with sequential circuits are covered by Jalaja et al.
From the viewpoint of complexity, Jiang and
Brayton, [9], synthesized the logic design. The
multiplier-less multiple constant multiplication
(MCM) technique has been retimed by Yagain and
Vijaya, [10]. Meher, [11], has used cut-set and
flexible retiming to enhance the CPD of traditional
direct form filters at the expense of more
sophisticated hardware. Carry increment adders
(CIA) and Vedic multipliers were used by Thakral
et al. [12], to apply an unfolding transformation to
an FIR filter. Wallace multiplier was used by Swati
and Himanshu, [13], to build the FIR filter.
According to [14] discusses two variations of
enhanced booth multipliers and efficient hybrid
adders.
For the implementation of the FIR filter, Rai et al.
[15], employed carry increment adders and Vedic
multipliers and improved the CPD and energy
efficiency at the expense of a larger filter area than
the traditional direct form FIR filter. Ting et al. G.
R. Hemantha et al. [16], proposed block RAM based
conversion model where critical path optimization is
difficult and may be implemented with retiming
technique. According to Meher and Park [17]
employed two-level pipelining to boost the
throughput of the FIR filter used in the adaptive
algorithm at the expense of a modest increase in
area and latency.
Meher [18] looked at several pipelining options for
DSP circuits. By pipelining to structural adders,
Mathias et al. [19], enhanced the CPD of transposed
form FIR filters at the price of more sophisticated
hardware. Using a modified transposed form FIR
filter, Pramod and Shahana, [20], created a high
throughput adaptive filter design. In [21], it is
addressed how the retiming transformation approach
may be used to create low-power VLSI designs.
Burhan et al [22], adopted Constant Matrix Vector
Multiplication (CMVM) instead of conventional
multiplication, where delay can be optimized by
using pipelining method. Using retiming and
redesigned CSLA-based adders, Pramod and
Shahana [23], created two types of high throughput
FIR filters.
The rest of this paper is structured as follows. The
Distributed Arithmetic Residue Number System is
explained in section 2. The proposed pipelining and
retiming methods were discussed in section 3. The
results and discussion with a comparison of the
performance of filters are done in Section 4. The
section 5 concludes the work proposed.
2 Distributed Arithmetic – Residue
Number System
One memory-based effective multiplier-less
architecture utilized in the design of FIR filters is
distributed arithmetic (DA) architecture. The bit
length of the input data determines how long it takes
to compute a design in DA architecture. High-speed
digital signal processing may be implemented
utilizing filters that are realized using the residue
number system (RNS). The RNS is a non-weighted
number system that simultaneously divides binary
values with a wider range into several smaller
numbers. By employing these little integers in the
math calculation, the critical route time is decreased.
RNS is an addition, subtraction, and multiplication
carry-free system. RNS is hence capable of
operating at high speed. The following graphic
displays the block diagram of an RNS-based FIR
filter, [24].
In the Residue Number System, the three main
components are Forward Conversion, Arithmetic
manipulations, and Reverse conversion as shown in
figure 1. The Chinese remainder theorem is used to
map back from residue to the binary number during
reverse conversion, and the binary to residue
conversion using modulo operation may be carried
out in the forward converter, [25].
Step– 1: Consider the RNS system to be {2n-1, 2n,
2n+1}.
Let n=3, then the RNS bases become {23-1, 23,
23+1}
= {7, 8, 9}.
Let the moduli set = {7, 8, 9}
Let the input be 7, then the equivalent binary is '111'
then the LUT value chosen from the table 1 is
A0+A1+A2 = 3+11+15 = 29.
Then the RNS Sequence generated will be
|29|7=1
|29|8=5
|29|9=2
i.e., the converted RNS Number is {1, 5, 2}.
Step– 2: Performing the related arithmetic
operations say if it is a filter operation, then
Hence x= {1, 5, 2}, let h = {8, 4, 2, 1}
WSEAS TRANSACTIONS on SYSTEMS and CONTROL
DOI: 10.37394/23203.2022.17.60