Quanitization reduces the amount of data. The simplest form of quantization is rounding off, or dropping everything to the right of the decimal point (3.5 becomes 3, 5.67483 becomes 5, etc.). Here, the quanitizer follows a power law. This means that there is a higher resolution for small values and large values are coded with less accuracy.

Also, noise shaping is built into the quantizer.