Journal of Physics: Conference Series

PAPER • OPEN ACCESS

Performance analysis of LDPC decoding algorithm

To cite this article: Fan Zhou et al 2020 J. Phys.: Conf. Ser. 1453 012026

View the article online for updates and enhancements.

This content was downloaded from IP address 185.201.128.151 on 04/03/2020 at 13:02

CISAI 2019 IOP Publishing
Journal of Physics: Conference Series 1453 (2020) 012026 doi:10.1088/1742-6596/1453/1/012026

Performance analysis of LDPC decoding algorithm

Fan Zhou1*, Linlin Niu2 and Binghe Tian3

1
School of Information Science & Engineering,Shenyang Ligong University,
Shenyang 110000, China
2
School of Information Science & Engineering,Shenyang Ligong University,
Shenyang 110000, China
3
School of Information Science & Engineering,Shenyang Ligong University,
Shenyang 110000, China
*
Corresponding author’s e-mail: zhoufan@sylu.edu.cn

Abstract. The LDPC code is the guarantee of reliable information transmission channel coding
technology, compared with other coding technology, LDPC code has excellent performance,
LDPC decoding algorithm performance is a key, in order to find out the performance of
decoding algorithm, compare the BF algorithm and WBF algorithm, BP algorithm and BP
algorithm LLR, MS algorithm, NMS algorithm performance, the results show that NMS
algorithm has lower complexity and better performance.

1. Introduction
With the development of modern communication technology, LDPC codes have become a hotspot in
coding research. Theoretical research and experiments show that LDPC codes have better performance
and approach the Shannon limit under certain conditions. More and more people are studying LDPC
codes[1]. Although LDPC decoding The algorithm has also been continuously improved. The LDPC
decoding algorithm with low complexity and good performance is facing great challenges. Therefore,
comparing and analyzing the advantages and disadvantages of different decoding algorithms has an
important role in practical applications.
Hard decision decoding algorithms and soft decision decoding algorithms are the current research
focus of decoding algorithms. Bit flip (BF) and weighted bit flip (WBF) algorithms belong to hard
decision decoding algorithms, and the amount of operations in the decoding process is low[2].
However, the decoding performance is poor. The most traditional soft decision decoding is the belief
propagation (BP) algorithm [3]. To solve the problem of high complexity of the BP algorithm,
proposed to replace the log of the probability message in the BP algorithm with the likelihood log ratio.
Likelihood Ratio Confidence Propagation (LLR BP) algorithm, proposed a minimum sum (MS)
decoding algorithm that updates the LLR BP algorithm and checks the nodes, and the minimum sum
(MS) algorithm performs a complex sum operation. The approximate value is used instead to reduce
the decoding complexity, but the decoding performance will be lost[4]. Jing hu Chen and
MPCFossorier use a correction factor to make up the minimum sum decoding algorithm, and propose
a normalized minimum sum (NMS) decoding algorithmLDPC representation methods.

2. LDPC representation methods

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
CISAI 2019 IOP Publishing
Journal of Physics: Conference Series 1453 (2020) 012026 doi:10.1088/1742-6596/1453/1/012026

LDPC is an abbreviation for Low-Density Parity Check Code. The biggest feature of LDPC code is
that the "1" element in the check matrix representing this code is far less than the "0" element.
c10 c9 , c8 , c7 , c6 represent information bit and c5 c4 , c3 , c2 , c1 represent the supervision bit. then the
code word is ：
c1  c 2  c3  c 4  0 1 1 1 1 0 0 0 0 0 0
  
 c1  c5  c 6  c7  0 1 0 0 0 1 1 1 0 0 0
 
 c2  c5  c 8  c 9  0 0 1 0 0 1 0 0 1 1 0 (1)
c  
 3  c6  c8  c10  0 0 0 1 0 0 1 0 1 0 1
c4  c7  c9  c10  0 0 0 0 1 0 0 1 0 1 1
There are two ways to define LDPC codes. One can be defined by a check matrix in mathematics,
and the other can be represented by Tanner diagram[5]. The Tanner diagram graphically represents the
information in the LDPC code. The relationship between bits and check bits is a grid graph composed
of check nodes and variable nodes. The Tanner graph is expressed as Va  Vb , E  . Va expressed as a set
of check nodes, Vb presents the set of Variable nodes. E epresents the set of connecting edges between
nodes, which corresponds to the "1" element in the test matrix. The test matrix in the above formula
has 5 rows 10 columns, then there are 5 test nodes and 10 variable nodes. Each row in the test matrix
has four elements. The "1" element represents a linear relationship between the information bits and
check bits at this position, which is reflected in Tanner. The graph shows that there is a connection
between the test node and the information node. The four "1" elements indicate that four edges are
connected. The Tanner graph corresponding to the test matrix in the above formula is as follows:
a1 a2 a3 a4 a5

v1 v2 v3 v4 v5 v6 v7 v8 v9 v 10
Figure 1. Tanner diagram corresponding to matrix H

3. LDPC decoding algorithms

LDPC decoding algorithm algorithms are divided into two categories according to different message
passing forms. One is the hard decision decoding algorithm in which the confidence message is in the
form of bit value, and the other is the soft decision decoding algorithm in which the confidence
message is the posterior probability. Hard decision decoding algorithms commonly include BF
algorithm, WBF algorithm, soft decision decoding algorithms include BP decoding algorithm, LLR
BP decoding algorithm, MS decoding algorithm, NMS decoding algorithm, Soft decision decoding is
an improvement on the basis of BP decoding algorithm[6]. In order to find the best decoding
algorithm, compare the performance of different decoding algorithms.

3.1. Hard decision decoding algorithm

Hard decision means that the algorithm determines whether the input signal is 0 or 1 by setting a
threshold value. The hard decision decoding algorithm includes BF algorithm and WBF algorithm.

3.1.1 BF algorithm

2
CISAI 2019 IOP Publishing
Journal of Physics: Conference Series 1453 (2020) 012026 doi:10.1088/1742-6596/1453/1/012026

Bit flip translation (BF) is a classic algorithm of hard decision algorithm. The BF algorithm records
the bits that do not satisfy the equation, and considers that the bits are likely to be interfered with to
the greatest extent[7]. The bit is flipped, and then the code word after the flip is used to continue the
iterative operation. The steps of the bit-flip decoding algorithm are as follows :
(1) Set the number of initial iterations k1 and its upper limit kmax , judge the information
y   y1 , y2 ,  , yn  , and obtain a decision sequence is:
1, y  0
zn  sgn( yn )   n n  1, 2, N (2)
0, yn  0
(2) The decision sequence zn is calculated with the hm , n elements of the check matrix, and the
calculation result is represented s   s1 , s2 ,, sn  . The formula is calculated as follows:
N 1
sm   zn hm ,n mod  2 , m  0,1 , M  1 (3)
n 0
If indicates that the code word translated is correct, the end of decoding, otherwise indicates that
there is a bit error.
(3) When the hard judgment bit zn does not meet the conditions, the number of bits that do not
meet the check equation is recorded, and the number f n is recorded as:
M 1
f n   sm hm ,n , n  0,1 , N  1 (4)
m0

If f n  T , then flippe zn and update the code word sequence z , and then continue to iterate
the calculation order.

3.1.2 WBF algorithm

Weighted bits flip (WBF) calculated the influence of different weights, highlights the importance of
weight of decoding algorithm, this method will be involved in each calibration equation of element
weight is calculated[8], and then the weighted sum of the corresponding position of the element
addition, turnover is one of the largest elements according to the weight of a bit of information, the
decoding algorithm steps are as follows:
(1) The weighted bit flip decoding algorithm is the same as the first two steps of the bit flip
decoding algorithm. The decision decoding sequence zn is obtained and substituted into sm  zn  H
The weight of each check formula is :
wm  min yi (5)
iN ( m )

Where, wm is determined by the adjacent nodes.

(2) Calculate the corresponding weights of all elements:
E  （2s m  1)wm，n   0, M  1 (6)
mN ( n）

(3) Compared with the above formula, if E is the least reliable, then flip the corresponding bit and
update the code word sequence , The flipped bit is substituted in sm  zn  H .If sm  0 ,then the
iteration operation is stopped. Otherwise, the iteration operation is continued and the above steps are
repeated until the condition is met or the maximum number of iterations is reached.

3.2. Soft decision decoding algorithm

The LDPC soft decision decoding algorithm uses the characteristics of the channel. The soft decision
decoding algorithm uses the posterior probability of the bit value to continuously pass between the

3
CISAI 2019 IOP Publishing
Journal of Physics: Conference Series 1453 (2020) 012026 doi:10.1088/1742-6596/1453/1/012026

variable node and the check node. The posterior probability value is continuously modified as the
number of iterations increases. Determine the bit value based on the magnitude of the probability
value[9]. Soft decision decoding algorithms include BP algorithm, LLR BP algorithm, MS algorithm,
NMS algorithm。

3.2.1. BP algorithm
BP algorithm is the most classical soft decision decoding algorithm. The decoding steps of BP
algorithm are as follows：
(1) Initial assignment. qba (0) is the probability of the information bit 0, and qba (1) is the
probability of the information 1, the probability formula is:
qba  0   pb  0  (7)
qba 1  pb 1 (8)

(2) Check node update. rab (0) is the probability of the check node be 0 and rab (1) is probability of
the check node be 1, the probability formula is expressed as:
1 1
rab 0     1  2qb'a 1 (9)
2 2 b'R \ b
a

1 1
rab 1    1  2qb'a 1 (10)
2 2 b'R \b
a

(3) Variable node update. qba  0  is probability of the variable node be 0, qba 1 is probability of
the variable node be 1, the formula is expressed as:
qba 0  kba pb 0  ra 'b 0 (11)
a 'Cb \ a
qba 1  kba pb 1  ra'b 1 (12)
a 'Cb \ a
(4) Decode the decision to determine the probability of the variable node. The probability
expression is:
qb 0  kb pb 0 rab 0  (13)
aCb
qb 1  kb pb 1  rab 1 (14)
aCb

3.2.2. LLR BP algorithm

The LLR BP decoding algorithm is developed on the basis of the BP decoding algorithm, which
reduces the complexity of the algorithm by transforming a large number of multiplication operations
into addition operations in the BP decoding algorithm, and the confidence messages to be delivered
are expressed by logarithmic likelihood ratio [10]. The workflow of LLR BP decoding algorithm is as
follows:
(1) Initial assignment. Representing the information received by the channel as a log-likelihood
ratio, the log-likelihood ratio of the variable nodes L  pb  is:
pb  0 
L  pb   ln (15)
pb 1

4
CISAI 2019 IOP Publishing
Journal of Physics: Conference Series 1453 (2020) 012026 doi:10.1088/1742-6596/1453/1/012026

(2) Check node update. The variable node is passed to the check node after processing, and the
log-likelihood ratio of the check node L  rab  is:
1 
Lrab   2 tanh 1  tanh Lqb' a  (16)
b'Ra \ b 2 
(3) Variable node update. The log-likelihood ratio of the variable nodes L qba  is:
q 0 
Lqba   ln ba  L pb    Lra 'b  (17)
qba 1 '
a C b \ a
(4) Decoding judgment. The log-likelihood ratio information of the decision variable node Lqb 
is:
q 0 
Lqb   ln b  L p b    Lrab  (18)
qb 1 a 'C b

If L qb   0 , then cˆb  0 , otherwise cˆb  1 , the code after the judgment is cˆ  (cˆ1，cˆ2，
cˆn ) , if
H  cˆT  0T ,then decoding is over.

3.2.3. MS algorithm
MS algorithm is on the basis of the LLR BP decoding algorithm, MS algorithm will use
approximation instead of complicated operation, MS algorithm and LLR BP algorithm steps are the
same, decoding steps difference is check node update[11]. The minimum and decoding algorithm
verifies that the node is updated as:
Lrab   sgnLqb'a   min| Lqb'a  |
 (19)
b'Ra \ b

3.2.4. NMS decoding algorithm

NMS algorithm takes measures to increase the normalization factor to make up for the performance
loss of MS algorithm. NMS decoding algorithm adds the normalization factor in the update of the
verification node of MS decoding algorithm. NMS algorithm verifies the node update step is :
Lrab    sgnLqb' a   min| Lqb' a  |
 (20)
b'Ra \ b

4. Simulation results and analysis

On the basis of the above description of the workflow of BP decoding algorithm, LLR BP decoding
algorithm, MS decoding algorithm and NMS decoding algorithm, the performance of the quasi-cyclic
LDPC decoding algorithm based on ieee802.16e is comprehensively compared and analyzed.
MATLAB is used as the simulation tool to analyze the performance of the four decoding algorithms
from the perspectives of complexity and bit error rate.
In order to compare the performance of the soft decision decoding algorithm and the hard decision
decoding algorithm, the code rate is 1/2, the code length is 576, and the channel environment is
AWGN channel simulation unencoded, BF, WBF algorithms and soft decisions in the hard decision
algorithm. The performance of the decoding algorithm BP algorithm, the bit error rate curves of
different decoding algorithms are shown in Figure 2.

5
CISAI 2019 IOP Publishing
Journal of Physics: Conference Series 1453 (2020) 012026 doi:10.1088/1742-6596/1453/1/012026

Figure 2. Performance curve of soft and hard decision decoding algorithm

The simulation results show that no matter what encoding algorithm is used, the bit error rate
performance of the encoded signal is much better than that of the unencoded signal. The hard decision
BF algorithm makes a hard decision on the signal with a low calculation amount, but does not take full
advantage of the characteristics of the channel The decoding effect on codewords affected by channel
noise is very poor, and the bit error rate is the highest. The WVBF algorithm makes full use of the
influence of each signal on the encoding and decoding weights, and improves the BF algorithm with
better performance than the BF algorithm. The BP algorithm Making full use of channel
characteristics and performing soft decision on signals, the decoding performance is the best.
Soft decision decoding algorithm is better than hard decision decoding performance, thus decoding
algorithm selects the soft decision decoding algorithm, further, in order to compare the performance of
different soft decision decoding algorithm, the simulation compare BP decoding algorithm, log-
domain BP decoding algorithm, the minimum and decoding algorithm, normalized the minimum and
the decoding principle of decoding algorithm, in order to find out the best decoding algorithm, the
IEEE802.16 e of quasi-cyclic LDPC code, on the basis of the bit rate is 1/2, code length respectively
take 576 , the simulation contrast four decoding algorithm of bit error rate curve is shown in Figure 3.

Figure 3. Performance curves of four decoding algorithms

The simulation results show that BP decoding algorithm, LLR BP decoding algorithm and NMS
decoding algorithm almost coincide in the code length of 576, indicating that the performance of these
three decoding algorithms is similar. The bit error rate curve of MS decoding algorithm is higher than
that of the three decoding algorithms, indicating that the performance of MS decoding algorithm is the
worst.

6
CISAI 2019 IOP Publishing
Journal of Physics: Conference Series 1453 (2020) 012026 doi:10.1088/1742-6596/1453/1/012026

5. Conclusions
Research in this paper, the LDPC decoding algorithm, compare the hard decision decoding algorithm
is analyzed and the performance of the soft decision decoding algorithm, the simulation results show
that the soft decision decoding algorithm of BP decoding algorithm is better than hard decision in BF,
the WBF performance, on the basis of BP decoding algorithm, the contrast analysis of the BP
decoding algorithm, LLR BP decoding algorithm, MS performance and complexity of decoding
algorithm, NMS decoding algorithm, combined with the two aspects of complexity and performance
considerations, NMS decoding algorithm has the best performance..

Acknowledgments
This work is supported by the National Defense Basic Scientific Research Planed Project(2018),
Natural Science Fund of Liaoning Province NO.20170540775,Program for Liaoning Innovative
Research Team in University, Science and Technology Foundation from Liaoning Education
Department No.LG201914.

References
[1] Chung,Y.,Yun,S. (2013) Experiments on MIMO-OFDM system combined with adaptive
beamforming based on IEEE 802.16e WMAN standard. Telecommunication Systems, 52:
1931-1944 .
[2] Yi,L.,Wei,H. (2017) Survey of Beyond-BP Decoding Algorithms: Theory and Applications,
39:1503-1514.
[3] Min,Z,Bai,B. (2016) Reliability-Based Joint Detection-Decoding Algorithm for Nonbinary
LDPC-Coded Modulation Systems. IEEE Transactions on Communications,6:2-14 .
[4] Zhang,R.(2016)Improved minimum sum decoding algorithm for LDPC codes. Information on
microcomputers ,30: 170-172.
[5] Liu,W.,Xu,Y.G.(2017)Research and implementation of LDPC coding and decoding.Electronic
world,12:125-127 .
[6] Li.,X.,Chen H.(2016) Hard-Information Bit-Reliability Based Decoding Algorithm for Majority-
Logic Decodable Nonbinary LDPC Codes .IEEE Communications Letters.,20:866-869.
[7] Zhou,H.,Feng,J.(2017)Improvement of LDPC code node residual confidence propagation
decoding.Application of electronic technology,43:107-111 .
[8] Li,S.(2018) Research on LDPC decoding based on LLR BP algorithm.Television
Technology,42:51-55 .
[9] Chen,Z., Bai.,J.C. (2018) LDPC Layered Correction Minimum and Decoding Algorithm for
IEEE802.16e Standard.Information Communication,2.:11-12 .
[10] Li,C.H., Liang,Q. (2017) Iterative large number logic LDPC decoding algorithm and its
quantization optimization based on binary decoding .Information of electronics and
information technology,4:873-880 .
[11] Guo,L.L(2017).FPGA-based IEEE 802.16e LDPC coding and decoding method .Automation
technology and application. Automation technology and applicationvol.,36:49-53.