Methods stated. Limits made visible.
Online Spectrum Analyzer separates standards-aligned measurements from automated engineering screening. This page documents what is calculated, how the result should be interpreted, and when another instrument is required.
How to read these results
FFT, sample peak, RMS, crest factor, and complete-file loudness are calculated from decoded audio samples.
Noise and mains-hum findings identify passages to inspect and confirm by listening.
Targets and reports support engineering review; they do not certify delivery compliance.
Spectrum and spectrogram
The spectrum is calculated from short blocks of decoded digital audio using a Hann window and a Fast Fourier Transform. Available FFT sizes range from 1,024 to 8,192 samples. A larger transform separates nearby steady frequencies more clearly, while a smaller transform places short events more precisely in time. The spectrogram repeats that calculation across the recording to display frequency vertically and time horizontally.
Displayed frequency resolution is constrained by the source sample rate, FFT size, window function, channel selection, and available signal energy. A narrow visual peak is evidence of energy near a frequency bin; it is not a calibrated acoustic sound-pressure measurement.
Digital level measurements
Peak and RMS readings are calculated from decoded sample values and expressed relative to digital full scale. A sample is treated as reaching the clipping threshold at an absolute value of 0.999 or above. Crest factor is the difference between sample peak and RMS level. These measurements describe the digital signal and do not provide dB SPL without a calibrated microphone and measurement chain.
Loudness and Loudness Range
Complete-file Pro diagnostics apply K-weighting and gating aligned with ITU-R BS.1770-5 for Integrated LUFS, Momentary loudness, and Short-term loudness. Loudness Range follows the distribution-based approach described by EBU Tech 3342. At 48 kHz the engine uses the specified K-weighting coefficients; other sample rates use a sample-rate-adapted implementation and are identified accordingly in the report.
Reference profiles compare the measured file with a selected target, but they are guidance rather than delivery certification. A platform may use different programme boundaries, album logic, codecs, or current acceptance rules.
True Peak
For 48 kHz sources, True Peak uses 4× reconstruction with a 48-tap FIR described in Annex 2 of ITU-R BS.1770-5. Other sample rates use a four-times reconstructed engineering estimate and are labelled as non-compliance estimates. True Peak can exceed sample peak because the reconstructed waveform between stored samples may rise higher than any individual sample.
Noise floor and mains-hum screening
The noise-floor result is a quiet-window RMS estimate supported by a spectral-flatness check. Mains-hum screening compares energy around 50 or 60 Hz and related harmonics with nearby frequencies over time. These detectors are intended to locate passages worth listening to; programme bass, sustained notes, room tone, filtering, or very short events can create false positives or hide a real fault.
Timestamped findings
Clipping events, elevated broadband-noise candidates, material DC offset, and persistent 50/60 Hz hum candidates are attached to positions in the decoded file timeline. Selecting a timestamp moves playback to that passage so the finding can be confirmed by listening and visual inspection. Findings are deliberately described as automated screening, not a diagnosis guaranteed for every source.
Validation scope
The included 48 kHz synthetic reference checks use tolerances of ±0.1 LU for loudness and ±0.1 dB for True Peak. Those test tolerances verify specific deterministic fixtures; they are not a universal uncertainty specification for arbitrary recordings, browsers, codecs, microphones, or playback systems.
When to use a calibrated tool
Use a certified meter or calibrated measurement chain when a broadcaster, regulator, laboratory procedure, workplace standard, or contractual delivery specification requires traceable compliance. Browser results are appropriate for private inspection, troubleshooting, comparison, documentation, and preliminary quality control when their stated scope is sufficient.