Fourier Transform Tool (FFT)
A tool for viewing the spectrum of experimental data. Paste your data below (or choose a file), then press “Transform”.
Data input
Settings
Results
Results appear here after transforming
Peaks
No peaks found (lower the threshold)
Numeric table
Inverse transform
Definitions
Symbols: N = number of data points, M = number of transform points (after zero padding), Δt = sampling interval, w[n] = window, y[n] = preprocessed data. Phase is referenced to the first sample (n = 0) at time 0.
Preprocessing (in this order): mean removal x'[n] = x[n] − mean(x) → window x''[n] = w[n]·x'[n] (only the N real data points) → zero padding (append M − N zeros). S = Σ w[n].
Forward: X[k] = Σ_{n=0}^{M−1} y[n]·exp(−i·2π·k·n / M) (no prefactor; same as fft in NumPy and MATLAB)
Inverse: x[n] = (1/M) Σ_{k=0}^{M−1} X[k]·exp(+i·2π·k·n / M)
Frequency: Δf = 1/(M·Δt), f[k] = k·Δf, ω[k] = 2π·f[k]. One-sided: k = 0…floor(M/2); two-sided: fftshift order.
Normalization on: X~[k] = g·X[k]/S (g = 2 for one-sided bins other than DC and Nyquist, g = 1 otherwise). Off: X~[k] = X[k]. Amplitude = |X~|, phase = atan2(Im, Re), power = amplitude².
dB: 20·log10(amplitude / ref). ref = 1, or ref = maximum amplitude when “Set the maximum to 0 dB” is on.
Peaks: using one-sided, normalized amplitudes, local maxima at or above the given % of the maximum, excluding nearby maxima (within the window’s main-lobe width), up to 10 in descending order. Frequencies are refined by log-parabolic interpolation.
Windows (n = 0…N−1, θ = 2πn/N; cosine-sum windows are periodic, triangular and Gaussian are symmetric):