A snare hit can have a clean attack and a rough decay. A spectrum averaged across the whole hit may hide that distinction, while a spectrogram can show where the roughness appears in time. Choosing between them starts with the sound you need to explain. Use the display that separates the suspect event from the musical detail you want to keep, then check the decision by listening.
Locate the sound before choosing the display
Find a short passage where the problem is easy to hear. Include a little audio before the event and enough afterward for the decay to finish. For a click before a kick, the lead-in matters; for a gritty cymbal tail, the seconds after the strike matter more. A selection containing an entire chorus gives a broad picture but may bury the specific event under other instruments.
Describe whether the problem is a short interruption, a steady tone or a changing texture. A narrow whistle held through a pause is different from a bright splash that develops after a snare. Listen at a comfortable level and note the exact moment. If you cannot consistently locate the problem with your ears, a bright patch in the display is weak grounds for an audio repair.
Keep an untouched copy of the original audio. Use the same passage for original and repaired comparisons, with matching playback levels and display settings. Automatic scaling can make a quieter repair look as bright as its source because the display stretches its reduced range. A fair visual comparison needs a fixed frequency range, level range and time scale.
A waveform remains useful alongside either spectral view. It shows the timing and overall shape of the attack, making it easier to spot a cut or an accidental fade. It does not separate frequencies. Let each display answer its own question: where the event starts, which frequencies dominate, and how those frequencies change as the sound dies away.
Use a spectrum when the tone is reasonably steady
A spectrum shows frequency content for an analysis window or an average of windows. It is useful when you want to compare a sustained note, identify a stable narrow tone or judge a broad tonal balance. Select a representative portion of the sound rather than combining its attack, sustain and silence by default. The result depends on what you feed into the analysis.
For a ringing note after a hit, inspect a section of the ring without the initial strike. A peak that persists across several nearby selections supports the idea of a steady resonance. A peak present only during the attack may belong to the instrument's ordinary strike. Cutting both equally with a permanent notch could weaken the musical attack to solve a problem that lives later in the decay.
Keep smoothing and averaging consistent between comparisons. Heavy smoothing can hide a narrow peak; long averaging can dilute a short event. Peak-hold displays preserve maximum values reached over time, which can be useful for finding occasional energy but can also make a vanished event appear current. Reset the hold before comparing another passage.
Do not assume every isolated peak is unwanted. Harmonics are part of pitched sounds, and a cymbal can have an uneven spectrum without being damaged. Ask whether the peak aligns with what you hear and whether it persists when nearby instruments stop. A spectrum is good at showing distribution across frequency; it cannot by itself establish which instrument or process caused that distribution.
Use a spectrogram when the decay changes
A spectrogram places successive frequency measurements along time, with colour or brightness representing level. A transient may appear as a short vertical spread, a sustained pitched component as a horizontal trace, and a decay as energy fading across time. These are useful patterns rather than rigid labels. Several sounds occurring together can overlap into a shape that does not identify a single source.
For a cymbal tail that turns grainy halfway through, zoom until the strike and later tail are separate on the time axis. Listen while following the corresponding region. Look for energy that pulses, breaks into changing bands or drops abruptly, but compare it with the audible texture. A tail can look irregular because of beating between partials and still sound entirely natural.
Extend the displayed level range enough to see quiet detail, then keep it fixed. Raising the display's sensitivity can reveal the noise floor and low-level reverberation, not newly created damage. If the repaired tail looks cleaner only because its level fell, listen for lost sustain. A dark patch is not a success when it marks music that should still be audible.
Inspect the channels separately when a problem seems one-sided. A combined display can conceal different left and right behaviour. Then return to stereo listening and a mono check where relevant. A noisy tail on one channel may make the image wander; a repair that changes its timing can affect the way the channels combine even if its individual display looks smoother.
Choose enough resolution for the event
Window length creates a practical tradeoff. A longer analysis window separates nearby frequencies more clearly but spreads information over a longer span of time. A shorter window places rapid events more precisely while separating close frequencies less well. Neither setting is universally more accurate for the question you are asking. A fast click and a low sustained ring need different emphasis.
Try two views of the same event rather than treating one setting as final. Use a shorter window to locate the attack or a sudden interruption, then a longer window to inspect the frequencies in the decay. Keep the selection and level scale unchanged. If the apparent streak moves or broadens with the analysis setting, some of its shape belongs to the measurement rather than the source.
| Sound you are checking | Useful first view | Second check |
|---|---|---|
| A steady whistle during a pause | Spectrum of the pause | Spectrogram to confirm how long it persists |
| A sharp click before a drum hit | Waveform and short-window spectrogram | Listening with the surrounding beat |
| A rough texture developing in a cymbal tail | Spectrogram across the whole decay | Spectrum of early and late selections separately |
| A broad change in vocal brightness | Spectrum of comparable sung passages | Level-matched listening and a time view of consonants |
More screen pixels do not create missing analysis detail. A large display can make an interpolated image look precise, while its underlying time or frequency bins remain unchanged. Zero padding can make a spectral curve look smoother without increasing the ability to resolve close components in a short recording window. Use the settings to support a decision, not to produce a more impressive picture.
Compare the attack and tail after a small repair
Work on a duplicate and treat only the region that needs attention when practical. If the unwanted texture starts after the hit, keep the initial attack outside the strongest treatment. Allow smooth transitions so the repair does not introduce a sudden change in noise or ambience. A hard boundary through a sustained tail can create a new click that was absent from the source.
Listen to the first instant of the sound separately from the decay, then listen to the whole phrase. A softened attack can make a percussion repair sound less harsh at the cost of rhythmic definition. A shortened tail can reduce grain while making the instrument unnaturally dry. Check the nearby bass, vocal or room sound too; in a full mix, the selected region contains more than the obvious drum.
Compare original and repaired versions with the same display configuration and matched listening level. Inspect whether the specific troublesome event changed, not whether the entire spectrogram looks darker. Then hide the display for a listening pass. It is easy to prefer an orderly image after spending several minutes staring at it, even when the music has lost air or timing.
Save a reversible export under a new name and verify it outside the editing session. Check the repaired hit, its following tail and the final ending of the track. Keep the source, session and settings so a narrower or gentler repair remains possible. The useful result is a less distracting sound with its musical role intact; the spectrum and spectrogram provide evidence about that result, while your listening establishes whether it works.
