Sec.11.5 Summary 801
Figure 11.13(d) shows S,(e!) with M = Sin Eq, (11.78). S(e/”) and two of the frequency-
scaled and translated copies of S(e/”) are indicated explicitly in Figure 11.{3(d). It is
clear that aliasing has been avoided and that all the information necessary to reconstruct
the original sampled real bandpass signal now resides in the discrete-time frequency
interval —1 <@ < x. A complex filter applied to s4[n] can transform this information
in useful ways, such as by further bandlimiting, amplitude or phase compensation, etc.,
or the complex signal can be coded for transmission or digital storage. This processing
takes place at the low sampling rate, and this is, of course, the motivation for reducing
the sampling rate.
The original real bandpass signal s,[n] can be reconstructed ideally by the following
procedure:
1. Expand the complex sequence by a factor M; i.e., obtain
nina {gia tens sSissatins xy
2. Filter the signal s,{n] using an ideal complex bandpass filter with impulse response
hi[n) and frequency response
Hi(e!”)
0, -m 0. It can be shown that such a
signal can be sampled with sampling rate 2x/T = AQ, directly yielding the complex
sequence se}.
11.5 SUMMARY
In this chapter, we have discussed a variety of relations between the real and imaginary
parts of Fourier transforms and the real and imaginary parts of complex sequences.
These relationships are collectively referred to as Hilbert transform relationships. Our
approach to deriving all the Hilbert transform relations was to apply a basic causality
principle that allows a sequence or function to be recovered from its even part. We
showed that, for a causal sequence, the real and imaginary parts of the Fourier trans-
formare related through a convolution-type integral. Also, for the special case when the
complex cepstrum of a sequence is causal or, equivalently, both the poles and zeros of
its z-transform lie inside the unit circle (the minimum-phase condition), the logarithm
of the magnitude and the phase of the Fourier transform are a Hilbert transform pair
of each other.
Hilbert transform relations were derived for periodic sequences that satisfy a
modified causality constraint and for complex sequences whose Fourier transforms
vanish on the bottom half of the unit circle. Applications of complex analytic signals to
the representation and efficient sampling of handpass signals were also discussed.