You are on page 1of 4

Turbo Code verses Convolutional Code

Yousef Hwegy and Manal Saad


Tripoli university
y_hwegy@yahoo.com
ma_saad08@yahoo.com

ABSTRACT to correct any errors in the received data


stream[2].
Shannon and Hamming in 1949 have given a powerful
techniques for achieving reliable data transmission 2. Convolutional Coding
using coding techniques.
In this paper, we analyze the performance of two Convolution code[1,2,3,4] are specified by three
coding techniques, which are the Convolutional code parameters (n,m,k) where, n=number of output
and the Turbo code. The Convolutional code with
bits, m=number of inputs bits, and k is the
Viterbi decoder soft and hard decisions, the
turbo code with random interleaver, and the channel is number of shift registers. The quantity k/n is
the AWGN, are all considered for the simulation in called the code rate which is a measure of the code
this work. efficiency. The constraint length of the code is
For the same message length and the same channel given by L=k(m-1).
distortion the coding techniques considered in this The constraint length of the code represents the
work are compared. This paper demonstrate that Turbo number of bits in the encoder memory that effects
code exhibit a better performance in comparison to the the generation of the n output bits. One important
Convolutional code especially at low signal to noise feature of Convolutional coding is the constraint
ratio. The Convolutional coding using Viterbi decoder length generator polynomial. This acts as trellis
soft decision provides coding gain of 2 dB in structure that provides coding gain, which does
comparison to Viterbi decoder hard decision. Also the
codec system exhibit a far better performance in
not available in other coding schemes.
comparison to non codec system that uses modulated
QPSK only.
2.1 Soft Decision Viterbi Decoding
Keywords: Turbo code, random interleaver, Viterbi
For rate binary Convolutional code system, the
decoder, Coding theory, Digital Signal Processing
demodulator delivers two code symbols at a time
1. Introduction
to decoder. With hard decision (2-levels)
Due to a wide applications of internet technology
decoding, each pair of received code symbols can
and multimedia digital communication,
be drawn on a plane, as shown in figure 1a, at
development of reliable and high speed
each corner of square. The corners are labeled as
transmission methods over wireless channels has
(0,0), (1,0), (0,0), (1,1), represented four possible
gained high importance in recent years. This
hard decision values that the two code symbols
demand has been accelerated by the emergence of
might have. While the 8-levels soft decision
large-scale, high-speed data transmission.
decoding, each pair of code symbols can be
In order to improve the performance and the
similarly represented on an equally spaced 8-
reliability of high-speed data transmission forward
levels by 8-levels, as a point from the set of 64
error correcting code (FEC) is used[1,2,3]. Error
points shown in figure 1b.
correction codes are generated in a FEC encoder
The primary difference between two kinds of
for the data stream and are sent to a receiver
Viterbi decision decoding (soft decision and hard
including a FEC decoder. The FEC decoder
decision), is that the hard decision algorithm uses
recovers the error correction codes and uses them
Hamming distance metric which is limited

ISBN: 978-0-9891305-4-7 2014 SDIWC 8


resolution, while soft decision uses Euclidean placed on the solid line. While The metric ? ,
distance metric with more resolution. representing the Euclidean distance between the
point (? , ? ) and the point (7,7) code symbol is
placed on the dashed line. With comparing to
figure 3.5, encoder trellis diagram as describes
hard decision, with the branch words transformed
from binary to octal, as describes soft decision[4].
Figure 1 (a) Hard decision plane (b) Soft decision plane
3. Turbo Codes
In figure 2 is the point (? , ? ), representing a pair Turbo Codes [5,6,7,8,9,10], or parallel
of noisy code-symbol values that might stem from concatenated Convolutional codes, represent
a demodulator. Then Euclidean distance between a way of concatenating two simple codes to
the noisy point (? , ? ) and noiseless point (0,0) on obtain an overall powerful code. After their
the plane as ? (? 0)2 +(? 0)2 = ? , while invention in 1993, The basic idea in Turbo
the Euclidean distance between the noisy point Coding is to concatenate two recursive
(? , ? ) and noiseless point (7,7) as systematic Convolutional codes in parallel
via an interleaver. The information
? (? 7)2 +(? 7)2 = ? . sequence is divided into blocks of certain
length.
The input of the first encoder is the
a

information block and the input of the


b second encoder is an interleaved version
of the information block. The encoded
sequence ,i.e., the codeword, corresponding
to that information sequence is then the
Figure 2 noisy point for soft decision information block itself, the first parity
block and the second parity block. The
Figures 3a and 3b show the first section of an
block diagram of the rate 1/3 Turbo Code
encoder and decoder Trellis diagram, where the
with 5/7 component Convolutional codes is
soft decision is performed with the use Euclidean
shown in Figure 1. To obtain a good
distance,
performance in Turbo Codes, the component
b
codes should be chosen as Convolutional
a
codes with feedback (recursive). Also the
Convolutional codes chosen should be
systematic. If not, then the non-systematic
Convolutional codes tend to be catastrophic.
A catastrophic error is defined as an event
whereby a finite number of code symbol
errors cause an infinite number of decoded
data bit errors. This happens when the feed
Figure 3 (a) Encoder Trellis section (b) Decoder Trellis forward and feedback generator polynomials
section of the constituent Convolutional encoder
have a common factor between them. It is
When the pair of soft decision code symbols with also common practice to choose identical
value (? , ? ) arrive at a decoder during the first component encoders.
duration time as shown above figure. The metric The interleaver can be chosen pseudo-
? , representing the Euclidean distance between randomly to obtain good performance. It
the point (? , ? ) and the point (0,0) branch word is follows a pseudo-random rule to reorder

ISBN: 978-0-9891305-4-7 2014 SDIWC 9


the bits. Hence Turbo Codes appear ran ran- 2- To investigate on the performance of the
dom to the channel, yet have enough Convolutional code with soft and hard decision ,
structure so that decoding can be the set up circuit shown in figure.6 is used.
physically realized. During the decoding Convolutional
nal coder rate with
process if the inputs are correlated then the specification(171,133,120,153) with number of
performance of the decoder deteriorates. memories =6 and Viterbi decoding.
The interleaver decor relates them to give a 3- Figure.7 shows the performance of the
good performance. Hence interleaver design Convolutional codes with soft and hard decision.
is an important aspect of Turbo codes. The 4- Figure.8 shows the performance of the Turbo
performance of Turbo Codes improves code verses the Convolutional code with respect to
when the interleaver length is increased, the un coded system.
but at the expense of greater delay.ay. This is
because for the large block size one has to
wait to get all the bits of the block and then
do the interleaving.

Figure5:: The simulation model

Figure 4:Rate
:Rate 1/3 Turbo Code wit 5/7 component codes

4. simulation and result

In order to evaluate the


he performance of the
different coding techniques over the AWGN
channel, the BER and SNR are considered as the
key performance metrics used in this study. Figure6:: Convolutional code with Hard/Soft decision over
AWGN Channel
Matlab simulator shown in Figure.5 and Figure
Figure.6
were designed and built to investigate the
performance of the different coding techniques
scenarios considered in this paper. 5. Conclusion
The simulation procedure could be summarized Figure.7 shows the Convolutional coding
using the following points: using Viterbi decoder soft decision provides
1- To investigate on the performance of the turbo coding gain of 2 dB in comparison to Viterbi
code, the set up circuit shown in figure.5 is used. decoder hard decision.
The signal from work station simulate the burst Figure.8 shows the superior performance of
error that introduced by the communication
ommunication the coded system with respect to the un coded
channel. system which is obvious and clear. However, the
ISBN: 978-0-9891305-4-7 2014 SDIWC 10
performance of the turbo code is far better than the
Convolutional code.
Hence, all of the goals as set forth in this
paper have been achieved. The verifications have
been met as shown in figure[7,8], and the
requirements have been completed. REFERENCES

[1]C. Berrou, A. Glavieux, and P. Thitimajshima, Near


Shannon limit error-correcting coding and decoding: Turbo-
codes, in Proc. ICC93, Geneva, Switzerland, May 1993,
pp. 10641070.

[2] Theodore S. Rappaport. Wireless Communications


Principles and Practice. 2nd Edition. NewJersey; Prentice
Hall PTR; 2002.

[3] . P. Agrawal, Q. Zeng, Introduction to Wireless and


Mobile Systems, Second Edition, Thomson, 2006.

[4] Frenger, P., P. Orten, and T. Ottosson, "Convolution


Codes with Optimum Distance Spectrum," IEEE
Communications Letters, vol. 3, pp. 317-319, November
1999.

[5] D. Divsalar and R. J. McEliece, The effective free


distance of turbo codes, Electron. Lett., vol. 32, no. 5, pp.
445446, Feb. 1996.

Figure7 Viterbi hard and soft decision over AWGN channel [6] S. Benedetto and G. Montorsi, Unveiling turbo-codes:
Some results on parallel concatenated coding schemes,
IEEE Trans. Inform. Theory,vol. 42, pp. 409428, Mar.
1996.

[7] Gao Yingzi, M.Reza Soleymai and U.Vilaipornsawai,


Turbo Coding for Satellite and wireless Communications,
2002.

[8] Battail, G. A Conceptual Framework for Understanding


Turbo Codes. IEEE Journal on Selected Areas in
0
Communications, 1998.
1
0 Un coded
[9] Fonseka JP. Applications of Turbo Codes in Satellite
Convolution code
Mobile Systems. Electronics letters January 1999.
turbo
1
-1 code [10] Berrou C, Glavieux A & Thitimajshima P. Near
0 Shannon Limit Error-Correcting Coding and Decoding :
BER Turbo Codes. IEEE 1993.

-2
1
0

-3
1
0 0 0. 1 1. 2 2. 3
5 5 in dB
SNR 5
Figure(8): The simulation results

ISBN: 978-0-9891305-4-7 2014 SDIWC 11

You might also like