A vibration measurement is a record of motion through a particular sensor, mounting arrangement, signal-conditioning chain, and sampling plan. Preserve those choices with the data so a later comparison can distinguish a machine change from a measurement change.
This guide helps test and reliability technicians plan accelerometer data acquisition. Begin with the question, expected operating state, frequency range of interest, sensor calibration, and acquisition-hardware manual. Installing sensors and making changes around machinery require the site's approved access and isolation procedures.
Match the sensor to the input
Many vibration accelerometers use piezoelectric sensing. An IEPE accelerometer includes electronics powered by a suitable constant-current source; a charge-output sensor needs a compatible charge amplifier or converter. NI's accelerometer measurement guide describes these arrangements and their conditioning needs.
Verify excitation current, compliance voltage, sensor bias, input range, coupling, and the calibrated sensitivity. Confirm that the complete sensor and input chain covers the required frequency and amplitude range. A connector that fits establishes only the mechanical connection.
On a narrow screen, scroll the table horizontally. Keyboard users can focus it and use the arrow keys.
| Part of the chain | Record | Comparison risk if it changes |
|---|---|---|
| Sensor | Model, serial number, calibration date, sensitivity, units, range, and axis. | A scaling or orientation change can look like a changed vibration level. |
| Mounting | Location photograph, axis direction, attachment method, surface preparation, and cable restraint. | The mechanical path and useful frequency response can change. |
| Conditioning | Excitation, coupling, gain, input range, filters, and settling behavior. | Clipping, attenuation, offset, or startup transients can enter the record. |
| Acquisition | Actual sample rate per channel, channel timing, record length, and trigger. | Time alignment and spectral interpretation can change. |
| Machine state | Speed, load, process condition, temperature, and relevant recent work. | Different operating conditions can dominate the comparison. |
Mounting stiffness affects the upper usable frequency range, as explained in PCB Piezotronics' mounting guidance. Use the sensor manufacturer's instructions and reproduce the location and orientation for trend measurements. For changing rotational speed, preserve a suitable speed reference if the analysis needs to relate vibration to shaft rotation.
Choose bandwidth, sampling, and record duration together
Sampling turns a continuous signal into a sequence. Frequencies outside the adequately filtered input band can fold into the recorded band as aliases. NI's anti-aliasing explanation describes why filtering and sampling must be designed together.
Use the acquisition device's specified passband, filter transition, and alias rejection at the chosen sample rate. Leave room for the filter transition above the highest frequency you need. A software filter applied after an aliased signal has been sampled cannot reconstruct which frequencies were originally present.
The frequency-analysis chapter of NI's archived LabVIEW Analysis Concepts manual gives the relationship Δf = sample rate ÷ sample count. The mathematics remains useful independently of that software version. Zero-padding adds displayed frequency points; additional actual observation time is what changes the underlying unpadded bin spacing.
Make the units and amplitude definition explicit
For a fictional calibrated sensitivity of 100 mV/g, a 250 mV peak signal corresponds to 2.5 g peak, assuming linear response, correct conditioning, and no clipping. For a pure sine wave only, that corresponds to about 1.77 g RMS. A mixed or impulsive waveform needs its RMS calculated from the actual data.
Keep peak, peak-to-peak, RMS, spectral amplitude, and power spectral density distinct in filenames and plots. If acceleration is converted to velocity or displacement, document the integration, filters, and frequency limits. Low-frequency offsets can become much larger errors after integration.
Look at the time record before trusting the spectrum. Flat-topped peaks suggest clipping; a long initial transient may reflect settling. NI documents one device-specific example of IEPE acquisition startup saturation. Use the actual hardware's behavior to define how much valid settled data the test needs.
Make the first acquisition a validation run
- Verify channel identities, calibrated scaling, and expected sensor bias or health indications.
- Acquire a short record at the documented machine state and inspect range, clipping, gaps, and settling.
- Confirm actual channel timing if phase or cross-channel comparisons matter.
- Save original samples, settings, units, and a mounting photograph before deriving plots.
- Repeat under equivalent conditions. Investigate differences in setup before assigning a machine fault.
- Record the analysis version, window, frequency band, averaging, and amplitude convention with each derived result.
The useful output is a comparable record with a defensible frequency and amplitude range. Interpretation then draws on machine design, speed, operating history, and appropriate diagnostic expertise. The predictive asset monitoring guide follows that next step from evidence to a reviewed maintenance response.
Keep a repeatable acquisition plan
Download the vibration acquisition record. It includes sensor and mounting fields, the sampling calculation, amplitude conventions, validation observations, and a place to identify the original data file. The examples are mathematical exercises, with no machine fault limits implied.