Tools › Electrical and Controls › Signal Processing › Signal Analysis (FFT and PSD) › Example
Random signal with a resonance at 120 Hz: PSD by Welch's method with its degrees of freedom
| Tool | Signal Analysis (FFT and PSD) |
| Source | Original case (illustrative: Gaussian white noise from a fixed seed passed through a two-pole resonator at 120 Hz with 5 % damping and scaled to unit rms), checked against scipy.signal.welch in test_signal_analysis.php; the degrees of freedom against a numpy white-noise simulation. |
| Level | Graduate |
Problem
A random vibration record of 4,096 samples at 1,000 samples a second has an rms of 1 g. Estimate its power spectral density, say how reliable the estimate is, and read the resonance from it.
Inputs to Enter
- 4,096 values in g, sample rate 1,000 Hz
- Window Hann; show the PSD; Simple mode (mean removed, automatic segment, 50 % overlap)
Answers
- Segments: K = floor((4,096 − 256) / (512 − 256)) = 15 of 512 samples; resolution 1,000 / 512 = 1.953 Hz
- Degrees of freedom ν = 2 K / c_ov = 30 / 1.052 = 28.5; random error √(2 / ν) = 26 %; the true PSD lies between 0.68 and 1.65 times each value with 90 % confidence
- PSD peak 0.0533 g²/Hz at the 119.1 Hz line (resonator at 120 Hz)
- Half-power width 10.4 Hz gives ζ = 10.4 / (2 × 119.3) = 0.044 (5 % was built in; one estimate with 26 % scatter)
- Area under the PSD: rms 0.9992 g against 1.0000 g from the samples
Notes
Open this case in the tool: /analysis-tools/electrical/signals/signal-analysis/?try=shaped-random. The amplitude spectrum of the same record shows a forest of lines near 120 Hz that changes with every record; the averaged PSD is the stable description of a random signal. The PSD is the input of Random Vibration Response.