If quantization in digital filters is performed by means of magnitude truncation or related methods, the various quantization noise signals are correlated with each other and with the signal to be transmitted. Also the resulting quantization noise at the filter output exhibits partial correlation with the output signal. In many cases the correlated part of this noise need not be considered as disturbance, because it can be viewed as emanating from a fictitious linear system excited by the same signal as the main system. After combination of these two systems the correlated noise contribution can be Interpreted as a weak linear distortion of the original transmission characteristic. In this paper the spectral power density of the uncorrelated noise only is determined, which can be ascribed to the nonlinear effects of quantization. In many practical cases, this uncorrelated noise carries a power which is substantially lower than that of the total quantization noise.
[1]
Bruno O. Shubert,et al.
Random variables and stochastic processes
,
1979
.
[2]
A bound on the temporal degree of coherence
,
1977
.
[3]
Mj Martin Bastiaans.
A Frequency-domain Treatment of Partial Coherence
,
1977
.
[4]
T. Claasen,et al.
Quantization noise analysis for fixed-point digital filters using magnitude truncation for quantization
,
1975
.
[5]
Leland B. Jackson,et al.
On the interaction of roundoff noise and dynamic range in digital filters
,
1970,
Bell Syst. Tech. J..
[6]
Hendrik Bosma,et al.
On the theory of linear noisy systems
,
1967
.
[7]
Athanasios Papoulis,et al.
Probability, Random Variables and Stochastic Processes
,
1965
.