Design and Implementation of Low Power DSP Core with Programmable Truncated Vedic Multiplier

IJSRD - International Journal for Scientific Research & Development| Vol. 2, Issue 07, 2014 | ISSN (online): 2321-0613
All rights reserved by www.ijsrd.com 561
Design and Implementation of Low Power DSP Core with Programmable
Truncated Vedic Multiplier
Mathew George1
Reneesh C. Zacharia2
1
M.Tech. Scholar 2
Assistant Professor
1
Department of VLSI&ES 2
Department of Electronics & Communication Engineering
1,2
MLMCE, Kottayam
Abstract— The programmable truncated Vedic
multiplication is the method which uses Vedic multiplier
and programmable truncation control bits and which
reduces part of the area and power required by multipliers
by only computing the most-significant bits of the product.
The basic process of truncation includes physical reduction
of the partial product matrix and a compensation for the
reduced bits via different hardware compensation sub
circuits. These results in fixed systems optimized for a given
application at design time. A novel approach to truncation is
proposed, where a full precision vedic multiplier is
implemented, but the active section of the truncation is
selected by truncation control bits dynamically at run-time.
Such architecture brings together the power reduction
benefits from truncated multipliers and the flexibility of
reconfigurable and general purpose devices. Efficient
implementation of such a multiplier is presented in a custom
digital signal processor where the concept of software
compensation is introduced and analyzed for different
applications. Experimental results and power measurements
are studied, including power measurements from both post-
synthesis simulations and a fabricated IC implementation.
This is the first system-level DSP core using a high speed
Vedic truncated multiplier. Results demonstrate the
effectiveness of the programmable truncated MAC
(PTMAC) in achieving power reduction, with minimum
impact on functionality for a number of applications. On
comparison with the previous parallel multipliers Vedic
should be much more fast and area should be reduced.
Programmable truncated Vedic multiplier (PTVM) should
be the basic part implemented for the arithmetic and
PTMAC units.
Keywords: Vedic multiplier, truncated multiplication,
Arithmetic unit, barrel shifter, PTMAC, flexible DSP
I. INTRODUCTION
The portable communication and computing devices and the
advance in mobile multimedia systems has made power
consumption critical to optimize in the design of digital
signal processing architectures. Advances in VLSI has been
one of the major driving forces for the growth of digital
signal processors(DSP), enabling the implementation of
complex algorithms in programmable DSP architectures and
fixed application specific hardware. Within DSP systems,
multipliers are among the most basic building blocks and
parallel implementations, required for high-speed
computation, represent the main component in terms of
power consumption of any hardware.. The optimization of
multipliers in terms of area, power and timing has been
extensively studied in the past. Full or direct multiplier
implementations of an NXN –bit multiplication yields a 2N -
bit product.
In order to keep the full accuracy of the system,
DSP architecture would need an ever-growing bit width that
would be impossible or impractical to implement. To avoid
this, results are usually truncated or rounded down to keep
results within the limits of the architecture bit width. In
order to reduce the parallel multiplier requirements in
systems where an exact result is not required, several
techniques that reduce power consumption at the expense of
lower precision have been presented in the literature. Such
techniques skip the implementation and/or disable parts of
the partial product matrix to trade energy spent by the
computational process for degradation on the output signal.
The architecture presented in this paper, a programmable
truncated Vedic multiplier (PTVM) using Vedic techniques.
The main objective of this thesis is to design and
implementation of low power dsp core that can execute and
manages the operations up to sixteen bit. Instead of using
conventional parallel multiplier, high speed Vedic multiplier
with programmable truncation is used, and also the parallel
shifter design using mux with a simple circuit is used. The
power consumption and speed are the factors considered in
this proposed work. The DSP processor should work on the
basis of some commands based on the instructions. In this
work some additional commands are also included for the
modification of the base paper.
II. LITERATURE SURVEY
Parallel multipliers provide high speed method for
multiplications, but require large area for VLSI
implementation. In most signal processing applications,
rounded product is required to avoid growth in word size.
Thus an important aim is to design a multiplier which
required less area and that is possible with the Vedic
truncated multiplier. In the wireless multimedia word, DSP
systems are ubiquitous Multiplication is the main operation
in many signal processing algorithms hence efficient
multipliers is desirable. A full-width digital n × n bits
multiplier computes the 2n bits output as a weighted sum of
partial products. A multiplier with the output represented on
n bits output is useful, as example, in DSP data paths which
saves the output in the same n bits registers of the input. A
Vedic truncated multiplier is an n × n multiplier with n bits
output. Since in a Vedic truncated multiplier the n less
significant bits of the full-width product are discarded, some
of the partial products are removed and replaced by a
suitable compensation function, to trade-off accuracy with
hardware cost. As more columns are eliminated, the area
and power consumption of the arithmetic unit are
significantly reduced, and in many cases the delay also
decreases
In this work, Urdhva tiryakbhyam Sutra is first
applied to the binary number system and is used to develop
digital multiplier architecture. This is shown to be very

Design and Implementation of Low Power DSP Core with Programmable Truncated Vedic Multiplier
(IJSRD/Vol. 2/Issue 07/2014/127)
similar to the popular array multiplier architecture. This
Sutra also shows the effectiveness of to reduce the NXN
multiplier structure into an efficient 4X4 multiplier
structures. Nikhilam Sutra is then discussed and is shown to
be much more efficient in the multiplication of large
numbers as it reduces the multiplication of two large
numbers to that of two smaller ones. The Multiplier
Architecture is based on the Vertical and Crosswise
algorithm of ancient Indian Vedic Mathematics.
A. Mathematical Basis of Truncated Multiplier
Multiplication is a common requirement in DSP systems
and plays an important role in terms of power consumption.
In those systems where it is not necessary to compute the
exact least significant part of the product, truncated
multipliers achieve power, area and timing improvements by
skipping the implementation of a part of the least significant
part of the partial product matrix. The partial product matrix
is split into two main regions, the least significant part(LSP),
which contains the least significant columns of the partial
product matrix, and the most significant part (MSP), that
includes the most significant columns of partial product
terms. Truncation allows a way of reducing the complexity
of the multiplier unit by discarding the lower parts of the
partial product matrix.
Fig. 1: Subdivision of the matrix of Partial Products, for
unsigned multiplier, n = 8. MSP (Most Significant Part) of
the PPs Matrix is constituted by the first n columns. LSP
(Less Significant Part) of the PPs Matrix is constituted by
the last n columns.
B. Programmable Truncated DSP
In order to analyze the benefits and drawbacks of
programmable truncation from an error and power saving
point of view, a lightweight DSP with a programmable
truncated multiplier was designed, implemented and tested.
The architecture presented in this paper is a custom-
designed DSP structure with minimum control logic and a
multiply and accumulate block as the core of the arithmetic
unit. PTMAC stands for programmable truncated multiply
and accumulate, and it allows the benefits of programmable
truncation to be explored, where different operation modes
can be defined with different power and accuracy
requirements. PTMAC can also be used as a platform to
obtain an estimation of the noise introduced andpower
savings for fixed truncation designs. Fig. 2 shows a
schematic view of the PTMAC architecture. The control unit
operates in a 5-stage pipeline that fetches and decodes
instructions from the program memory and loads data from
the two data memory blocks into the arithmetic unit.
Fig. 2: PTMAC top level diagram
The following is a description of the main
characteristics of the proposed architecture:
1) Control Unit
The Control Unit in PTMAC is a small 5-stage pipeline that
fetches and decodes instructions, and controls the data flow,
arithmetic unit and the IO ports. It allows the user to access
one or both of the data memory blocks while reading
instructions and includes a zero-overhead-loop for the
instruction memory. Circular buffers are included for
efficient access of both memory blocks, again to maximize
the use of the Arithmetic Unit. PTMAC has the ability to
perform one-instruction hardware loops, that can reduce a
significant amount of accesses to the program memory and
to facilitate the storage of intermediate results. PTMAC
includes four general purpose registers within its Control
unit.
2) Custom Instruction Set
The set of instructions implemented for PTMAC has been
designed to maximize the usage of the Arithmetic Unit,
optimizing the power reductions achievable by the
programmable truncated multiplier. Instructions are 32 bits
wide and allow access to one or both memory blocks in a
single instruction. The instruction set is split into different
functional areas.
3) Arithmetic Instructions
It is focused on the arithmetic capabilities of the unit.
Maintaining the MAC as busy as possible during operation
time is key to achieving both efficient implementation of
processing routines, and benefits derived from the
programmable truncated multiplier. The range of
instructions dedicated to arithmetic operations includes
ADD for addition, MULT for multiplication and MAC for
multiply-and-accumulate operations.
4) Flow Control Instructions
Flow Control Instructions implemented on PTMAC allow
the device to perform relative and absolute jumps in the
program and data memory blocks in an independent fashion
5) Looping Instructions
Looping instructions have been implemented aimed at a
reduction in flow-control instructions, maximizing the duty
of the arithmetic unit and achieving faster DSP subroutines.
6) Data Flow Instructions
Data Flow instructions, described in Table IV, make
reference to instructions for storing and loading data and
also inputting and outputting samples to/from the system.

7) Memory Blocks
Memory on the PTMAC architecture is formed by a 1024 X
32 b Program Memory block and two 512 X16 b Data
Memory blocks. All three memories can be accessed on a
single clock cycle, as they are directly connected to the
control unit. Arithmetic instructions such as multiply and
accumulate can be performed in a single clock cycle,
reading data from both data memory blocks while fetching
the next instruction.
8) Arithmetic Unit
The Arithmetic Unit is the core and most distinctive block
of the PTMAC. It consists of a multiply and accumulate
structure with a 16-bit programmable truncated multiplier, a
40-bit adder and accumulator, and a 40-bit barrel shifter for
scaling and rotating the accumulated value. A block diagram
of the Arithmetic Unit is displayed in Fig. 3.The
programmable truncated multiplier operates as a standard
16-bit two’s complement multiplier, with the distinction of
having an extra control input matrix
Fig. 3: Arithmetic unit diagram
III. SCOPE OF THE PROPOSED WORK
Multiplication is frequently required in digital signal
processing. Parallel multipliers provide a high-speed method
for multiplication, but require large area for VLSI
implementations. Thus an important design goal is to reduce
the area requirement of the rounded output multiplier. This
paper presents a method for vedic multiplication which
computes the products of two n-bit numbers and truncate the
lower bits using truncation control bits.Truncated multipliers
can be used in finite impulse response (FIR) and discrete
cosine transforms (DCT). The truncated multiplier shows
much more reduction in device utilization as compared to
standard multiplier. Significant reduction in FPGA
resources, delay, and power can be achieved using truncated
multipliers instead of standard parallel multipliers when the
full precision of the standard multiplier is not required.
Based on this paper the modification done on the
multiplier, the multiplier used in this proposed work should
be high speed vedic multiplier with programmable
truncation.
IV. PROPOSED WORK
In this paper the modification should be planned to done
using high speed Vedic multiplier with programmable
truncation. Multiplication is an important fundamental
function in arithmetic operations. Multiplication-based
operations such as Multiply and Accumulate(MAC) and
inner product are among some of the frequently used
Computation- Intensive Arithmetic Functions(CIAF)
currently implemented in many Digital Signal Processing
(DSP) applications such as convolution, Fast Fourier
Transform(FFT), filtering and in microprocessors in its
arithmetic and logic unit . The multiplication dominates the
execution time of most DSP algorithms, so there is a need of
high speed multiplier.
In this work, Urdhva tiryakbhyam Sutra is first
applied to the binary number system and is used to develop
digital multiplier architecture. This is shown to be very
similar to the popular array multiplier architecture. This
Sutra also shows the effectiveness of to reduce the NXN
multiplier structure into an efficient 4X4 multiplier
structures.. The Multiplier Architecture is based on the
Vertical and Crosswise algorithm of ancient Indian Vedic
Mathematics.
A. 4x4 Bit Vedic Multiplier
Algorithm for 4 x 4 bit Vedic multiplier Using Urdhva
Tiryakbhyam (Vertically and crosswise) for two Binary
numbers
CP = Cross Product (Vertically and Crosswise)
X3 X2 X1 X0 Multiplicand
Y3 Y2 Y1 Y0 Multiplier
--------------------------------------------------------------------
H G F E D C B A
---------------------------------------------------------------------
P7 P6 P5 P4 P3 P2 P1 P0 Product
Fig. 4: Line diagram for multiplication of two 4 - bit
numbers.
B. Proposed 16x16 Bit Vedic Multiplier
Fig. 5: 16X16 Bits Vedic Multiplier

Fig. 6: Truncation control bit added to output and discarded
LSP (trunc_cntrl=1111111100000000)
The paper emphasizes an efficient 40-bit MAC
architecture a and results are presented .The efficiency in
terms of area and speed of proposed MAC unit architecture
is observed through reduced area, low critical delay and low
hardware complexity. The proposed MAC unit reduces the
area by reducing the number of multiplication and addition
in the multiplier unit. Increase in the speed of operation is
achieved by the hierarchical nature of the Vedic multiplier
unit. The PTMAC consist of 16 bit programmable truncated
vedic multiplier, 40bit SQRT carry select adder, 40 Bit
barrel shifter and 40 bit accumulator unit. PTMAC is a small
5-stage pipeline that fetches and decodes instructions, and
controls the data flow, arithmetic unit and the IO ports. It
allows the user to access one or both of the data memory
blocks while reading instructions and includes a zero-
overhead-loop for the instruction memory. Circular buffers
are included for efficient access of both memory blocks,
again to maximize the use of the Arithmetic Unit. PTMAC
has the ability to perform one-instruction hardware loops,
that can reduce a significant amount of accesses to the
program memory and to facilitate the storage of
intermediate results.
V. TOOL USED
In this work, while the algorithms have been implemented in
Verilog and logic simulation done in Xilinx 14.2. After gate-
level synthesis from high level behavioral and/or structural
RTL HDL codes, basic schematics have been optimized for
speed and area using Xilinx family of device SPARTAN 3E
Speed Grade : -4.
VI. SIMULATION RESULTS
Fig. 7: RTL schematic for control unit including PTVM
Fig. 9: Simulation result of control unit.
Fig. 10: Simulation result of modified DSP core.
VII. COMPARISON WITH PROPOSED WORK AND PREVIOUS
WORK
Proposed Work Area Delay(Ns)
Programmabe
truncated vedic
multiplier
No. of 4 input LUT’s
-14%
No. of occupied
slices - 15%
No. of bonded IOBs-
87%
45.153
40 bit CSA
-2%
No. of occupied
slices - 2%
No. of bonded IOBs -
65%
18.084
Previous Work
Programmabe
truncated parallel
multiplier
-16%
No. of occupied
slices - 17%
No. of bonded IOBs-
87%
65.614

40bit riple carry adder
-1%
No. of occupied
slices - 2%
No. of bonded IOBs -
112% overmapped
54.206
VIII. CONCLUSION
In this paper, a new architecture to perform truncated
multiplication has been presented and compared against
previous implementations. It has been shown that Vedic
truncation implementation is much faster and area is
reduced than the other parallel multiplier. A number of real
algorithms implemented on the device show the actual
benefits of using truncated multiplication in a practical
scenario. The carry select adder is used here for the fast
addition and it is proved. The barrel shifter completely
designed as single module using mux and it should be more
fast and less power. From the above comparisons and
analysis it is clear that the modified DSP core and its
modules are better than the previously implemented works
and also understood that the modified DSP core is a low
power DSP core with programmable truncated Vedic
multiplier. Since the the power comparison of Modified
DSP block with previous implemented DSP block is done
using Xpower Analyser of Xilinx ISE14.2.
REFERENCES
[1] Manuel de la Guia Solaz, and Richard Conway,”
AFlexible Low Power DSP With a
ProgrammableTruncated Multiplier”, IEEE
Transactions On Circuits And Systems—I: Regular
Papers, Vol. 59, No. 11,November 2012.
[2] http://www.altera.com/products/fpga.html
[3] Delay and Power Comparison of Adder Topologies”,
International Journal of VLSI design &
Communication Systems(VLSICS) Vol.3, No.1,
February 2012.
[4] Low-Power and Area-Efficient Carry Select Adder
IEEE TRANSACTIONS ON VERY LARGE
SCALE INTEGRATION (VLSI) SYSTEMS, VOL.
20, NO. 2,FEBRUARY 2012
[5] Design And Implementation Of Low Power
Multiplier Using Vedic Multiplication Tech-nique,”
International Journal of Computer Science and
Communication (IJCSC) Vol. 3, No. 1, January-June
2012.
[6] VHDL Implementation of Fast 32X32 Multiplier
based on Vedic Mathematics, International Journal of
Engineering Technology and Computer Applications
Vol.2, No.1, Apr 2012
[7] Performance Evaluation of Squaring Operation by
Vedic Mathematics”, IETE Journal of Research,
57(1), Jan-Feb 2011.
[8] R.UMA,Vidya Vijayan, M. Mohanapriya, Sharon
PaulArea, Delay and Pow-er Comparison of Adder
Topologies”, International Journal of VLSI design &
February 2012 DOI 10.5121/vlsic.2012.3113 153
[9] Performance Evaluation of Squaring Operation by
Vedic Mathematics”, IETE Journal of Research,
57(1), Jan-Feb 2011.
[10]R.UMA,Vidya Vijayan, M. Mohanapriya, Sharon
PaulArea, Delay and Pow-er Comparison of Adder
Topologies”, International Journal of VLSI design &
February 2012 DOI 10.5121/vlsic.2012.3113 153
[11]Jagadguru Swami Sri Bharati Krisna Tirthaji
Maharaja,“Vedic Mathematics Sixteen Simple
Mathematical Formulae from the Veda”, Delhi 1965.

Design and Implementation of Low Power DSP Core with Programmable Truncated Vedic Multiplier

More Related Content

What's hot (16)

Similar to Design and Implementation of Low Power DSP Core with Programmable Truncated Vedic Multiplier (20)

More from ijsrd.com (20)

Recently uploaded (20)

Design and Implementation of Low Power DSP Core with Programmable Truncated Vedic Multiplier