Vibration Sensors

What is a Vibration Sensor

A vibration sensor is a transducer that converts the oscillatory mechanical motion of a structure into a proportional electrical signal. In rotating machinery, that motion is the symptom: a healthy bearing, an undamaged gear, and a balanced rotor each produce a characteristic vibration signature, and any change in that signature, well before audible noise or heat, signals a developing fault. The vibration sensor is therefore the primary input to condition monitoring, the discipline that lets a plant move from run-to-failure or fixed-interval maintenance toward predictive maintenance based on the real state of the machine.

Vibration is described by three interchangeable kinematic quantities related by differentiation: displacement, velocity, and acceleration. Displacement (how far the surface moves, in micrometres or mils) dominates at low frequency and large amplitude. Velocity (how fast it moves, in mm/s) is the balanced energy metric across the mid band and is the basis of broadband severity standards. Acceleration (how quickly velocity changes, in g or m/s2) emphasises high-frequency content and is the most sensitive to early bearing and gear defects. A sensor is built to measure one of these directly, and the others are obtained by integration in the analyser. Integration is mathematically clean from acceleration toward velocity and displacement, but double integration amplifies low-frequency noise, which is one reason direct proximity probes still exist for shaft displacement.

Structurally, the dominant industrial vibration sensor is the piezoelectric accelerometer. It contains a seismic (proof) mass coupled to a piezoelectric crystal element; when the housing accelerates, the mass exerts a force on the crystal, which generates an electric charge proportional to that force, and therefore to acceleration. Above the sensor base sits the signal chain: in an IEPE sensor, a built-in microelectronic amplifier converts the high-impedance charge into a low-impedance voltage; in a charge-mode sensor, that amplifier is external. A second major family, the eddy-current proximity probe, contains no moving parts at all: a coil driven at radio frequency induces eddy currents in a nearby conductive shaft, and the resulting change in coil impedance is demodulated by a driver (Proximitor) into a voltage proportional to the gap.

The industrial history runs from the 1950s, when piezoelectric accelerometers and charge amplifiers entered aerospace and laboratory use, to the late 1960s, when Bently Nevada commercialised the eddy-current proximity probe for turbomachinery shaft monitoring, to the 1970s introduction of integrated-electronics (IEPE / ICP) accelerometers that put the amplifier inside the sensor and made two-wire field wiring practical. From the 2000s, MEMS (Micro-Electro-Mechanical Systems) accelerometers shrank the sensing element to chip scale and, paired with wireless radios and edge analytics, opened low-cost, battery-powered monitoring of secondary assets that were never economic to wire individually.

In application scale, vibration sensing covers an enormous dynamic range: from sub-micrometre shaft orbits on a 3000 rpm steam turbine, to a few mm/s on a well-maintained pump, to tens of g of high-frequency energy radiating from a spalled rolling-element bearing, up to thousands of g of shock on a crusher or a drop test. No single sensor spans that range; the engineering task is to match the measured quantity, sensitivity, and frequency band to the specific failure mode and machine class being protected.

Sensor Types and Measured Quantity

The first selection decision is not brand or price, it is which physical quantity to measure, because that fixes the sensor family. Industrial practice recognises three primary types, each tuned to a frequency band and a failure mode. The table below compares the core characteristics; the discussion that follows explains where each one belongs.

Accelerometers are the workhorse of plant vibration monitoring. Because acceleration emphasises high-frequency content, an accelerometer is the most sensitive instrument for the earliest stages of rolling-element bearing degradation and for gear-mesh defects, which appear as energy in the kilohertz region long before they show up as overall velocity. A general-purpose industrial accelerometer such as the PCB Piezotronics 352C33 offers 100 mV/g sensitivity over a 0.5 Hz to 10 kHz frequency range, which spans both the low-speed imbalance fundamentals and the high-frequency bearing tones.

Velocity sensors and 4-20 mA velocity transmitters exist because the international severity standards are written in velocity. A velocity transmitter typically integrates an internal accelerometer and outputs overall RMS velocity directly, with ranges such as 0 to 25 mm/s RMS or 0 to 50 mm/s RMS over the 10 to 1000 Hz band, matching ISO 20816 evaluation for pumps, fans, and motors. The loop-powered 4-20 mA output lets a PLC, DCS, or variable-frequency drive trend machine health and trip on a threshold without a dedicated analyser, which is why velocity transmitters are common on balance-of-plant assets.

Eddy-current proximity probes occupy a separate world. They are non-contact, mounted through the bearing housing to observe the shaft directly, and they measure relative displacement between probe tip and shaft surface. Crucially, they read both the dynamic vibration and the static average gap, so they capture a steady shaft position offset that an accelerometer fundamentally cannot. This is why they are mandated by API 670 on large machines with sleeve or journal (fluid-film) bearings, where the casing barely moves but the shaft can orbit dangerously inside the oil film. A standard Bently Nevada 3300 XL 8 mm system delivers 7.87 mV/um (200 mV/mil) and is the reference design for this duty.

MEMS accelerometers are the newest family. A micromachined silicon proof mass and capacitive or piezoelectric pickoff give a small, low-cost, low-power device with usable response down to DC. Early MEMS noise floors limited them to coarse monitoring, but modern parts have closed much of the gap, and their integration with wireless radios and battery power makes them the practical choice for blanket-coverage monitoring of the many secondary assets that were never worth hard-wiring.

Sensing Technologies and the IEPE Signal Chain

Underneath the type label sits the sensing physics and, just as important, the signal chain that delivers the signal to the analyser intact. Four sensing technologies dominate, and within the piezoelectric family the choice between IEPE and charge mode is the single most consequential wiring decision. The table compares the four technologies before the text explains the IEPE chain in detail.

Piezoelectric IEPE (ICP) is the de-facto industrial standard. The acronym stands for Integrated Electronics Piezo-Electric; ICP is the equivalent trademark, and CCLD, Isotron, and DeltaTron are other vendor names for the same two-wire constant-current scheme. A miniature charge amplifier inside the sensor converts the crystal's high-impedance charge into a low-impedance voltage, conventionally 100 mV/g. The sensor is powered and read over a single coaxial pair: the instrument supplies a constant current, commonly between 2 and 20 mA with 4 mA typical, and the sensor returns its signal riding on a positive DC bias of roughly 8 to 12 V. Because the output is low impedance, IEPE rejects electromagnetic and radio-frequency interference far better than raw charge and tolerates long cable runs and ordinary connectors, which is exactly why it took over the plant floor.

Piezoelectric charge mode keeps only the crystal in the sensor and moves the amplifier outside. With no semiconductor electronics inside, the element survives temperatures of 250 degrees Celsius and well beyond, making charge mode the correct, and often the only, choice for gas-turbine casings, exhaust manifolds, and steam-line monitoring. The cost is fragility of the signal path: the high-impedance charge output is corrupted by cable flexing (triboelectric noise) and by contamination at the connector, so charge mode demands a dedicated external charge amplifier, low-noise coaxial cable, and scrupulous connector cleanliness. For any ambient-temperature machine, IEPE is simpler and cheaper and should be preferred.

MEMS sensors form a silicon proof mass on a chip and detect its deflection capacitively or piezoelectrically. They respond down to DC, which makes them suitable for low-frequency and tilt measurements that AC-coupled piezoelectric IEPE cannot reach, and they are cheap and tiny enough to embed by the thousand. Research-grade piezoelectric MEMS designs have reported voltage sensitivities from a few mV/g to well over 100 mV/g depending on beam geometry and proof mass, illustrating how strongly MEMS performance depends on the specific structure rather than the technology label. In the field, MEMS is the enabler of wireless, battery-powered condition monitoring on secondary assets.

Eddy-current sensing is fundamentally different: it is inductive, not piezoelectric. A coil at the probe tip is driven at radio frequency and generates an alternating magnetic field; when the conductive shaft approaches, induced eddy currents alter the coil impedance, and an external driver demodulates this into a voltage proportional to the gap. The system measures both the slowly varying average gap (static shaft position) and the dynamic gap variation (vibration), with no contact and no wear. Its frequency response runs from DC to roughly 1 kHz, which is ample for shaft vibration on machines turning at thousands of rpm, and its output is calibrated in mV per micrometre of displacement.

Mounting, Frequency Range and Standards

A vibration sensor is only as good as its mechanical coupling and the standard you judge its readings against. Two engineering facts dominate this chapter: mounting sets the upper usable frequency, and the evaluation standard sets the pass-fail line. Both are routinely underestimated by buyers who focus only on the sensor's catalogue frequency range.

Mounting and the mounted resonance. An accelerometer has a high natural resonance on its own, but once attached to a machine the effective resonance, called the mounted resonance, drops, because the stiffness and mass of the mounting interface join the system. The flat, usable band conventionally extends only to about one third of the mounted resonance; above that, the response rises sharply toward the resonant peak and the reading is no longer trustworthy. Mounting method therefore directly determines how much high-frequency content (bearing and gear faults) you can actually measure. The table summarises typical mounted resonance and the resulting usable band.

The practical rule: stud mounting is the reference and the only method suitable for permanent installations and for any diagnostic that needs 5 to 10 kHz, such as rolling-element bearing or gear-mesh analysis. Adhesive mounting (epoxy or a glued mounting pad) is the choice for curved or thin housings where drilling and tapping a stud hole is impractical. Magnetic mounting trades bandwidth for speed and is acceptable for route-based overall-level surveys, with flat magnets outperforming two-pole magnets on usable band. Handheld probes are for the quickest of spot checks only; their poor and inconsistent coupling makes trended data meaningless.

The ISO 20816 / 10816 severity standards. The international framework for judging overall machine condition is the ISO 20816 series, which consolidates and supersedes the older ISO 10816 (bearing-housing vibration) and ISO 7919 (shaft vibration) series, with ISO 20816-3:2022 replacing ISO 10816-3:2009 and ISO 7919-3:2009 into a unified evaluation. The criterion is broadband velocity RMS measured on the bearing housing in the 10 to 1000 Hz band. The standard sorts results into four zones: Zone A for newly commissioned machines, Zone B acceptable for unrestricted long-term operation, Zone C unsuitable for continuous long-term operation and requiring remedial action, and Zone D dangerous, where damage can occur. ISO 20816-3 applies to industrial machine sets above 15 kW with operating speeds from 120 to 15000 rpm; the exact mm/s boundaries depend on the machine group and on whether the support is rigid or flexible, so the boundary must be read from the applicable sub-part rather than memorised as one number. Because so many existing plant procedures still cite ISO 10816-3 limits, the older designation remains in everyday use even as the standard moves to ISO 20816.

API 670 and calibration. For turbomachinery (turbines, centrifugal and axial compressors, large pumps on fluid-film bearings), the governing document is API 670, the machinery protection system standard, which mandates non-contact eddy-current shaft probes and specifies their mechanical configuration, linear sensing range, accuracy, and temperature stability; a 3300 XL 8 mm 5-metre system is built to comply with API 670. Sensor calibration itself is governed by the ISO 16063 series, the international methods for the calibration of vibration and shock transducers, which underpins the traceable sensitivity figure printed on every accelerometer calibration certificate.

Key Specification Parameters

A vibration sensor datasheet can list two or three dozen lines, but only a handful drive selection. The parameters below are the ones to lock down before requesting a quote; each is explained in the order an engineer actually reasons through them.

Sensitivity (mV/g for accelerometers, mV/um for proximity probes). Sensitivity is how much electrical output the sensor produces per unit of input. For accelerometers it is given in mV/g, where 1 g equals 9.80665 m/s2; 100 mV/g is the general-machinery default because it balances signal level against measurement range. Higher sensitivity such as 500 mV/g improves resolution on quiet, low-speed machines but reduces the maximum measurable amplitude before the output clips; lower sensitivity such as 10 mV/g extends the range for high-shock duties like crushers and presses. For eddy-current probes, sensitivity is in mV/um, with 7.87 mV/um (200 mV/mil) the API 670 reference value.

Frequency range and amplitude range. The frequency range is the band over which sensitivity stays within a stated tolerance (often plus or minus 5 percent or plus or minus 3 dB). Confirm the low end can reach your shaft fundamental (a 600 rpm machine has a 10 Hz fundamental) and the high end, after mounting derating, covers your highest fault frequency. The amplitude range (for example plus or minus 50 g) must exceed the worst-case peak, including transient shock, with margin so the IEPE output does not rail against its bias-limited swing.

Output and powering. For IEPE sensors, the key numbers are the required constant-current excitation (typically 2 to 20 mA) and the output DC bias (typically 8 to 12 V); the analyser or signal conditioner must supply the former and accommodate the latter. For 4-20 mA velocity transmitters, the relevant figures are the velocity full scale (such as 0 to 25 or 0 to 50 mm/s RMS) and the integration band (10 to 1000 Hz). For proximity-probe systems, the driver output scaling and the required supply (commonly -24 V DC) define the interface.

The next table is a key-specifications comparison of representative, real series across the three sensor families, so the spec lines above appear in concrete form rather than in the abstract.

Linearity and resolution. For proximity probes, linearity is the deviation of the gap-to-voltage curve from a straight line over the calibrated range; a 3300 XL system holds linearity within plus or minus 1.27 um (0.05 mil) over the middle 80 percent of its range, with resolution around 0.38 um (0.015 mil). For accelerometers, the equivalent figures are amplitude linearity and the broadband noise floor, which sets the smallest vibration you can resolve, an important number for low-speed and incipient-fault detection.

Environmental and electrical robustness. The operating temperature range determines whether IEPE or charge mode is viable. Ingress protection (IP rating) governs washdown and outdoor survival; sealed welded-hermetic stainless housings are standard for harsh duty. Maximum shock rating (for example several thousand g) protects the sensor against handling and process transients. Where the machine sits in a classified area, the hazardous-area certificate (ATEX, IECEx, CSA) is a hard gate, not an option.

Selection Decision Factors

To turn the preceding chapters into a specific part number, follow the decision sequence below. As with most instrumentation, the costly mistakes come not from one wrong line on the datasheet but from deciding too early at the wrong level. These eight steps make a workable RFQ template.

1. Measured quantity and machine class: Decide acceleration, velocity, or displacement first. Fluid-film (sleeve or journal) bearings on large turbomachinery call for eddy-current proximity probes per API 670; rolling-element bearings and gearboxes call for accelerometers; overall trending and threshold tripping call for velocity transmitters.
2. Frequency band of the target fault: Map the failure mode to a band. Imbalance and misalignment live near the running speed (often 10 to 200 Hz); bearing defect tones and gear mesh live in the kHz region. Choose a sensor whose flat band, after mounting derating, covers that fault frequency.
3. Sensitivity versus range: Select sensitivity (100 mV/g default, 500 mV/g for low-level machines, 10 mV/g for high-shock) so the worst-case peak stays well inside the output swing while still resolving the smallest fault of interest.
4. Signal chain and temperature: Use IEPE for ambient machinery (simple two-wire, noise-tolerant). Switch to charge mode only when temperature exceeds the IEPE electronics limit of roughly 120 to 150 degrees Celsius, accepting the external charge amplifier and low-noise cabling.
5. Mounting method: Stud mount for permanent installs and any high-frequency diagnostic; adhesive for curved or thin housings; magnetic only for route surveys; never a handheld probe for trended data. Confirm the mounted resonance keeps your highest fault frequency below one third of it.
6. Output and integration: Match the interface to the receiving system. A continuous monitor wants raw IEPE or a proximity driver into an analyser; a PLC or DCS that only needs an alarm wants a loop-powered 4-20 mA velocity transmitter scaled to ISO 20816 severity.
7. Environment and certification: Set IP rating for washdown and outdoor exposure, temperature range for the mount location, and the hazardous-area certificate (ATEX, IECEx, CSA) per the area classification. Verify the certificate against the exact part number, since approval is per variant.
8. Calibration and total cost of ownership: Require an ISO 16063 traceable calibration certificate, plan periodic recalibration, and weigh the installed cost (sensor, cable, conditioner, mounting hardware, labour) against the avoided cost of an undetected bearing or rotor failure, which is the entire economic justification for the sensor.

One last dimension that buyers overlook is serviceability and ecosystem fit: connector and cable availability, compatibility with the existing analyser or monitoring rack, spare-part lead time, and vendor calibration-lab access. For turbomachinery this usually means staying within the Bently Nevada 3300 XL and 3500 ecosystem mandated by API 670 projects; for general plant monitoring, PCB Piezotronics, Wilcoxon, Hansford Sensors, and CTC supply interchangeable 100 mV/g IEPE accelerometers and 4-20 mA transmitters with broad ATEX and IECEx coverage, so the practical question becomes which one your maintenance team can stock, calibrate, and replace fastest over the next decade.