How Loud Is 100 Decibels? Where the Criteria Get Short

OSHA's Table G-16 lists sound levels against permitted daily durations. 100 dBA appears against 2 hours. Above it, 105 dBA appears against 1 hour and 110 dBA against half an hour; below it, 95 dBA appears against 4 hours.

That progression is the 5-decibel exchange rate at work. It is a regulatory choice, and it is the reason the OSHA table looks generous at high levels compared with the equal-energy arithmetic used elsewhere.

NIOSH publishes a different pairing entirely: 85 dBA over 8 hours with a 3-decibel exchange rate. Because the rate is smaller, allowed durations shrink much faster as level rises. At 100 dB the gap between the two frameworks is dramatic.

It is easy to assume a 100 dB peak on a display is covered by the averaging table. It is not. OSHA handles impulsive or impact noise separately, with a ceiling of 140 decibels peak sound pressure level.

That separation exists because averaging and peak detection answer different questions. An average describes accumulated energy over a period; a peak describes the largest instantaneous pressure. Instruments must be built to capture the second, and consumer audio paths generally are not.

So a single loud bang and a steady 100 dB machine are assessed under different provisions, by different instruments, against different numbers.

Numbers near 100 do appear in the published record, but they arrive with conditions attached that most content strips off. NIOSH's power tools database, for example, reports A-weighted sound power levels of roughly 103 to 113 dBA for corded circular saws under load, measured to ANSI S12.15 and ISO 3744 — sound power, not the level at an operator's ear.

Type-approval work gives another example. Under UN Regulation 28, a Class II motor vehicle horn is tested on a bench at 2 metres, and once fitted to a vehicle it must meet a minimum measured at 7 metres in front. Both figures describe a compliance test, not a level in traffic.

The pattern repeats across every family of sources: the figure is real, and it is inseparable from the quantity and geometry that produced it.

Two things go wrong at high levels on consumer hardware. Automatic gain control reduces input sensitivity as sound rises, compressing exactly the events you are trying to capture. And the microphone and preamplifier can clamp, so the digital signal stops tracking the acoustic one at all.

Our meter reports the browser's own audio-processing state and warns when gain control or noise suppression is active, because in that state calibration is void and the reading is not proportional to sound pressure. Uncalibrated it shows an A-weighted digital level in dBFS; after a local calibration it shows an estimated dBA value.

For anything consequential at this level — an occupational assessment, a complaint, a compliance record — the instrument has to be a calibrated meter or dosimeter with a documented procedure.