Hindawi

Wireless Communications and Mobile Computing

Volume 2018, Article ID 1398191, 7 pages
https://doi.org/10.1155/2018/1398191

Research Article
An Alternative Approach Obtaining a Normalization
Factor in Normalized Min-Sum Algorithm for Low-Density
Parity-Check Code

In-Woo Yun , Hee-ran Lee, and Joon Tae Kim

Department of Electronic Engineering, Konkuk University, Seoul, Republic of Korea

Correspondence should be addressed to Joon Tae Kim; [email protected]

Received 11 June 2018; Revised 29 August 2018; Accepted 23 September 2018; Published 17 October 2018

Academic Editor: Mohammed El-Hajjar

Copyright © 2018 In-Woo Yun et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The min-sum algorithm (MSA) for decoding Low-Density Parity-Check (LDPC) code is an approximation algorithm that can
greatly reduce the computational complexity of the belief propagation algorithm (BPA). To reduce the error between MSA and BPA,
an improved MSA such as normalized min-sum algorithm (NMSA) that uses the normalization factor when updating the check
node is used in many LDPC decoders. When obtaining an optimal normalization factor, density evolution (DE) is usually used.
However, not only does the DE method require a large number of calculations, it may not be optimal for obtaining a normalization
factor due to the theoretical assumptions that need to be satisfied. This paper proposes a new method obtaining a normalization
factor for NMSA. We first examine the relationship between the minimum value of variable node messages’ magnitudes and
the magnitudes of check node outputs of BPA using the check node message distribution (CMD) chart. And then, we find a
normalization factor that minimizes the error between the magnitudes of check node output of NMSA and BPA. We use the least
square method (LSM) to minimize the error. Simulation on ATSC 3.0 LDPC codes demonstrates that the normalization factor
obtained by this proposed method shows better decoding performance than the normalization factor obtained by DE.

1. Introduction to the minimum value selector. However, MSA has large

performance degradation due to the approximation. To
Low-Density Parity-Check (LDPC) codes [1] proposed by improve decoding performances of MSA normalized min-
Gallager in the 1960s have been used for error correcting sum algorithm (NMSA), which normalizes the minimum
codes for many communication standards showing the error value of the variable node messages, has been introduced [3].
correcting capability close to the Shannon limit. However, When selecting a normalization factor in the NMSA,
LDPC codes burdens the hardware since it must perform density evolution (DE) is usually used [4]. The DE is an
many floating-point operations repeatedly during decoding. analytical tool used for LDPC codes with message passing
Therefore, numerous researches have been carried out to decoders. It is used to calculate the average probability density
reduce the computational complexity of LDPC decoder. One function (pdf) of messages along the edges for a given
of the topics of the researches was to reduce the com- code ensemble and decoding algorithm. If the code length
putational complexity of the belief propagation algorithm and the number of iterative decoding approach infinity, DE
(BPA), known as an optimal message passing decoder for predicts the minimum signal-to-noise ratio (i.e., threshold)
LDPC codes in the binary symmetric channel. The BPA for convergence of the bit error probability to zero [5].
involves great number of logarithmic and multiplicative However, this method requires the assumption that the
operations when updating the check node. In an effort to specific LDPC code is a tree structure and the algorithm
reduce the computational complexity of the BPA, min-sum used for decoding should satisfy a symmetric condition.
algorithm (MSA) has been proposed [2]. MSA approximates Considering that the LDPC code is rarely a tree structure and
the exponential term of the check node update function the approximation algorithm does not satisfy a symmetric
2 Wireless Communications and Mobile Computing

condition, a normalization factor derived from DE may from the variable nodes connected to 𝐶𝑚 except the message
not be optimal. Additionally, performing DE needs a lot of from 𝑉𝑛 . The sign of the check node output is calculated in
computation when obtaining pdfs of messages [6]. the same way as BPA. The magnitude of check node output
For this reason, we introduce an alternative approach is processed by selecting the minimum value of variable
obtaining a normalization factor. The main idea of the node messages’ magnitudes excluding the node that passes
proposed method is to find a factor that minimizes errors the processed message. In contrast to BPA, this algorithm
between check node outputs of BPA and NMSA by observing only needs 2𝑑𝑐 floating point comparison to calculate the
the distribution of check node messages of BPA with a graph magnitudes of check node outputs for one check node.
called check node message distribution (CMD) chart. Since
󵄨󵄨 󵄨󵄨
the BPA uses all the input messages of a check node when 𝐿𝑀𝑆𝐴
𝑚𝑛 = ∏ sign (𝐿 𝑛󸀠 𝑚 ) 󸀠min (󵄨󵄨𝐿𝑛󸀠 𝑚 󵄨󵄨) (3)
𝑛 ∈𝐶 \𝑛 𝑚
processing a check node output, it is very hard to predict 𝑛󸀠 ∈𝐶𝑚 \𝑛
the output using only a minimum input message. However,
if the signal-to-noise ratio (SNR) and the number of iteration To compensate the large difference between the check
is fixed, we can easily see the trend of the error by plotting node output messages of BPA and MSA, the NMSA scales
the CMD chart with the minimum value of input messages’ the selected minimum value as in (4). By using a proper
magnitudes as the x-axis and the magnitudes of the check normalization factor, the NMSA can achieve performances
node output of BPA as the y-axis. After observing the CMD very close to the BPA [4]. This algorithm needs one additional
chart, the LSM is used to find a normalization factor that floating-point multiplier per one check node compare to
minimizes the error. MSA, but it can greatly improve the decoding performance.
The rest of this paper is organized as follows. In Section 2.1 󵄨 󵄨
𝐿𝑁𝑀𝑆𝐴
𝑚𝑛 = 𝛼 ⋅ ∏ sign (𝐿 𝑛󸀠 𝑚 ) 󸀠min (󵄨󵄨󵄨𝐿𝑛󸀠 𝑚 󵄨󵄨󵄨) (4)
we briefly introduce the check node update equation for var- 𝑛 ∈𝐶 \𝑛 𝑚
𝑛󸀠 ∈𝐶𝑚 \𝑛
ious decoding algorithms of LDPC codes in the LLR domain.
In Section 2.2 we explain how to derive a normalization
factor using the CMD chart and the least squares method 2.2. The Method of Obtaining the Normalization Factor Using
(LSM). Lastly, we show simulation result with some of the the CMD Chart. In the previous section, we mentioned that
code rates on ATSC 3.0 LDPC codes that the normalization the MSA can significantly reduce the number of computa-
factor derived from the proposed methods shows superior tions for obtaining the magnitudes of a check node message
decoding performance over factor using the DE method in by approximating the transcendental function to a minimum
NMSA decoding. selecting function. Also, NMSA can bridge the gap between
the check node output of BPA and MSA by scaling the
minimum value. In this chapter, we visualize how NMSA
2. Materials and Methods bridges the gap by showing an example of a CMD chart that
2.1. Check Node Update of LDPC Codes. BPA, MSA, and we propose in this paper. This chart can be further used to get
NMSA are identical during the decoding procedures but a normalization factor. CMD chart is short for check node
not during the updating of check nodes [7]. Therefore, in message distribution chart, which plots the magnitudes of
this paper, we show only the check node update process check node output messages as the y axis, and the minimum
value of input variable node messages’ magnitudes connected
for each algorithm. If the 𝑛𝑡ℎ variable node is 𝑉𝑛 , and 𝑚𝑡ℎ
to a check node excluding 𝑛𝑡ℎ variable node (min(|𝐿𝑛󸀠 𝑚 |)).
check node is 𝐶𝑚 , the check node message of BPA 𝐿𝐵𝑃𝐴 𝑚𝑛 Figure 1 shows an example of the CMD chart simulated
is computed using (1) and (2) where 𝐿 𝑛󸀠 𝑚 denotes variable
using (n, j, k) = (𝑛, 𝑑V , 𝑑𝑐 ) = (20, 3, 4) regular LDPC code
node messages passing from 𝑉𝑛󸀠 to 𝐶𝑚 . The notation 𝑛󸀠
where n denotes number of encoded bits and 𝑑V denotes
denotes variable nodes connected to 𝐶𝑚 , excluding 𝑛, and the
the number of check node connected to one variable node.
notation 𝑑𝑐 denotes the number of variable nodes connected
The parity check matrix was constructed using the Gallager
to 𝐶𝑚 . All algorithms covered in this chapter including BPA
method [1]. In this simulation, SNR=-5[dB] AWGN was
calculate the sign of 𝐿𝐵𝑃𝐴
𝑚𝑛 by multiplying all signs of variable added to the BPSK mapped signal. After demapping, 20
node messages except the node that passes the message. The log-likelihood ratios (LLRs) were set as the initial value of
magnitudes of 𝐿𝐵𝑃𝐴 𝑚𝑛 is calculated using Φ(𝑥) function in variable node messages and check node output messages
(2), and this function consists of a transcendental function. are obtained using each decoding algorithm. Data points on
This algorithm needs the 2𝑑𝑐 floating point addition and Figure 1 are the magnitudes of a check node output messages
2𝑑𝑐 Φ(𝑥) operation to calculate the magnitudes of 𝑑𝑐 check for the same (min(|𝐿𝑛󸀠 𝑚 |)). To explain in detail, we pick out
node output messages per one check node. one check node then follow the message calculation and
plotting process. One check node 𝐶𝑚 receives four variable
󵄨󵄨 󵄨󵄨 node messages 𝐿𝑛𝑚 = [1.49, 0.97, −0.40, 0.52]. Then, the
𝐿𝐵𝑃𝐴
𝑚𝑛 = ∏ sign (𝐿 𝑛󸀠 𝑚 ) Φ ( ∑ Φ (󵄨󵄨𝐿 𝑛󸀠 𝑚 󵄨󵄨)) (1)
𝑛󸀠 ∈𝐶𝑚 \𝑛 𝑛󸀠 ∈𝐶𝑚 \𝑛 check node processes output messages using BPA using Eq.
(1) and Eq. (2). The four check node output messages are
𝑥 𝐿𝐵𝑃𝐴
Φ (𝑥) = − ln (tanh ( )) (2) 𝑚𝑛 = [−0.04, −0.06, 0.14, −0.11]. Meanwhile, the minimum
2 values of the variable node messages are min(|𝐿𝑛󸀠 𝑚 |) =
Equation (3) is used to update the check node using MSA. [−0.40, −0.40, 0.52, −0.40] and this vector is equal to the
𝐿𝑛󸀠 𝑚 denotes a column vector whose elements are messages check node output message vector of MSA (𝐿𝑀𝑆𝐴 𝑚𝑛 ). We can
Wireless Communications and Mobile Computing 3

0.9 Carlo simulation. We first generated the random bits and

modulated using QPSK. We then calculated the check node
0.8
messages, assuming that one check node is connected to
0.7 three variable nodes (𝑑𝑐 = 3). In Figure 2(a), the simulation
|Check mode messages|

shows that values of check node output messages of two

0.6
algorithms are very similar in a high SNR. In the lowest SNR,
0.5 as shown in Figure 2(c), the differences are relatively large.
0.4
This change of distribution in the different SNRs is caused
by the characteristics of error between BPA and MSA. The
0.3 check node update function of the BPA can be divided into
0.2 two terms, which are sign computing and minimum value
selecting term, and error term as in Eq. (5) [8]. Since the sign
0.1 computing and the minimum value selecting term is equal
0 to the check node update function of the MSA, the error
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 between the check node output messages of the BPA and
 
min(Lnm  ) MSA can be calculated using the error term. This error can
BPA NMSA(≈0.26) have values between ln((1 + 𝑒−|∞| )/(1 + 𝑒−|0| )) ≈ −ln(2) and
MSA example data ln((1 + 𝑒−|0| )/(1 + 𝑒−|∞| )) ≈ ln(2). In other words, regardless
of the values of variable node messages, the error is bound
Figure 1: CMD chart according to 𝑑𝑐 = 3, 𝐸𝑠 /𝑁0 = −5[dB]. to give a value between −ln(2) and ln(2). In conclusion, the
initial variable node messages have large values in a high
SNR environment, so min(|𝐿𝑛󸀠 𝑚 |) is also high, resulting in
plot the data points (min(|𝐿𝑛󸀠 𝑚 |), |𝐿𝐵𝑃𝐴 the bounded error showing a relatively small value com-
𝑚𝑛 |) as shown in Figure 1
(∗). We can easily notice that the check node output messages, pared to min(|𝐿𝑛󸀠 𝑚 |). Therefore, the data points of BPA are
obtained using MSA, are very different from the messages distributed closely to that of MSA in a high SNR. Using
obtained using BPA because data points (I, ◻) are distant this information, we can understand that relatively small
from each other. Furthermore, we can guess that this causes normalization factor will improve decoding performance
a large performance gap between the two algorithms. We can in low SNR, and MSA shows great performance in high
also see how NMSA improves the performance in the CMD SNR.
chart. In the case of setting the normalization factor as 0.26, 𝐿𝐵𝑃𝐴 󵄨󵄨󵄨 󵄨󵄨󵄨 󵄨󵄨󵄨 󵄨󵄨󵄨
𝑚𝑛3 = sign (𝐿 𝑛1 𝑚 ) sign (𝐿 𝑛2 𝑚 ) min (󵄨󵄨𝐿 𝑛1 𝑚 󵄨󵄨 , 󵄨󵄨𝐿 𝑛2 𝑚 󵄨󵄨)
the check node output message vector of NMSA is 𝐿𝑁𝑀𝑆𝐴 𝑚𝑛 =
[−0.10, −0.10, 0.14, −0.10]. NMSA is expected to achieve (5)
1 + 𝑒−|𝐿 𝑛1 𝑚 +𝐿 𝑛2 𝑚 |
performances close to that of BPA with the error being + ln ( )
1 + 𝑒−|𝐿 𝑛1 𝑚 −𝐿 𝑛2 𝑚 |
reduced from [0.36, 0.34, 0.38, 0.29] to [0.06, 0.04, 0.01, 0.01].
In this way, we find the normalization factor which best Figure 3 shows the CMD chart according to 𝑑𝑐 . As shown
approximates the check node output of BPA. To obtain the in the figure, when 𝑑𝑐 becomes larger, the output value of
appropriate normalization factor, we first examine how the the check node becomes smaller. This can be explained by
appropriate normalization factor varies with dc and SNR in the characteristics of Φ(𝑥) function of (2) which is used
Section 2.2.1. Then we propose a new method to minimize the to update the check node with BPA. The function Φ(𝑥)
error between BPA and NMSA by using LSM in Section 2.2.2. has characteristics, which is that input and output of the
function are defined only for positive real values and that
2.2.1. Check Node Message Distribution Chart. The CMD they are inversely proportional. Since Φ(|𝐿 𝑛󸀠𝑚 |) from (1) is
chart shows the magnitudes of check node messages of a always positive, when they are summed, the large 𝑑𝑐 yields
specific decoding algorithm against the minimum value of a small output of the check node. Therefore, for the NMSA
|𝐿𝑛󸀠 𝑚 |. In this chapter, we introduce more examples of CMD to approximate the BPA well, in lower SNR or larger 𝑑𝑐
chart and explain the characteristics of check node output conditions, the normalization factor should have a small
messages of BPA decoder using the chart. We first show how value.
the distribution of the data points in CMD charts changes
in a different SNR. Then, we show how the distribution 2.2.2. Obtaining Normalization Factors Using the Least Square
changes when the number of variable nodes connected to Method. The LSM is used to derive the parameters of the
one check node (𝑑𝑐 ) changes. The distribution also changes curve, which minimizes the square sum (L2 norm) of the
as the iteration proceeds. However, since the processing of shortest distance between the curve and the given data
iteration can be considered as an increase in SNR, we skip points. It is used to remove noise from data points or to
the subject and cover it in detail in Section 2.2.2. These characterize the data set. We use this method to obtain a
examples will help us to get a sense of, which values of normalization factor that minimizes the error between the
normalization factor will improve the decoding performance. check node messages of BPA and NMSA. We use data points
Figure 2 shows CMD chart for different SNR (𝐸𝑠 /𝑁0 ). (min(|𝐿𝑛󸀠 𝑚 |), |𝐿𝐵𝑃𝐴
𝑚𝑛 |) for 𝑛
󸀠
∈ 𝐶𝑚 \ 𝑛, and 𝑚 ∈ [1, 𝑀]
The check node messages are obtained through the Monte where 𝑀 is the number of the check node. If we define a
4 Wireless Communications and Mobile Computing

5
12 2.5

10 4

|Check node Message|

2

|Check node Message|

|Check node Message|

8 3 1.5
6
2 1
4
1 0.5
2

0 0 0
0 2 4 6 8 10 12 0 1 2 3 4 5 0 0.5 1 1.5 2 2.5
     
min(Lnm  ) min(Lnm  ) min(Lnm  )

BPA BPA BPA

MSA MSA MSA
(a) (b) (c)

Figure 2: CMD charts according to 𝐸𝑠 /𝑁0 . (a) 𝐸𝑠 /𝑁0 = 5 [dB]. (b) 𝐸𝑠 /𝑁0 = 0 [dB]. (c) 𝐸𝑠 /𝑁0 = −5 [dB].

5 5 5

4 4 4
|Check node Message|

|Check node Message|

|Check node Message|

3 3 3

2 2 2

1 1 1

0 0 0
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
     
min(Lnm  ) min(Lnm  ) min(Lnm  )

BPA BPA BPA

MSA MSA MSA
(a) (b) (c)

Figure 3: CMD charts according to 𝑑𝑐 . (a) 𝑑𝑐 = 3, (b) 𝑑𝑐 = 6, and (c) 𝑑𝑐 = 9.

set of min(|𝐿𝑛󸀠 𝑚 |) as the column vector 𝑋 and a column for 𝑙th iteration and 𝑋(𝑙) and 𝑌(𝑙) denote vectors for data points
vector |𝐿𝐵𝑃𝐴
𝑚𝑛 | as the vector 𝑌, the normalization factor can be (min(|𝐿𝑛󸀠 𝑚 |), |𝐿𝐵𝑃𝐴 th
𝑚𝑛 |) at 𝑙 iteration.
obtained from equation (6) as follows.
−1
−1
𝛼(𝑙) = (𝑋T(𝑙) 𝑋(𝑙) ) 𝑋𝑇(𝑙) 𝑌(𝑙) (7)
𝑇 𝑇
𝛼 = (𝑋 𝑋) 𝑋 𝑌 (6)
Figure 5 shows a computer simulation block diagram
Figure 4 is a graph showing the bit error rate of four for obtaining a set of normalization factors 𝛼(𝑙) . In the
different decoding algorithms when the maximum number simulation, we get vectors 𝑋(𝑙) and 𝑌(𝑙) assuming that the
of iterations is set to 1. The simulation is conducted on the (𝑙 − 1)𝑡ℎ iteration is correctly decoded using BPA and find
ATSC 3.0 LDPC code rate of 2/15 and the frame size of 𝛼(𝑙) that has the least square sum with the given data points.
64800 while using QPSK mapping. An average value for the Since the decoding tends to proceed well, irrespective of the
normalization factor is derived through repetitive simulation decoding algorithm in high SNR, the SNR to get 𝛼(𝑙) is set
and used for decoding in each SNR. As shown in Figure 4, to a minimum error free SNR for the BPA algorithm so that
the decoding performance of the proposed method is close the check node messages can be well approximated in a low
to that of BPA than the DE method [9], especially in low SNR. The simulation result of obtaining the set of 𝛼(𝑙) for
SNR. each iteration is shown in Figure 6. The maximum iteration
When the number of iterations is higher than one, we number is set as 40 in the simulation. In early iterations,
need to obtain an appropriate 𝛼 for each iteration. So we the magnitudes of LLR values of the variable node messages
rewrite (6) to (7) where 𝛼(𝑙) denotes a normalization factor have small values, which means that the messages have high
Wireless Communications and Mobile Computing 5

BER
10−1

−10 −9 −8 −7 −6 −5 −4 −3 −2
Es /N0 [dB]

BPA proposed- NMSA

[9]-NMSA MSA

Figure 4: Bit error rate when the number of maximum iterations is set as 1.

L n m LBPA
mn
Check node update using BPA

|·|


LSM

 
Get min(L n m )

Figure 5: Block diagram of a procedure obtaining 𝛼 with LSM.

0.9

0.8

0.7
(l)

0.6

0.5

0.4

0.3
0 5 10 15 20 25 30 35
l

Code rate = 2/15

Code rate = 3/15

Figure 6: Trend of appropriate normalization factors over the number of iterations.

6 Wireless Communications and Mobile Computing

−3

−3.5 −2.5

−4
−3
Es /N0 [dB]

Es /N0 [dB]
−4.5

−5 −3.5

−5.5
−4
−6
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
l l

Code rate = 2/15 Code rate = 3/15

(a) (b)

Figure 7: Error free 𝐸𝑠 /𝑁0 when 𝛼(𝑙) used for decoding. (a) 2/15 and (b) 3/15.

a bit error probability. So, the CMD chart shows similar method needs fixed number (i.e., the maximum iteration
distribution as Figure 2(c), and the appropriate normalization number) of 𝛼 verifications to derive a single alpha for the
factor is comparatively small. As the iteration progresses, the entire iteration, whereas the DE method needs full search of
magnitudes of LLR values of the variable node messages get alpha in range from 0 to 1 [11].
larger and their distribution becomes closer to the CMD chart
of Figure 2(a). At the completion of iterative decoding, the
normalization factor converges to 1 and the NMSA becomes 3. Results and Discussion
equal to the MSA.
In this section, we show computer simulation results on
By using the set of normalization factors derived from
ATSC 3.0 LDPC codes that compare the BPA, NMSA with
this simulation, we can obtain good decoding performance
the normalization factor proposed in [9] ([9]-𝛼 NMSA), and
by changing the normalization factor for each iteration.
NMSA with the single normalization factor proposed in this
However, for the sakeof computational efficiency, we need
paper.
to derive a single normalization factor for all iterations.
Figure 8 shows the simulation result, when the maximum
When the 𝛼(𝑙) obtained from early iterations is used as
iteration is set as 40 on both ATSC 3.0 LDPC 2/15 and 3/15
a single normalization factor, the error free SNR is high
codes for a code length of 64800, when layered decoding
because it does not properly approximate BPA as iteration
and QPSK mapping are used. As shown in Figure 8(a), the
proceeds. On the other hand, if the 𝛼(𝑙) obtained from the
BPA showed the best performance that -6.35 [dB] of Es /𝑁0
end of iterations is selected as a single factor, the message
is required to have BER under 10−6 . The NMSA which uses
is exaggerated at the beginning of decoding, and as a result
the normalization factor derived in [9] (𝛼 = 0.63) (◻) has
the error free SNR is high. Therefore, it is appropriate to
a performance gap of 0.75 [dB] compared to BPA. Whereas
use an intermediate normalization factor. So, we derived the
the NMSA, using the proposed 𝛼(≈ 0.5538), showed a
best normalization factor for the entire iteration through
performance gap of 0.50 [dB] (). The simulation on code
simulation. The simulation result is shown in Figure 7. In
rate 3/15 also showed that the NMSA using the proposed 𝛼 (≈
the case of code rate 2/15, Figure 7(a), the normalization
0.5576)showed better performance by 0.15 [dB] compared to
factor obtained from the eighth iteration 𝛼(8) ≈ 0.5538
the NMSA that uses the normalization factor derived in [9]
showed the lowest error free SNR. Likewise, in the case of
(𝛼 ≈ 0.63) (◻) as shown in Figure 8(b).
code rate 3/15, Figure 7(b), the normalization factor obtained
from the fourth iteration 𝛼(4) ≈ 0.5576 showed the lowest
error free SNR. As a result, we specify the normalization 4. Conclusions
factor 𝛼(8) ≈ 0.5538 as a single coefficient at code rate
2/15 and 𝛼(4) ≈ 0.5576 as a single coefficient at code rate In this paper, we introduced the CMD chart and proposed
3/15. a new method to obtain a single normalization factor using
The proposed method can efficiently obtain normaliza- the chart and LSM. In addition, we showed that the NMSA
tion factor for each iteration by using CMD chart and LSM. using the normalization factor derived from the proposed
This method is much easier to implement than DE method method showed far superior decoding performance than the
since it does not require infinitely repeated pdf calculation to NMSA employing the normalization factor of the conven-
get a single normalization factor [10, 11]. Also, the proposed tional scheme [9]. Furthermore, the proposed method can be
Wireless Communications and Mobile Computing 7

100 100

10 −1 10 −1

10 −2 10 −2
Bit error rate

Bit error rate

10 −3 10 −3

10 −4 10 −4

10 −5 10 −5

10 −6 10 −6
−7 −6.5 −6 −5.5 −5 −4.5 −4 −3.5 −5 −4.5 −4 −3.5 −3 −2.5 −2
Es /N0 [dB] Es /N0 [dB]

BPA BPA
[9]- NMSA(= 0.63) [9]- NMSA(= 0.63)
proposed- NMSA(≈0.5538) proposed- NMSA(≈0.5576)
MSA MSA
(a) (b)

Figure 8: Bit error rate of code rate (a) 2/15 and (b) 3/15.

used to obtain the correction coefficients of other decoding IEEE Transactions on Communications, vol. 50, no. 3, pp. 406–
algorithms that approximate the BPA algorithm. 414, 2002.
[4] J. Chen and M. P. C. Fossorier, “Density evolution for two
improved BP-based decoding algorithms of LDPC codes,” IEEE
Data Availability Communications Letters, vol. 6, no. 5, pp. 208–210, 2002.
No data were used to support this study. [5] T. J. Richardson, M. A. Shokrollahi, and R. L. Urbanke,
“Design of capacity-approaching irregular low-density parity-
check codes,” IEEE Transactions on Information Theory, vol. 47,
Disclosure no. 2, pp. 619–637, 2001.
[6] A. Anastasopoulos, “A comparison between the sum-product
In-Woo Yun and Hee-ran Lee are co-first authors. and the min-sum iterative detection algorithms based on den-
sity evolution,” in Proceedings of the IEEE Global Telecommuni-
catins Conference, GLOBECOM’01, pp. 1021–1025, San Antonio,
Conflicts of Interest TX, USA, November 2001.
[7] R. W. Munn, “An introduction to LDPC codes,” in CRC Hand-
The authors declare that they have no conflicts of interest.
book for Coding and Signal Processing for Recording Systems, B.
Vasic, Ed., vol. 4, CRC, Boca Raton, FL, USA, 2004.
Acknowledgments [8] F. Guilloud, E. Boutillon, and J. Danger, “𝜆-Min Decoding Algo-
rithm of Regular and Irregular LDPC Codes,” in Proceedings
This research was supported by the KIAT (Korea Insti- of the 3nd International Symposium on Turbo Codes & Related
tute for Advancement of Technology) grant funded by the Topics, pp. 451–454, Brest, France, 2003.
Korea Government (MOTIE: Ministry of Trade Industry [9] S. Myung, S.-I. Park, K.-J. Kim, J.-Y. Lee, S. Kwon, and J. Kim,
and Energy) (No. N0001884, HRD program for Embedded “Offset and Normalized Min-Sum Algorithms for ATSC 3.0
Software). LDPC Decoder,” IEEE Transactions on Broadcasting, vol. 63, no.
4, pp. 734–739, 2017.
[10] X. Wei and A. N. Akansu, “Density evolution for low-density
References parity-check codes under Max-Log-MAP decoding,” IEEE Elec-
tronics Letters, vol. 37, no. 18, pp. 1125-1126, 2001.
[1] R. G. Gallager, “Low-density parity-check codes,” IEEE Trans-
actions on Information Theory, vol. 8, pp. 21–28, 1962. [11] J. Chen, A. Dholakia, E. Eleftheriou, M. P. C. Fossorier, and X.-
Y. Hu, “Reduced-complexity decoding of LDPC codes,” IEEE
[2] M. P. C. Fossorier, M. Mihaljevic, and H. Imai, “Reduced Transactions on Communications, vol. 53, no. 8, pp. 1288–1299,
complexity iterative decoding of low-density parity check codes 2005.
based on belief propagation,” IEEE Transactions on Communi-
cations, vol. 47, no. 5, pp. 673–680, 1999.
[3] J. Chen and M. P. C. Fossorier, “Near optimum universal belief
propagation based decoding of low-density parity check codes,”
 International Journal of

Rotating Advances in
Machinery Multimedia

The Scientific
Engineering
Journal of
Journal of

Hindawi
World Journal
Hindawi Publishing Corporation Hindawi
Sensors
Hindawi Hindawi
www.hindawi.com Volume 2018 http://www.hindawi.com
www.hindawi.com Volume 2018
2013 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018

Journal of

Control Science
and Engineering

Advances in
Civil Engineering
Hindawi Hindawi
www.hindawi.com Volume 2018 www.hindawi.com Volume 2018

Submit your manuscripts at

www.hindawi.com

Journal of
Journal of Electrical and Computer
Robotics
Hindawi
Engineering
Hindawi
www.hindawi.com Volume 2018 www.hindawi.com Volume 2018

VLSI Design
Advances in
OptoElectronics
International Journal of

International Journal of
Modelling &
Simulation
Aerospace
Hindawi Volume 2018
Navigation and
Observation
Hindawi
www.hindawi.com Volume 2018
in Engineering
Hindawi
www.hindawi.com Volume 2018
Engineering
Hindawi
www.hindawi.com Volume 2018
Hindawi
www.hindawi.com www.hindawi.com Volume 2018

International Journal of
International Journal of Antennas and Active and Passive Advances in
Chemical Engineering Propagation Electronic Components Shock and Vibration Acoustics and Vibration
Hindawi Hindawi Hindawi Hindawi Hindawi
www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018 www.hindawi.com Volume 2018