Audio Bit Depth Dynamic Range Calculator | PCM Dynamic Range, SNR and LSB Voltage

Turn a PCM bit depth and reference voltage into dynamic range, ideal SNR, quantization level count, 1 LSB voltage and quantization noise RMS in one view. Presets for CD, DAT, studio and 32-bit float map each result to a real format with a single tap.

💡 About this tool

"Is 24-bit really worth it over 16-bit, and by how many dB?" is one of those questions that comes up every time you spec an ADC, pick a recording format, or argue about it in a forum thread. The math is short — dynamic range is 20·log10(2^N) dB and ideal SNR is 6.02·N + 1.76 dB — but converting that into LSB voltages and noise floors by hand gets old fast.

Feed in just the bit depth and Vref, and the tool shows dynamic range, quantization SNR and the level count as three headline cards, then fills in 1 LSB voltage, quantization noise RMS, the max signed-integer amplitude and full-scale voltage in a detail grid. LSB voltage auto-scales across V / mV / µV / nV, so a 24-bit step doesn't drown in zeros. The presets run from 8-bit telephone audio up to the 32-bit float buses inside a DAW, so you can line up the formats side by side instead of recomputing each one.

🧐 Frequently Asked Questions

How much does one extra bit buy me? About 6.02 dB of dynamic range per bit. Going from 16-bit to 24-bit adds 8 bits, or roughly 48 dB of extra headroom against the quantization floor.

Where does the "+1.76 dB" in the SNR formula come from? It is the correction term for a full-scale sine wave: the ratio of the sine's RMS value to the RMS of a uniformly distributed quantization error. The full expression, 6.02·N + 1.76 dB, is the ideal ceiling set purely by bit depth.

Why is quantization noise RMS shown as 1 LSB / √12? That is the standard deviation of an error assumed uniform over ±0.5 LSB. It is the undithered ideal; real converters add dither and analog circuit noise on top.

Why does 32-bit float report an absurd dynamic range? The tool applies the same 20·log10(2^N) integer formula, so 32 bits reads about 192.66 dB. That is the integer-model figure; real 32-bit float buys range through its exponent, so this formula actually understates it. Treat the number as a headroom indicator for internal processing rather than a literal converter spec.

What should I put for Vref? The peak voltage at full scale. If you don't know your converter's value, leave it at 1 V and read the LSB voltage as a relative figure.

📚 Why bit depth is mostly about headroom, not audibility

A practical point that gets lost in "16 vs 24" threads: 16-bit's ~96 dB already exceeds the noise floor of most listening rooms, so the audible benefit of 24-bit at playback is thin. The real win is on the capture and mixing side. Recording at 24-bit lets you set conservative input levels with plenty of headroom, so a quiet take doesn't crowd the quantization floor when you later push gain. And 32-bit float on the mix bus exists for the same reason — internally you can sum dozens of tracks well past 0 dBFS without clipping, then pull the master back down, because the format carries the level in its exponent.