Engine Sound Analyzer.
Listen live. Inspect the frequencies. See what changed.
Listen closer.
Your microphone or an audio file. Everything is processed here, in your browser.
Files up to 60 seconds / 20 MB. Stereo analysis uses channel 1; playback keeps both channels.
Ready. Your microphone is off.
Optional recording
Recording is optional and stays in this browser. Download a clip if you want to keep it.
Playback does not move the charts: imported and recorded clips show a whole-clip analysis.
Measurement settings & microphone details
Hann window · no spectral smoothing · DC removed. Larger windows separate nearby tones better but cover more time. Change settings before starting a new measurement.
Audio is processed locally in this browser. Microphone hardware or browser processing can affect measurements.
Optional engine references RPM, cylinder count & harmonics
Live analysis works with every field blank. Manual RPM is a reference, not synchronized order tracking.
Engine & harmonic references
Harmonics use your Hz reference, otherwise the firing reference, shaft reference, or strongest peak—in that order. The strongest peak is not necessarily the fundamental.
No RPM reference. General frequency analysis remains available.
Firing reference assumes a conventional evenly firing engine. Uneven firing, cylinder deactivation and unusual operating modes can differ. References outside the plot range are listed but not drawn.
The sound, made visible.
Start a measurement to see numerical results.
Waveform
Digital amplitude / timeFrequency spectrum
Tonal dBFS / frequency (Hz, log scale)Spectrogram
Frequency (Hz, log scale) / audio-clock timeSample-clock frames when supported; sampled fallback is labeled. Skipped frames interrupt history. Plot range: 5 Hz to Nyquist. File view includes every analysis frame. Dark areas are quieter; bright bands show stronger frequencies.
Measurement quality
Start analysis to collect quality observations.
Quality describes the signal and capture, not machine health. A live snapshot is one FFT window; stability describes up to three preceding seconds.
Measurement notes, repeatability & export
Repeat the same condition
Keep position, distance, RPM, load and gain consistent. Wait for stable observations, then collect 2–6 separate snapshots. Avoid overlapping FFT windows. A high spectral-shape match does not establish machine health.
INSUFFICIENT DATA · Capture at least two separate measurements.
Pair-by-pair measurement differences
How comparison quality works
Shape overlap runs from 0 (disjoint retained power) to 1 (identical normalized spectral shape). Gain is removed from this metric; RMS differences remain separate. Only matching frequency grids and window support are compared. A heuristic good match needs overlap ≥ 0.95, RMS difference ≤ 1.5 dB, band shares within 5 percentage points, stable quality observations and compatible context. Unknown conditions remain unverified. These are engineering screening thresholds, not validated diagnostic limits.
Physical-device validation log
Manual engineering log; nothing here claims a device has already been tested. Perform microphone tests from a safe stationary position. Do not enter serial numbers, names or other personal information.
Browser capture evidence appears after microphone start. No device identifiers are collected.
Reported base latency concerns output buffering, not microphone-to-display latency. See the known-signal validation workbench for software-only checks.
Exports current measurement, A/B, repeats and your validation notes. No raw audio. Data remains local; reload clears this tab’s measurements.
Microphone off. No recording is needed for live measurement.
Harmonic structure
Start analysis to inspect harmonics.
Relative levels use the strongest matched harmonic as 0 dB. Matching a reference does not identify the source of a sound.
| Harmonic / target | Found | Level / relative |
|---|
Persistent components
Repeated spectral peaks, observed over time. Movement is the change since first seen; persistence is not diagnostic confidence.
- Waiting for live measurements.
Significant peaks
Up to six separated peaks above a conservative relative threshold. These are spectral features, not a diagnosis.
- No measurement yet.
Frequency-band energy
Share of total AC spectral power. Bands below 20 Hz or above 20 kHz are excluded, so values may total less than 100%.
- No measurement yet.
Before. After. Side by side.
Capture a measurement while live, without recording or stopping. Use the same microphone, position, gain, FFT settings and operating conditions.
No snapshots saved. Snapshots stay in this tab and are lost on reload.
Comparison quality
| Context | A | B / change |
|---|
Pink: A · Blue: B · tonal dBFS / Hz (log scale)
| Measurement | A | B |
|---|
A reference is a starting point.
Start with live sound
Start the microphone and inspect the waveform, spectrum and spectrogram immediately. Capture A and B without recording or stopping. Optional recording and audio files use the same DSP core as the General Sound Analyzer.
RPM references are calculated, not measured
Shaft frequency is RPM ÷ 60: 1800 RPM gives 30 Hz. Dashed markers show 0.5×, 1×, 2×, 3× and 4× that reference. Manual RPM is not synchronized order tracking, and matching a marker does not establish which part made a sound.
Firing frequency depends on the engine
For a conventional evenly firing four-stroke engine, the reference is RPM ÷ 60 × cylinders ÷ 2. For a two-stroke, it is RPM ÷ 60 × cylinders. Uneven firing, cylinder deactivation and unusual layouts or operating modes can differ.
Harmonics show structure
A supplied reference defines 1×–5× harmonic targets. Nearby spectral peaks are matched within one FFT bin. Very low references may be unresolved by the current window. A strong higher harmonic or missing fundamental does not automatically identify a fault. Relative levels are referenced to the strongest matched harmonic; matched-band power is not a distortion or health score.
Make comparisons repeatable
Keep the same phone position and distance, similar RPM and load, and the same analysis settings. Avoid wind and avoid touching or blocking the microphone. Phone automatic gain and noise processing can change levels. A/B warns when recorded conditions differ; unrecorded conditions still need your attention.
Digital levels, honest limits
Levels are digital dBFS, not calibrated acoustic dB SPL. Phone microphones are not laboratory microphones. This tool cannot guarantee fault detection, establish engine health, or replace professional vibration analysis. Brief impacts or rapid changes may require more specialized acquisition and analysis.
Keep clear of the machinery
Keep phones, hands, cables, clothing and other objects clear of belts, fans, pulleys, moving components and hot exhaust parts. Measure from a safe, stationary position. Do not reach into machinery to improve a reading.
Your audio stays in this browser
No microphone or audio data is uploaded. This page has no analytics or cloud analysis. Snapshots remain in memory only. Stop, hide the tab or leave the page to release the microphone. Save an optional recording with Download clip if you want to keep it.