You are on page 1of 6

A Comparative Study on 16-bit Digital Adders

Rahul Sajeeth Mohan Joseph James Jose K Joyes Mrs. Niji Mathews
Dept.of Electronics and Dept.of Electronics and Dept.of Electronics and Asst. Professor
Communications Engineering Communications Engineering Communications Engineering ECE Department
Viswajyothi College of Viswajyothi College of Viswajyothi College of Viswajyothi College of
Engineering & Technology Engineering & Technology Engineering & Technology Engineering & Technology
Vazhakulam, Muvattupuzha, Vazhakulam, Muvattupuzha, Vazhakulam, Muvattupuzha, Vazhakulam, Muvattupuzha,
Kerala, India Kerala, India Kerala, India Kerala, India
rahul.sajeeth@gmail.com josephjames253@gmail.com josekjoyes101@gmail.com niji@vjcet.org

Abstract - This report compares the area, delay, II. TYPES OF ADDERS
complexity (in terms of gate count), and the power
A. Ripple Carry Adder
consumption of various 16-bit serial and parallel
The simplest form of adder is the ripple-carry
adders. The adders that were chosen to be studied
adder[2]. An n-bit ripple carry adder consists of n one-bit
includes serial adders like Ripple Carry Adder, Carry
full adders connected in succession. The carry "ripples"
Look-Ahead Adder, and Carry Select Adder, and
from the least significant bit to the most significant bit and
parallel prefix adders like Kogge-Stone, Brent-Kung,
hence it gets its name. Since, the carry in at any stage
Ladner-Fischer, and Han-Carlson. The adders were
depends on the carry out from the previous stage, the
designed using the Verilog hardware description
delay of ripple carry adders is O(N) and increases linearly
language and then simulated, synthesised, and
as the size of the operand is increased. Thus it is essential
virtually implemented using the Xilinx Vivado Design
to look for other alternatives.
Suite. The performance comparision of the adders was
done in Cadence SoC Encounter and all the adders
were virtually implemented on the ZedBoard Zynq
Evaluation and Development Board.
Keywords - Ripple Carry Adder, Carry Look-
Ahead Adder, Carry Select Adder, Kogge-Stone
Adder, Brent-Kung Adder, Ladner-Fischer Adder,
Han-Carlson Adder Fig 1. 4-bit Ripple Carry Adder

I. INTRODUCTION B. Carry Select Adder


Carry select adders consists of two ripple carry
Adders play an important role in almost all digital
adders and a multiplexer. Addition of two n-bit numbers is
systems[1], and the choice of adder changes from one
done with two adders: one time with the assumption that
design to another based on a few performance parameters
the carry in is zero, and the other assuming that the carry
such as
is one[3]. After the two results are calculated, the correct
Power Consumption
sum and the correct carry is selected with the multiplexer
Power Dissipation
once the correct carry of previous stage is known. The
Speed of Operation
number of bits in each carry select block can be uniform,
Area Occupied
or variable. In the uniform case, the optimal delay occurs
Circuit Complexity
for a block size of √N, where N is the size of the operands.
Cost etc.
The carry select adders are simple, but rather fast
Since a single adder is not able to exhibit all of the
compared to the ripple carry adders having a delay of
above mentioned parameters, certain adders usurp others,
O (√N).
based on the requirement of the user. Adders have become
a requisite component in most cognitive systems. Usage of C. Carry Look-Ahead Adder
adders in any circuit, reduces the number of transistors The higher delay in the case of ripple carry adders is
used. Digital adders are mostly used in a computer's ALU due to the carry chain[4]. In carry look-ahead adders, the
to compute addition. Digital calculators use adders for carry signals are calculated in advance, based on the input
arithmetic addition. Microcontrollers use adders in signals. For any bit position i, a carry will be generated if
program counters, timers, etc. Every device that uses some the corresponding input bits are ‘1’ or if the carry-in to
kind of increment, decrement, or arithmetic process that bit was a ‘1’ and at least one of the input bits is ‘1’.
contains adders. From this, a recurrence relation is derived that expresses
carry in to any bit position in terms of the relevant addend E. Brent-Kung Adder
and augend digits and some lower-adder carry. This can Brent-Kung[6] is a simple parallel adder with regular
result in considerable gain in speed. design. It takes into account the problem of connecting the
gates in an economical and regular way to minimize chip
area and design costs along with delay and complexity. It
uses the same idea as Ladner-Fischer but is not directly
applicable because they ignored fan-out restriction. Brent-
Kung assumes existence of gates that compute a logical
function of two inputs in constant time and an output
signal can be divided into two signals in constant time.

Fig 2. 4-bit Carry Select Adder

Fig 5. 16-bit Brent-Kung Adder

F. Ladner-Fischer Adder
Ladner-Fischer[7] provides a general method for
deriving efficient parallel solutions to the fixed-length
version of any problem solved by a finite state transducer.
The Ladner-Fischer adder computes prefixes for the odd
numbered bits and uses one more stage to ripple into the
even positions. Cells at high-fanout nodes must still be
sized or ganged appropriately to achieve good speed.
Fig 3. 4-bit Carry Look-Ahead Adder

D. Kogge-Stone Adder
Kogge-Stone[5] uses a technique called recursive
doubling in an algorithm for solving a large class of
recurrence problems on parallel computers. Recursive
doubling involves the splitting of the computation of a
function into two equally complex sub functions whose
evaluation can be performed simultaneously. The Kogge-
Stone adder has log 2N stages and a fanout of 2 at each
stage. This comes at the cost of many long wires that must
be routed between stages. Fig 6. 16-bit Ladner-Fischer Adder
G. Han-Carlson Adder
Han-Carlson[8] is a new graph representation for
prefix computation that leads to the design of a fast, area-
efficient binary adder. The new graph is a combination of
Brent-Kung and Kogge-Stone for prefix computation, and
its area is close to known lower bounds on the VLSI area
of parallel prefix graphs. HanCarlson adders perform
Kogge-Stone on the odd numbered bits, and then use one
more stage to ripple into the even positions. They claim it
to be the fastest possible area-efficient VLSI adder.
Fig 4. 16-bit Kogge-Stone Adder
Fig 9. Synthesis Result of Ripple Carry Adder

Fig 7. 16-bit Han-Carlson Adder

III. SYNTHESIS AND SIMULATION RESULTS

This section deals with the synthesis and simulation


results of each of the adders mentioned in Section II.
Synthesis and simulation were both done on the Xilinx
Vivado Design Suite software. Two testbenches were
considered for all the adders to check the design code and
obtain the simulation results. The testbenches used were,
a = 1010 1010 1010 1010
b = 1111 0000 1111 0000
cin = 0
a = 1010 1010 1010 1010
b = 1111 0000 1111 0000
cin = 1
The expected results for testbench case 1 were,
sum = 1001 1011 1001 1010
cout = 1
The expected results for testbench case 2 were, Fig 10. Synthesis Result of Carry Select Adder
sum = 1001 1011 1001 1011
cout = 1
The simulation results are shown in Fig. 8. After
successful simulation of the designs, they were
synthesised onto the ZedBoard Zynq Development and
Evaluation Board. The synthesis results are shown in
Figs. 9, 10, 11, 12, 13, 14, 15.

Fig 11. Synthesis Result of Carry Look-Ahead Adder

Fig 8. Simulation Results


IV. PERFORMANCE ANALYSIS RESULTS
This section deals with the performance analysis
results of the various adders mentioned in Section II.
Performance analysis of the adders was done using
Cadence SoC Encounter software. The result of the
performance analysis is shown in Tables I, II, III, IV.

TABLE I
PRE- & POST-APR DELAY COMPARISION

Fig 12. Synthesis Result of Kogge-Stone Adder TABLE II


GATE COUNT COMPARISION

TABLE III
AREA OF LAYOUT COMPARISION

Fig 13. Synthesis Result of Brent-Kung Adder

TABLE IV
POWER CONSUMPTION COMPARISION

Fig 14. Synthesis Result of Ladner-Fischer Adder

Fig 15. Synthesis Result of Han-Carlson Adder


V. CONCLUSIONS REFERENCES

Among all the adders studied, the ripple carry adder [1] ‘Digital Design : Principles and Practices’, John F. Walkerly,
Third Edition, Published by Pearson Education, 2003
was found to have the highest delay and the lowest gate
[2] ‘Ripple Carry Adder Design Using Universal Logic Gates’,
count, which is to be expected as the ripple carry adder is Patel Chandrahash and C.S. Veena, Department of ECE,
the simplest form of an adder. The delay of the carry Technocrats Institute of Technology, Bhopal, MP INDIA,
select adder and the carry look-ahead adder was found to Research Journal of Engineering Sciences, Vol. 3(11), 1-5,
November (2014), ISSN 2278 – 9472
be considerably lower than that of the ripple carry adder.
[3] ‘Design of High-Speed Ladner-Fischer Based Carry Select
This is to be expected as both these adders are upgrades Adder’, Pakkiraiah Chakali, Madhu Kumar Patnala,
made upon the simple ripple carry adder. However, both International Journal of Soft Computing and Engineering
these adders have more complex circuits since it uses two (IJSCE), ISSN: 2231-2307, Volume-3, Issue-1, March 2013
[4] ‘Area Efficient Design Analysis of Carry-Look Ahead
blocks of ripple carry adders to perform addition and they Adder’, Anku Bala, Rajesh Mehra, International Journal of
also have a multiplexer to choose the correct sum and Computer Applications (0975 – 8887), Volume 119 – No.20,
carry. Parallel-prefix adders were found to have June 2015
significantly lower delays than their serial adder [6] P. M. Kogge and H. S. Stone, “A Parallel Algorithm for the
Efficient Solution of a General Class of Recurrence
counterparts. Among the parallel adders, the Kogge-Stone Equations,” IEEE Transactions on Computers, Vol. C-22,
adder was found to have the least delay. However, the No. 8, pp. 783-791, August 1973
Kogge-Stone adder occupies the most area and also has [7] R. P. Brent and H. T. Kung, “A Regular Layout for Parallel
Adders,” IEEE Transactions on Computers, Vol. C-31, No.
the highest complexity amongst the parallel adders. This is
3, March 1982
to be expected as the Kogge-Stone adder has the most [8] R. E. Lardner and M. J. Fischer, “Parallel Prefix
parallel prefix operations. The Brent-Kung adder occupies Computation,” Journal of the ACM, Vol. 27, No. 4, pp. 831-
the least area because it has very few parallel prefix 838, October 1980
[9] T. Han and D.A. Carlson, “Fast Area-Efficient VLSI
operations, however, it is found to have the worst delay
Adders,” 1987 IEEE 8th Symposium on Computer
amongst the parallel adders, as it has the most Arithmetic, pp.49-56, May 1987
computation stages. The Han-Carlson and Ladner-Fischer
adders have delays and areas between that of the Kogge-
Niji Mathews was born in Kerala,
Stone and Brent-Kung adders. The delay of Han-Carlson
India. She obtained her Bachelor
and Ladner-Fischer adders is almost the same even though
of Technology (BTech) from
the Han-Carlson adder has an extra computational stage.
Cochin University in 2004. She
The area occupied by the Han-Carlson adder is slightly
did her post-graduation from Anna
less than the Ladner-Fischer adder.
University, Coimbatore in 2010.
As a result, we can conclude that, the Kogge-Stone
Currently, she is working as Asst. Professor,
adder should be used if delay needs to be minimised with
Department of Electronics and Communication
no constraint on area. The Brent-Kung adder should be
Engineering, Viswajyothi College of Engineering
used if area needs to be minimised but some delay is
and Technology, Vazhakulam, India. Her special
tolerable. If delay and area both needs to be optimised,
fields of interest include VLSI and image
then the Han-Carlson adder needs to be used. These
processing.
results support the statement that the Han-Carlson adder is
the fastest possible area efficient adder.
Rahul Sajeeth Mohan was born
in Dubai, UAE. He is currently
doing his Bachelor of Technology
(BTech) in Electronics &
Communications Engineering at
Viswajyothi College of Engin-
eering and Technology, Vazhakulam, India.
Joseph James was born in
Kerala, India. He is currently
doing his Bachelor of Technology
(BTech) in Electronics &
Communications Engineering at
Viswajyothi College of Engin-
eering and Technology, Vazhakulam, India.

Jose K Joyes was born in Kerala,


India. He is currently doing his
Bachelor of Technology (BTech)
in Electronics & Communications
Engineering at Viswajyothi
College of Engineering &
Tec- India.
hnology, Vazhakulam,

You might also like