Compression size estimate

Compares an uncompressed PCM stream with a constant bitrate such as 128 or 320 kbps. Variable bitrate will not match the estimate exactly.

These pages do arithmetic and show the formula next to the result. They do not measure your room. If a decibel figure is involved, the formula tells you whether the ten-log or the twenty-log form is in use, and what the reference is. Plug in the sample values first. The worked number is how you check that you understand the unit before you trust your own input.

Amplitude decibels are twenty times the log of a ratio. Power decibels are ten times the log of a ratio. Doubling a voltage, or a sound pressure, is about 6.02 dB. Doubling a power is about 3.01 dB. People mix these and then wonder why a “3 dB” change sounded like a small step or a large one. The calculator asks which ratio you mean. The listen button plays a tone and the same tone about 6 dB down so the amplitude case is not only a formula.

dBu and dBV are voltages. dBu references about 0.7746 volts. dBV references 1 volt. +4 dBu, a common professional line level, is about 1.23 volts. Consumer gear often talks about −10 dBV, which is a different voltage and a different convention. Adding the two numbers because they are both “decibels” is how interfaces get the gain structure wrong. Convert to volts if you need to compare them.

Wavelength is speed divided by frequency. Speed of sound in air is roughly 343 metres per second at 20 °C, and it rises by about 0.6 metres per second for each degree. A 1 kHz tone is about 34 centimetres long at room temperature. A 50 Hz tone is several metres long, which is why a subwoofer does not image like a tweeter and why “aiming” bass like a flashlight does not work.

File size for uncompressed audio is sample rate times bit depth times channel count times seconds, divided by eight to get bytes. Three minutes of 44.1 kHz, 16-bit stereo is about 30 megabytes before any header. A compressed estimate is duration times bitrate. 320 kbps for those three minutes is far smaller, and a variable-bitrate file will only be near the estimate. Headers make very short files look slightly long.

Nyquist frequency is half the sample rate. That is the highest frequency the digital stream can represent, not the highest frequency your speakers can make, and not the highest frequency you can hear. 44.1 kHz was chosen to cover the human range with room for a filter. 48 kHz is the video world’s habit. 96 kHz does not repair a clipped microphone.

Latency in a buffer is samples divided by sample rate. A 256-sample buffer at 48 kHz is about 5.3 milliseconds. That is only the buffer you typed in. Converters, a computer’s extra safety buffer, and a Bluetooth radio sit on top. The sample calculator is how you read a driver setting. The Bluetooth page is how you estimate the rest, badly but honestly.

Channel tools that change a file do it locally. Stereo to mono averages the two sides, which is the correct simple downmix and also the reason a polarity error cancels. Swap, left-only, and right-only are there so you can hear a side without a second app. The download is 16-bit WAV. If the browser cannot decode the file you picked, it says so. It does not send the file anywhere to be decoded later.

Sound pressure level is twenty log ten of a pressure over 20 micropascals. The page will turn pascals into dB SPL and back. A distance estimate uses the free-field rule, about six decibels quieter each time the distance doubles, referenced to one metre. Rooms, walls, and bass modes break that rule. If you need the level of a venue, use a meter. This is the formula the meter is implementing, not a substitute for one.