Sound Measurement Tips: How to Get Accurate Decibel Readings

A sound‑level meter only reports what it captures at the microphone. The quality of the captured signal depends on where you place the meter, which weighting and time response you choose, how you handle the device, and how you average the result. Get those right and a calibrated phone meter agrees with a Class 2 reference to within ±2 dB on typical environmental sources. Get them wrong and the same phone gives you readings that are off by 10 dB or more — not because the meter is bad, but because the measurement was wrong.

This page collects the practical techniques that move readings from "rough estimate" to "trustworthy." It assumes you have already done the calibration procedure once for the device. None of these tips compensate for an uncalibrated meter — they extract reliable numbers from a calibrated one.

Microphone placement

The microphone is the only thing that "hears" the room. The closer it is to the source, the more reflective surfaces it sees, and the more turbulent the air around it, the more those factors dominate the reading over the actual sound you care about.

Distance from the source

For an idealized point source in free field, the inverse‑square law predicts that doubling the distance reduces the SPL by 6 dB. Real indoor measurements rarely match that perfectly because of room reflections, but the trend holds:

For occupational measurements, the rule is measure where the worker's ear is. For environmental measurements, choose the location that matters: bed pillow level for sleep noise, working surface for office noise, dance floor center for venue measurements.

If you are publishing a number, always note the distance. "Lawn mower: 85 dBA at 5 m" is a complete statement; "lawn mower: 85 dBA" is ambiguous and roughly useless to someone trying to estimate the level at a different distance.

Free field vs reflective field

A perfectly free field has no reflective surfaces. An anechoic chamber approximates one; an open meadow on a windless day is the next best thing. Most rooms are highly reflective, especially small hard‑surfaced rooms like bathrooms, kitchens, and elevator cabs.

In a reflective room, the SPL at most points is dominated by the reverberant field, not the direct sound from the source. This has two practical consequences:

• The direct‑field 6 dB‑per‑doubling rule breaks down a few meters from the source.
• The reading at one position can shift by 3 – 6 dB if you move the mic 0.5 m, because of standing‑wave nulls and antinodes.

To suppress this, average over multiple positions in the room (at least three, separated by ~1 m), and report the mean.

Avoid handling, pockets, and surfaces

The microphone in your phone or laptop picks up vibration as easily as it picks up airborne sound. Holding the device in your hand introduces finger noise and clothing rustle that can add 5 – 15 dB at low frequencies. Setting the device on a hard table introduces ground reflection.

Practical placement that works on every device:

• Set the device on a folded soft cloth (a microfiber cloth, a t‑shirt) on a table at chest height.
• Orient the microphone toward the source. Most phones have the mic at the bottom edge, but it varies — verify with a clap test if unsure.
• Once the device is placed, don't touch it during the measurement.

For very long measurements, a small tripod with a smartphone clamp holds the device free of any surface and provides the cleanest reading.

Frequency weighting

The meter applies a frequency weighting before computing the SPL. Three choices are commonly used; they give very different numbers for the same sound.

The default for almost every published reference (NIOSH REL, OSHA PEL, WHO community guidelines, ISO 1996) is A‑weighting. Unless you have a specific reason — usually because the source is dominated by frequencies below 200 Hz — leave the meter on A.

If you are measuring a venue with significant subwoofer content and your A‑weighted reading seems too low for what you can feel in your chest, switch to C and report both. The difference between A and C is itself diagnostic: a 20 dB gap (C is 20 dB higher) means the source is dominated by bass; a 5 dB gap means it is mid‑frequency.

The full math behind the weighting curves is on the science page.

Time weighting

Time weighting is an exponential averaging applied before the value is displayed. Three settings are standard:

• Fast (125 ms time constant) — captures speech‑rate variation and short events without flickering. The default for nearly everything.
• Slow (1 s time constant) — smoother readings for steady ambient noise. Use when you want a stable number to log and the source is steady.
• Impulse (35 ms attack, 1.5 s decay) — designed to catch transients (gunshots, hammer strikes, balloon pops). Reads markedly higher than Fast for short events.

A common mistake is leaving the meter on Slow when measuring an intermittent or impulsive source. Slow underreports peaks because the 1‑second time constant doesn't have time to "settle" to the peak before the event passes. Use Fast unless you specifically want the smoothing.

For sources with frequent strong transients — workshop tools, sporting events, rifle ranges — Impulse is the most representative weighting, and it is the one some standards (ISO 9612 for occupational measurements with prominent impulses) require.

Multi‑sample averaging

A single 5‑second reading captures only the noise during those 5 seconds. For variable sources, that snapshot is rarely representative of the longer‑term exposure that matters for health.

Two averaging strategies cover most cases:

Arithmetic mean (Avg)

The Avg statistic on our meter is the arithmetic mean of all displayed values. For approximately steady noise, this is a fine estimate of the ambient level — for example, a fan running at 55 dBA continuously will have Min, Avg, and Max all within 2 dB of each other.

Equivalent continuous level (Leq)

For variable noise, the appropriate quantity is the equivalent continuous A‑weighted level, or LAeq. This is the steady SPL that would deliver the same total acoustic energy over the measurement period as the actual variable signal. Mathematically:

LAeq,T = 10 × log10( (1/T) × integral( 10^(LA(t)/10) ) dt )

The browser meter does not compute Leq directly today, but for reasonably steady or non‑extreme noise, the Avg statistic is within 1 dB of LAeq. If you are publishing a number that needs the energy‑equivalent definition (e.g., for a noise‑complaint measurement), use a Class 2 SLM with built‑in Leq integration.

How long to measure

Shorter samples on variable sources risk catching atypical conditions — either a quiet pocket or a peak event — and reporting them as representative.

Eliminating contamination

Most large measurement errors come from things that are not the source you intended to measure. Identify and eliminate them in this order:

1. Wind on the microphone. Even a light breeze creates pressure fluctuations that read as 60 – 80 dBA at the mic. Outdoor measurements without a windscreen are unreliable above ~1 m/s of wind. A foam ball over the mic helps; for serious outdoor work, use a furry ("dead cat") windshield.
2. Handling and cable noise. Discussed above. Set the device down on cloth and don't touch it.
3. HVAC, fridge, computer fan in the same room. Background noise below 10 dB lower than the target is fine to ignore. Background between 10 and 6 dB lower than the target requires a correction (subtract about 1 – 2 dB from the combined reading). Background less than 6 dB lower than the target makes the measurement unreliable — either eliminate the background or use a different location.
4. Operator presence. Your breathing, clothing rustle, and footsteps are surprisingly loud at close range. Stand back from the meter; if you have to be near it, freeze.
5. Rain on the device. Even fine drizzle taps the device with audible energy. Use a covered location for outdoor measurements.

Specific scenarios

A few common situations have their own quirks worth knowing.

Outdoor traffic monitoring

ISO 1996‑2 requires measurement at 7.5 m from the road centerline, at 1.2 m height, with a windscreen. For informal home use, choose a position close to the receiver of interest (the bedroom window, the patio chair) and average for at least 10 minutes during representative traffic conditions.

Workplace machinery

Measure where the operator's ear is during normal use. If the operator moves between positions, take the time‑weighted average across positions weighted by the time spent at each. For variable‑duty equipment (intermittent stamping presses), measure across at least one full cycle.

Concerts, clubs, and venue volumes

C‑weighting reads more representatively than A‑weighting because the sound is bass‑heavy. Measure at multiple positions (front, middle, back, sides) — venue sound is rarely uniform. The hearing health page has guidance on what protector NRR you need for different exposure durations.

Baby room overnight

Use Slow time weighting and average over at least one full hour during the time of night you care about. The WHO bedroom recommendation of 30 dBA LAeq during sleep is the reference. White‑noise machines, if used, should produce no more than 50 dBA at the crib position — many machines on max volume far exceed that.

Restaurant or office acoustics

Measure during peak occupancy. Empty rooms read 10 – 20 dB quieter than full rooms. The published "comfort" thresholds (50 dBA office, 70 dBA restaurant) apply to the in‑use condition.

Limitations of phone measurements

A calibrated phone meter is a screening tool. It is excellent for mapping spaces, identifying problem areas, and giving non‑specialist users visibility into their own exposure. It is not a Class 2 sound‑level meter, and a few situations require one:

• Compliance documentation. OSHA inspections, workers' compensation claims, formal noise‑complaint cases all require a calibrated SLM (and usually a calibrator, paperwork, and a chain of custody for the readings).
• Levels above ~95 dBA. Most phone microphones clip in this range, under‑reporting the actual level. A Class 2 SLM is rated to ≥130 dB.
• Frequency analysis. Our meter shows a real‑time FFT that is useful for diagnosis but is not a 1/3‑octave or octave‑band analyzer. Standards that specify octave‑band measurements need dedicated equipment.

For everything else, a calibrated phone, careful placement, and the right weighting and time response will produce numbers you can trust.

Pulling it together

A reliable measurement is the result of all of the above being right at the same time. As a checklist for any measurement that matters:

• Calibrated this device against a known reference (procedure)
• Microphone placed at the listener's position, on a soft surface, not handled during the read
• A‑weighting (or C‑weighting if specifically measuring bass)
• Fast time weighting (or Impulse for transient sources)
• Wind and handling controlled (windscreen if outdoor, hands off the device)
• Background noise at least 6 dB below the target, ideally 10 dB
• Averaged over a duration appropriate to the source (at least 30 s for steady, 10+ min for variable)
• Source distance and weighting reported alongside the number

Done well, this gives a reading that is both accurate and reproducible — two qualities you will want every time the number you report has consequences.