Tools · Signal analysis
FFT setup planner
Before you hit record: what resolution you'll get, how long each average takes, and whether you'll be able to separate the frequencies you care about.
Separation check (optional)
The relationships
Δf = Fmax / lines Tblock = 1 / Δf = lines / Fmax
fs = 2.56 · Fmax N = 2.56 · lines
Ttotal = Tblock · (1 + (navg − 1)(1 − overlap))
The 2.56 factor gives the anti-aliasing filter room to roll off above Fmax, and it's why analysers sample at 2.56× rather than the Nyquist minimum of 2×.
Resolution is set purely by how long you record. To separate two close peaks you need the bin spacing, widened by the window's effective noise bandwidth, to be several times smaller than their difference. A common rule of thumb is a gap of at least 4 bins with a Hann window. Zoom FFT or a lower Fmax with more lines both achieve this.
Method & sources
What it calculates
- Δf = Fmax / lines; block time T = lines / Fmax; sample rate fs = 2.56·Fmax; block size N = 2.56·lines.
- Total time = T·(1 + (navg − 1)(1 − overlap)).
- Effective bandwidth = ENBW × Δf, with ENBW 1.5 (Hann), 3.77 (flat top) and 1.0 (rectangular) bins.
Assumptions & limits
- The 4-bin separation rule is a practical rule of thumb for Hann, not a hard limit.
- Analysers with non-2.56 sampling ratios or zoom FFT behave differently.
Verified: 1000 Hz Fmax, 800 lines, 4 averages at 50% overlap gives Δf = 1.25 Hz and 2.0 s total, checked by hand.
Sources
- Harris, F. J. (1978). On the use of windows for harmonic analysis with the discrete Fourier transform. Proceedings of the IEEE 66(1), 51-83 ↗
- Gade, S. and Herlufsen, H. (1987). Use of weighting functions in DFT/FFT analysis, parts I and II. Brüel & Kjær Technical Review 3 and 4 ↗
- Randall, R. B. (2011). Vibration-based condition monitoring: industrial, aerospace and automotive applications. Wiley