Understanding Aliasing in Vibration Analysis
Aliasing is a signal-processing error that can corrupt the digital analysis of vibration data. It happens when a signal is sampled at a rate too low to capture its highest-frequency components, so those high frequencies “fold down” and impersonate lower frequencies in the resulting FFT spectrum. The outcome is false peaks that never existed in the real machine — peaks that can lead to a serious misdiagnosis. Understanding aliasing, and the safeguard that prevents it, is fundamental to trusting any digital vibration spectrum.
1. Definition: What is Aliasing?
When an analyser digitises a vibration signal it does not record a continuous curve; it records a sequence of discrete samples — snapshots taken at a fixed time interval. If those snapshots are spaced too far apart relative to how fast the signal is changing, the analyser literally cannot tell a fast wave from a slow one. The few points it captures of a high-frequency component can be joined into a perfectly plausible low-frequency sine wave. That phantom low frequency is the alias, and once it appears in the spectrum it is indistinguishable from a genuine vibration at that frequency.
2. The Nyquist Theorem and Sampling Rate
To understand aliasing you must first understand the Nyquist theorem (the Nyquist–Shannon sampling theorem). This foundational principle of digital signal processing states:
To accurately represent an analogue signal in digital form, the sampling frequency (Fs) must be at least twice the highest frequency component (Fmax) present in the signal.
This minimum sampling rate (2 × Fmax) is called the Nyquist rate. Turned around, the highest frequency that a given sampling rate can faithfully measure is half of it: Fmax = Fs / 2. That ceiling is the Nyquist frequency. Any real frequency above the Nyquist frequency cannot be represented honestly and will instead be reflected back below it. In practice the chosen Fmax also sets the resolution of the analysis together with the number of FFT lines — a relationship you can explore with an FFT Resolution Calculator when planning a measurement.
3. How Does Aliasing Occur?
Imagine a high-frequency vibration being measured by a digital analyser taking discrete samples at a fixed rate:
- If the sampling rate is high enough — well above the Nyquist rate — the analyser captures enough points per cycle to reconstruct the waveform accurately.
- If the sampling rate is too low, the analyser misses what happens between samples. The handful of points it does capture connect into a completely different, lower-frequency sine wave. That false low frequency is the alias.
A concrete example: suppose a signal contains a real 900 Hz component but the analyser’s Fmax is set to 500 Hz, which corresponds to a sampling rate of 1000 Hz. The 900 Hz content lies above the 500 Hz Nyquist frequency and cannot be measured correctly. It is aliased and reappears at Fs − 900 = 1000 − 900 = 100 Hz. An analyst scanning the spectrum could easily mistake that 100 Hz peak for a 1× running-speed vibration, or for a real defect, and chase a fault that does not exist. Worse still, the high-frequency culprits — bearing impacts, gear-mesh energy, electrical noise — are often the very signals an analyst most wants to trust.
4. Preventing Aliasing: The Anti-Aliasing Filter
It is impossible to know in advance all the high-frequency content a signal might carry — ultrasonic noise, sharp impacts, radio-frequency interference and electrical pickup can all intrude. Simply hoping the sampling rate is high enough is therefore not a safe strategy.
The solution used in every modern digital vibration analyser is the anti-aliasing filter: a steep low-pass filter placed in the signal path before the analogue-to-digital converter (ADC). It works like this:
- The user sets the desired maximum frequency, Fmax, for the analysis.
- Based on that Fmax, the analyser automatically sets the anti-aliasing filter’s cut-off frequency just above Fmax.
- The analogue sensor signal passes through the filter, which removes or strongly attenuates everything above the cut-off.
- Only the filtered, clean signal reaches the ADC for sampling.
Because the filter strips out the high frequencies the chosen sampling rate cannot handle before sampling takes place, it makes aliasing physically impossible. A real filter cannot cut off infinitely sharply, which is why the cut-off is set a little below the Nyquist frequency to leave a guard band on its skirt. The anti-aliasing filter is one of the most critical elements of any analyser, ensuring the resulting FFT is a true and faithful picture of the machine’s vibration within the selected range. Note that this filtering must be analogue and must precede digitisation — applying digital filtering after the ADC cannot undo an alias, because by then the false frequency is already locked into the data.
5. Practical Implications for the Analyst
For the engineer in the field, the lesson is to respect the instrument’s frequency settings. Choosing Fmax too low to keep good resolution on low-order peaks can hide important high-frequency information; the anti-aliasing filter will protect you from false peaks, but it cannot show you energy you have filtered away. Reliable instruments handle this automatically — a portable analyser such as the Balanset-1A applies anti-aliasing in hardware ahead of its ADC, so the spectra it presents for diagnostics and the 1× amplitude-and-phase it uses for balancing are free of aliased artefacts across its working range. The practical takeaways: set Fmax high enough to cover the highest fault frequency you care about, trust that a properly designed analyser will not alias, and treat any unexplained low-frequency peak with healthy suspicion until you have ruled out other causes.
