Page 219 - Fundamentals of Radar Signal Processing
P. 219

replications (k = 3, specifically) is centered at ω = 2π radians. The periodicity

               of  the  spectrum  of  a  discrete-time  signal  therefore  guarantees  that  there  is  a
               replica  centered  at ω = 0 as well; this replica is the final desired spectrum.
               Thus, the real and imaginary outputs of the decimator are the desired I and Q
               signals. Another digital I/Q system that uses the spectrum replicating properties
               of decimation to advantage is described in Rice and Wu (1982).
                     The success of the decimation operation in eliminating the need for a final

               complex frequency translation depended on the proper relationship between the
               bandwidth and center frequency of the signal, and the decimation factor. This is
               the  major  reason  for  choosing  the  IF  to  be β  instead  of β/2  (or  some  other
               permissible value), and the sampling frequency as 4β instead of 3β (or some
               other value).
                     Rader’s digital I/Q architecture has reduced the number of analog signal
               channels from two to one, making the issues of oscillator quadrature and gain

               and phase matching completely moot, while also providing a natural opportunity
               to filter out DC biases introduced by the remaining analog mixer. Furthermore,
               the two A/D converters required at the output of the conventional quadrature
               receiver  to  enable  subsequent  digital  processing  have  been  reduced  to  one.
               There are two major costs to these improvements. The first is an increase by a
               factor  of  four  in  the  A/D  converter  speed  requirement,  from β  samples  per

               second  for  conventional  baseband  sampling  to  4β  samples  per  second  for
               Rader’s system; this may be difficult at radar signal bandwidths. The second is
               the  introduction  of  the  need  for  high-rate  digital  filtering,  which  is
               computationally expensive (although Rader’s efficient filter design lessens this
               cost).
                     Figures 3.22 and 3.23 sketch the processor conceptual architecture and the
               relevant signals of the second digital I/Q architecture (Shaw and Pohlig, 1995).

               In this case, analog frequency translation is used to shift the signal spectrum to a
               lower IF than used by Rader, namely 0.625β. The signal is then A/D converted
               at  a  rate  of  2.5β  samples  per  second,  resulting  in  the  signal             having  a
               spectrum  centered  at ω  =  π/2  as  shown  in Fig.  3.23.  An  explicit  complex
                                                   n
               modulation by exp(+jπn/2) = j  then shifts one of the sidebands, in this case the
               lower one, to baseband, resulting in the spectrum shown in Fig. 3.23d. Clearly
                    is complex as a result of this complex modulation.

















               FIGURE 3.22   Conceptual architecture of Lincoln Laboratory system for digital
   214   215   216   217   218   219   220   221   222   223   224