Performance bounds for unequally punctured terminated convolutional codes

IJRET: International Journal of Research in Engineering and TechnologyeISSN: 2319-1163 | pISSN: 2321-7308
__________________________________________________________________________________________
Volume: 03 Issue: 01 | Jan-2014, Available @ http://www.ijret.org 228
PERFORMANCE BOUNDS FOR UNEQUALLY PUNCTURED
TERMINATED CONVOLUTIONAL CODES
Nilofer.Sk
Assistant professor, ECE, Mallareddy engineering college for women, AP, India
Abstract
The main objective of this paper is to discuss about the performance of the punctured convolution codes. Generally the punctured
convolution codes are designed for a protection purpose in CDMA.In this paper a systematic method to construct performance upper
bound for a punctured convolution code with an arbitrary pattern is proposed. At the decoder side perfect channel estimation is
assumed and to represent the relation between input information bits and their corresponding Hamming weight outputs a weight
enumerator is defined. Finally the simulations results provided very well agree with their predicted results with the upper bounds.
Keywords:Convolution code, upper bound, weight enumerator, rate matching.
---------------------------------------------------------------------***----------------------------------------------------------------------
1. INTRODUCTION
Efficiency of spectrum utilization is a very important factor in
wireless communications. Several multiple access schemes are
proposed to share radio spectrum which is finite and
constrained. Therefore available spectrum is efficiently used
by sharing the spectrum simultaneously among several
individual users and this process must be done without severer
degradation. And most of the wireless communications
requirehigh spectral efficiency. Therefore for achieving high
spectral efficiency definitely there is a need for high rate codes
which maintains the system requirements such as SNR, BER
and FER (frame-error rate) etc. punctured convolutional codes
are high rate codes which are used to achieve high spectral
efficiency [1]. Puncturing is a well known and most suitable
technique to achieve high data rates. In the process of
puncturing a CC (Convolution Code) can be extracted from
mother code with less code rate. At receiver side in two
subsequent separate processing steps equalization and
decoding are performed [1]. Whenever the equalization is
completed bit probability ½ is inserted before performing
decoding when the symbols punctured. To generate higher
data rates some of the coded symbols are punctured
periodically in convolutional punctured codes and the process
of periodic puncturing is not an optimum method. In this
project, initially based on the transfer function of the Viterbi
decoded convolution code and assumption of that input
sequence is infinite with equal error rates the convolutional
bounds will be obtained. However center position bits contains
higher error rate than the end position bits. Similarly if the
frame length decreases gradually then the bound of BER
becomes loose and FER (Frame Error Rate) upper bound was
designed for terminated trellis code. But it is difficult to
achieve upper bounds for terminated CCs using conventional
transfer function.
A systematic method is proposed to obtain the performance
upper bounds for a terminated CC. for obtaining performance
upper bounds a weight enumerator is defined to represent the
relation between input information bits and their
corresponding Hamming weight outputs. A modified trellis
diagram is proposed to compute the weigh enumerator [4]-
[5].Remaining of this paper arranged as: section II clearly
discusses about system model of the proposed method. Section
IV consists of brief explanation about proposed bounds where
as section IV consists of analysis of results. Finally in section
V we conclude this paper.
2. SYSTEM MODEL
This part of the paper presents a system model for terminated
CC. we assume an arbitrary puncturing. Assume a frame with
1L information bits with 2L zeros as tail bits. Therefore the
input of the convolutional encoder is 21 LL  bits and these
inputs bits are encoded at a rate of 1/n.
Therefore after the process of convolutional encoding the
number of convolutional coded bits is
)LL.(nN 21s  (1)
sN is the total number of convolutional coded symbols.
For generating required length of frame and matching rate
some of the coded symbols are punctured. The symbol
position which is going to puncture isdefined using puncturing
pattern as
]p,p,p[P sN21  (2)

__________________________________________________________________________________________
jp = 0, jth
coded is symbol is punctured
If after convolutional process pN symbols are punctured then
the effective code rate er is defined as
p21
1
e
N)LL.(n
L
r

 (3)
In practical communications it is not possible to puncture the
coded symbols periodically due to the restrictions on number
of symbols. Therefore for getting a required rate matching
aperiodical puncturing is required.
3. PROPOSED UPPER BOUNDS
As discussed previously with an arbitrary puncturing pattern
of a terminated CC average BER of selected input symbol and
upper bounds on FER are established in this section. To obtain
the proposed bounds weight enumerator is defined. Therefore
to obtain upper bounds single error event code words are
considered for weight enumerator.
The weighting vector is defined as
]ww,w[W 1L21  (4)
wj= 1, means that jth
bit is considered for weight enumerator.
Therefore the weight enumerator of the convolutional code for
a puncturing pattern P with weighing vector W is defined as
   
i j
ji
j,ip,w DB.p,wcD,BT (5)
)p,w(c j,i : Total number of composed code words.
J: hamming weight of the output.
i: hamming weight of the input
The average Bit Error Rate of the bits can be obtained from
the weight enumerator as
   



freedj i
jj,ib P).p,w(c.i
w
1
p,wP (6)
|w|: w weighing vector’s hamming weight.
dfree: code free distance.
Therefore the upper bound on the average BER is obtained by
allowing only one vector for the weighing vector in equation 6
 



freedj i
jj,i
1
b P).p,w(c.i
L
1
P (7)
Similarly upper bound on FER is obtained by using weigh
enumerator with an all zero vector for w
 



freedj i
jj,iFE P).p,w(cP (8)
3.1Procedure for Weight Enumerator for Proposed
Bounds
This method is purely based on trellis diagram of the
terminated CC and chooses a CC encoder with the 2M
number
of states and the trellis depth t is St (St=0,1,2…..2M
-1) and for
the terminated CC the modified trellis diagram is as:
 At trellis depth‘t’, if any branch is merging into the ‘0’
state then define another state St=2M
and connect every
new state to a trellis depth t-1 with a gain of 1.
 Similarly if any branch is diverging 0 states then connect
0 states from t-1 to t.
 More than two branches are merged using an adder.
The branch gain between the St-1 and St states is defined as
GSt-1, St (B, D) and it modified as per puncturing pattern as
)S,S(y)S,S(x
SS
t1tt1t
t1t
DB)D,B(G 


(9)
B: input Hamming weights of x(St-1,St)
D: input Hamming weights of y(St-1,St)
For the state StFst(B, D) is the temporary polynomial. From the
trellis depth 1 the weigh enumerator is obtained by initializing
Fs0 (B, D) to 1 if S0 is 0.
)D,B(G.)D,B(F)D,B(F t1t
tS1t
tt S,S
S
1SS 


 (10)
tS
 : set of the states in trellis depth t-1
Therefore now find the computation for the trellis depth t+1
by increasing all the trellis states at depth at t by 1. And repeat
this for the trellis depth L1+L2.
Repeat this iteration until the trellis depth L1 + L2. After the
completion of iteration the required weight enumerator is
)D,B(2F M
LS 21L


__________________________________________________________________________________________
Here the proposed modified trellis diagram method consists
the of (L1+L2).2M
complexity additions. This proposed new
approach is applicable to almost all practical terminated CCs
decoded with a Viterbi decoder.
For understanding the proposed approach consider an
modified trellis diagram for a rate ½ CC C1 with L1=5 and
L2=2 as shown in figure 1 and the generation matrix of C1 is
[x2
+1, x2
+x+1]
Fig 1: Modified trellis diagram for convolutional code C1.
For obtaining the upper bound on average BER assume that all
the puncturing and weighing vectors are set to 1 which means
that coded symbols are not punctured and the weight
enumerator is obtained as
958473625
1,1 DBDB5DB8DB7BD5)D,B(T  (11)
By using above equation (11)the upper bound on average BER
is
98765b PP4P8.4P8.2PP  (12)
The transfer function for the CC C1is computed as
 7362515
DB4DB2DB)BD21/(BD)D,B(T
(13)
The conventional upper bound on BER is computed for this
code as
 8765b P32P12P4PP
(14)
Therefore the upper bound for considered example and
simulation results are shown in figure 2. And it is clear that the
proposed bound is tighter than conventional bound. Thereforer
the by setting the corresponding ith
component wi of weighing
vector, upper bound for the ith
information bit is calculated as
wi=1 for j=i, oterwise 0
Fig 2: BER performance of convolutional code C1.
For the terminated CC C2, for the purpose of rate matching the
weight enumerator is obtained before the puncturing and it is
as


206205
20418318218
1,1
DB21DB42
DB99DB53DB54BD29)D,B(T
(15)
Due to the puncturing the performance of the CC degraded.
With the end position symbols are punctured with the
puncturing pattern p2
. The weight enumerator of the CC after
the puncturing is defined as

__________________________________________________________________________________________




127126
125124123122121
117116115114113
11211103102
p,1
DB2DB4
DBDB2DB5DB2DB3
DBDBDB3DBDB4
DB3D5DBDB2)D,B(T 2
(16)
Therefore by comparing equation 15 and 16 it is clear that
distance is decreased from 18 to 10 with the puncturing.
Therefore the upper bound on FER and average BER are as
)P99P51P7(
29
1
P
P22P14P3P
121110b
121110FE




Therefore by applying the proposed method on CC C2, the
performance results are shown in Fig 3-Fig 6. Here C2 is
terminated convolutional code in the second part of HS-SCCS.
And the coded output Hamming distances are less than
dmin+10, where mind is the minimum distance between the
codewords.
4. RESULTS
Fig.3: BER performance of C2 for different bit positions
(before end puncturing).
Above figure represents the simulation results for CC C2 and
upper bound on BER before end puncturing. And at the center
position bits the error rate is 4 times larger than those of end
position bits. At Eb/N0 = 2.0 dB and 4.0 dB, the 15-th
information bit error rate is 4.3 and 4.0 times larger than that
of the first information bit, respectively
Fig. 4:BER performance of C2 for different bit positions (after
end puncturing).
Similarly fig 4 represents the simulation results of CC C2 and
upper bound on BER after end puncturing. From the results it
is clear that depending on the position of the bit in the frame
the variation of the BER is small with low Eb/N0. Since the
minimum distance of the end position bits are decreased
compared to center position bits the BER of the center
position bits is low.
Fig 5:BER performance of C2 before end puncturing.

__________________________________________________________________________________________
Figure 5 shows simulation results for the initial and 15th
information bits in C2 before the puncturing. The performance
difference present between initial and 15th
bit is 0.6dB.And
with the increment in Eb/N0 the performance difference is
maintained regularly.
Fig.6:BER performance of C2 after end puncturing.
Similarly figure 6 represents simulation results for the
information bits of 17th
and 28th
positions. Here the 17th
bit
shows the best BER performance where as 28th
bit shows the
worst BER performance. At low Eb/N0 both the bits are similar
to each other. When compared to 17th
bit, the 28th
bit
performance increases more slowly with increment in Eb/N0.
CONCLUSIONS
Using this project performance of upper bounds for a
terminated CC with arbitrary puncturing is proposed and
weight enumerator is proposed to compute proposed bounds
which define the relation between the input information bit
and their corresponding Hamming code outputs. Using the
coefficients of weight enumerator the upper bounds on
average BER and on FER are computed. Here the proposed
method is based on modified trellis diagram by considering
the puncturing pattern and finally the results show that the
proposed bounds are tighter to the actual performance of
terminated CCs.
REFERENCES
[1]. A. J. Viterbi, “Convolutional codes and their performance
in communication systems," IEEE Trans. Commun. Tech., vol.
COM-19, no. 5, 751-772, Oct. 1971.
[2]. R. J. McEliece, The Theory of Information and Coding - A
Mathematical Framework for Communication. Boston, MA:
Addison-Wesley, 1977.
[3]. C. Berrou, A. Glavieux, and P. Thitimasjshima, “Near
Shannon limit error-correcting coding and decoding: turbo
codes," in Proc. IEEE International Conf. Commun., 1993, pp.
1064-1070.
[4]. C. Berrou, “The ten-year-old turbo codes are entering into
service,"IEEE Commun. Mag., vol. 41, no. 8, pp. 110-116,
Aug. 2003.
[5]. 3GPP, “Multiplexing and channel coding (FDD)," 3GPP
TS25.212, v. 6.8.0, June 2006.
[6]. J. Cain, G. Clard, and J. Geister, “Punctured convolutional
codes ofrate (n − 1)/n and simplified maximum likelihood
decoding," IEEETrans. Inform Theory, vol. 25, no. 1, pp. 97-
100, Jan. 1979.
[7]. J. Hagenauer, “Rate compatible punctured convolutional
codes andtheir applications," IEEE Trans. Commun., vol. 36,
no. 4, pp. 389-400,Apr. 1988
[8]. Y. Yasuda, K. Kashiki, and Y. Hirata, “High rate
punctured convolutional codes for soft decision viterbi
decoding," IEEE Trans. Commun., vol. 32, no. 3, pp. 315-319,
Mar. 1984.
[9]. G. Caire and E. Viterbo, “Upper bound on the frame error
probability of terminated trellis codes," IEEE Commun.Lett.,
vol. 2, no. 1, pp. 2-4, Jan. 1998.
[10]. K. M. Rege and S. Nanda, “Irreducible FER for
convolutional codes with random bit puncturing: application
to CDMA forward channel," in Proc. IEEE Veh. Technol.
Conf., 1996, pp. 1336-1340.
[11]. H. Moon, “Improved upper bound on bit error rate for
truncatedconvolutional codes," IEE Electron. Lett., vol. 34,
no. 1, pp. 65-66,Jan. 1998
[12]. H. Moon and D. C. Cox, “Improved performance upper
bound forterminated convolutional codes," IEEE Commun.
Lett., vol. 11, no. 6,pp. 519-521, June 2007.
[13]. H. Moon and D. C. Cox, “Performance upper bounds for
terminated convolutional codes," in Proc. IEEE Wireless
Commun.Networking Conf., 2003, pp. 252-256.
BIOGRAPHIES
Ms .Shaik. Nilofer is working as an Assistant
Professor in Mallareddy engineering College
for women, Hyderabad. She has 3
yearsteaching experience at UG level.

Performance bounds for unequally punctured terminated convolutional codes

More Related Content

What's hot (20)

Viewers also liked (18)

Similar to Performance bounds for unequally punctured terminated convolutional codes (20)

More from eSAT Journals (20)

Recently uploaded (20)

Performance bounds for unequally punctured terminated convolutional codes