The Machine Made 900 Hz. The FFT Found 300 Hz.

A vibration signal entered an ADC at 900 Hz, but the sampled spectrum showed a convincing 300 Hz peak. The machine did not create both.

1. Start With the Sample Rate

The ADC takes 1,200 samples each second. Always record the sampling rate before interpreting any frequency-domain result.

2. Mark the Nyquist Limit

Nyquist frequency is half the sample rate. At 1,200 samples per second, the highest representable frequency is 600 Hz.

3. Compare the Input

The 900 Hz component sits above the 600 Hz Nyquist limit. Those samples cannot uniquely preserve its original frequency.

4. Calculate the Fold

Here, the alias is |900 − 1,200| = 300 Hz. The sampled points fit a slower tone even though the analog input is faster.

5. Reject the False Diagnosis

Do not label the 300 Hz peak a machine subharmonic yet. First prove that the acquisition chain could capture 900 Hz correctly.

6. Raise the Sampling Rate

The theoretical minimum is above 1,800 samples per second. Use practical margin so filtering and real hardware do not sit at the limit.

7. Add an Anti-Alias Filter

Band-limit the analog signal before the ADC. A higher sample rate cannot protect the measurement from unwanted out-of-band energy by itself.

Trust the Spectrum Only After Sampling Passes

Check the rate. Mark Nyquist. Calculate aliases. Filter first. Then decide whether a spectral peak belongs to the machine or the measurement.