Quantization noise

The uncertainty introduced by rounding the sample amplitudes to discrete levels can be viewed as adding quantization noise to the signal. The amount of this 'noise' decreases with increasing amplitude resolution. It can be expressed by the Signal-to-noise ratio (SNR, unit decibel [dB]):

SNR = 6·Nbits - 7.3 Note that Nbits is the number of bits used to represent a typical signal, which is less than the output resolution of the A/D-converter.

The amplitude resolution must be such that the quantization noise is smaller than the signal-to-noise ratio of the input signal, which is given by the amplifier and the input signal itself.

Example: In a system with a 14 bit A/D-converter, a typical signal with 50 µV amplitude could be represented with 256 digital steps (8 bits). For this signal, SNR = 40.7 dB.
A small signal with 3 µV amplitude would then use only 16 steps (4 bits), and SNR = 16.7 dB.

 

 

 

©1998 BITS OF SLEEP (Demo) : Human sleep investigation: Signal analysis basics.