Turn measurements into frequency insights for better decisions. Detect periodic signals, noise, and harmonics effortlessly. Download tables, charts, and reports in seconds today here.
Enter evenly sampled values. Use commas, spaces, or new lines. For irregular sampling, resample before analysis.
This example combines a 1-cycle signal with a smaller harmonic component.
| Time | Value |
|---|---|
| 0.00 | 0.15000 |
| 0.10 | 0.54143 |
| 0.20 | 0.82970 |
| 0.30 | 1.07241 |
| 0.40 | 0.63414 |
| 0.50 | -0.15000 |
| 0.60 | -0.54143 |
| 0.70 | -0.82970 |
| 0.80 | -1.07241 |
| 0.90 | -0.63414 |
For evenly sampled values x[n], the discrete Fourier transform at bin k is:
The (one-sided) periodogram estimates spectral power as:
If windowing is enabled, samples are multiplied by a window w[n]. With window power correction, power is divided by U = (1/N) Σ w[n]² to keep levels comparable.
Tip: If the dominant frequency is near 0, consider removing a trend or mean, then recalculate.
A periodogram summarizes how signal variance distributes across frequencies. It converts evenly spaced measurements into spectral power, helping you spot repeating cycles, dominant harmonics, and broadband noise. This calculator reports dominant frequency, estimated period, ranked local peaks, and run metadata such as NFFT, resolution, and normalization so analysts can reproduce outputs consistently.
Reliable spectra start with clean inputs. Remove obvious recording errors, align sampling intervals, and keep units consistent. Mean removal suppresses the DC spike, while linear detrending reduces low‑frequency bias from drift. If variance changes sharply over time, analyze shorter segments and compare results. When data are irregularly sampled, resample or aggregate before using a standard periodogram; otherwise frequency estimates may be misleading.
Frequency locations depend on sampling frequency and the selected transform length. With NFFT points, frequency resolution is Δf = Fs/NFFT, so higher NFFT sharpens peak separation. Zero padding increases NFFT without adding information, but it interpolates the spectrum, making peak locations easier to read. The one‑sided spectrum reports bins from 0 to the Nyquist limit, Fs/2, which is the highest interpretable frequency for real‑valued data. Keep expected signal frequencies well below Nyquist to reduce aliasing risk.
Finite samples create leakage, where energy from a tone spreads into nearby bins. Hann, Hamming, and Blackman windows trade narrower peaks for reduced leakage; rectangular keeps maximum resolution but leaks most. Window power correction rescales levels using U = (1/N)Σw² so runs remain comparable across window choices. Choose PSD normalization when you need power per hertz, for example when comparing sensors with different sampling rates, or when integrating over a frequency band.
After calculation, confirm that the strongest peak is not an artifact of trend, DC, or a single outlier. Compare the top peaks’ periods with domain expectations, then rerun using a different window or padding factor to test stability. Peak listing uses local maxima within your chosen frequency range, so adjust min and max frequency to focus interpretation. Export CSV for audits, or PDF for reporting, and record parameter settings with each result. For traceable governance.
It estimates a one‑sided or two‑sided periodogram from evenly spaced samples, then reports dominant frequency, period, peak power, and a ranked list of spectral peaks within your selected range.
Yes. Standard periodograms assume constant sampling intervals. If timestamps are irregular, resample, interpolate, or aggregate onto a uniform grid before analysis to avoid biased frequency estimates.
Hann is a strong default for reducing leakage while keeping good resolution. Hamming can improve sidelobe behavior for some signals, Blackman reduces leakage further, and rectangular offers maximum resolution but leaks most.
For real‑valued signals, negative‑frequency power mirrors positive frequencies. A one‑sided spectrum combines them and doubles non‑DC, non‑Nyquist bins, making charts and totals easier to interpret.
Zero padding increases the frequency grid density, which can make peaks appear smoother and peak locations easier to read. It does not add new information or improve true frequency resolution.
Raw power scales with sampling frequency and bandwidth. PSD divides by sampling frequency, giving power per hertz, which is better for comparing spectra across datasets or integrating power over a frequency band.
Important Note: All the Calculators listed in this site are for educational purpose only and we do not guarentee the accuracy of results. Please do consult with other sources as well.