A decibel is not a unit of loudness in the way a meter is a unit of length. It is a way of compressing enormous ratios into numbers human ears and brains can compare. When sound power doubles, the level rises by about three decibels. When it increases tenfold, the level rises by ten decibels. Most people describe a ten decibel increase as sounding roughly twice as loud, which tells you something important: both our hearing and our measurement scales are logarithmic. Sonarish displays approximate sound pressure level in decibels and, for longer sessions, LAeq — the A-weighted equivalent continuous level that environmental regulations often cite.

Sound pressure level compares acoustic pressure to a reference pressure of twenty micropascals, the conventional threshold of human hearing at one kilohertz. The formula uses twenty times the base-ten logarithm of the pressure ratio because pressure appears squared in energy calculations. That detail matters when you read specifications: power uses a multiplier of ten in the logarithm, pressure amplitude uses twenty. Mixing the two formulas when converting between datasets is a silent way to be wrong by six decibels.

Why frequency weighting exists

Your ear does not treat all frequencies equally at moderate listening levels. Low rumble and very high hiss can carry substantial acoustic energy while sounding less intrusive than midrange content at the same unweighted level. A-weighting, standardized in IEC 61672, applies a frequency-dependent filter that approximates human sensitivity. Energy below about five hundred hertz and above about four kilohertz is attenuated before the level is computed. A motorcycle passing at eighty decibels on an unweighted Z-scale might read near sixty-five decibels A-weighted even though the physical pressure waveform is unchanged — the filter reflects how most listeners perceive it.

City noise ordinances, workplace exposure limits, and neighbor disputes frequently reference A-weighted metrics such as LAeq over fifteen minutes or LAFmax for the fastest responding peak. Comparing a raw unweighted spectrum from a hobby app to a statute written in dBA without applying the same filter is a common amateur mistake. Sonarish applies A-weighting in its energy integrator so session statistics align with how regulations are written, while still offering unweighted views for machinery diagnosis where low-frequency tones carry diagnostic meaning.

LAeq and why peaks lie

Instantaneous peak level is dramatic and often misleading. A single door slam can flash a high number while contributing almost nothing to the energy that disturbs sleep over an entire night. LAeq over interval T is the constant level that would deliver the same total A-weighted energy as the actual fluctuating signal across that interval. Mathematically, you square the A-weighted pressure samples, average them over time, take the square root implicitly through the logarithm, and express the result in decibels relative to the reference pressure.

Sonarish lets you choose integration windows — one minute for quick checks, fifteen minutes for neighborhood noise logs, custom spans for factory walk-throughs — and displays running LAeq alongside short-term percentiles. That design choice reflects real use: comparing whether an air conditioner grew louder over three years, not chasing the single loudest sample a gust of wind created.

Phone microphones and honest labeling

Professional class-one sound level meters cost thousands of dollars because their capsules, preamps, and calibration chains are traceable to national standards. Phone microphones are optimized for voice calls and voice assistants, not courtroom acoustics. We apply a factory offset calibration using reference tones and offer an optional field check — clap at known distance, compare to a trusted meter if you have one — to tighten typical error into roughly plus or minus two to four decibels. That is sufficient to answer "is this HVAC unit louder than last month" and insufficient to prove compliance in a legal dispute. The UI always labels readings as approximate SPL.

Digital full scale, written dBFS, measures how close a recording is to the maximum value the analog-to-digital converter can represent. It says nothing about absolute acoustic pressure in a room. Voice Memos showing minus six dBFS tells you the recording is hot, not that your apartment is sixty decibels A-weighted. Never compare dBFS exports to municipal noise law written in dBA. Our FFT post covers spectral views; machine baseline comparison shows how calibrated level differences reveal mechanical change over months.