Abstract: A block adaptive quantization device processes blocks of input data samples to determine an aggregate or average value which is applied to each of a number of exponent look-up tables to derive an exponent value dependent on the extent of compression required. The exponent value is split by an exponent splitter into a remainder value and a shift value. The remainder value is used to select one of a number of divide look-up tables to which the unsigned data sample is applied, and the shift value is used to apply a variable bit shift to the data sample. Each divide look-up table comprises the same group of divide look-up tables. The combination of the four look-up tables in the group, together with the variable bit shift defines a series of code tables made up of a group of divide look-up tables repeated at six dB intervals. The remainder value selects one of the code tables within the group, and the bit shift operation determines the selected one of the repeated groups.