Turn tones, chirps, or noise into clear spectrograms. Adjust window size to balance time detail. Reveal hidden frequencies and track changes across time precisely.
Spectrogram Analysis Article
A spectrogram visualizes how frequency content evolves over time. It is built by slicing a signal into short frames, transforming each frame to the frequency domain, and stacking the results. This is widely used in acoustics, vibration testing, radar, and transient diagnostics.
For sample rate fs, window size N, hop H, and NFFT, each frame returns bins up to fs/2. Frequency spacing is Δf = fs/NFFT, and time spacing is Δt = H/fs. These two spacings are the core “data knobs.”
With fs=8000 Hz, N=512 spans 512/8000 = 0.064 s. If H=128, frames arrive every 0.016 s. With NFFT=1024, Δf≈7.8125 Hz. Increasing NFFT improves bin spacing; increasing N improves peak sharpness but blurs fast changes.
Hann and Hamming windows reduce spectral leakage from frame edges and usually give clean, stable plots. Blackman suppresses sidelobes more strongly, which helps isolate weak components near strong tones, but it widens peaks. Rectangular windows are fastest but often leak more energy.
A steady sine appears as a horizontal band at its frequency. A linear chirp forms a diagonal ridge, where slope reflects sweep rate. Two tones show two parallel bands. Broadband noise fills many bins with a mottled texture, and its “flatness” depends on scaling and windowing.
Linear magnitude preserves proportional amplitudes, which is useful for comparing absolute levels. Decibel scaling compresses large values and reveals weak components. A practical working range is 60–90 dB of dynamic range, allowing faint harmonics or resonances to remain visible beside dominant tones.
CSV export records magnitude values with associated time and frequency coordinates. Downsampling reduces size by skipping frames or bins while retaining overall structure. The PDF report captures the plot plus the analysis settings, enabling reproducible comparisons across experiments and consistent documentation.
Confirm that the peak frequency matches known tone inputs and that chirp endpoints align with selected start and end frequencies. If features smear in time, reduce hop size. If peaks look broad, increase window size or NFFT. If plots appear blocky, extend duration for more frames.
A larger window contains more cycles, narrowing spectral peaks. That increases frequency resolution but averages changes over a longer time span, so fast transients may appear smeared.
Hop size controls frame spacing. Smaller hops create more frames and smoother time tracking, but increase computation and file size. Larger hops are faster but may miss short events.
Pick a power of two at least as large as the window. Larger NFFT provides finer bin spacing and nicer plots, but it does not add new information beyond the window’s content.
Use decibels when your signal has a wide dynamic range. It helps reveal weak harmonics, secondary tones, or faint resonances that can be hidden by strong components in linear scale.
That is spectral leakage. If the tone does not fit an integer number of cycles in the window, energy spreads. Hann or Hamming windows reduce leakage compared to rectangular windows.
Paste a list of numeric values separated by commas, spaces, or new lines. The tool reads them in order as samples. Ensure your chosen sample rate matches how the data was recorded.
Increase the time and frequency downsample values. This skips frames and bins while preserving overall structure. For fine measurements, keep downsampling at 1 and export fewer seconds.
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.