Sei sulla pagina 1di 27

Telecommunication Systems

https://doi.org/10.1007/s11235-019-00630-3

A survey on channel coding techniques for 5G wireless networks


Komal Arora1,2 · Jaswinder Singh3 · Yogeshwar Singh Randhawa4

© Springer Science+Business Media, LLC, part of Springer Nature 2019

Abstract
Although 4G (fourth generation) i.e. LTE (long term evolution) systems are now in use world-wide. But today’s 4G systems
have some challenges left such as spectrum scarcity and energy efficiency. The prime objectives of near-by-future 5G (fifth
generation) wireless communications are reliability, higher data rate, higher bandwidth, high spectrum efficiency, higher
energy efficient and that too at lower latency. Channel coding tend to increase the reliability of the wireless communications
system by adding extra bits in a controlled fashion and is considered to be most persuasive element of communication system.
4G LTE Turbo Codes have already been replaced by LDPC (low density parity check) Codes in many of the standards includ-
ing mMTC (massive machine type communication), D2D (device to device communication) and URLLC (ultra-reliable low
latency reliable communications). LDPC Codes and Polar Codes are securing much more observation because of their inher-
ent advantages of excellent bit-error-rate performance, fast encoding and decoding procedures; which make them the strong
contenders for 5G Channel Codes too. This paper provides the broad survey and comparison of the LDPC and Polar Codes
along with their advantages and drawbacks which will aid in further improvement of the next generation wireless networks. In
order to enlighten future research possibilities in this direction, issues addressed by distinct researchers have been explored too.

Keywords  LDPC · Polar · Channel coding · 5G

Abbreviations NTT Nippon telegraph and telephone


LTE Long term evolution TACS Total access communications system
IoT Internet of things IS-95 Interim standard 95
LDPC Low density parity check PDC Pacific digital cellular systems
SNR Signal to noise ratio GPRS General packet radio service
FEC Forward error correction EDGE Enhanced data rate for GSM evolution
mMTC Massive machine type communication UMTS Universal mobile telecommunication
D2D Device to device communication systems
URLLC Ultra-reliable low latency reliable HSPA High speed packet access
communications CDMA code division multiple access
AMPS Advanced mobile phone systems BSC Base station controller
RNC Radio network controller
* Komal Arora
TDMA Time division multiple access
komal03may@gmail.com GSM Global system for mobile
Jaswinder Singh
FDMA Frequency division multiple access
j_singh73@rediffmail.com WCDMA Wideband code division multiple access
Yogeshwar Singh Randhawa
UMTS Universal mobile telecommunication system
ysrandhawa@hotmail.com HSPA High speed packet access
EvDO Evolution data optimized
1
IKG Punjab Technical University, Jalandhar, India QPSK Quadrature phase shift keying
2
Chandigarh Group of Colleges, Landran, Mohali, India OFDMA Orthogonal frequency division multiple
3
BCET (Affiliated to IKG Punjab Technical University), access
Gurdaspur, India SC-FDMA Single carrier frequency division multiple
4
Lyallpur Khalsa College (Affiliated to IKG Punjab Technical access
University), Jalandhar, India

13
Vol.:(0123456789)
K. Arora et al.

S-OFDMA Scalable orthogonal frequency division mul- 1 Introduction


tiple access
BDMA Beam division multiple access In today’s life, intelligent systems are flourishing in enor-
FBMC Filter bank multiple carrier multiple access mous fields and aid in resolving difficulties that arise in
BCH Bose–Chaudhuri–Hocquenghem Codes diverge areas related to environment cleanliness like waste
LT Luby transform codes management, smart roads, air pollution control, forest fire
UWB Ultra wide band communications detection; security purposes like smart phone detection,
SE Spectral efficiency earth quake early detection and many more. Innovative
NBLC Non-binary LDPC codes Wireless Networks are the prime foundation behind these
PCM Parity check matrix trending applications and their exertion in these systems
BG Bi-partite graph are greatly influenced by emanating technologies i.e. IoTs
RLDPC Regular LDPC codes (Internet of things) and are an integral part of 5G. The main
IRLDPC Irregular LDPC Codes challenge these systems are facing is the energy manage-
CP Closed path ment [1, 2]. As most of the energy is wasted to retransmit
CG Connected graph the corrupted packets, one solution is to use variable size
SG Sub graph of frames. If the medium of transmission is not good (i.e.
ISG Induces sub-graph fading effects are very prone), then larger frames are used
TC Trapping cycle else smaller frames can be utilized as smaller frames have
ETC Elementary trapped cycle less tendency of error. Another effective solution is the use
ML Maximum likelihood of highly efficient error-correction-codes [1–5].
LR Likelihood ratio Channel coding is considered to be the most essential but
LLR Log likelihood ratio crucial component of today’s wireless communication sys-
AWGN Additive white Gaussian noise tems [1]. The prominent Channel Coding theorem by Shan-
BDC Binary discrete channels non affirms the feasibility of codes that can led the reliable
SC Successive cancellation transmission of information via a clattery channel if and
SSC Simplified successive cancellation only if the transmission rate is less than or equal to channel
LSC List successive cancellation capacity. The significant facts that came out of this theorem
CRC​ Cyclic redundancy check include: 1. Codes need to be selected in a random fashion. 2.
MPA Message passing algorithm Joint equipartition property should be defined amidst trans-
AMA Addition–multiplication algorithm mitted and receiving sequences. 3. Decoding Algorithm
MS Min–sum algorithm needs to be optimum or sub-optimum so as to minimize the
WB Weighted bit-flicking probability of error [3]. Codes (Turbo and LDPC codes)
BS Boot-strapping devised in previous two decennaries, promised to fulfill this
WC Weighing-coefficient channel coding criterion. Achieving channel capacity that
QAMA Q-ary addition–multiplication algorithm too with lower encoder-decoder intricacy is the prime chal-
EMSA Elongated minimum–sum algorithm lenge of channel coding techniques. Turbo Codes were dis-
TEMSA Trellis-based-EMSA covered to be a capacity-approaching code by Glavieux et al.
DP Deviation paths [4] but these codes had very high encoding and decoding
FPMSA Fixed path minimum sum algorithm intricacies. The exemplary realization of Turbo codes made
GF Galois field them the main contenders of FEC (Forward Error Correc-
BER Bit error rate tion) Codes for LTE and other 4G standards. LDPC Codes
RW Re-weighing re-devised by Mackay et al. [5, 6] as a capacity-achieving
RS Re-scheduling code with better performance that too with lower decod-
RBPA Residual-belief-propagation-algorithm ing intricacy via parallel decoding-structure implementa-
NBPA Node-wise-belief-propagation-algorithm tions [7, 8]. Main drawback of LDPC Code is its encoding
LBPA Layered-belief-propagation-algorithm intricacy which is further eradicated by the aid of several
DSOC Deep Space optical communications design methods [8–10].In addition to the above advantages
HSC Helicopter satellite communications of LDPC Codes; optimization in this field is carried out by
PPMBPC PPM based Poisson channel the density-evolution-methods [7, 11] and external-informa-
QC-LDPC Quasi-cyclic LDPC Codes tion-transfer functions [12–14] and LDPC has come out as
MCS Mobile-satellite communication system one of the major channel coding contender for 5G.
MIMO Multiple input multiple output

13
A survey on channel coding techniques for 5G wireless networks

The near-by-future technologies i.e. 5G demand for (General Packet Radio Service)) systems. EDGE(Enhanced
higher reliability, higher data rates (about 1000 times as Data Rate for GSM Evolution) technology (2.75G) further
compared with 4G) which has led to the requirement of geared up the data rates up to 200 kbps. The research on
more-efficient channel coding techniques. Till now, LDPC 3G started in 1980s and continued for 15 years. 3G systems
(Low density Parity Check) Codes have replaced turbo codes revolutionized the wireless technology by combining mobile
in the upcoming wireless technologies such as mMTC (mas- phone, laptop PC and TV. 3G features included global roam-
sive machine type communication) [15], D2D (device to ing, sending/receiving e-mails, high speed web, navigation,
device communication), URLLC (ultra-reliable low latency video conferencing and TV streaming. UMTS (Universal
reliable communications) [16] etc. Some advancements are Mobile Telecommunication systems), HSPA (High speed
still required in the designing of LDPC Codes in terms of Packet Access), CDMA2000, EvDO (Evolution data opti-
variable block lengths and rates so as to make them more mized) are amongst 3G standards. 4G systems relied upon
felicitous for 5G. Further, the emergence of polar codes adaptive digital signal processing and smart antenna tech-
[3, 17–19] in this field introduces new dignified perimeter nologies. The salient features of 4G include variable band-
known as polarization so as to attain channel capacity in dis- widths, enormously high data rates, reduced delay (as there
crete binary channels without memory. This process involves is no BSC (Base station controller) or RNC (Radio network
the use of polarization transform that converts the independ- controller)). Table 1 gives the detailed description of various
ent channels into variable-reliability channels. Iterative use parameters for all generations [21].
of this transform shows an eloquent change in the reliability
of these channels which make the already-reliable channel 1.2 5G channel coding: a challenge
to become ‘more-reliable’ and clattery channel to become
further ‘less-reliable’. So, the data-bits are send over ‘more- Since 5G is considered to incorporate the key characteris-
reliable’ channels and rest-bits (frozen bits) are send over tics of higher transmission rate of around 10 Gbps, delay
‘less-reliable’ channels. less than 1 ms, higher energy efficiency and higher spec-
tral efficiency as compared to 4G. Hence the 5G channel
1.1 Expedition from 1G to 5G coding techniques must adhere to these requirements too.
Figure 1 shows its pictorial representation. Following points
Marconi et al. [20] an electrical engineer, disengaged the describes the Channel Coding Challenges for 5G wireless
path of wireless communications through the invention Networks.
of wireless telegraphy in 1880s. Furthermore, H. Hertz
proved the wave nature of radio waves. These inventions 1. Design of Shorter Code-word lengths and parallel
led to the deployment of first wireless voice transmission decoders 5G networks demand for higher data raterang-
set-up in 1915. First generation cellular systems were ini- ing from 100 Mbps to 10 Gbps and Channel Coding
tially established in 1980s. AMPS (advanced mobile phone adds extra bits, which could decrease 5G transmission
systems) technology was firstly deployed by Bell Labs in rate requirement. To overcome this challenge, M. Khan
1970s. Other systems supported by 1G were Japanese NTT et al. illustrated the use of LDPC codes along with punc-
(Nippon Telegraph and Telephone), European TACS (Total turing methods which reduces the code-word lengthto
Access communications system). These systems faced sev- increase throughput [22]. As the data after reception
eral drawbacks including security threats, severe distortions, must pass through channel-decoder, so if N = 10 itera-
imperfect hand-overs, huge phones, inferior voice quality tions are needed for convergence and if processing
and poor capacity. These systems used analog modulation units in the channel decoder are working at frequency
schemes (Frequency modulation) and do not incorporate any of f = 400 MHz, then (N*throughput/f) i.e. 500 paral-
type of FEC coding schemes. In 1990s, 2G systems aimed lel processing units are needed so as to attain 20Gbps
to support good quality voice calls were devised and these maximum data rate and adding 500 processing units. F.
systems work on digital technologies. These systems incor- Kahn et al. described the inherent parallelism of LDPC
porate the ethical features of less battery power consump- decoder that can aid in escalating the transmission rate
tion, improved voice quality, lesser noise, good security and to some extent. Further the detection and demodulation
better error correction-detection capabilities. These systems can be combined to form a Joint detector and demodula-
used convolutional codes and parity check codes as their tor and which can also work at shorter-code words [22,
channel codes. Amongst 2G systems, GSM (Global System 23].
for Mobile) systems) are dominant and in use world-wide. 2. To minimize channel decoding delay For 4G technolo-
Other 2G systems were IS-95 (Interim standard), IS-136 and gies, the round trip delay is around 15 ms. The foreseen
PDC (Pacific Digital Cellular systems). The need to sup- applications for 5G (which include Google glass, will
port data-services has led to deployment of 2.5G (GPRS have latency of around 1 ms [24].

13

Table 1  Expedition from 1G to 5G
Expedition Systems Access technol- Data rate Modulation Frequency band Band width Channel coding Switching Application
from 1G to ogy techniques

13
5G

1G (1980’s) Advanced Frequency Divi- 2.4 kbps Frequency 800 MHZ 30 kHz NA Circuit Voice
mobile sion Multi- Modulation
phone ple Access (FM)
systems (FDMA)
(AMPS)
2G (1990’s) Global Time Divi- 10 kbps Gaussian 850/900/1900 MHz 200 kHz Convolutional Circuit Voice + Data
System sion Multi- Minimum Encoding
for Mobile ple Access shift Keying
Commu- (TDMA) (GMSK)
nication
(GSM)
Code Divi- 10kbps
sion Multiple
Access
2.5G (2000) General Time Divi- 50kbps 200 kHz Cyclic Circuit Voice + Data
Packet sion Multi- Code + Punc-
Radio Ser- ple Access tured Convo-
vice (GPRS) (TDMA) lutional Code
2.75G Enhanced TDMA 200kbps GMSK + Phase 200 kHz Cyclic Code Circuit/
(2003) Data Rates Shift Keying +Punctured Packet
for GSM (8-PSK) Convolutional
Evolution Code
(EDGE)
3G (2005) Wideband Code Divi- 384 kbps Quadrature 800/850/900/18,001,900/2100 MHz 5 MHz Convolu- Circuit/ Voice + Data + Video
Code Divi- sion Multi- Phase Shift tional + Turbo Packet Calling
sion Multi- ple Access Keying Codes
ple Access (CDMA) (QPSK)
(WCDMA)/
Universal
Mobile
Telecom-
munications
Systems
(UMTS)
CDMA2000 CDMA 384 kbps 1.25 MHz
(Code
Division
Multiple
Access)
K. Arora et al.
Table 1  (continued)
Expedition Systems Access technol- Data rate Modulation Frequency band Band width Channel coding Switching Application
from 1G to ogy techniques
5G

3.5G (2008) High speed CDMA 5-30 Mbps QPSK/16- 800/850/900/1800/1900/2100 MHz 5 MHz Turbo Codes Packet Voice + Data + Video
Packet QAM Calling
Access (Quadrature
(HSPA) Amplitude
Modulation)
Evolution data CDMA 5-30 Mbps QPSK/16- 800/850/900/1800/1900/2100 MHz 1.25 MHz Turbo Codes Packet
optimized QAM)
(EVDO)
3.9G (2013) Long Term Orthogonal 100– QPSK, 1.8 GHz, 2.6 GHz 1.4 MHz to Fountain Codes, Packet Online gaming + High
Evolution Frequency 200 Mbps 16QAM, 20 MHz Turbo Codes Definition T.V.
(LTE) Division Mul- 64QAM
tiple Access
(OFDMA)/
SC-FDMA
(single Carrier
A survey on channel coding techniques for 5G wireless networks

Frequency
Division Mul-
tiple Access)
World-Wide Scalable 3.5 GHZ and 5.8 GHz 3.5 MHz and Packet Online gaming + High
Interoper- Orthogonal 7 MHz in Definition T.V.
ability for Frequency 3.5 GHz band
multiple Division Mul- and 10 MHz
Access tiple Access in 5.8 GHz
(Fixed (S- OFDMA)/ band
WiMAX)
4G (2015) Long Term Orthogonal 3Gbps for QPSK, 1.8 GHz and 2.6 GHz 1.4 MHz to Fountain Codes, Packet Online gaming + High
Evolution- Frequency downlink 16QAM, 20 MHz Turbo Codes Definition T.V.
Advanced Division Mul- 1.5 Gbps 64QAM
(LTE-A) tiple Access for Up-link
(OFDMA)/
SC-
FDMA(single
carrier fre-
quency Divi-
sion Multiple
Access)

13
K. Arora et al.

Online gaming + High

tion Video +Virtual


Ultra-High Defini-
Higher Transmission Rate (1000

Reality Applica-
Definition T.V.
times more than4G)
Application

tions
High Spectrum
Lower
Delay(<1ms)
5G Efficiency
Switching

Fountain Codes, Packet

Packet
Channel coding

convolutional
High Energy Efficiency (100
Turbo Codes

Turbo, Polar,

non-binary
binary and

LDPC and times compared with 4G)


tail-biting
techniques

codes
7 MHz,5 MHz,

Fig. 1  5G requirements
Band width

10 MHz
3.5 MHz,

60 GHz

Physical layer delay is reduced to around 50microsec-


onds in5G. So, this figure of delay is shared among all the
components present in physical layer. As few components
2.3 GHz, 2.5 GHz and 3.5 GHz

like synchronization block may have higher delay require-


ments. So, the channel decoder must possess the lower
Index –Modula- 1.8, 2.6 GHz and expected

delay by operating their processing units in parallelism. To


achieve a data rate of 10 Gbps, for a block of data with
length L = 10 Kbits, there will be delay of D = L/(Through-
Frequency band

30-300 GHz

put) = 1 µs which will be sufficient to attain an overall delay


of 1 ms. Several decoding algorithms are being designed
which could lower the decoding delay e.g. Belief Propaga-
tion decoding methods.
tion Schemes

3. Need of deliberately lower decoding error rate As 5G


Modulation

utilizes massive MIMO technology [21] that deals


with the introduction of large number of antennas and
enormous base-stations. The focus here is that the cost
should not increase as compared with the data rate pro-
(expected)
10–50 Gbps
Data rate

vided. In addition, as the data rate need to be increased


100–200
Mbps

by 100 times. So, their cost and power per bit must fall
by 100 times i.e. 5G channel coding techniques must
ensure less cost and must be energy efficient too [25].
Division Mul-

(S- OFDMA)/
Access technol-

multiple car-
rier (FBMC)
tiple Access
Orthogonal

sion Multi-

As 5G will shift their spectrum to mm wave frequencies


Filter bank
ple Access
Frequency

Beam Divi-

(BDMA)/

multiple

[26] so license cost for this spectrum should also be less.


Scalable

access

Liang et al. [27] utilized low-density parity-check codes


ogy

for non-coherent UWB communication systems so as to


improve energy efficiency. Energy Efficiency in decod-
5G-NR (New
World-Wide
Interoper-
ability for

ing process is measured by the count of decoded bits per


multiple

Wimax)
(Mobile
Access

Radio)
Systems

nano-Joule of the energy wasted in decoding.


Table 1  (continued)

4. Need of spectrum efficient channel coding Recent stand-


ards utilize the collaborative use of re-transmission
from 1G to
Expedition

protocol and error-correction-codes for lowering end-


5G (2020)

to-end delay essential for high-speed applications. But


this requires the need of additional bit in sending the
5G

13
A survey on channel coding techniques for 5G wireless networks

acknowledgement of the received ensemble data. Fur- Amongst the above listed codes in Table 2, Turbo, LDPC
ther each ensemble block consists of several chunks of and Polar Codes are the capacity approaching codes. Further
encoded blocks. If anyone chunk of encoded data gets amongst these codes, we limit our survey to the design of
corrupted, the whole ensemble block need to be re-send LDPC and Polar Codes only because of their lower delay,
which in turn lowers the spectrum efficiency (SE) [28]. efficient encoding and lower bit-error-rate performance that
As 5G trends focus on increasing data-rate hundred make them suitable to be used in 5G too. Also, it has been
times, so there are more chunks in one ensemble block observed that there exist few survey papers that compare
which leads to further reduction in the spectral efficiency. LDPC and Polar Codes on the basis of their encoding, decod-
To overcome this challenge, several techniques have been ing algorithms, performance, research issues and application
implemented [28, 29] that re-transmit selective chunks areas. The main contribution of this paper is enlisted as:
of corrupted data but at the cost of increased overhead.
Yang et al. [24] proposed a method which involves the • In this paper, the detailed description of 5G Channel
use of some interlinked chunks of encoded data which Coding techniques (LDPC, Polar coding) are elaborated.
improves the problem of spectrum efficiency and reduces This aids in the energy efficient implementation of 5G
delay too. Also, Non-Binary-LDPC Codes (NBLC) [30– networks which is one of the key requirement of 5G net-
32] have been proposed to escalate spectrum efficiency. works.
Also, it has been noticed that polar Codes [19] used in • Design of LDPC and Polar Codes with their essential
conjunction with 5G waveforms too help to increase parameters and different methods to construct its parity-
throughput as compared with turbo codes. check-matrix are elaborated in detail along with the dif-
ferent decoding techniques.
In addition, As 5G relies on the use of mm waves which • Decoding Algorithms involved and the various chal-
were earlier considered not appropriate for radio commu- lenges faced in the channel encoding techniques for both
nication but now due to the recent advancements in tech- algorithms are also enlisted along with their other areas
nologies, millimeter waves can be used for short-distance of application.
communication(i.e. for femto cells). 60 GHz band is idle • Further, bit-error-rate performance comparison of LDPC
which is proposed to be used in 5G which can lead to and Polar Codes has been done and their suitability based
increase in bandwidth (in terms of bps/Hz) and hence spec- on particular parameter is concluded in comparison table.
trum efficiency. [32].
1.4 Organization
1.3 Motivation
This paper is organized in the following way: Sect. 1 speci-
As described in the above section, enormous challenges exist fies the introduction and expedition from 1G to 5G in wire-
in the implementation of channel coding techniques so as to less communication. In this section, an outline of the channel
meet 5G requirements which highly motivated us to present coding methods used in these generations till date along with
the survey of various channel coding techniques. Further, the the challenges and motivation is described. Sections 2 and
work published by Shannon [33] has made “Channel cod- 3 elaborates the design and decoding algorithms of LDPC
ing” a vital element to build a capacity-approaching com- Codes and Polar Codes respectively. In Sect. 4, the issues
munication system. In his study, he clarified that it is viable addressed by different researchers and future directions in
to achieve error-less transmission provided the transmission Channel Coding have been discussed. Simulation Results for
rate should be lesser than channel capacity. Till now, enor- different LDPC and Polar Codes are presented and compared
mous channel codes have been devised based upon applica- in Sect. 5. Further, the application areas for LDPC Codes
tion demands and with advances in technologies. Channel and Polar Codes are explored in Sect. 6. Finally, conclusion
codes are broadly classified as block codes and trellis codes. along with the suitability of codes for different applications
Block Codes are the codes which scrupulously rely upon is included in Sect. 7.
the mathematics. These codes yield a block of bits as out-
put using pre-defined algorithms. Examples of block codes
include Hamming codes, Bose–Chaudhuri–Hocquenghem 2 Design of LDPC codes
(BCH), LDPC codes. Trellis codes [34] are a class of chan-
nel codes in which the present output relies not only upon LDPC Codes were first devised by Gallager. In recent
the current block of input bits but also upon the previous applications, LDPC Codes having short block lengths are
block of input bits. Examples of such codes include: Con- in use instead of earlier stated large block LDPC Codes by
volutional Codes, Turbo codes. Brief description of these the inventor. LDPC Codes can be depicted in two peculiar
algorithms is illustrated in Table 2. ways: one using parity check matrix (PCM) i.e. matrix

13

Table 2  Classification of channel coding methods


References Methods Description Advantages Drawbacks Decoding Algorithm

13
[89–94] Algebraic Codes Hamming Codes This type of coding works by This aids in lowering of These codes suffer with Maximum Likelihood
Golay Codes searching for the code-word BER the major drawback of Algorithm
that maximizes the hamming increased intricacy level
Reed Muller Codes
distance for large code-word
Bose–Chaudhuri–Hoc- lengths.
quenghem (BCH)
codes
[95–99] Probabilistic codes Convolutional codes These codes involved the use of These codes somehow Decoding and encoding Viterbi Decoding
linear feedback shift registers. improves the bit-error complexities are high
The researchers, at that rate but with restrictions
instant, noticed that the algo- on decoding and encod-
rithm behind code construc- ing complexities
tion need to be modified so as
that their practical enactment
can be made simple
Concatenated codes These codes involve the The usefulness of such Encoder complexity is still Decoding algorithm varies
concatenation of two linear codes is that these codes very high according to inner and
block codes. Indeed, one have lesser decoding outer encoder used
concatenated code with inner complexity
encoder as Reed-Solomon and
outer as convolutional codes
was introduced by NASA
[4] Turbo Codes Turbo codes exhibit randomness Because of the associ- These codes incorporate Iterative soft decoding
in their design method which ated parallelism in the very high decoding algorithms
permits for effective decoding architecture, the data rate intricacy. Also, Bit-Error-
algorithms. These codes are a is very high Rate is very high at lower
class of parallel concatenated values of Signal-to-Noise
codes with convolutional Ratio
encoders as its element-
encoders
[100, 101] Low Density Parity Check (LDPC) Codes LDPC Codes are the codes LDPC Codes can provide Encoding Intricacy is very
whose parity check matrix higher data rate along high
consists of lesser number of with excellent bit-error-
ones rate performance
K. Arora et al.
A survey on channel coding techniques for 5G wireless networks

representation and other using graphical depiction i.e. bi-

Decoding. Belief propaga-


partite graph (BG) as described in Sect. 2.1 and 2.2.

Successive-Cancellation,
List Decoding, Stack
Decoding Algorithm

2.1 Matrix representation of LDPC codes

Consider N is code word length, K is the information bits


and M represents the parity bits, Code Rate,CR = NK > N−MN
 .

tion
Consider ′ x′ represents input bits to the LDPC encoder then
LDPC code-word
[ ]can be represented as:C = xG where
GK×N = IK PK×N−K is the generator matrix, IK is identity

Decoding latency is more


matrix of dimension K × K and P is parity matrix of dimen-
Encoding and decoding
intricacies are high

sion K × (N − K) . Parity Check Matrix can be illustrated as:


H(N−K)×N = PT IN−K  . Also because of orthogonality between
two matrices: HGT = 0 . PCM can be constructed by the aid
Drawbacks

of various methods listed in Table 3.

Definition 1  (Sporadic LDPC Codes) PCM of LDPC


Codes is sporadic i.e. lesser count of finite-valued elements.
huge data with effective

intricacies are optimum


These codes can work on

encoding and decoding

Pre-dominantly, finite-valued elements are 1. If instead of


Encoding and decoding

‘1’ some other digits are chosen, then such codes are speci-
fied as non-binary LDPC Codes. These codes are described
over Galois field GF(2m ). As an example for GF(4) field,
infinite-length coded sequence procedures
Advantages

parity check matrix incorporate elements {0, 1, 2, 3} [35].

Definition 2  (Regular and Irregular LDPC Codes) If


Count of finite-valued elements of rows in PCM is Nr and
a finite-length input sequence

promise to provide Shannon-

encoding and decoding com-


code-rate. With these codes,

can be generated with aid of

sion same or meagerly more


Fountain Codes depict a class

Polar codes are the codes that


from a sub-group of dimen-

capacity achieving codes in

for column is Nc  , then in regular LDPC codes


as compared with the input

binary channels with lower


recovery of input sequence
in a way that promises the
of codes having variable

(RLDPC), Nr = Nc while it is distinct for irregular LDPC


(IRLDPC) Codes and For RLDPC, Rate of Code
CR = 1 − Nr  . In case of IRLDPCs, the degree of bit vertices
N
c

dbmax
and par ity ver tices are 𝛬(a) = 𝛬i ai−1 and
Description

sequence,

plexities

i=d
bmin
dpmax

𝛾(a) = 𝛶i a  , where dbmax and dbmin ( dpmin anddpmax )
i−1
i=d
pmin

depicts the highest and lowest values of degrees of bit ver-


Fountain Codes: Luby transform codes (LT

tices (parity vertices); 𝛬i denotes the links corresponding to


degree-i bit vertices and 𝛶i depicts the links for degree-i
parity vertices.

Definition 3  (Systematic and Non Systematic PCM) Con-


sidering non-systematic PCM as ′ A′ and assuming its sys-
codes), Raptor Codes

tematic representation as ′ H ′ . Bi-partite graph is the pictorial


representation of PCM. It has two pairs of vertices. First pair
contains ‘ N ′ bit vertices b1 , b2 , … bN and second pair con-
[71, 72, 102–108] Polar Codes

t ains ′ M ′ par ity ver tices p1 , p2 … pM  . Consider


Methods

⎡1 1 1 0 0 1 1 0 0 1⎤
⎢1 0⎥
Table 2  (continued)

0 1 0 1 1 0 1 1
⎢ ⎥
A = ⎢0 0 1 1 1 0 1 0 1 1⎥
⎢0 1 0 1 1 1 0 1 0 1⎥
References

⎢1 0 ⎥⎦
⎣ 1 0 1 0 0 1 1 1
[101]

13

13
Table 3  Classification of Parity Check Matrix construction methods
Algorithm Algorithm used for Description Features
Random or Structured
Codes

Davey and Mackey Construction [52] Random In this method, PCM of dimension is generated by ran- These codes refrain the occurrence of length-4 cycle
domly constructed rows one after other taking care of two Because of its random-code design, the encoding is in-efficient
conditions: (a) At every step, maximum count of ones in
a column must be Nc (b) At every step maximum count of
ones in a row must be Nr
Gallager Construction [53] Random Gallager devised a probabilistic design method in which Encoding process is less complex
dimensions ( N , M) of parity-check-matrix are selected in Length-4 cycles can be avoided only for few specific row
such a way that N = l.Nc and M = l.Nr and l is an integer combinations
such that NN = NM = l.
c r
Cyclic-Shift Matrices [109, 110]. Random The Parity-check matrix in this method is constructed with Decoding complexity can be reduced by using tree-based
help of circularly shift of individual columns by some permutation matrices
integer
Partial geometry [111, 112] Structured For designing a partial-geometry system, the situations Low error floor
need to be checked are as follows; for pre-defined integers
𝛼 ≥ 2, 𝛽 ≥ 2 and 𝛾 ≥ 1:
(1) two points must lie on a single line solely.
(2) every point must lie on 𝛼 lines.
(3) every line must cross 𝛽 points.
(4) in case, a point p doesnot lie upon line E, then there exists
𝛾 lines, every line crosses through p and one point upon E
Graphs [75, 113–117] Structured Bifurcating row and column operations aid in the reduction of Reduction in error-floor
number of cycles in bi-partite graph. In addition bit flipping
and progressive-edge-growth methods help in increasing
girth
Combinatorial designs [36] Structured In this method, a group of bits are assigned to a subset of Efficient encoding
check-nodes conditioned on parameters like bit-count cor-
responding to each check-node
Protograph Based LDPC Codes [37] Structured Constructed using repetition of the same base matrix BER is less
K. Arora et al.
A survey on channel coding techniques for 5G wireless networks

Definition 4  (Vertices, Links and Degree) Assume a non-


directed
{ graph }as G(V, L) where V = {v1,v2 , … vk } and
L = l1 , l2 , … lm are the k vertices and m links of G respec-
tively. The link between two vertices ( vi , vj ) is represented
by vi vj orvj vi where 1 ≤ i ≤ k and 1 ≤ j ≤ m.Number of links
corresponding to a particular vertex is called the ‘degree’ of
that vertex and is depicted by d(v) . The maximum value of
d(v) for a particular graph is denoted by D(G) and minimum
value is denoted by d(G).

Definition 5  (Chain) ‘Chain’ between two verti-


ces vi and vk+1 is defined as series of vertices and links
v1 , l1 , v2 , l2 , … vk , lk , vk+1 while vertices or edges need not to
be different.

Definition 6  (Path) ‘Path’,P is defined as finite/infinite set


Parity of links which exist between a set of vertices that are dif-
vertices ferent from each other. ‘Length’ of Path is described by the
number of links.

Definition 7  (Open and Closed Path) ‘Open path’ is the


name given to path in case its vertices are distinct else path
Bit vertices is known as ‘closed path’(CP ) or cycle.

Fig. 2  Bi-partite graph corresponding to parity check matrix A Definition 8  (Girth) ‘Girth’ is the word which describes
the shortest CP.

Definition 9  (Connected Graph) If there exists ′ P′ between


every node, the graph is labeled as ‘Connected Graph’ ( CG).

Definition 10  (Tree)‘Tree’ (T) is the name given to ′ CG′


without CP . ‘Rooted Tree’ is ′ T ′ in which one particular
vertex is labelled as ‘Root’. A vertex is called leaf if d(v) is 1.

Fig. 3  ISG and normal graph Definition 11  (Leafless Graph) Leafless Graph is CG with
d(G) greater than or equal to two.

Bi-partite graph corresponding to A is shown in Fig. 2 Definition 12  (Sub-Graph(SG)) If v1 , v2 and v3 are vertices
[35]. A(j, i) = 1 indicates a link in the bi-partite graph. of a particular graph G , which too are the vertices of some
Length of closed-path for bi-partite graph can be even- other graph W  , then W is the SG of G.
number only. In bi-partite graph, CP never exceed the value
of four. Definition 13  (Induced Sub-Graph (ISG)) If two vertices of
SG are connected by a link and that link is also present in G ,
2.2 Preliminaries related to graph theory that particular Graph is labelled as ISG as depicted in Fig. 3.
and bi‑partite graph (BG) for LDPC Codes The set of parity vertices with odd and even degrees in ISG
are depicted by do and de respectively. Its length is depicted
LDPC Code performance critically depends upon the Bi- by number of bit-vertices (x). All ISGs with same length and
partite graph design. The cycles associated with Bi-partite same number of do (say y) belong to common class (x,y) [42].
graph design have a greater impact upon error-floor too
[36–41]. Hence some basic definitions related with graph Definition 14  (Trapping Cycle(TC)) It depicts the ISG of
theory are defined below: a particular BG for LDPC Code.

13
K. Arora et al.

Iterative Decoding Algorithms (Message Passing Algorithms) Steps to be followed:

Step2:Parity Vertex Updation: Utilizing this information send by


Based upon Firm (Hard) Decision
parity vertices, each bit-vertex checks for its parity-check equation.
In case all equations are satisfied, the algorithm stop else every
Bit Flicking Algorithm bit-vertex transmits the messages to their connected parity-check
vertices depicting the equation is satisfied or not.
Step3:Bit Vertex Updation: In case, parity check vertices receive huge
Based upon Flexible (Soft) Decision
number of notifications of “un-satisfied”, the parity-check vertex
flicks its present value; else the bit-value is preserved. In case, the
Using Maximum Posteriori Algorithm limit for allowed number of iterations is reached, the algorithm
(Addition-Multiplication Algorithm) stops and non-convergence is reported.

Using Min-Sum Algorithm

2.3.2 Addition multiplication algorithm


Weighted Bit Flicking

Modified Weighted Bit Flicking This algorithm involves the utilization of decoding algo-
rithms that perform in a dispersed manner for the decod-
Boot-strapped Bit Flicking ing of the received symbols [6, 43–45]. Such algorithms
are based upon maximum-likelihood (ML) algorithm. This
Gradient Descent Algorithms
method is the most intricate one but its bit-error-perfor-
mance is the best. To reduce the intricacy, min-sum method
Set Partition Algorithms
has been devised which is closest approximation to ML.
Min-sum method provides a good balance between bit-
Fig. 4  Decoding algorithms for LDPC codes error-rate and intricacy. This method is used by weighted
bit-flicking algorithm, modified bit-flicking algorithm and
improved modified bit-flicking algorithm. Iterative methods
Definition 15  (Elementary Trapped Cycle(ETC)) It depicts involve the use likelihood ratio (LR) or log-likelihood ratio
TC having d value as one/two. Normal Graph for an ETC (LLR). Practically, LLR is less intricate.
is acquired from ISG by eliminating parity vertices of ISG.
2.3.3 Weighted bit‑flicking algorithm
2.3 Decoding algorithms
Assume a LDPC code-word of W of dimension N is send over
Based upon firm, flexible or hybrid decisions, decoding a Additive White Gaussian
{ } Noise (AWGN) channel. Its parity
algorithms for LDPC Codes are broadly classfied into two check matrix H = hij where 1 ≤ i ≤ M and 1 ≤ j ≤ N con-
major categories described in Fig. 4. Flexible(soft)-decision tains M rows and N columns. If BPSK modulation is used and
decoding algorithms provide significantly good performance received signal is r = s + w  ; where sj = (−1)Wj   , where
but at the cost of complexity which can obstruct lower-delay j = 1, 2, 3 … N  . Consider noise vector consists of i.i.d. (sta-
applications. For such cases, firm(hard) decision decoding tistically-independent and identically-distributed) elements
algorithms can be useful with somewhat compromise in and their variance value is 𝜎w2 with power-spectral-density of
bit-error-rate. N0
 . If Es denotes signal-energy for each bit and Rc denotes
2
rate for this code, then 𝜎w2 = 2R (E1 ∕N ) . The sign for every
2.3.1 Bit‑flicking algorithm c s 0
received-bit value gives an estimation value e for bit-transmit-
ted [46]. If e is the estimated value( for (code-word
)) W
This algorithm relies upon hard decisions. Three steps acquired by firm-decision on ej = 1 − sgn rj ∕2 ; where
involved in this method are described below: ( ) ( )
sgn rj = 1ifrj ≥ 0;elsesgn rj = −1.
Steps to be followed:

Step1:Initiation: Initiate each parity vertex with the value of bit


acquired after passing through the channel and pass these values to
its connected bit vertices

13
A survey on channel coding techniques for 5G wireless networks

Steps involved in Weighted bit-flicking algorithm 3 Design of polar codes


{ }
Step1: Calculate the syndrome vector Si  , where 1 ≤ i ≤ M using Polar Codes are emanating codes that were introduced in
Si = ehi mod2 . In case Si is a null vector for all values of i  , stop the
t

procedure.
2009 by Arıkan [19]. These codes are the foremost codes
Step2: For each
to attain the channel capacity for binary discrete chan-
{ bit (fixed) vertex 1 ≤ i}≤ M  , calculate:
(i)
rmin
| |
= min |rj | ∶ 1 ≤ j ≤ N, hij = 1 nels (BDC) without memory. Despite this, lower encoding
| | and decoding complexities make these codes to gain more
Step3: Corresponding to each parity (variable) vertex 1 ≤ j ≤ N  , attention. These codes are derived from Reed–Muller (RM)
compute
∑ error
� vector as:
Ej =
� (i)
2Si − 1 rmin
Codes and promise to provide Shannon capacity.
i∶hij =1

Step4: Flick jth bit corresponding to the maximum value of Ej 3.1 Polar coding
Step5:Repeat step no. 1­ st, ­3rd and ­4th, till are syndrome bits become
zero or limit for iteration is attained. Polar Coding can be represented in two different manners:
1. Bi-step Procedure. 2. Wedge Product.

2.3.4 Boot‑strapping or modified weighted bit‑flicking 3.1.1 Bi‑step procedure


algorithm
Polar encoder works in an iterative way and the polarization
Here we define received element rj and its corresponding process is a bi-stage phenomenon where first stage involves
| |
vertex j is considered “non-reliable”, if |rj | < 𝛽 , (where 𝛽 is channel-mixing and second consists of channel-isolation.
| |
a fixed value calculated initially) else reliable. Bit-vertex is
called as ‘reliable’ w.r.t. a non-reliable parity-vertex, in case 1. Channel mixing Considering ‘ K  ’ bits need to be send
if entire parity-vertices connected to that bit-vertex are reli- over ‘ B ’ binary discrete channels (BDC) without mem-
able. This algorithm is just similar to one iteration of min- ory. The channel mixing is done by taking two inputs
sum or max–sum algorithm [43]. first and combining them as depicted in Fig. 5.

Steps involved in Modified Weighted bit-flicking algorithm Consider B ∶ 𝜒 → Z depict a BDC without memory,
where 𝜒 = {0, 1} depicts input symbol and z denotes
Step1: Decoding process is started by recognizing and removing
entire non-reliable parity vertices. output symbol with state-conversion probabilities
Step2: Allocate new rectified values and reliability values to the B(z|b), z ∈ Zandb ∈ 𝜒 .The channel B2 ∶ b2 → z2 is hereby
removed parity-vertices via the reliable bit-vertices. If rj represents

formed using Eq. (1) as:
the rectified value for a removed parity vertex j as ( )
� ∑ ∏ � � �� B2 b1 , b2 |z1 , z2 = B(z1 |b1 ⊕ b2 )B(z2 |b2 ) (1)
rj = rj + sgn(rj ).minj� ∈U(i)�j �rj �
i∈R(j) j� ∈U(i)�j � �
where R(j) represents the near-by reliable parity vertices and U(i)�j The channels formed in Fig. 5 and using two more input
depicts all the near by vertices of bit-vertex i other than j. bits, new channel B4 ∶ b4 → z4 is created with the aid of
Step3: In case, no reliable bit-vertex is connected to removed parity Eq. (2).This process progresses till the formation of BK
vertex, the received value is re-allocated to that vertex.
channel.
Step4: Corresponding to the entire reliable parity vertices rj is
unchanged i.e.rj = rj . Algorithm then continues with the traditional
� ( ) ( )
B4 b41 |z41 = B2 z21 |b1 ⊕ b2 , b3 ⊕ b4 B2 (z43 |b2 , b4 ) (2)
weighted bit-flicking algorithm.
If 𝜋4 represents permutation matrix. Let r1N depict random
In addition to above traditional algorithms, many algo-
vector of set { (r1 , r2 , … rN }) and ri depict the sub-set{ri , … rj }
j
rithms have been devised by researchers [45, 47–50]. Low-
.So 𝜋4 is r1 = r1 , r2, r3 , r4 . (
4
)
intricate LDPC decoding algorithm based upon look-up-
tables abstain the need of some mathematical-operations ( The m a p p i n g) b41 frominputofchannelB4 → x14
totheinputofchannelB4 can also be illustrated as: x14 = b41 G4
thereby reducing the complexity. The algorithms known as
box-minus/plus algorithms further aid in reducing decoding ⎡1 0 0 0⎤
⎢1 0 1 0⎥
complexity. Extended Min-Sum Algorithm and trellis-based where (G)4 = ⎢ ⎥ is the generator matrix. In general-
algorithms further reduce decoding complexity and increase ⎢1 1 0 0⎥
⎣1 1 1 1⎦
throughput too. Further in this direction partition based min-
ized form, channel mixing process can be illustrated as:
sum algorithm have been introduced [51].

13
K. Arora et al.

Fig. 5  Formation of channel in polar encoding

3.1.2 Using wedge product

The above steps of channel mixing and channel isolation can


be combined[ in one step using kernel of length
]
10
K  = 2.T = and T ⊗p depicts pth wedge product of
11
T ⊗ T ⋯ ⊗ T ( p times).
Generator matrix can also be defined in terms of T as
described below:

(G)K = ⊓̃ K T ⊗p = T ⊗p ⊓̃ K (5)
Fig. 6  Channel isolation process step in polarization
where ⊓̃ K denotes bit-rescinding matrix.
Again considering Fig. 5, the detection procedure on the
( ) ( )
BK bK1 |zK1 = BK bK1 |zK1 (G)K for all zK1 ∈ ZK , bK1 ∈ 𝜒 K . receiver side initiates the decoding process with the pre-
(3) diction of b1 from channel outcomes. Then using b1 along
with the channel outputs, more reliable detection of b2 can
be done. This reliability can be quantified by a parameter
2. Channel isolation The second step involves the isolation
known as Bhattacharya reliability metric.
of channels so as to form ′ K ′ different channels. Figure 6
below depicts the formation of 4 different channels.
Definition 16  (Reliability Metric) Consider R(B) denote
the Bhattacharya reliability
√ metric of channel B and is rep-
In isolation process, BK channel is converted back into
resented as: R(B) = B(z�0)B(z�1).

K ′ individual channels defined by BiK ∶ 𝜒 → ZK × 𝜒 i−1 ,
1 ≤ i ≤ K with state-conversion probabilities:
Example  Considering the probability of error is 0.5 , then
( K i−1 ) ∑ the capacity of clattered BDC channel (shown in Fig.
B(i) z1 , b1 |bi =
1
B (zK |bK ) ( a)) )is
K
2K−1 K 1 1 (4) I(B) = 1 − 0.5 = 0.5. By Mixing two input bits b0 , b1  ,
bKi+1 ∈𝜒 K−i
a joint channel ( B, B) is designed with ouput bits ( z1 , z2 ) .
( )
where zK1 , bi−1 depicts the outcome of channel B(i) and bi Capacity of this joint channel is 2I(B) . Using the chain rule
1 K
represents input. of mutual information, the joint channel can be spilt into two
sub-channels:B+ (depicted by blue line) and B− (depicted by

13
A survey on channel coding techniques for 5G wireless networks

green line). Thus, Channel splitting and channel isolation 4 Issues addressed by different researchers
processes convert the ‘same reliability mutually-exclusive and future directions in Channel Coding
channels’ into ‘variable-reliability
( ) polarized channels’ with field
joint capacity as I B+ + I(B− ) = 2I(B) and Here mutual
information
( ) of sub-channels is illustrated by the inventor [] Various issues addressed by different researchers are
as:I B+ = 2I(B) − I(B)2 and I(B− ) = I(B)2 . With the aid described below and their tabular description is depicted
of these expressions, it has been proved that after polariza- in Table 5.
tion, B− channels(less-reliable) channels have capacity lesser
than channel B and B+ channels(more-reliable) channels have 4.1 Encoding complexity
capacity more than channel B . As the code-length K grows,
this process occurs in an iterative way and two cusps of polari- LDPC Codes are categorized into two types: (a) Computer
zation channels (one that are highly reliable and others that are generated random LDPC codes which are more intricate
less-reliable) are created. This reliability improvement can be to decode due to their randomized structures. (b) Well-
illustrated with the aid of the following Fig. 7. designed structured codes. These codes have lesser encod-
ing intricacy because of their structured design. QC (Quasi-
Polar Coding can be done in systematic, non-systematic Cyclic) codes [52] are one of these structured codes whose
and joint ways. Considering the above figure and K = 4; if encoding process includes the use of simple shift registers.
the input bits are designated to b2 , b4 (i.e. more raliable chan- Their error performance is identical with random codes but
nels); such type of polar coding belong to the class of ‘non- because of lesser intricacy, these codes are gaining attention
systematic coding’. If the input bits are designated to x2 , x4 , as compared with random LDPC Codes [35, 53].
type of polar coding is ‘systematic’. Systematic polar coding Researchers have used variants of parity check matrices to
has performance better as compared with non-systematic ensure lesser complexity of operations involved in encoding
polar-coding in terms of bit-error-rate. and decoding process. One such parity check matrix given
by Li et al. [54] uses circular shift of the rows and is able
3.2 Decoding algorithms to provide linearity in complexity that relies upon the count
of parity bits. In another attempt to reduce the complexity
Various decoding algorithms can be used depending upon Huang et al. [55] used matrix operations and with the aid of
the application requirements are enlisted in Table 4.

Stage 2

(a) K=

Less Reliable Channel

More Reliable Channel (c) K=

(b) K=

Fig. 7  Reliability improvement using polarization process

13

Table 4  Description of decoding algorithms for polar encoding


Method Description Features:

13
Successive Cancellation (SC) Decoding [102] Interference caused by the past bits got cancelled in each successive step. Complexity: O(KlogK) where K represents length of code-word
Amongst the branches incorporated with message bit, this algorithm Performance is not good in case of large code-word lengths
involves selecting a branch with higher probability. In case of false deter-
mination of a single bit, the whole procedure fails. This algorithm is not
optimal one
Simplified Successive-Cancellation (SSC) Repeated computations by SC decoding are reduced by using three catego- Less complex. Complexity is reduced up to twenty times
Decoding [102] ries: rate 0, rate 1 and rate r
LSC(List Successive Cancellation) [71] Consider X (i) represent a group of contenders in the ith step in decoding and Complexity: O(MKlogK)
Reduction in dimension of list will degrade the performance of
list. Following are the steps of this algorithm: polar codes
|X (i) | is the dimension of X (i) . If M is maximum permitted dimension of the
| |

1. Initiation: Consider an empty vector as the single contender initially Quite burdensome to calculate the ideal value for dimension of
present in the list and fix its probability to one list
2. Estimation of bits: Bits are guessed in succession with indices i = 1,2,
…K if the K is the length of code-word
(2a) Inflation: Corresponding to each contender present, i-lengthened
sequences are generated and decoded as bit 0 and bit 1 respectively

M candidates with higher probabilities leaving rest


(2b) Candidacy: If the ||X (i) || is lesser than M  , go to next step else retain only

(2c) Elimination: If these probabilities are below a threshold, the candidates


are eliminated
3. Confirmation: Encode each contender present in list and calculate their
likelihood functions and probabilities
SCS (Successive Cancellation using Stack) Consider d denote the maximum value of depth for stack and t denotes the This decoding works similar to Viterbi decoding in a stack
[103] depth of stack at particular instant: Their performance matches with SC decoding and these codes
1. Initiation: Assume t = 0 . Calculate the corresponding cadent of this are less intricate too
single-vertex track and store this value with in the stack simultaneously Memory occupation is slightly more
increment the value of t . This algorithm is not useful if stack-size is short
2. Retrieving from stack: Retrieve the track pi1 = p1 , p2 , p3 , … pi  ,
( )
i ∈ {1, 2, … , K) denotes the i arms of the tree. Corresponding to a track,
the arm at ith level depict the bit having value of pi . Decrement t by 1
3. Expanding: In case
i+1
the track to two new tracks
( the decoded) bit is non-reliable, expand the present
p1 , p2 , … , pi , 0 and = p1 , p2 , … , pi , 1 and compute the recent metrics
(track by p1 = ) p1 , p2 , (… , pi , 0  ; else expand
)
for these tracks
4. Saving: Corresponding to a data bit, if t > d − 2, remove that track
from stack and decrement t by 1. For reliable bit, save the expanded

pi+1
1
= pi1 , 0 in stack and increment t by 1
tracks in( stack) and increment t by two. In case of non-reliable bit, save

5. Arranging: Rearrange tracks in stack starting from upper to lower value


with decreasing value of metric
6. Conclusion: If upper track in stack reaches leaf node of tree, method
terminates and gives the upper path as the resultant sequence. Else jump
to step 2
K. Arora et al.
Table 4  (continued)
Method Description Features:

CRC (Cyclic Redundancy Check) decoding Concatenation of CRC encoder before polar encoding and CRC decoder This decoding method can be concatenated with other listing
[104, 105] after polar decoding reduces the bit-error rate further and makes it even algorithms but that will increase the complexity too
better than turbo codes. In addition to above five steps, sixth step is modi-
fied and one more is included
Conclusion: If upper track in stack reaches leaf node of tree, go to next step
CRC based conclusion: If this step is also passed, then the upper path is
concluded as resultant sequence and method terminates; else jump to step
2 depicted in above method. But if the number of iteration reaches the
maximum limit, the algorithm fails
Hybrid Decoding [106] It combines the listing and stacking algorithms. It works in two different This algorithm has bit-error rate performance similar to SC
operating states: continue and delay mode. This works similar to the stack- decoding and too with lesser complexity
decoding algorithm but transits between these two states
While searching for the optimal track with the help of stack, it simultane-
ously checks for the upper limit of depth. In case, upper limit is reached,
this algorithm waits and stops finding tracks further and delay mode is on.
In delay mode, this method further expands the lowest-length track till all
the candidatures have achieved the same dimension. This mode is identical
A survey on channel coding techniques for 5G wireless networks

to List decoding. Here, number of tracks are limited to a maximum limit of


M  . Then it slides to the continue step till K dimension track arrives at the
top of stack
Sorting and Non-sorting Algorithms aided List Choosing M optimal tracks is a challenge in case of list decoding algo- Intricacy for Direct sorting: O(KM)
Decoding [72] rithms. This challenge is resolved by different sorting strategies of bubble Intricacy for Bubble sorting: O KM 2
( )
and quick sorting. Also, non-sorting algorithm of direct-selection reduces Intricacy for Quick sorting: O(KMlogM)
the complexity to linear
Belief Propagation [107] This method which is based upon soft decisions incorporates several pros as Reduced Delay and Higher data rate
compared with successive-cancellation method. This provides higher data
rate due to parallel structure also with lower-delay

13

Table 5  Issues involved in channel coding techniques


Issues Algorithms used Advantages Shortcomings

13
Encoding complexity Quasi-Cyclic (QC) LDPC Codes [35, 52, 53] Lesser Intricacy BER performance is marginal
Variants of Parity Check Matrices Circular shift of rows and columns Lesser Complexity (21% lesser than High Error Floor
[54] conventional LDPC Codes)
Double Diagonal Matrices [55] Less complex along with adaptable
code rates
Shorter Code Words [2] No iterations are required in decod- Error detection is limited to two bits
ing and correction to 1 bit only.
Decoding Algorithms and Complex- Proximities in the parity vertices updates [46] Computing load is less Decoding cycles are more
ity Weighted bit-flicking (WB) and boot-strapping (BS) methods [57, 58] Less Decoding Complexity Reliability is also less
Modified Weighted bit-flicking (WB) method [56] Better Reliability as compared with WC depends upon SNR
conventional WB method and
iterations are also reduced by half
Trellis-based-EMSA (TEMSA) [55] Inherent parallelism increases Complexity is dependent upon parity
throughput vertices update
Fixed path minimum sum algorithm (FPMSA) [55] No need to search all deviation paths Reliability is less
Using Max-log-Map Algorithm in Decoding of LDPC Codes [118] To reduce decoding complexity, Reliability is less
logarithmic operations are used so
as to convert multiplications into
additions
Reliability Re-weighing (RW) Methods [61] Improvement in Speed and BER Increases computational load
Re-scheduling (RS) Methods [61] Residual-Belief-Propagation-Algo-
rithm (RBPA)
Node-wise-Belief-Propagation-
Algorithm (NBPA)
Layered-Belief- Propagation-Algo-
rithm (LBPA)
Hybrid LDPC Codes with two encoding stages [61] BER less Two stages (sparse Generator matri-
ces and sparse Parity check matri-
ces) are implemented to achieve
better bit-error-rate but complexity
increases
Non-bipartite LDPC Codes [62] Remarkable good performance …
codes to be used in 5G Networks
Compromise between reliability and LDPC Coding with Iteratively Reliable Decoding [65] This method provides compromise between complexity and performance.
decoding complexity The soft outputs of belief propagation algorithm for each iteration are
used as reliability outputs
New message passing Algorithm [119] In case if girths are present in the LDPC Codes, conventional approach
of decoding does not provide optimal performance. So, a new approach
is proposed which schedule the updating process of messages at each
symbol node in accordance with length of girths
K. Arora et al.
A survey on channel coding techniques for 5G wireless networks

double diagonal matrices, it is possible to reduce complexity


by around 21% than conventional LDPC Codes along with
adaptability in code rates. But suffer from an ineluctable
drawback of error-floor (EF). Low complexity codes have
been proposed by Alabady et al. [2] which use shortened
code-words that are suitable to be used in IoT based applica-
tions. These types of codes have lesser bit-error-rate and pro-
Shortcomings

vide higher coding gain than LDPC codes. One of the major
advantages is that no iterations are required in decoding pro-
Complex

cess. Error detection is limited to 2 bits and correction to 1


bit only. Arikan proved that Systematic Codes work well as
compared to Non-Systematic codes. But this improvement
has to pay for in terms of complexity. Systematic codes need
Highly Secure as well as reliable
Low bit error rate even at large

large storage requirement and computations.

4.2 Decoding complexity
Reduction in EF

Several methods have been illustrated for decoding of LDPC


High Security

Codes in literature. Parameters considered for decoding pro-


Advantages

cess include block Length, decoding speed. It has been dis-


SNRs

covered that for large-block-lengths, Message passing Algo-


rithm (MPA) works great but the appearance of CP in shorter
LDPC detroits the proficiency of the codes. Dynamic sched-
Polar Codes transmission using channels that are unreliable for undesired
Tri-diagonal matrices and avoiding double-degree vertices in their parity

uling methods work well for short-block-length codes. Also,


some MPAs requires exclusively large number of decoding
cycles for lower-delay and energy-efficient applications.
Examples of these include AMA (Addition–Multiplication
Algorithm) and Min–sum (MS) algorithm. The diminution
in computing load can be done by the aid of proximities to
the decoding cycles in AMA. However, this method if used
with MS leads to the deterioration in BER. MS algorithm
LDPC Coded Algorithm for encryption [68]

performance can be improved by the aid of proximities in the


Polar codes for forbidden channels [69]

parity vertices updates. Applying this method will reduce the


Trapping cycles are analyzed [37–42]

calculation tariff but the count of decoding cycles remains


Protographed LDPC Codes [67]

unaltered [46].
LDPC Codes traditionally use addition–multiplication
approach (AMA) for decoding which are quite complex
check matrices [66]

but for fast and highly energy efficient systems, intricacy of


Algorithms used

AMA may not be accepted. Decoding algorithm’s decisions


may be firm, flexible or combination of both [56]. Enormous
user [3]

methods have been suggested in this area which include


weighted bit-flicking (WB) and boot-strapping (BS) [57, 58]
algorithms. BS method avoids the lesser accurate informa-
tion to travel through the bipartite graph used for decoding
Avoidance of cycles in LDPC Code

[59]. Zhang et al. [56] proposed a method that is based upon


one of the variants of WB method which made use of both
parity and message bits for computing the accuracy for each
bit instead of conventional method that was based upon the
Table 5  (continued)

reliability–checking of only parity bits. Reliability of this


Construction

method is better than the conventional WB method but suf-


fers from a drawback of dependency of weighing-coefficient
Error floor

Security

(WC) upon SNR. This method works with lesser intricacy


Issues

for longer code-word lengths. With proper selection of WC,

13
K. Arora et al.

this method requires half iterations as compared with con- with improved constellation diagram are designed using Gal-
ventional WB method. lager mapping contribute a remarkable good performance
NBLDPC (Non-Binary LDPC Codes) have potency to for Rayleigh Fading Channels.
replace binary LDPC Codes but these codes suffer with
the problem of decoding complexity. Many solutions [55, 4.4 Energy efficiency and inflexibility
59] have been devised to overwhelm this problem. Q-ary
addition-multiplication algorithm (QAMA) is considered to Kernel construction method makes the polar code inflex-
be best decoding algorithm for NBLDPC Codes and com- ible in terms of its length because of its length dependency
plexity is dependent upon parity vertices update. Elongated on the size of kernel. This issue is addressed by Korada
minimum-sum algorithm (EMSA) chooses most accurate et al. [63] using variable-length BCH kernels. Further in
values ( mn ) from a length-q set of values. The complexity is this direction, puncturing method devised by Wang et al.
of the order of dmn log2 mn , where d is the degree of parity [18] found magnifical attention as it also aids in increasing
vertices. One more algorithm for reducing decoding com- energy efficiency of system which is a key requirement for
plexity is trellis-based-EMSA (TEMSA) which seeks for the 5G networks. Also iterative decoding algorithms can lead
most accurate nodes in each trellis row and also seeks for to increase in delay and wastage of energy. So, in this direc-
the deviation paths (DPs) in the trellis. Inherent parallelism tion to reduce the number of iterations and thereby to make
makes this method more efficient in terms of throughput. energy-efficient system, sub-factor graph has been proposed
Fixed path minimum sum algorithm (FPMSA) by Huang [51]. Also it has been noticed that multi-level codes can aid
et al. [55] has also been devised which has the advantage of in spectrally efficient design of system [64].
not searching for DPs as it assumes static paths and is able
to decode with complexity of the order of d . 4.5 Error floor

4.3 Reliability Rate-Compatible-LDPC Codes suffers from the problem of


Error Floor [65]. This problem is addressed by Wu et al.
Finite Field LDPC Codes with Galois Field GF(q) where [66] using tri-diagonal matrices and avoiding double-degree
q > 2 using Belief Propagation Algorithm, have been earlier vertices in their parity check matrices; these codes provide a
employed to scale down BER and in turn inflate reliability. fairly low bit error rate even at large SNRs (signal-to-noise
By using GF(q) where q > 2, i.e. GF(4) and GF(8), slight ratios).In addition, It has been observed that joined source-
improvement in BER is achieved as compared with conven- channel encoding [67] in which Protographed LDPC Codes
tional turbo codes. Also, to achieve capacity near-by-Shan- are considered to be optimum one; too suffered from the
non capacity, irregular parity check matrices are used [60]. problem of higher EF. Utilizing the principle that source-
Altering the decoding method (using maximum Likelihood encoder entropy is lesser than source-decoding threshold
decoding algorithm) further aids to improve reliability but at (SDT) and SDT varies from channel-decoder threshold
the expense of intricacy [61].To increase computing speed value, Source codes are designed to provide lower EF values.
of the decoder, re-weighing (RW) and re-scheduling (RS)
techniques have been adapted. RW methods utilize weighing 4.6 Security
coefficients for the updates in parity vertices while RS meth-
ods use message-status shared via bi-partite graph in order It is very tedious task to task to acquire reliability and secu-
to find the update which can provide maximum efficiency. rity together. In security issues too, these codes have gained
Among RS methods, Residual-Belief-Propagation-Algo- substantial attention. Shahbaz et al. [68] proposed an algo-
rithm (RBPA), Node-wise-Belief-Propagation-Algorithm rithm using LDPC Codes to provide encryption to data. The
(NBPA) and layered-Belief- Propagation-Algorithm (LBPA) message bits are encoded and some bits are punctured ran-
have been introduced. RBPA takes the maximum remainder domly. Joint decoding is used as decoding process. Their
value calculated from the subtraction of shared messaged performance is judged for same and different transmission
between vertices and messages. LBPA utilizes the latest power levels for and verified that they provide small secu-
message in bipartite graph and updates are provided seri- rity gap. Polar Codes can too be used for encrypting data
ally. Using variable scheduling methods led to improvement by sending over channels which are considered to be non-
in the speed and BER but simultaneously there is increase reliable for un-desired users and reliable for desired ones
in computational load as more calculations are required for and vice versa [59, 61–63]. But in case of polar codes with
computing remainders. To increase decoding speed with less finite code-words, polarization process does not work per-
complexity, a new algorithm is proposed by Healy et al. in fectly. Here, most of the channels resulted out to be neither
[62]. Further in this direction, non-bipartite codes have been exactly reliable nor un-reliable. So, a new approach of polar
devised. These LDPC codes along with space diversity and encoding with Ex-OR and repetition encoding is proposed

13
A survey on channel coding techniques for 5G wireless networks

Fig. 8  FER comparison of 10 0
Polar Codes with Successive List decoding[52]
LDPC codes with polar codes
Protograph based LDPC Codes with base matrix2[102]
Polar Codes with Successive cancellation decoding[57]
Protograph based LDPC Codes with base matrix1[102]

10 -1

FER
10 -2

10 -3
0.5 1 1.5 2 2.5 3 3.5 4
Eb/N0 (dB)

by Kim et al. [69]. In this work, using these operations of successive decoding (list size = 4) [71] provides a gain of
Ex-OR and repetition coding, the forbidden channels (that about 0.67 dB over 3GPP Polar codes with successive can-
are neither reliable nor un-reliable) are converted into good cellation decoding [72] corresponding to frame-error-rate of
channels of polarization [70]. 10−2 . Simulation Parameters for this comparison are enlisted
in Table 6.
4.7 Avoidance of cycles in LDPC code construction Comparison of LDPC Codes: Fig. 9 compares some of
the existing Binary and Non-Binary LDPC Codes in terms
The study of cycles is found beneficial to predict EF so that of their Frame-Error-rate (FER) performance with Rates
an optimum decoding algorithm can be discovered and EF ½. LDPC Codes [52] used belief-propagation (BP) as the
can be minimized. Hashemi et al. [57] suggested that LDPC decoding algorithm. Further [73] improves the Frame-Error-
code’s error floor depends upon the trapping cycles (TC) Rate by employing (BP)-list erasure decoders at the cost of
involved in their bi-partite graphs. In case of RLDPCs, the increased complexity. Such decoders provide gain of 0.8 dB
most detrimental effects are caused by ETC (Elementary for FER of 0.1. Also, Fig. 9 shows that Irregular LDPC
trapping cycles) and in particular leafless-TCs are the main Codes perform better as compared with regular LDPC Codes
offenders amongst RLDPCs. [74]. Non-binary LDPC Codes perform better as compared
with binary LDPC Codes in terms of error rate and are band-
width efficient too. But the major drawback is of complexity
5 Simulation results which is addressed in [37]. Boutillon [75] reduced complex-
ity by the use of optimized values of elements in non-binary
In this section we have done the simulation analysis for LDPC Codes over GF (64) and GF (256) instead of taking
LDPC and Polar Codes. In this section, the simulations are random values.
carried out for Protograph based LDPC Codes and Polar Comparison of Polar Codes: Fig. 10 compares some of
Codes as specified by 3GPP standard document for 5G and the existing Polar Codes in terms of their Frame-Error-rate
comparisons are done on this basis. Expansion factors can (FER) performance. Polar Codes by Hashemi et al. [76, 77]
be E = a ∗ 2j where the value of a = {2, 3, 5, 7, 9, 11, 13, 15} devised an improved way of successive list decoding and
, j = 0, 1, 2, … ja and ja = {7, 7, 6, 5, 5, 5, 4, 4} . Correspond- this is further improved by Polar Codes in [78] devised by
ing to every value of E, the proto-matrix can have elements Chen et al. Different Segments have been taken for designing
ranging from {−1, 0, 1, … E − 1}. Maximum value of E for Polar Codes. Polar Codes 1 are constructed using two seg-
5G standard is kept as 384. ments each containing 16 bits. Polar Codes 2 are designed
Figure 8 shows the frame-error-rate comparison of LDPC with two segments containing 4 and 7 bits, Polar Codes 3
Codes and Polar Codes. It has been found that the Proto- have been constructed with 3 segments and each segment
graph-based LDPC Codes constructed using base matrix1 contains 8 bits while Polar Codes 4 used four segments. It
[67] has a gain of about 0.3 dB to attain frame-error-rate of has been observed that Polar Codes 2 performed well and
10−2 over Protograph-based LDPC Codes constructed using provide coding gain of 0.5 dB over Polar Codes 3 to achieve
base-matrix2 [67]. In addition, 3GPP Polar codes with list frame-error-rate of 10−3 [78].

13
K. Arora et al.

Table 6  Simulation parameters Channel encoding 3GPP LDPC codes 3GPP polar codes

Code-word length 1024 1024


Message length 512 512
Rate ½ ½
Decoding algorithm Min-sum decoder (For LDPC Codes) Successive cancellation decoding and
successive list decoding
Channel environment AWGN channel AWGN channel
Constellation BPSK BPSK
Performance parameters Frame error rate, signal-to-noise ratio Frame error rate, signal-to-noise ratio

6 Application areas

1. Deep Space Optical Communications (DSOC) DSOC


based technology is under development and is also a pro-
ject led by NASA. This technology promises to enhance
the performance around hundred folds as compared with
the on-going RF (radio-frequency) technology with no
impact on power requirements. For such power-efficient
systems, PPM based Poisson channel (PPMBPC) is
supposed to be the best candidate. For PPMBPC, Non-
Bipartiite LDPC (NBLDPC) outperforms as compared
with the earlier used RS-Codes by 3 dB but suffers with
the drawback of having lesser capacity. To remove this
drawback, BLDPC codes are considered which suffered
from performance degradation in ambient-noise condi-
Fig. 9  Comparison of LDPC codes tions. For this, LDPC Code design is modified which
ensures good performance [46] but at cost of complexity.

Fig. 10  Comparison of existing 10
0
Polar Codes1 [76]
polar codes
Polar Codes1 [78]
Polar Codes2 [77]
-1
10 Polar Codes2 [78]
Polar Codes3 [76]
Polar Codes3 [78]
-2
10 Polar Codes4 [77]
Polar Codes5 [78]
FER

-3
10

-4
10

-5
10

-6
10
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
E bN0 dB

13
A survey on channel coding techniques for 5G wireless networks

Fig. 11  Multi-beam MCS uplink with MIMO receiver

2. Helicopter Satellite Communications (HSC) HSC is found un-licensed (secondary) users can utilize the spectrum if
to gain notable attention now-a-days because of their utili- the licensed users are not transmitting over a particular fre-
zation in enormous fields e.g. relieving operations in case quency band. LDPC Codes aid in detecting errors caused
of fire or floods and airy surveillance. In the implementa- during secondary user data transmission. Also, Azeem et al.
tion of HSC, there occurs major problem of blockage of [81] compared the efficiency for shorter and longer block-
communication due to the regular rotation of rotary wings length LDPC Codes. It has been found that shorter-block
of helicopter. The modified LDPC codes are being used to codes provide lesser complexity and lesser delay as com-
resolve this problem. Check-hybrid codes are constructed pared to longer-block codes. This technique can be further
from QC-LDPC codes using two-level algorithm. The first extended to cooperative sensing techniques [82, 83]
step involves the multiplication of data sequence with the 5. UWB Non-Coherent Communications Non coher-
generator matrix of QC LDPC Code. In second step com- ent detection gains a lot of attention because of their
ponent-code is designed. The decoding algorithm used is simplicity and non-requirement of estimating channel
MPA (Message-passing-algorithm). MAP algorithm if coefficients which make the system power efficient too.
used, decreases the decoding complexity [79]. Liang illustrated that using these codes, bit-error-rate
3. Mobile-Satellite Communication System (MCS) The declines as compared to Reed–Solomon or convolutional
transmission channel from Earth station to a space- codes but with increased computational complexity. For
craft in MCS consisting of multiple beams is basically ultra-wide-band applications, short-length LDPC codes
a MIMO channel. To lessen the interference amongst are generally used to ensure lesser latency [34].
these beams an effective decoding algorithm using 6. WiMax Networks LDPC Codes find their applications
LDPC codes is ascertained by Yang et al. [80]. With the in enormous wireless standards including Wi-Max
aid of this algorithm, there is declination in both calcu- (Worldwide interoperability for microwave access) [84],
lation load and BER. Amongst MIMO detectors, turbo WLAN (Wireless Local Area Networks) [85], Wi-Fi
detector performing iteratively is considered to be the (Wireless Fidelity) [86] and wireless sensor networks
optimal detector. The linear SISO detector used in iter- [87, 88]. Tsatsaragkos et al. [86] used LDPC Codes
ative-turbo-detector (ITD), if replaced by approximate- with min-sum decoding for Wi-Fi applications. The
message-passing detector; has been proven to provide ever-increasing high-speeds in 5G enforced the usage of
less BERs and lower complexities [80]. QC-LDPC Codes in Wi-Max networks. Recently, Wang
et al. [84] invented a high-rate efficient encoder with
Figure 11 shows MCS consisting of ′ m ’ beams. Consid- salient features of high encoding speed, minimum hard-
ering ‘ n′ subscribers are provided with service simultane- ware usage. LDPC Code used is dual-diagonal along
ously and LDPC Encoder is used to encode ′ k′-bit sequence with lower-triangular form structure. Such parallel-high
of data bits of each user (denoted by di for ith user) and pro- speed LDPC structures can also be utilized in IEEE
vides code-word of length ‘ w ’ bits ( denoted by wi for ith user) . 802.11n and IEEE 802.22 standards.
These wi bits are mapped by QAM modulator and outputs Xi
sequence. Xit specifies all the transmitted symbols for i th user
at symbol period t  , Xi is denoted by Xit.Here yi represents data 7 Discussions and conclusions
after beamforming and d̂ i represent detected signals.
Channel Coding is considered to be the most crucial ele-
4. Cognitive Radio Networks Cognitive Radio Networks aim ment in the wireless communications and thus an area of
towards achieving a spectrally-efficient system in which the prime interest for scientists and researchers. This paper gives

13
K. Arora et al.

Table 7  Suitability of codes

Parameter Suitable channel codes

Bit-error-rate performance Short block size (~ 40) Polar codes


Moderate (~ 400) and large block size (~ 6144) LDPC codes
Complexity With increase in List size of decoder, complexity of Polar Codes
increases. Non Binary LDPC Codes provide high spectral
efficiency but complexity is high
Chip area required LDPC codes require lesser area for lower code-rates
Data channels LDPC codes
Control channels Polar codes
Decoding latency Polar codes using parallel-list and adaptive-list decoders
Similarity Non Binary LDPC Codes and Polar Codes using successive list
decoding (list size of 1024) perform identically

the review of channel coding techniques till in use amongst 9. Luby, M. G., Mitzenmacher, M., Shokrollahi, M. A., Spielman,
different generations. A detailed survey of channel coding D. A., & Stemann, V. (1997). Practical loss-resilient codes. In
Proceedings of 29th annual ACM symposium on theory of com-
techniques for 5G is done elaborating key requirements for puting (pp. 150–159).
5G networks in terms of transmission rate, spectral effi- 10. Richardson, T., & Urbanke, R. (2001). Efficient encoding of low-
ciency, energy efficiency and data rate. To meet these essen- density parity-check codes. IEEE Transactions on Information
tial requirements, channel coding techniques named LDPC Theory, 47(2), 638–656.
11. Richardson, T., Shokrollahi, M., & Urbanke, R. (2001). Design of
Codes and Polar Codes are elaborated. In addition research capacity approaching irregular low-density parity-check codes.
challenges faced in the implementation of these techniques IEEE Transactions on Information Theory, 47(2), 619–637.
i.e. complexities in terms of encoding or decoding, error 12. ten Brink, S. (2001). Convergence behavior of iteratively decoded
floor, security, cycles impact on decoding are described. As parallel concatenated codes. IEEE Transactions on Communica-
tions, 49(10), 1727–1737.
there exist tradeoffs amongst these features; hence depend- 13. Brink, S., Kramer, G., & Ashikhmin, A. (2004). Design of low-
ing upon their particular application, encoding and decoding density parity-check codes for modulation and detection. IEEE
algorithms are enlisted too. Further based on the survey, Transactions on Communications, 52(4), 670–678.
the suitability of codes for particular parameter has been 14. Ashikhmin, A., Kramer, G., & Brink, S. (2004). Extrinsic infor-
mation transfer functions: Model and erasure channel properties.
enlisted in Table 7. IEEE Transactions on Information Theory, 50(11), 2657–2673.
15. Prayogo, G. K., Putra, R., Prasetyo, A. H., & Suryanegara, M.
(2018). A 5G new radio LDPC coded NOMA scheme supporting
References high user load for massive MTC. In International conference on
information technology and electrical engineering (ICITEE) (pp.
170–174).
1. Turjman, F. A. (2017). Cognitive caching for the future sensors 16. Wu, X., Jiang, M., Zhao, C., Ma, L., & Wei, Y. (2018). Low-rate
in fog networking. Pervasive Mobile Computing, 42, 317–334. PBRL-LDPC codes for URLLC in 5G. IEEE Wireless Commu-
2. Salah, A. A., & Turjman, F. A. (2018). Low complexity parity nications Letters., 7(5), 800–803.
check code for futuristic wireless networks applications. IEEE 17. Sharma, A, & Salim, M. (2017). Polar Code: The Channel
Access, 6, 18398–18407. Code contender for 5G scenarios. In International conference
3. Niu, K., Chen, K., Lin, J., & Zhang, Q. T. (2014). Polar codes: on computer, communications and electronics (Comptelix) (pp.
Primary concepts and practical decoding algorithms. IEEE Com- 676–682).
munications Magazine, 52(7), 192–203. 18. Wang, R., & Rongke, L. A. (2014). Novel puncturing scheme
4. Berrou, C., & Glavieux, A. (1996). Near optimum error correct- for polar codes. IEEE Communication on Letters, 18(12),
ing coding and decoding: Turbo-codes. IEEE Transactions on 2081–2083.
Communications, 44, 1261–1271. 19. Arıkan, E. (2009). Channel polarization: A method for construct-
5. MacKay, D. J. C., & Neal, R. M. (1997). Near shannon limit ing capacity-achieving codes. IEEE Transactions on Information
performance of low density parity check codes. Electronics Let- Theory, 55(7), 3051–3073.
ters, 33, 457–458. 20. Marconi, G. (1899). Wireless telegraphy. Journal of the Institu-
6. MacKay, D. J. C. (1999). Good error-correcting codes based on tion of Electrical Engineers, 28(139), 273–290.
very sparse matrices. IEEE Transactions on Information Theory, 21. Sofi, I. B., & Gupta, A. (2018). A survey on energy efficient 5G
45(2), 399–431. green network with a planned multi-tier architecture. Journal of
7. Richardson, T., & Urbanke, R. (2001). The capacity of low-den- Network and Computer Applications, 118, 1–28.
sity parity-check codes under message-passing decoding. IEEE 22. Khan, F. (2009). LTE for 4G mobile broadband: Air interface
Transactions on Information Theory, 47, 599–618. technologies and performance. Cambridge: Cambridge Univer-
8. Spielman, D. (1996). Linear-time encodable and decodable error- sity Press.
correcting codes. IEEE Transactions on Information Theory,
42(6), 1723–1731.

13
A survey on channel coding techniques for 5G wireless networks

23. Khan, M. N., Gilani, S. O., Jamil, M., Rafay, A., et al. (2018). 44. Kocarev, L., Lehmann, F., Maggio, G., Scanavino, B., et al.
Maximizing throughput of hybrid FSO-RF communication sys- (2006). Nonlinear dynamics of iterative decoding systems:
tem: An algorithm. IEEE Access, 6, 30039–30048. Analysis and applications. IEEE Transactions on Information
24. Yang, L., Xie, Y., Yuan, J., Cheng, X., & Wan, L. (2018). Theory, 52, 1366–1384.
Chained LDPC codes for future communication systems. IEEE 45. Forney, G. D. (1997). On iterative decoding and the two-way
Communications Letters, 22(5), 898–901. algorithm. In Proceedings of international symposium on turbo
25. Andrews, J. G., Buzzi, S., Choi, W., et al. (2014). What will 5G codes and related topics brest, France (pp. 12–25).
be. IEEE Journal on Selected Areas in Communications, 32(6), 46. Matuz, B., Paolini, E., Zabini, F., & Liva, G. (2017). Non-binary
1065–1082. LDPC code design for the Poisson PPM channel. IEEE Transac-
26. Kong, L., Khan, M. K., Wu, F., Chen, G., & Zeng, P. (2017). tions on Communications, 65(11), 4600–4611.
Millimeter-wave wireless communications for IoT-cloud sup- 47. Vatta, F., Soranzo, A., & Babich, F. (2018). Low-complexity
ported autonomous vehicles: Overview, design, and challenges. bound on irregular LDPC belief-propagation decoding thresh-
IEEE Communications Magazine, 55(1), 62–68. olds using a Gaussian approximation. Electronics Letters, 54(17),
27. Liang, Z., Zang, J., Yang, X., Dong, X., & Song, H. (2017). Low- 1038–1040.
density parity-check codes for noncoherent UWB communica- 48. Clevorn, T., & Vary P. (2004). Low-complexity belief propaga-
tion systems. China Communications, 14(7), 1–11. tion by approximations with lookup-tables. In Proceedings of
28. Hung-Ta, P., Han, Y. S., & Chu, Y. J. (2011). New HARQ scheme 5th international ITG conference on source and channel coding
based on decoding of tail-biting convolutional codes in IEEE (SCC), Erlangen, Germany (pp. 211–216).
802.16e. IEEE Transactions on Vehicular Technology, 60(3), 49. Wadayama, T., Nakamura, K., Yagita, M., et al. (2010). Gradient
912–918. descent bit flipping algorithms for decoding LDPC codes. IEEE
29. GPP TSG-RAN WG1 #86. (2016). Discussion on outer coding Transactions on Communications, 58(6), 1610–1614.
on eMBB data. LG Electronics. 50. Sundararajan, G., Winstead, C., & Boutillon, E. (2014). Noisy
30. Peng, R. H., & Chen, R. R. (2006). Application of non-binary gradient descent bit-flip decoding for decoding LDPC codes.
LDPC codes for communication over fading channels using IEEE Transactions on Communications, 62(10), 3385–3400.
higher order modulations. In Proceedings of IEEE global com- 51. Huang, Q., Song, L., & Wang, Z. (2017). Set message-passing
munications conference (GLOBECOM). decoding algorithms for regular non-binary LDPC codes. IEEE
31. Feng, D., Xu, H., Zheng, J., & Bai, B. (2018). Nonbinary LDPC- Transactions on Communications, 65(12), 5110–5122.
coded spatial modulation. IEEE Transactions on Wireless Com- 52. Davey, M. C., & MacKay, D. J. C. (1998). Low density parity
munications, 17(4), 2786–2799. check codes over GF(q). Information theory workshop.
32. Chen, X., & Wang, C. L. (2012). High-throughput efficient 53. MacKay, D. J. C., Wilson, S. T., & Davey, M. C. (1999). Com-
non-binary LDPC decoder based on the simplified min-sum parison of constructions of irregular Gallager codes. IEEE Trans-
algorithm. IEEE Transactions on Circuits and Systems, 59(11), actions on Communications, 47(10), 1449–1454.
2784–2794. 54. Li, Z., Chen. L., Zeng, L., Lin, S., & Fong, F. H. (2006). Efficient
33. Shannon, C. E.(1948). A mathematical theory of communication. encoding of quasi-cyclic low-density parity-check codes 54(1),
Bell System Technical Journal, 27:379–423, 623–656. 71–81.
34. Calderbank, A., & Mazo, J. (1984). A new description of trel- 55. Huang, Q., Tang, L., He, S., Xiong, Z., & Wang, Z. (2014). Low-
lis codes. IEEE Transactions on Information Theory, 30(6), complexity encoding of quasi-cyclic codes based on Galois fou-
784–791. rier transform. IEEE Transactions on Communications, 62(6),
35. Lin, S., & Costello, D. J. (2004). Error control coding: Fun- 1757–1767.
damentals and applications (2nd ed.). Upper Saddle River, NJ: 56. Zhang, J., & Fossorier, M. P. C. (2004). A modified weighted
Prentice-Hall. bit-flipping decoding of low-density parity-check codes. IEEE
36. Xiao, H., & Banihashemi, A. H. (2009). Error rate estimation of Communications Letters, 8(3), 165–167.
low-density parity-check codes on binary symmetric channels 57. Nouh, A., & Banihashemi, A. H. (2002). Bootstrap decoding of
using cycle enumeration. IEEE Trans. Communications, 57(6), low-density parity-check codes. IEEE Communications Letters,
1550–1555. 6(9), 391–393.
37. Karimi, M., & Banihashemi, A. H. (2014). On characterization of 58. Oh, J., & Ha, J. (2018). A two-bit weighted bit-flipping decoding
elementary trapping sets of variable-regular LDPC codes. IEEE algorithm for LDPC codes. IEEE Communications Letters, 22(5),
Transactions on Information Theory, 60(9), 5188–5203. 874–877.
38. Halford, T. R., & Chugg, K. M. (2006). An algorithm for count- 59. Li, E., Gunnam, K., & Declercq, D. (2011). Trellis based
ing short cycles in bipartite graphs. IEEE Transactions on Infor- extended Min-Sum for decoding non-binary LDPC codes. In
mation Theory, 52(1), 287–292. Proceedings of IEEE international symposium on wireless com-
39. Asvadi, R., Banihashemi, A. H., & Attari, M. A. (2011). Lower- munication systems (pp. 46–50).
ing the error floor of LDPC codes using cyclic liftings. IEEE 60. Davey, M. C., & MacKay, D. J. C. (1998). Low density parity
Transactions on Information Theory, 57(4), 2213–2224. check codes over GF(q). Ireland: ITW.
40. Karimi, M., & Banihashemi, A. H. (2012). Efficient algorithm for 61. Sason, I., & Shamai, S. (2000). Improved upper bounds on the
finding dominant trapping sets of LDPC codes. IEEE Transac- ensemble performance of ML decoded low density parity check
tions on Information Theory, 58(11), 6942–6958. codes. IEEE Communication Letters, 4(3), 88–91.
41. Hashemi, Y., & Banihashemi, A. H. (2015). On characterization 62. Healy, C., Shao, Z., Oliveira, R. M., et al. (2018). Knowledge-
and efficient exhaustive search of elementary trapping sets of aided informed dynamic scheduling for LDPC decoding of short
variable-regular LDPC codes. IEEE Communications Letters, blocks. IET Communications, 12(9), 1094–1101.
19(3), 323–326. 63. Korada, S. B., Soglu, S., & Urbanke, R. (2010). Polar codes:
42. Tanner, R. (1981). A recursive approach to low complexity codes. Characterization of exponent, bounds, constructions. IEEE
IEEE Transactions on Information Theory, 27(5), 33–47. Transactions on Information Theory, 56(12), 6253–6264.
43. Ryan, W. E., & Lin, S. (2009). Channel codes: Classical and 64. Mughal, S., Yang, F., Xu, H., et al. (2018). Coded cooperative
modern (1st ed.). Cambridge: Cambridge University Press. spatial modulation based on multi-level construction of polar

13
K. Arora et al.

code. Telecommunications Systems. https​://doi.org/10.1007/ 85. Uthansakul, M., & Uthansakul, P. (2011). Experiments with a
s1123​5-018-0485-6. low-profile beamforming MIMO system for WLAN applications.
65. Fossorier, M. P. C. (2001). Iterative reliability-based decoding of IEEE Antennas and Propagation Magazine, 53(6), 56–69.
low-density parity check codes. IEEE Journal on Selected Areas 86. Tsatsaragkos, I., & Paliouras, V. (2018). A reconfigurable LDPC
in Communications, 19(5), 908–917. decoder optimized for 802.11n/ac applications. IEEE Transac-
66. Wu, X., Jiang, M., & Zhao, C. (2017). A parity structure for tions on Very Large Scale Integration, 26(1), 182–195.
scalable QC-LDPC codes with all nodes of degree three. IEEE 87. Pellenz, M. E., Souza, D. R., & Fonseca, M. S. P. (2010). Error
Communications Letters, 21(9), 1913–1916. control coding in wireless sensor networks. Telecommunication
67. Chen, C., Wang, L., & Liu, S. (2018). The design of protograph Systems, 44(1), 61–68.
LDPC codes as source codes in a JSCC system. IEEE Commu- 88. Tsai, H.-C. (2019). Iterative multiuser detector-decoding for
nications Letters, 22(4), 672–675. nonbinary LDPC coded multicarrier MFSK systems. Telecom-
68. Shahbaz, S., Akhbari, B., & Asvadi, R. (2018). LDPC codes over munication Systems, 70(2), 309–320.
Gaussian multiple access wiretap channel. IET Communications, 89. Hamming, R. W. (1950). Error detecting and error correcting
12(8), 962–969. codes. Bell System Technical Journal, 29(2), 147–160.
69. Kim, M., Kim, B. H., & Ahn, J. K. (2017). Secure polar cod- 90. Golay, M. J. (1949). Notes on digital coding. In Proceedings of
ing with REP and XOR coding. IEEE Communications Letters, the IRE (vol. 37, p. 657).
21(10), 2126–2129. 91. Muller, D. E. (1954). Application of boolean algebra to switch-
70. Shao, S., Hailes, P., Wang, T.-Y., Wu, J.-Y., Maunder, R. G., ing circuit design and to error detection. Electronic Computers,
Al-Hashimi, B. M., & Hanzo, L. (2019). Survey of turbo, LDPC Transactions of the IRE Professional Group, 3, 6–12.
and polar decoder ASIC implementations. IEEE Communica- 92. Reed, I. (1954). A class of multiple-error-correcting codes and
tions Surveys & Tutorials. the decoding scheme. Information Theory, Transactions of the
71. Chen, K., Niu, K., & Lin, J. R. (2012). List successive cancel- IRE Professional Group, 4(4), 38–49.
lation decoding of polar codes. IEEE Electronics Letters, 48(9), 93. Bose, R. C., & Ray-Chaudhuri, D. K. (1960). On a class of error
500–501. correcting binary group codes. Information and Control, 3(1),
72. Kong, B. Y., Yoo, H., & Park, I. C. (2016). Efficient sorting 68–79.
architecture for successive cancellation list decoding of polar 94. Reed, I. S., & Solomon, G. (1960). Polynomial codes over certain
codes. IEEE Transactions on Circuits and Systems II: Express finite fields. Journal of the Society for Industrial and Applied
Briefs, 63(7), 673–677. Mathematics, 8(2), 300–304.
73. Bocharova, I. E., Kudryashov, B. D., Skachek, V., & Yakimenka, 95. Costello, D. J., & Forney, G. D. (2007). Channel coding: The road
Y. (2018). BP-LED decoding algorithm for LDPC codes over to channel capacity. Proceedings of the IEEE, 95, 1150–1177.
AWGN channels. IEEE Transactions on Information Theory, 96. Elias P. (1955). Coding for two noisy channels. In Proceedings
65(3), 1677–1693. Th. Theory (pp. 61–76).
74. Boutillon, E. (2018). Optimization of non binary parity check 97. Viterbi, A. J. (1967). Error bounds for convolutional codes and an
coefficients. IEEE Transactions on Information Theory, 65(4), asymptotically optimum decoding algorithm. IEEE Transactions
2092–2100. on Information Theory, 13(2), 260–269.
75. Khazraie, S., Asvadi, R., & Banihashemi, A. H. (2012). A PEG 98. Bahl, L., Cocke, J., Jelinek, F., & Raviv, J. (1974). Optimal
construction of finite-length LDPC codes with low error floor. decoding of linear codes for minimizing symbol error rate. IEEE
IEEE Communications Letters, 16(8), 1288–1291. Transactions on Information Theory, 20(2), 284–287.
76. Hashemi, S., Balatsoukas-Stimming, & Giard, P. (2016). Par- 99. Consulative Committee for Space Data Systems (CCSDS).
titioned successive-cancellation list decoding of polar codes. (1984).Telemetry channel coding. Silver Book.
In Proceedings of IEEE international conference on acoustics, 100. Gallager, R. G. (1963). Low-density parity-check codes. Ph.D.
speech and signal process. Shanghai, China (pp. 957–960). thesis, Dep. Electrical Eng., M.I.T, Cambridge.
77. Hashemi, S., Mondelli, M., & Hamed, S. (2018). Decoder par- 101. Arıkan, E. (2008). A performance comparison of polar codes
titioning: Towards practical list decoding of polar codes. IEEE and Reed–Muller Codes. IEEE Communications Letters, 12(6),
Transactions on Communications, 66(9), 3749–3759. 447–449.
78. Chen, P., Bai, B., Ren, Z., Wang, J., & Sun, S. (2019). Hash-polar 102. Liang, H., Liu, A., Zhang, Y., & Zhang, Q. (2017). Analysis
codes with application to 5G. IEEE Access, 7, 12441–12455. and adaptive design of polar coded HARQ transmission under
79. Wang, P., Yin, L., & Lu, J. (2018). Efficient helicopter-satellite SC-list decoding. IEEE Wireless Communications Letters, 6(6),
communication scheme based on check-hybrid LDPC coding. 798–801.
Tsinghua Science and Technology, 23(3), 323–332. 103. Niu, K., & Chen, K. (2012). Stack decoding of polar codes. Elec-
80. Yang, Y., Wang, W., & Gao, X. (2018). AMP Dual-turbo iterative tronics Letters, 48(12), 695–696.
detection and decoding for LDPC coded multibeam MSC uplink. 104. Chen, K., Niu, K., & Lin, Jiaru. (2013). Improved successive
China Communications, 15(6), 178–186. cancellation decoding of polar codes. IEEE Transactions on
81. Azeem, M. M., Khan, A. B., & Azeem, U. (2017). Application of Communications, 61(8), 3100–3107.
short erasure correcting codes for cognitive radio. In IEEE 86th 105. Li, B., Shen, H., & Tse, D. (2012). An adaptive successive can-
vehicular technology conference (VTC-Fall). cellation list decoder for polar codes with cyclic redundancy
82. Azmi, M. H., & Leib, Harry. (2018). Multichannel cooperative check. IEEE Communications Letters, 16(12), 2044–2047.
spectrum sensing that integrates channel decoding with fusion- 106. Niu, K., & Chen, K. (2012). CRC-aided decoding of polar codes.
based decision. IEEE Transactions on Aerospace and Electronic IEEE Communications Letters, 16(10), 1668–1671.
Systems, 54(4), 1998–2014. 107. Elkelesh, A., Ebada, M., Cammerer, S., & ten Brin, Stephan.
83. Soliman, Samir S., & Song, Bongyong. (2017). Fifth genera- (2018). Belief propagation list decoding of polar codes. IEEE
tion (5G) cellular and the network for tomorrow: Cognitive and Communication Letters, 22(8), 1536–1539.
cooperative approach for energy savings. JNCA, 85(1), 84–93. 108. OnurDizdar, Erdal Arıkan. (2016). A high-throughput energy-
84. Wang, X., Ge, T., Li, J., & Su, C. (2017). Efficient multi-rate efficient implementation of successive cancellation decoder for
encoder of QC-LDPC codes based on FPGA for WIMAX Stand- polar codes using combinational logic. IEEE Transactions on
ard. Chinese Journal of Electronics, 26(2), 250–255. Circuits and Systems, 63(3), 436–447.

13
A survey on channel coding techniques for 5G wireless networks

109. Jiang, S., Mo, F., Lau, C. M., & Sham, C. W. (2018). Tree- Jaswinder Singh  obtained his
permutation-matrix based LDPC codes. IEEE Transactions on Bachelor’s degree in Electronics
Circuits and Systems, 65(8), 1019–1023. & Communication Engineering
110. Kim, K. S., Lee, S. H., Kim, Y. H., & Ahn, J. Y. (2004). Design from Guru Nanak Dev Univer-
of binary LDPC code using cyclic shift matrices. Electronics sity, Amritsar. Then he obtained
letters, 40(5), 325–326. his Master’s degree in Electron-
111. Tasdighi, A., Banihashemi, A. H., & Sadeghi, M. R. (2017). ics & Communication Engineer-
Symmetrical constructions for regular Girth-8 QC-LDPC codes. ing from PTU Jalandhar and
IEEE Transactions on Communications, 65(1), 14–22. Ph.D. degree in Electronics &
112. Diao, Q., Li, J., Lin, S., & Blake, I. F. (2016). New classes of Communication Engineering
partial geometries and their associated LDPC codes. IEEE Trans- from Guru Nanak Dev Univer-
actions on Information Theory, 62(6), 2947–2965. sity, Amritsar in 2003 and 2011
113. Elsanadily, S., Mahran, A., & Elghandour, O. (2018). Classifica- respectively. Currently he is an
tion-based algorithm for bit-flipping decoding of GLDPC codes Associate Professor in the
over AWGN channels. IEEE Communications Letters, 22(8), Department of Electronics and
1520–1523. Communication Engineering at
114. He, X., Zhou, L., & Du, J. (2018). PEG-like design of binary BCET, Gurdaspur. His area of specialization is Optical CDMA and his
QC-LDPC codes based on detecting and avoiding generating research interests are in Optical CDMA systems & Networks, Optical
small cycles. IEEE Transactions on Communications, 66(5), Communications systems and Networks.
1845–1858.
115. Jiang, X. Q., Hai, H., Wang, H. M., & Lee, M. H. (2017). Con- Yo g e s h w a r S i n g h R a n d ‑
structing large Girth QC protograph LDPC codes based on PSD- hawa  received the B.E. degree
PEG algorithm. IEEE Access, 5, 13489–13500. in electronics and communica-
116. Jiang, X., Xia, X. G., & Lee, M. H. (2014). Efficient progressive tion engineering from BBSBEC
edge-growth algorithm based on Chinese remainder theorem. (Baba Banda Singh Bahadur
IEEE Transactions on Communications, 62(2), 442–451. Engineering College), Fatehgarh
117. Gruner, A., & Huber, M. (2012). New combinatorial construction Sahib in 2003, M.B.A. in HR
techniques for low-density parity-check codes and systematic from PTU University, Jalandhar
repeat-accumulate codes. IEEE Transactions on Communica- in 2007, M.Tech in Electronics
tions, 60(9), 2387–2395. and Communication Engineering
118. Wei, X., & Akansu, A. (2001). Density evolution for low-density From Beant College, Gurdaspur
parity-check codes under Max-Log-MAP decoding. Electronics in 2010 and Ph.D in Design and
Letters, 37(18), 1125–1126. Analysis of Low power SRAM
119. Mao, Y., & Banihashemi, A. H. (2001). Decoding low-density from State University (SU)
parity-check codes with probabilistic scheduling. IEEE Com- Rajasthan. He is currently work-
muincation on Letters, 5(10), 414–416. ing as a Director in Lyallpur
Khalsa College of Engineering, Jalandhar, India. His research area
Publisher’s Note Springer Nature remains neutral with regard to include polarization modeling, simulation of optical fiber in electronics
jurisdictional claims in published maps and institutional affiliations. and communication and VLSI. He is a member and active reviewer of
many international journals.

Komal Arora  has received the


B.E. degree in electronics and
communication engineering
from Punjabi University, Punjab
India, in 2008, M.Tech degree in
electronics and communication
engineering from Thapar Uni-
versity, India, in 2011 and cur-
rently pursuing the Ph.D. degree
in electronics and communica-
tion engineering from IKG Pun-
jab Technical University, Jaland-
har, India. She is currently an
Assistant Professor in the School
of Electronics and Electrical
Engineering, CGC, Landran,
Mohali, India. She is currently involved in research work on channel
encoding and decoding algorithms of LDPC and Polar Codes in 5G.
She is also working on Cognitive Radio networks and MATLAB tools
for wireless communication. Her research interests include the emerg-
ing channel encoding techniques in 5G wireless communication
network.

13

Potrebbero piacerti anche