Page 543 - DSP Integrated Circuits
P. 543
528 Chapter 11 Processing Elements
(a) Express the word length in the WRQM in terms of the coefficients a;.
an
(b) How is the hardware cost and throughput related to WRQM d W^?
11.16 A fourth-order digital filter is realized in cascade form using second-order
sections in direct form II. The second-order sections are implemented using
distributed arithmetic. How many PEs are required? How many words are
required in the ROMs? Sketch the major parts of the implementation.
11.17 The filter structure in
Figure PI 1.17 shall be
implemented using
distributed arithmetic.
(a) Sketch the major parts
of the implementation.
(b) How many PEs are
required? How many
words are needed in the
ROMs?
(c) Determine the contents
in the ROMs.
(d) Show that all
computations can be
performed in parallel if
the structure is
pipelined. How is the
transfer function
affected?
Figure PI 1.17 Parallel form
11.18 The filter structure shown
in Problem 6.6 shall be
implemented using vector-multipliers (distributed arithmetic). Determine
first the difference equations in computable order. Sketch the most
important parts in such an implementation. Determine the number of
words in the ROMs.
11.19 (a) Describe how the so-called interpolated FIR filter (IFIR) in Figure
PI 1.19 can be implemented using distributed arithmetic. How many
PEs are required?
(b) Determine the contents in the ROMs.
11.20 The lattice WDF shown in Figure P11.20 shall be implemented using PEs
(distributed arithmetic) that compute inner products with fixed coefficients.
The adaptor coefficients a; have a word length of 6 bits while the data have
18 bits (two's-complement representation).
(a) Determine the number of inner product operations required per
sample.
(b) Determine the best placement of the delay elements—i.e., minimize the
number of words in the ROMs.
(c) Determine the maximal sample frequency when the PEs and the shift
registers can be clocked at least at 400 MHz.
(d) Sketch a block diagram for an implementation where the memories are
implemented as circulating shift registers.

