elastic, reports changes in optical path length and gives acoustic and high-resolution strain sensing. Brillouin scattering, shifted by the acoustic velocity of the glass, reports strain and temperature together through a frequency immune to loss. Raman scattering, shifted by molecular vibration, reports temperature alone and is the only one of the three with no cross-sensitivity to resolve.

The interrogation architectures follow from the ranging problem. Time-domain methods trade pulse energy for spatial resolution and reach tens of kilometers, frequency-domain methods trade range for resolution and reach micrometers over tens of meters, stimulated Brillouin analysis buys signal strength at the cost of requiring both fiber ends, and phase-sensitive OTDR buys waveform fidelity at the cost of managing fading. In every case the four performance quantities are locked together by the finite energy the fiber returns, and the honest way to read a datasheet is to ask which operating point each number describes.

What decides success in the field, however, is usually not the interrogator. It is whether the cable transmits the strain it should transmit and rejects the strain it should reject, whether the loose-tube reference and the strain-coupled fiber share a thermal environment, whether the optical path is free of reflective connectors, and whether the distance axis has been mapped to real positions. The optics set the ceiling; the mechanical and installation engineering decide how close a real system gets to it.

Related Topics