QUADVIONVIBRATION · NVH · DAQ · INSIGHTS Request quote
FFT vibration analysis: resolution, window and averaging
FFT analysis guide

FFT vibration analysis: resolution, window and averaging

FFT analysis converts a time record into frequency components. The spectrum is only meaningful when the sampling rate, record duration, frequency resolution, window and amplitude scaling are appropriate for the measurement.

01

Frequency resolution

Frequency resolution is the sample rate divided by FFT length. A smaller frequency spacing requires a longer record. The acquisition must remain sufficiently stable during that record.

  • Sample rate
  • FFT length
  • Record duration
  • Displayed frequency span
02

Window selection

A window reduces leakage when the record does not contain an exact integer number of signal cycles. Hann is a practical general-purpose choice. Rectangular is appropriate only when the signal is periodic within the record or transient analysis requires it.

  • Hann window
  • Hamming window
  • Rectangular window
  • Coherent-gain correction
03

Averaging

Averaging reduces random variation and makes stable spectral components easier to see. The selected method should match the objective and must not hide short-duration events.

  • Linear averaging
  • Exponential averaging
  • Peak hold
  • Overlap
04

Amplitude scaling

The spectrum must state whether amplitudes are RMS, peak or peak-to-peak and whether the display is linear or logarithmic. Engineering units and reference values must remain consistent.

  • RMS spectrum
  • Peak spectrum
  • Acceleration, velocity or displacement
  • dB reference
05

Interpreting peaks

A spectral peak should be related to speed, bearing geometry, gearmesh, electrical frequency or structural resonance before a fault conclusion is made.

  • Running speed and harmonics
  • Sidebands
  • Bearing frequencies
  • Gearmesh frequency
  • Structural modes

Frequently asked questions

Engineering questions answered clearly.

What FFT size should I use?

Choose FFT size from the required frequency resolution and the record duration the machine can hold stable.

Why does a peak spread across several lines?

Spectral leakage, modulation, speed variation and the selected window can spread energy across frequency bins.

Does zero padding improve true resolution?

Zero padding increases the number of displayed points but does not add new information or improve the true resolving ability of the acquired record.

When should peak hold be used?

Peak hold is useful for capturing the maximum amplitude seen at each frequency during a changing or transient measurement.

Project enquiry

Need a vibration, DAQ or machine-monitoring proposal?

Send requirement ↗