1Source and timestamps
Drop an audio file here or press Load demo to try a synthetic song with a known boom at B1 (61.7 Hz) and a 2.8 kHz harshness.
Waveform: click to seek · drag to pan · Ctrl + wheel to zoom · drag the t1/t2 shading to move a window · double-click it to loop (again to stop). − / + nudge a window by the step; a playing loop follows it. Numbered lines on the spectrogram are the ranked suspects.
2Differential spectrum
Spectrum at t1 and t2 dBFS
Differential Δ = dB(t2) − dB(t1) dB
Click any band in either chart to play it in the dock’s current mode (Shift+click plays a tone). Right-click a band (or press F on a focused chart) to find it in the song: its energy is drawn over the waveform and its loudest moments are listed. Numbered markers are the ranked suspects.
Band table (every value in the charts)
3DSP EQ Action Table
| # | Centre freq | Q (est.) | Width | Surge | Prominence | Score | PEQ start | Way | Audition at t2 | |
|---|---|---|---|---|---|---|---|---|---|---|
| No analysis yet. | ||||||||||
Solo plays only that narrow band of the audio around t2 (sweep the dock's frequency to pin the boom). Tone plays a pure sine/warble at the centre frequency. EQ A/B loops t2 with the suggested cut so you can hear whether it fixes the problem. Q is measured at half the spike's prominence (the peaking-EQ bandwidth definition); “≥” means the spike is as narrow as the window can resolve. Cuts are starting points; confirm them by ear.
Combined EQ response dB
| On | Way | Type | Freq (Hz) | Q | Gain (dB) | Note |
|---|
Recurring frequencies evidence
| Centre | ISO band | Hits | Songs | Sources | Avg Q | Avg score | Confidence |
|---|
All log entries
| Centre (bias-corrected) | Type | Marks | Spread |
|---|
Q suggester · harsh ranges
| Harsh range | Peak | Suggested PEQ | Readings |
|---|
Estimated response and correction dB
Workflow
- Connect this computer to the DSP (USB, optical, AUX or Bluetooth input) and set a normal listening level. Keep the Output slider at −6 dB or lower.
- Play the song from the player. When it booms or bites, press 2: t2 lands on that moment (your reaction time is compensated). At a passage that sounds clean, press 1.
- Pick a t1 with similar instruments and loudness. If t2 is simply louder overall, switch Δ mode to Level-matched.
- Read the Action Table. Press Solo on each suspect: the one that reproduces the boom on its own is your culprit. Sweep the dock's frequency slider while it plays to pin the loudest point.
- Press EQ A/B and toggle A/B: if the cut removes the problem without hollowing the sound, add it to the EQ Builder.
- Use Find to see that band's energy across the whole song and jump to its loudest moments. They should boom too.
- Repeat with 2–3 other songs and log the suspects. Frequencies that recur in the Resonance Log belong to your system or room, not the music.
- Enter the EQ Builder bands in your DSP software on the way shown (Sub / Midbass / Midrange / Tweeter), on both left and right channels unless the problem is clearly one-sided.
No-microphone tuning tips
- Cut, don't boost. Resonances are peaks; narrow boosts to fill holes rarely work in a car or room and cost amplifier headroom.
- Low-frequency booms (40–160 Hz) are usually room or cabin modes. They need moderate Q (2–6) and −3 to −8 dB. Longer windows (1–2 s) resolve them better.
- Harshness (2–6 kHz) is often a tweeter reflection (dash, desk, side wall) or breakup. It shows up in some vocals and cymbals more than others, so check several songs.
- Find freq (dock, EQ A/B mode) automates boost-and-sweep: a +9 dB, Q 6 boost steps ±½ octave, one loop pass per step, up then down. Press Space where it sounds worst; the up and down marks are averaged, which cancels your reaction delay.
- Find Q steps the cut from narrow (Q 8.6) to wide (Q 1.4). Press Space at the first one that sounds fixed: the narrowest cut that works does the least collateral damage. Then trim the depth until the problem just returns and add 1–2 dB back.
- Reflections (harsh "sh"/"s", glassy highs): use L only / R only in the header to hear which side causes it, and EQ that side's tweeter (and mid) only. The mix beside them (Mono, Left channel, Right channel) plays one signal on both sides, so headphones put it in both ears, the way you hear one car speaker. Above ~2 kHz a few centimetres of head movement shift reflection peaks, so prefer Q 2–4 over very narrow cuts.
- Harsh range → Q: in Sweep Hunter press 1 where the harshness starts, 2 at its worst and 3 where it stops, on the up and the down pass. The suggested Q makes the cut span exactly that range. Read the hearing note too: a constant-level sweep sounds loudest where your ear is most sensitive (about 2.5–4 kHz), so a "peak" there may partly be your hearing. Tick Equal loudness to take that out: the sweep is turned down where your ear is more sensitive, never boosted.
- Two devices: under Outputs in the header, pick a second device B (e.g. headphones while A is the car) to hear both at once. Line them up with the delay on a track with a sharp beat, then switch A ⇄ B with S to compare the same moment on each.
- Rattles marked in Sweep Hunter are mechanical: fix them with damping and foam, not EQ.
- Ear-Match works best at a fixed, moderate volume, eyes closed, with short breaks. Repeat it and average the two runs.
- Re-check the final EQ at low and high volume: the ear's bass sensitivity changes with level (ISO 226).
Keyboard
Space play/pause (marks while Find freq / Find Q or Sweep Hunter runs) · 1/2 set t1/t2 at playhead (during a sweep 1/2/3 mark a harsh range's start / peak / end) · ↑/↓ in a t1/t2 field nudge it · A A/B loops · S switch output device A ⇄ B (the key can be changed under Outputs) · Enter analyse · F on a focused chart: find that band in the song · Esc stop all · +/−/0 zoom · ←/→ nudge playhead
What the numbers mean
- dBFS: a full-scale sine in a band reads 0 dBFS (AES17). Band levels integrate the Welch PSD over the band edges.
- Surge: band level at t2 minus t1. Prominence: t2 band level minus the mean of the ±4 neighbouring 1/12-octave bands.
- Score = Surge + Prominence × gate. The gate rises from 0 to 1 as the surge reaches +6 dB, so steady content that was already loud at t1 can't outrank what changed. Switch to the plain sum in Settings.