Browser sound level meter with A, C and Z weighting, slow and fast response, equivalent level, and a check against Brazil's NR-15 exposure limits.
A sound level meter does not simply measure how loud a sound is. It applies two filters before arriving at a number, and the choice of those filters is what makes a reading comparable to a standard or not. The first is frequency weighting. Human hearing is far less sensitive to bass at moderate volumes, so an air-conditioner rumble that physically carries plenty of energy bothers people less than a conversation of the same intensity. The A curve corrects the measurement to approximate that perception, which is why practically every occupational noise standard requires dB(A). The C curve preserves bass and is used for peaks, and Z corrects nothing. This tool applies weighting bin by bin across the spectrum, summing powers, which approximates a real meter better than passing the signal through a single filter. The second is time response. NR-15 explicitly requires the slow response circuit, an exponential smoothing with a one-second time constant. The fast response, at 125 milliseconds, reveals transients the slow one hides. The smoothing is done on the energy scale rather than in decibels, because the average of a logarithm is not the logarithm of the average, and getting that wrong would underestimate peaks. The number that actually matters for exposure is not the instantaneous one but Leq, the energy average over time. It is logarithmic, meaning a single short burst weighs far more than several quiet minutes. Percentiles complete the picture: L90 describes the steady background of a room and L10 describes the recurring peaks, and two places with the same Leq can be very different to live with depending on the gap between those two.
These references help contextualize formulas, standards, APIs and limitations used on this page. They do not replace professional validation when a result has legal, financial, medical or operational impact.
As a comparison, yes; as an absolute value, not without calibration. Phone microphones respond reasonably well across the speech range and clearly show that one room is noisier than another. What they lack is a certified reference: each model saturates at a different level, so the absolute number can be off by 5 to 10 dB. They also tend to clip before 110 dB, which underestimates very loud environments.
Tap to turn on the microphone. Audio is analysed live on your device: it is never recorded or uploaded.
A weighting mimics the ear reduced sensitivity to low frequencies and is required by every occupational noise standard, Brazil's NR-15 included.
A phone microphone is not a certified instrument and each model saturates at a different level, so the absolute value depends on this adjustment. If you have access to a real sound level meter, place both next to a constant source and move the slider until the numbers agree. Without that, treat the reading as relative: it compares rooms well, but the absolute value can be off by 5 to 10 dB.
This tool is educational and does not replace a calibrated sound level meter. No reading here is valid for a technical report, expert evidence, legal proceedings, hazardous-work classification or early-retirement claims, all of which require an instrument with traceable calibration and a qualified professional.
Audio is analysed live inside the browser and discarded frame by frame. Nothing is recorded, nothing is saved, and no sound leaves your device.