Lesson 09 · Decoding 8-bit samples - Build a WAV/PCM Toolkit

Decoding 8-bit samples

Eight-bit WAV audio is the one depth stored unsigned, with silence sitting at 128 instead of 0. Today you decode it by removing that bias, a small twist that catches everyone once.

The goal

Decode 8-bit unsigned samples into signed integers centered on zero.

Start here - the target

TO DO

Scenario: 8-bit unsigned bytes decode by subtracting the 128 bias

Giventhe four data bytes 0x80 0x00 0xFF 0x81

Whenthey are decoded as 8-bit unsigned samples

Thenthe samples are [0, -128, 127, 1]

And0x80 (128) decodes to 0 - the unsigned midpoint is silence

Background

Eight-bit PCM is the odd one out: unlike every larger depth, it is stored unsigned. A byte ranges 0 to 255, and the format puts silence at 128, the midpoint, rather than at 0. So a raw byte of 128 means a zero-amplitude sample, 255 is the most positive, and 0 is the most negative. This is called a biased or offset representation.

To bring 8-bit samples into the same signed, zero-centered world as every other depth, you subtract 128: byte 128 becomes 0, byte 255 becomes 127, byte 0 becomes -128. Now the values live in -128 to 127, exactly the signed 8-bit range, and the same mixing and gain math will work across all depths. Forget the bias and every 8-bit file reads as a loud DC offset - constant near-maximum instead of silence at rest.

Make it work

// 8-bit WAV is UNSIGNED: subtract the 128 bias to center on zero
func decode8(b []byte) []int {
  var out []int
  for _, x := range b {
    out = append(out, int(x)-128)
  }
  return out
}

CheckpointDONE

You can decode 8-bit unsigned samples with the bias removed. Commit and stop here.