The 30-second answer: sound flickers, averages do not
Speech syllables, passing cars, clattering dishes — everyday sound arrives in bursts, and each burst moves the live number. Swings of 10–20 dB within seconds are completely normal on the instantaneous readout; they mean the meter is fast, not broken. Professionals never quote that flickering digit. They quote time-weighted statistics: Slow (a one-second smoothing that calms the display), Leq (the energy average over the whole session — the number that goes in reports), and Maximum/Peak (the loudest event, for bangs and barks).
Rule of thumb for our online decibel meter: watch the live number, quote the Leq. If the Leq itself keeps drifting after a minute, the sound genuinely comes and goes — extend the session or split it into "with event" and "background" lines instead of chasing one number.
One minute of traffic, three ways to read it
The same 60 seconds of street sound through three lenses. The instant trace jumps 23 dB; the Slow trace breathes gently; the Leq settles on one quotable number.
Fast vs Slow vs Leq vs Max: the comparison table
Five numbers, five jobs. Fast and Slow are time weightings (how quickly the display follows sound — 125 ms vs 1 s exponential averaging, per IEC 61672). Leq is an energy average over your whole session. Maximum and Peak remember the loudest moment.
| Reading | How it behaves | Quote it when… |
|---|---|---|
| Instant / Fast | Follows every burst, jumps 10–20 dB | Never quote — watch only, for finding sources |
| Slow | 1-second smoothing, breathes gently | Describing typical level live ("sitting near 65") |
| Leq | Energy average, stabilises over 30–60 s | Always — exposure, comparisons, logs, complaints |
| Maximum | Loudest Slow-weighted moment held | Bangs, barks, pass-bys, sleep disturbance events |
| Peak | Loudest instantaneous pressure, dBC/dBZ | Impacts and blasts — fireworks, gunshots, slams |
One subtlety worth knowing: Leq is an energy average, not an arithmetic one — loud seconds count more than quiet ones, which is exactly why it matches hearing risk and annoyance better than "the number I saw most". Feed it into the exposure calculator with your duration to get a dose percentage, and read the Leq meter guide for the full method.
Worked example: turning a flicker into a log line
A real-feeling 60-second session outside a café on a busy street, phone on the table, mic uncovered:
| What you saw | Value | What to do with it |
|---|---|---|
| Live number swinging | 55–78 dBA | Ignore as a quote — it only shows traffic comes in bursts |
| Slow hovering | ≈ 62–68 dBA | Fine for conversation ("typically mid-60s") |
| Leq, 60 s | 64 dBA | The quotable number — log it with place and duration |
| Maximum (bus pass-by) | 78 dBA | Note alongside Leq — explains the annoyance |
Final log line: "Café terrace, 3 m from kerb: 64 dBA Leq / 78 dBA max, 60 s, Slow." One line like that is worth more than ten screenshots of flickering digits — and repeats from the same spot should agree within a few dB. Compare against everyday sounds and the full chart to sanity-check.
The 60-second stable-reading routine
1. Freeze the setup. Phone at ear height where people are, mic pinhole uncovered, case clear, yourself silent — your own voice adds 10–20 dB and your grip adds handling thumps. 2. Set Slow response and dBA. Slow calms the display; dBA matches perceived loudness and every exposure limit (add a dBC pass when bass dominates — see dBA vs dBC). 3. Record 60 seconds without moving. Watch Leq converge; note Leq, minimum (the background), and maximum (the events). 4. Repeat once. A second 60-second run from a nearby spot should land within ~3 dB — if it does, you have a trustworthy number; if not, something pulses nearby, so log both lines and say so.
If readings still misbehave, suspect the usual culprits before the meter: a covered mic, automatic gain control re-enabled by the phone (check the Device panel), wind outdoors (+10 dB artefacts unshielded), or calibration drift between devices — covered in the phone-and-PC guide, are apps accurate, and what affects accuracy.
Frequently asked questions
Why does my decibel meter keep changing every second?
Should I use Fast or Slow response?
What is Leq and why should I quote it?
My two phones show different readings — which is right?
How long should I measure for a stable reading?
Sources and further reading
- IEC 61672-1 — Electroacoustics: sound level meters; Fast (125 ms) and Slow (1 s) time weightings, Leq definition.
- NIOSH — Criteria for a Recommended Standard: Occupational Noise Exposure (Leq-based dose, 3 dB exchange rate).
- EU Physical Agents Directive 2003/10/EC — Leq-based action values and exposure limits.