Published by PitchChanger · Updated · How these guides are made
How to use this workflow
State the amplitude convention
Use normalized signed integer PCM from −1 to 1−step. Do not mix this with an unsigned or floating-point representation.
Calculate the spacing
For 16 bits, step = 1/32768. For 24 bits, step = 1/8388608.
Choose output depth deliberately
Use the delivery requirement and retain an editing master. More bits in a new file cannot undo earlier clipping or a low-resolution source.
The extra code bits change spacing, not full-scale headroom
16-bit spacing is 0.000030517578125; 24-bit spacing is 0.00000011920928955078125. The second grid is 256 times finer because eight additional bits multiply the code count by 256.
At a normalized input of 0.1, nearest rounding to 16-bit gives code 3277, decoded as 0.100006103515625. The quantization error for this one idealized sample is +0.000006103515625. It does not predict the noise or audibility of a music file.
Swipe or scroll sideways to read all columns.
| Integer depth | Code count | Normalized step |
|---|---|---|
| 8-bit signed example | 256 | 0.0078125 |
| 16-bit | 65,536 | 0.000030517578125 |
| 24-bit | 16,777,216 | 0.00000011920928955 |
Dither and floating point require separate assumptions
For ideal nearest rounding without overload, the error magnitude is at most half a step. Dither deliberately changes error statistics; a single rounding example is not a dither demonstration or an audibility test.
Audio Converter offers 16- and 24-bit lossless settings. It has no user-facing dither selector. Browser processing may use floating-point buffers internally, whose spacing is not the fixed integer grid shown here. Inspect the final output metadata rather than inferring file bit depth from an internal buffer.