What Makes a Vibration Reading Trustworthy

No instrument measures millimetres per second directly. An accelerometer converts motion into a small voltage, that voltage rides down a cable, through a signal conditioner, into a DAQ, and only then does software divide by one stored number — the sensor's sensitivity, in millivolts per g — to turn electricity back into engineering units.

That one number is the entire bridge between physics and your report. If the sensitivity stored in the software does not match the sensitivity the sensor actually has today, every value downstream — RMS, spectrum peaks, alarm comparisons — is scaled by the same silent error, with nothing on screen to warn you.

Because the conversion is a straight division, a sensitivity error does not add noise or look strange — it rescales everything cleanly. A sensor believed to be 100 mV/g that actually outputs 90 mV/g makes every reading come out 10 percent low — the true level is about 11 percent higher than the number on screen, optimistic in exactly the wrong direction; and the spectrum shape stays perfectly plausible.

The figure below shows one identical electrical signal converted with three assumed sensitivities. Same machine, same instant, three different truths — and only one of them crosses the alarm line. This is why a sensitivity check is not paperwork; it decides whether a machine gets stopped.

Sensitivity is not carved in stone. Sensing elements age and drift slowly with time and temperature cycling, and a single hard knock — a dropped sensor, an over-range shock, a magnet mount slammed onto a machine — can shift it in one instant. Connectors loosen, and cable damage adds its own errors: flexing cheap cable under vibration generates triboelectric noise that rides on top of the real signal, which is why low-noise and armored constructions exist.

None of this announces itself. A sensor that drifted out of tolerance keeps producing clean, confident-looking waveforms. The illustrative graph below shows the pattern: slow drift you can live with inside a tolerance band, then one drop event that silently pushes the sensor outside it between checks.

Even a perfectly calibrated sensor is only honest over part of the frequency axis. Every mounting method — stud, adhesive pad, magnet — forms a spring-mass system with its own resonance, and readings near that resonance are amplified, not measured. A rigid stud mount keeps the flat, trustworthy region widest; a magnetic mount is faster to place but brings the resonance down into frequencies you may actually care about.

The practical rule: know your mounting's usable band, and treat peaks that sit near the mounting resonance with suspicion. This is also why a calibration result travels with a stated frequency — a sensitivity verified at a reference frequency describes the flat region, not the resonance.

The fix is a comparison, not faith. In back-to-back calibration, a reference accelerometer of known sensitivity and the sensor under test are mounted on the same vibrating surface, so both experience identical motion. The ratio of their outputs directly yields the test sensor's sensitivity — no anechoic chamber, no absolute measurement of motion needed. This is exactly what the TIERA T-Calibro does on the bench, exciting both sensors at 159.2 Hz — that is 1000 rad/s (1000 ÷ 2π = 159.155 Hz), a conventional reference frequency for comparison calibration under the ISO 16063 method family — and logging each run — date, operator, conditions, before and after sensitivity — into a certificate its software generates.

But a comparison only proves as much as the reference is worth, which is where traceability comes in: an unbroken chain of comparisons linking your reading back to a national measurement standard. T-Calibro's reference sensor ships with a factory calibration certificate establishing its sensitivity against a national-standard traceable chain. Whoever calibrates your sensors, ask for the same evidence: a certificate for their reference, the traceability chain behind it, and the measurement uncertainty — an accreditation logo on a brochure is not the same as a traceable certificate for the actual reference used on your sensor.

A common baseline is a yearly cycle — TIERA's own PhonoVibe DAQs ship with a factory calibration certificate carrying one year of validity — plus an event-driven check any time a sensor is dropped, over-ranged, runs hot, or simply starts disagreeing with its neighbours. With a bench calibrator in-house, a five-minute sensitivity check before a critical survey costs almost nothing; the drift chart earlier shows why the between-checks gap is where trouble hides.

A representative scenario, not a specific customer: a reliability team trends a fan bearing monthly at around 3.9 mm/s, just under their 4.5 mm/s alert. The route sensor is dropped on a walkway; nobody logs it. The next three surveys read about 3.4 mm/s — the team sees improvement while the bearing is actually degrading. A routine back-to-back check finally shows the sensor reading about 30 percent low; corrected, the trend was above the alert line for two months. Nothing in the data looked wrong — only the comparison against a reference exposed it.

If this explainer raised more questions than it answered, that is the right instinct. TIERA publishes free primers at 101.tieraonline.in — Accelerometer & DAQ 101 covers how IEPE sensors, sensitivity, and acquisition actually work, and Measurement Setup 101 covers mounting, cabling, and configuration. TIERA 101 is a free primer series, not an accredited certification.

For teams that need formal, examined competence, TIERA's TCAT programme (see /services) builds from these fundamentals to full measurement practice, with proctored examinations conducted at exams.tieraonline.in. And if you want the comparison workflow on your own bench, the T-Calibro product page describes the back-to-back system and its record-keeping software in detail.

The comparison this post describes is a bench instrument, not a metrology lab. The T-Calibro mounts a reference accelerometer and your sensor on the same vibrating surface at 159.2 Hz (1000 rad/s) — a conventional reference frequency for back-to-back comparison under the ISO 16063 method family — and its software logs every run (date, operator, conditions, before and after sensitivity) and exports the certificate. Its reference sensor ships with a factory calibration certificate establishing its sensitivity against a national-standard traceable chain: the same evidence this post just told you to demand from anyone who calibrates your sensors.

The rest of the chain matters too. Every PhonoVibe DAQ ships with its own factory calibration certificate carrying one year of validity, and TIERA stocks the cabling and mounting hardware that stops a good sensitivity being spoiled on the way to the DAQ — low-noise coaxial cable against triboelectric noise, armored cable for runs near moving machinery, and stud, triaxial, and magnetic mounts.