Calculation of the spectrum of error at the output of an analog-digital converter

Analog-digital conversion is investigated for a normal random process at a finite number of quantization levels. It is found that the error spectrum is considerably wider than the signal spectrum, which is chosen as 2. For instance, at 2-delta-f of approximately 2-6, the exponential correlation function gain reaches 10; for a Gaussian correlation function it is 5. It is concluded that output linear filtering leads to a significant gain in the singal-to-noise ratio.