Acoustic metrics: Leq, LAF, Lmax, Ld, Le, Ln and percentiles.

An acoustic report rarely contains a single number, and with good reason: each metric answers a different question. The night’s average, the current reading, the event’s peak and the background noise are four coexisting data points rather than versions of one figure. Choosing the wrong metric leads to fruitless arguments about regulations.

Here we unpack Leq / LAeq, LAF, LAS and time weightings, Lmax and Lpeak, Ld, Le and Ln, and acoustic percentiles L10, L50 and L90, ending with the important question: when to use each metric. If you are still unsure what a decibel measures or how levels add, first review decibels, the logarithmic scale and crest factor.

Leq / LAeq

Leq is the equivalent level: the constant level delivering the same energy as the varying sound measured over the entire period. It is an energy average rather than an arithmetic average, so peaks weigh more than troughs and long bursts more than isolated hits. That is why it summarises whole hours in one comparable number.

LAeq is the same with A-weighting applied: the curve emulating the ear’s response at moderate levels. Almost all legal limits and venue meters use LAeq because it correlates best with what people perceive.

Live, a set’s LAeq correctly answers “How much noise have we made in total?” As an average, it says nothing about peaks: a flat night and one with two huge drops can finish with the same LAeq. It therefore always needs company.

LAF, LAS and time weightings

When sound changes quickly, the instrument must decide how fast to follow it. That speed is time weighting, and the letters come from the English names: F for Fast, with a 125-millisecond time constant; S for Slow, with a 1-second constant; and I for Impulse, much faster and intended for impacts. IEC 61672-1 sets these constants and sound-level-meter accuracy classes.

Read LAF and LAS as A-weighted level with a fast or slow time constant. LAF reacts almost in real time and monitors rises and falls as they happen; LAS smooths the reading into a stable value you can follow without constant movement. In a booth, LAF warns and LAS confirms.

The same logic applies to frequency weighting: A for environmental noise and comparison with perception, C when bass matters, and flat Z when you want the uncorrected signal.

Lmax and Lpeak

Lmax is the highest level reached within the measurement period, respecting the configured weighting and time constant. It is the ceiling of what happened and the figure used to check whether a specified limit was exceeded at any point.

Lpeak goes further: it captures the narrowest acoustic peak, of minimal duration, without time-constant smoothing. By definition, Lpeak is equal to or greater than Lmax. The difference indicates how transient the sound is: a larger gap means a more sharply peaked event.

There are two practical uses. First, capturing the exact peak moment: a closing hit, an isolated shout, or the drop near the stage. Second, understanding that peak-only measurement cannot judge sustained nuisance, just as measuring only dBFS digitally says nothing about perceived volume.

Ld, Le and Ln

When regulations divide the day into periods, period levels appear: Ld for daytime, Le for evening and Ln for night. In British practice, for example, daytime runs from 07:00 to 19:00, evening from 19:00 to 23:00, and night from 23:00 to 07:00. There is also a combined level for the whole set of periods.

Night-time Ln is most often cited when discussing entertainment venues and nearby homes: it covers the period with stricter criteria because nuisance and interference with sleep carry more weight. A venue may easily comply during the day and sit close to the limit in the early hours.

A reading note: Ln with a lowercase n and a number represents a percentage of time, while Ld, Le and Ln as a classification are clock periods. These are different conventions sharing a letter, and badly labelled reports have caused more than one argument.

Acoustic percentiles L10, L50 and L90

An average hides structure; percentiles restore it. L10 is the level exceeded for 10 % of the measurement time, L50 the median, and L90 the level exceeded for 90 % of the time: the measurement’s relative quiet. All use n for the percentage of time exceeded.

The usual reading is that L10 describes intense moments, L50 the middle state and L90 background noise. A room with high L90 has a high noise floor that never falls; a road with L10 far above L90 has traffic peaks clearly standing above a constant background.

For venue measurement, the value lies in the difference: close L10 and L90 mean uniform, continuous sound; a large gap means activity with quiet gaps between. On a DJ dancefloor, that separation is precisely the distance between a full track and silences between tracks.

When to use each metric

Let the question decide rather than habit:

• Leq / LAeq: summarising a whole period and comparing exposures. The metric for reports and aggregate limits.
• LAF and LAS: observing and reacting during measurement. The live-meter metrics.
• Lmax and Lpeak: locating ceilings and peaks, or checking an isolated hit.
• Ld, Le and Ln: demonstrating compliance per period when the criterion separates times of day.
• L10, L50, L90: describing noise character, background, median and intense moments, without flattening it all into an average.

Several coexist in a venue: the night’s LAeq for exposure, each section’s Lmax for peaks, and percentiles where background characterisation is needed. Broadcast loudness metrics, such as EBU Tech 3341 and 3342, follow the same philosophy but measure programme perception: 0.4- and 3-second windows, averaging with threshold gating, and a loudness range calculated between distribution percentiles. The difference in approach, environmental noise versus content loudness, is covered in sound-level and noise measurement.

Mistakes when reading an acoustic report

Comparing different metrics. LAeq and Lmax are not alternatives: one averages the period’s energy and the other gives the ceiling. Putting them together without explaining this leads to false conclusions.

Ignoring weighting and the time constant. Without knowing whether a number is A or Z, fast or slow, you do not know what you are comparing. The same event measured with LAF and LAS gives different figures, both correct.

Confusing Ld, Le and Ln with percentiles. A time period and a percentage of time exceeded are different things. They share letters rather than meaning.

Judging nuisance only by the average. LAeq may be acceptable while an isolated Lmax causes discomfort; the acoustic equivalent of looking only at RMS and ignoring crest factor.

Measuring in the wrong place. Metrics describe what happened at the microphone’s location. Beyond the measurement point, there is the reverberant field and extra distance from the source, as discussed with the direct and reverberant fields.

Next step

You can now read an acoustic report. Next, move into digital audio: the Nyquist theorem, dither and noise shaping, and why the buffer and clock determine whether the chain sounds clean. Continue with Digital audio: Nyquist, dither, noise shaping, buffer, clock, Mid/Side and phase correlation.