Vibration Sensor Types: Accelerometers and Beyond

The three main vibration sensor types are the piezoelectric accelerometer, the velocity sensor, and the eddy-current proximity probe. They differ in what physical quantity they measure, the frequency range they cover, and whether they touch the machine. Pick the wrong one and you either miss the fault you were hunting or measure something that is not the machine's real problem.

This matters because the accelerometer has become so dominant that it gets used by reflex, and reflex is wrong on a meaningful share of machines. A big turbine or compressor on fluid-film bearings shakes its shaft inside a heavy casing that barely moves, an accelerometer on the case sees almost nothing while the shaft is orbiting toward a rub. Knowing what each sensor measures, and where each one is blind, is what keeps you from monitoring the wrong signal. This guide covers how the three types work, their trade-offs, and how to choose.

How does a piezoelectric accelerometer work?

A piezoelectric accelerometer works by squeezing a crystal. Inside the sensor a small seismic mass sits on a piezoelectric crystal; when the machine vibrates, the mass pushes on the crystal, and the crystal generates an electric charge proportional to the acceleration. That charge is the signal. Because acceleration emphasizes high frequencies, accelerometers excel exactly where early bearing and gear-mesh faults live, and because they are solid-state with no moving coils, they are rugged and reach high frequencies that other sensors cannot.

Most industrial accelerometers today are the IEPE type, also called ICP or voltage-mode, meaning a tiny amplifier is built into the sensor to convert the crystal's high-impedance charge into a low-impedance voltage that travels well down ordinary cable. A common sensitivity is 100 mV/g, a general-purpose choice. The one thing accelerometers do poorly is very low frequency: below a few hertz their output falls off and noise dominates, so they are a weak choice for slow shaft motion. Because acceleration integrates cleanly to velocity, a single accelerometer plus the analyzer's math covers both high-frequency detail and the mid-band velocity used to judge overall health, the reason it is the default sensor in vibration analysis.

What is a velocity sensor and when do you use one?

A velocity sensor measures how fast a surface is moving, and you use it where you want velocity directly, often at lower frequencies, and in rugged legacy installations. The classic design is electrodynamic: a coil moves relative to a magnet inside the sensor, and the motion generates a voltage proportional to velocity, no external power needed. A second, modern form is an accelerometer with a built-in electronic integrator, sometimes called a velometer, which outputs velocity while keeping an accelerometer's ruggedness.

Velocity sensors were the workhorse before IEPE accelerometers became cheap and reliable, and many are still in service on machines specified around them. Their strength is that velocity is the parameter the severity standards use, so a velocity sensor reads out directly in the mm/s that decide machine health. Their weaknesses are a limited frequency range, the moving-coil type rolls off at both ends and cannot reach the high frequencies of early bearing faults, and, in the electrodynamic design, moving parts that wear. On new installations most engineers now fit an accelerometer and integrate to velocity in software, keeping the wide range and reserving true velocity sensors for cases where their specific ruggedness or direct output is wanted.

What is a proximity probe and why is it different?

A proximity probe, or eddy-current probe, is a non-contact sensor that measures the displacement of a shaft relative to its bearing housing, and it is different because it watches the shaft itself, not the casing. Mounted through the bearing housing at a small gap from the shaft, it drives a high-frequency signal that induces eddy currents in the shaft surface; the strength of those currents varies with the gap, so the probe's output tracks how far the shaft moves. Crucially, its response reaches all the way down to 0 Hz (DC), so it measures both the static shaft position and slow orbital motion that an accelerometer cannot see.

This is the sensor for large machines on fluid-film (sleeve) bearings, turbines, big compressors, generators. In those machines the heavy casing barely moves while the shaft floats on an oil film and can orbit dangerously; a casing accelerometer would report almost nothing. The proximity probe measures the quantity that actually matters there: shaft position and motion inside the clearance. It is why permanent machinery-protection systems on critical turbomachinery are built around proximity probes, typically two per bearing at 90 degrees to capture the shaft orbit, feeding trip and alarm logic. It measures displacement, in micrometers or mils, the low-frequency parameter covered in vibration analysis basics.

How do the sensor types compare?

They compare across four axes that decide the choice: what they measure, their frequency range, contact versus non-contact, and what they are best at. No sensor wins on all four, which is why a mature program uses more than one type.

How do you choose the right vibration sensor?

You choose by working through what you need to measure, not by defaulting to the accelerometer. A short sequence gets you to the right sensor for the machine in front of you.

1. Decide casing or shaft. If the machine's problem is shaft motion inside fluid-film bearings, turbines, large compressors, you need a proximity probe on the shaft. For rolling-element-bearing machines where the casing transmits the vibration, a casing sensor is right.
2. Set the frequency you must see. High-frequency bearing and gear faults demand an accelerometer. Slow shaft motion and static position demand a proximity probe. Mid-band overall health can use either an accelerometer or a velocity sensor.
3. Pick the parameter and let it confirm the sensor. Displacement points to a proximity probe, acceleration to an accelerometer, direct velocity to a velocity sensor or an integrating accelerometer.
4. Match sensitivity to amplitude. Choose a sensitivity suited to the expected vibration level so the signal is neither buried in noise nor clipped. 100 mV/g suits general use; low-level or high-level machines want different sensitivities.
5. Account for the environment. High temperature, chemicals, or hazardous areas narrow the choice; pick a sensor rated for the conditions and, for hard-to-reach or hot points, consider a permanent sensor over a route reading.
6. Choose portable or permanent. Route data collection favors a handheld accelerometer; critical or fast-failing machines favor permanently mounted sensors, increasingly low-cost MEMS accelerometers, streaming continuously.
7. Plan the mount. The mount sets the usable frequency ceiling as much as the sensor does, so decide the mounting method alongside the sensor, per accelerometer mounting methods.

Where do MEMS and wireless sensors fit?

MEMS accelerometers fit the fast-growing continuous-monitoring layer. A MEMS sensor is a tiny accelerometer etched into silicon, cheap enough to leave permanently on machines that could never justify a wired system. Their frequency range and noise floor once trailed traditional piezoelectric sensors, but they have improved to the point that they cover a large share of general monitoring, especially when paired with wireless transmission and battery power. That combination is what makes it practical to instrument dozens or hundreds of second-tier machines that a walk-around route visits only occasionally.

The shift they enable is from periodic route readings to always-on data, which changes what a program can catch. A fault that develops between monthly route visits slips past a portable program but not a continuous one. This is the same move toward streaming machine data covered in machine monitoring and it is central to modern condition-based maintenance and predictive maintenance. The sensor choice and the monitoring architecture are two halves of the same decision.

Where sensor choice fits your reliability program

Choosing the right vibration sensor is a small decision that quietly sets the ceiling on everything downstream. The finest analysis software cannot recover a fault the sensor never captured, an accelerometer on a turbine casing, or a velocity sensor asked to find a high-frequency bearing tone, throws the information away before it reaches the analyzer. Match the sensor to what you measure, the frequency you need, and the environment, and the rest of the program has good data to work with.

The harder problem, as always, is turning readings from many sensor types into one coherent picture of a machine. A proximity-probe trip level, a route accelerometer trend, and a wireless MEMS alarm all describe the same asset but usually live in different systems. Pulling them into one operational view alongside the machine's maintenance history and run hours is where machine monitoring platforms like Harmony fit, so a rising vibration trend from any sensor lands next to the context that makes it actionable. It layers onto the systems you already run, with no rip-and-replace. See how the platform works and for the diagnostic step that follows the reading, vibration spectrum analysis and bearing defect frequencies.