Reverb Tails and the Noise Floor: Interpret the Last Part Carefully

A reverb tail can continue mathematically after it becomes hidden by background noise. RT60 describes a sixty-decibel relative drop, not absolute silence. The apparent ending depends on the starting level, noise floor, frequency content, playback level and the method used to detect or fit the decay.

This explains why the same response can seem longer after a gain increase, or shorter in a noisy recording, without necessarily changing its underlying relative decay rate. Separate the response’s envelope from the conditions under which you observe it.

Use a stated level reference

For a recorded digital signal, dBFS refers to digital full scale. For an acoustic measurement, sound-pressure levels and microphone calibration involve different references. Do not mix a digital noise-floor number with an acoustic level as if both used the same zero.

The original arithmetic example below stays entirely within a hypothetical digital level scale. It is not a room sound-pressure measurement, hearing assessment or equipment specification. Its purpose is to explain a slope and a fixed observation threshold.

The RT60 formula guide explains relative decay. The reverb calculator provides musical timing starts, not a microphone recording or a noise-floor measurement.

An ideal tail crossing a fixed floor

Assume a response starts at −10 dBFS and follows a straight decay of thirty decibels per second. Its sixty-decibel relative drop takes two seconds. A background floor at −70 dBFS is sixty decibels below the start, so the ideal envelope reaches that floor after two seconds.

Now reduce the response by twenty decibels without reducing the independent background noise. It starts at −30 dBFS, only forty decibels above the same floor. At thirty decibels per second, the crossing occurs after 40/30, approximately 1.333 seconds. Its ideal relative RT60 remains two seconds.

Same ideal slope, different visibility above the floor
Starting levelFloorAvailable dropIdeal crossing time
−10 dBFS−70 dBFS60 dB2.000 s
−30 dBFS−70 dBFS40 dB1.333 s
−40 dBFS−70 dBFS30 dB1.000 s

If you turn up the entire noisy recording afterward, both response and recorded noise rise together. That does not restore the lost signal-to-noise ratio. The comparison above explicitly changes the response relative to independent noise; it is not a claim that ordinary playback gain removes noise.

Why a late decay fit can become unreliable

As the response approaches the noise floor, the measured curve no longer follows only the desired decay. A fitting region that extends too far into noise can produce a misleading slope. A longer fitting range is useful only when the data support it.

T20 and T30 refer to different selected decay ranges, not guaranteed accuracy levels. The actual-versus-predicted guide explains how measurement conditions differ from model assumptions. Inspect the underlying curve and reliability information rather than accepting every reported number.

A room RT60 prediction does not know the recording’s noise floor. Entering geometry and absorption coefficients cannot prove that a particular recorded response contains a clean thirty-decibel fitting span.

Truncation can remove information

Cutting an impulse response early removes late samples. That may be appropriate for a deliberate creative edit, but it changes the response being analyzed. A shorter file is not evidence that the physical room’s decay improved.

Likewise, a fade can make a render end smoothly while altering its envelope. Keep the unedited response when investigating a room, and document any windowing or noise compensation used by measurement software. Different analysis methods should not be compared as if only the room changed.

The tail export guide concerns delivery range and ending choices. In production, an intentional fade can be musically useful. In measurement, hiding an unwanted tail is not a substitute for describing it accurately.

Frequency and listening context change audibility

A bright portion can become masked while low-frequency energy persists. A dense mix can hide a tail that is obvious during solo playback. Conversely, a quiet gap may reveal noise or late energy that was inaudible under the source.

The damping guide explains frequency-dependent decay, while the reverb EQ guide distinguishes spectral shaping from changing the engine’s internal losses. Filtering can change what is audible without making every frequency decay at the same rate.

Do not use the point where you personally stop hearing a solo tail as a universal acoustic threshold. Playback level, room noise and hearing vary. A useful report states the method and conditions rather than converting an informal listening impression into a precise physical claim.

A practical tail-checking sequence

  1. Identify the level reference and whether the task is production or measurement.
  2. Keep an unchanged source response or render for comparison.
  3. Inspect the late region for noise, truncation and changing slope.
  4. Check relevant frequency bands instead of only one broadband envelope.
  5. For production, define an intentional ending and verify the actual exported file.

The DAW settings checklist helps preserve return routing, automation and render conditions. A tail that disappears in one export may reflect an omitted return or range limit rather than any change to the reverb algorithm.

Frequently asked questions

Does RT60 mean the sound is silent after that time?

No. It describes a relative sixty-decibel decay under a stated model or analysis. An ideal exponential response continues below that level, and audibility depends on noise and playback conditions.

Can I reveal a buried tail by turning up the recording?

You also turn up the recorded noise. Playback gain does not recreate signal-to-noise information that was not captured. A better measurement setup may be needed for a reliable late-decay analysis.

Does a quieter source shorten the room’s RT60?

Not in an ideal linear model. It can reduce the usable decay range above noise and make the observed tail seem shorter. Real systems may add other level-dependent behavior.

Is a fade the same as a shorter decay setting?

No. A fade applies an external time-varying gain, while a decay setting changes the response according to the engine. Similar audible endings do not prove identical internal behavior.

Can background noise make a fit look longer?

It can distort a late slope, but the direction and severity depend on the analysis method and data. Inspect the curve and reliability indicators instead of assuming one universal bias.

Are the example crossing times measured results?

No. They are calculated illustrations using a fixed thirty-decibel-per-second slope and stated digital levels. They do not characterize a real room, recording device or listener.

Scroll to Top