Start with the parameter, not the sensor

The most common mistake in vibration monitoring is starting with a sensor catalogue. The sensor is the last decision. The first is: what do I actually want to measure?

Vibration is one physical motion, but it can be expressed three ways, and each emphasises a different part of the frequency spectrum.

Parameter	What it measures	Units (common)	Strong at
Displacement	How far the part moves	µm, mils (peak-to-peak)	Low frequency
Velocity	How fast it moves	mm/s RMS, in/s	Mid frequency
Acceleration	How quickly velocity changes	g, m/s²	High frequency

The relationship is one of differentiation. Velocity is the rate of change of displacement; acceleration is the rate of change of velocity. Each step shifts the emphasis upward in frequency, which explains almost every parameter-selection rule in the field.

What this means in practice:

• A slow-turning machine at 30 rpm has a 0.5 Hz fundamental. In acceleration units that fundamental is essentially invisible — you measure it as displacement, which is why large slow machines such as low-speed compressors, cooling tower fans and kiln drives use displacement or proximity probes.
• A motor or pump running at 1,500–3,600 rpm puts its most useful severity information in the velocity band, which is why velocity is the parameter behind the ISO machine vibration severity standards.
• A rolling element bearing defect produces impacts that excite frequencies from several kHz into the tens of kHz. In velocity those appear as a barely visible rise; in acceleration they stand out clearly.

The trap: acceleration and velocity are not better or worse than each other — they are different magnifying glasses. A machine can look acceptable in velocity while a bearing is already failing, and look alarming in acceleration because of a noisy gearbox that has run that way for a decade. Pick the parameter sensitive to the fault you are trying to catch.

The three sensor types you will actually meet

1. Piezoelectric accelerometer (IEPE)

The workhorse. A seismic mass presses on a piezoelectric crystal; vibration generates a charge, which the built-in electronics convert to a voltage. This is the IEPE (Integrated Electronics Piezo-Electric) type, and it is the sensor behind most condition-monitoring programmes.

Take the Wilcoxon 786A as a concrete reference point, since its published figures are useful for calibrating your expectations:

Specification	Value
Sensitivity	100 mV/g
Sensitivity tolerance	±5 %
Frequency response (±3 dB)	0.5 Hz – 14,000 Hz
Full-scale range	80 g peak
Resonance frequency	30 kHz
Mounting thread	1/4-28 UNF tapped hole
Connector	MIL-C-5015, 2-pin, top exit
Temperature range	−55 °C to +120 °C
Weight	90 g
MTBF	25 years

Two things worth noting. First, the resonance frequency of 30 kHz sits more than twice the top of the usable range — the general rule is that the mounted resonance should be at least twice your highest frequency of interest, or the resonance peak contaminates your data. Second, that 14 kHz figure is the sensor's limit, not your measurement's; mounting decides what you actually get.

2. Velocity sensor / 4–20 mA transmitter

Many plants do not want spectra. They want a number they can trend and an alarm they can wire into the DCS. That is what a loop-powered transmitter provides: a 4–20 mA output proportional to overall vibration, where 4 mA is no vibration and 20 mA is full scale.

Taking the Wilcoxon PC420 and PCC420 families as the reference, these come in two flavours:

Output	Full-scale options
Acceleration, RMS or calculated peak	5 g, 10 g, 20 g, 50 g
Velocity, RMS or calculated peak	0.5 / 1.0 / 2.0 / 3.0 / 5.0 in/s (12.7 / 25.4 / 50.8 / 76.2 / 127 mm/s), plus metric 10 / 20 / 25 / 50 mm/s

Connector options are a 2-pin MIL-C-5015 or a 4-pin M12. Plant control systems already accept 4–20 mA, so a vibration transmitter joins pressure, temperature and flow in the same historian without specialist software. The limitation is equally simple: you get one overall number, not a spectrum, so you cannot diagnose which fault it is — only that something changed.

3. Proximity probe (displacement)

An eddy-current probe measures the gap between probe tip and shaft surface directly, giving true shaft-relative displacement. It is the standard for large turbomachinery and any machine where you must see the shaft itself rather than the bearing housing.

Choosing between them

If you need to...	Choose
Diagnose a specific fault from a spectrum	IEPE accelerometer
Trend overall machine health with no analyst	4–20 mA velocity transmitter
Feed a plant DCS or SCADA alarm	4–20 mA transmitter (acceleration or velocity)
Measure slow shaft motion on a large machine	Proximity probe (displacement)
Catch early rolling-element bearing faults	IEPE accelerometer, acceleration units
Follow ISO severity standards for a motor or pump	Velocity — IEPE accelerometer or 4–20 mA velocity output
Protect a remote or unmanned asset with minimal wiring	4–20 mA transmitter

Mounting: where most measurement error is born

This is the part that quietly destroys data. The sensor's frequency response is measured under ideal calibration conditions. What you get on the machine depends entirely on how you attach it, because the mounting method adds its own stiffness and mass — and stiffness sets the frequency ceiling.

Wilcoxon publish a clear ranking across four basic methods, and the numbers are worth memorising:

Mounting method	Frequency capability	Permanence	Best for
Threaded stud	Highest — best method, recommended above 10 kHz	Permanent	Permanent installations, harsh environments, high-frequency work
Adhesive / cementing pad	Approaches stud mounting when done correctly	Permanent or long-term	Where the machine cannot be drilled
Magnet	Reduced — flat magnets about 5–10 kHz mounted resonance; 2-pole magnets about 3–7 kHz	Temporary	Route-based walkaround data collection
Probe tip	Lowest and least reliable; avoid below 10 Hz	Instant	Difficult-to-reach points, trial measurements only

Stud mounting — the reference standard

Stud mounting most closely reproduces the calibration condition and gives the widest dynamic range. It requires drilling and tapping a small hole, and the preparation matters more than the hardware:

• Spot face and clean the surface before tapping
• Keep hole depth less than the mounting surface thickness so you do not break through
• Check the hole is perpendicular, so the sensor seats flat
• Apply the correct torque — for a typical 1/4-28 stud, 30 in-lbs is generally about right
• Use a coupling fluid such as silicone grease, machine oil or petroleum jelly. This increases mounting stiffness and measurably improves frequency response at almost no cost

That last point is the cheapest accuracy improvement available, and it is routinely skipped.

Adhesive mounting — the practical alternative

If the machine cannot be drilled, a cemented pad is the next best thing and approaches stud-mounted performance. Surface preparation is again decisive: remove all paint and debris, grind the area reasonably flat, and use a hard adhesive (cyanoacrylate types are common). Some adhesives also electrically isolate the sensor, a useful side benefit where ground loops are a concern.

Magnetic mounting — convenient, and capped

Magnets are for route-based collection, where speed matters more than absolute accuracy. The poorer contact and added mass create a new, lower mounted resonance, truncating the usable range. Flat magnets give better contact than 2-pole magnets, which is why the figures differ. Two rules:

• Use flat magnets on flat surfaces and 2-pole magnets on curved or irregular surfaces
• Question large spikes in magnetic-mount spectra — check whether they are real before raising a work order

Probe tips — last resort

Handheld probe tips reach awkward places but are the least reliable method: data is frequently distorted, longer probes introduce more error, and below about 10 Hz the machine's own movement makes it hard to hold the sensor in contact. If a probe-tip reading looks alarming, verify it before acting.

The rule that ties it together: the usable range of your measurement is the lower of the sensor's frequency response and the mounting method's frequency capability. A 14 kHz sensor on a magnet is a 5–10 kHz measurement.

What mounting does to your data

Symptom	Likely mounting cause
Unexplained peak in the 3–10 kHz region	Magnetic mount resonance — verify with a stud or adhesive mount
High-frequency detail absent that you expected	Magnet, or poor surface preparation on a stud or pad
Repeatable but low reading	Insufficient mounting torque, or no coupling fluid
Reading varies between visits	Handheld probe tip — switch to a fixed point

Frequency range: putting it all together

The frequency range you need is set by the machine, not by the sensor catalogue. Work backwards from the faults you want to catch.

Speed / fault	Frequency of interest	Parameter	Practical mounting
Large slow machine, 30–600 rpm	Low — tens of Hz and below	Displacement	Proximity probe, or stud-mounted with extended low-frequency response
Standard motor or pump, 1,500–3,600 rpm	10 Hz – 1 kHz	Velocity	Stud or adhesive pad — very achievable
Gearbox mesh	Several hundred Hz to several kHz	Velocity and acceleration	Stud mounting — do not use a magnet
Rolling element bearing defect	1 kHz – 20 kHz+	Acceleration, with enveloping	Stud mounting, high-frequency sensor
Anything you want to trend without an analyst	Overall, any band	Velocity (ISO severity)	Stud or 4–20 mA transmitter

Three selection rules fall out of this table:

1. The sensor's resonance must be at least twice your highest frequency of interest — otherwise the resonance peak sits inside your measurement band and you will chase a fault that does not exist.
2. The mounting must not be the limiting factor. For bearing and gear work in the kHz range, stud mounting is not optional.
3. Match the parameter to the fault. If you can only afford one permanent sensor per machine, a velocity output gives the broadest coverage for general severity; if bearing failures are your history, buy acceleration.

Buying vibration sensors and matching what you have

Ruilianxin supplies Wilcoxon accelerometers, 4–20 mA vibration transmitters and accessories to overseas industrial buyers, sourced through original channels. The Wilcoxon brand page lists the representative models most often requested — including the 786A general-purpose accelerometer and the PC420VR velocity transmitter series — and the product centre covers the wider range.

To get the right sensor first time, send four things: the machine speed and the faults you are trying to catch; the frequency range you need to cover; the mounting method available; and the output your system accepts (IEPE/dynamic voltage, or 4–20 mA).

More selection and cross-reference guidance for imported industrial components is in the Knowledge Base.

FAQ

Q1. Should I measure vibration in velocity or acceleration?

It depends on the fault. Use velocity for general condition monitoring of motors, pumps and fans, and to compare readings against ISO severity standards — it covers the mid band where running-speed faults such as unbalance and misalignment show up clearly. Use acceleration for high-frequency faults, especially rolling element bearings and gear mesh, and when you plan to apply enveloping or shock-pulse techniques. If you can only instrument a machine once, velocity is the safer default; if your failure history is bearing-related, buy acceleration.

Q2. Does the mounting method really change the measurement?

Yes, and it is often the largest single source of error. The mounted sensor's usable frequency range is set by the stiffness of the mounting, so a magnetic mount can cap you at roughly 5–10 kHz even when the sensor itself is rated to 14 kHz. For permanent installations, stud mounting is the reference method; where drilling is impossible, a correctly applied cementing pad approaches it. Magnets belong on route-based rounds, and probe tips should be avoided unless there is no alternative.

Q3. My accelerometer is rated to 14 kHz. Why does my data stop at 6 kHz?

Almost certainly the mount rather than the sensor. A flat magnet mount has a mounted resonance of about 5–10 kHz and a 2-pole magnet about 3–7 kHz, so the mounting truncates the range well below the sensor's own figure. Poor surface preparation or insufficient torque on a stud mount does the same. Switch to stud or adhesive mounting, clean and flatten the surface, apply the correct torque, and add a coupling fluid — then re-measure before concluding anything about the machine.

Q4. What does a 4–20 mA vibration sensor give me that an accelerometer does not?

Integration and reach, rather than more information. A 4–20 mA transmitter delivers one overall vibration number proportional to the loop current, so it can be wired into a plant DCS or PLC alongside pressure, temperature and flow, and trended without specialist analysis software — which makes continuous protection affordable across many machines. What it does not give you is a spectrum: it tells you that something changed, not what. For diagnosis you still need a dynamic output and an analyser.

Q5. How does the sensor's resonance frequency affect my choice?

Treat the resonance frequency as a boundary rather than a feature. The mounted resonance should be at least twice your highest frequency of interest, so the resonance peak stays outside your measurement band. A sensor with a 30 kHz resonance comfortably covers a 14 kHz window. If you need 10 kHz and your sensor resonates at 12 kHz, the peak falls inside your range and produces a false indication.