Received March 3, 2019, accepted March 18, 2019, date of publication March 28, 2019, date of current version

April 11, 2019.

Digital Object Identifier 10.1109/ACCESS.2019.2907970

Synchronization Procedure in 5G NR Systems

AYMEN OMRI 1 , (Member, IEEE), MOHAMMED SHAQFEH 1 , (Member, IEEE),
ABDELMOHSEN ALI2 , (Member, IEEE), AND HUSSEIN ALNUWEIRI1 , (Senior Member, IEEE)
1 Department of Electrical and Computer Engineering, Texas A&M University at Qatar, Doha, Qatar
2 Department of Electrical and Computer Engineering, Concordia University, Montreal, QC H3G 1M8, Canada
Corresponding author: Aymen Omri ([email protected])
This work was supported by the Qatar National Research Fund (a member of Qatar Foundation) through the National Priorities Research
Program (NPRP) under Grant 8-1545-2-657. The statements made herein are solely the responsibility of the authors.

ABSTRACT Similar to all mobile communication networks, synchronization in the time–frequency domain
is a fundamental step that allows a fifth-generation (5G) new radio (NR) user equipment (UE) to properly
receive and transmit its data. Due to the wide range of frequencies that are defined for the 5G NR systems,
the corresponding synchronization procedure becomes critical and presents many challenges, especially
for the applications that would need accurate oscillators to reduce the large values of the frequency
offset. In this paper, we present and detail the 5G NR physical layer. Then, we describe the required
synchronization procedure for 5G NR. And finally, we present the main challenges and issues within the
5G NR synchronization.

INDEX TERMS 5G NR systems, beam management, physical layer, frequency offset, time offset, synchro-
nization procedure.

I. INTRODUCTION inter-symbol interference (ISI) [4]–[7]. The time offset is

Fifth generation (5G) new radio (NR), the mobile communi- due to the transmission delay, where the transmitted signal
cation standard presented by the 3rd generation partnership reaches the receiver delayed in time [8], [9]. In this case,
project (3GPP) as the 3GPP Release 15 [1]–[3], presents the receiver does not know when the transmitter sent a new
a major improvement of the long term evolution-advance burst. Usually, the normalized timing offset is considered,
(LTE-A) standard, where the main focus is on enhanced which is equal to the number of samples between the trans-
mobile broadband, ultra-reliable and low latency communica- mitted signal and the received signal [4], [6]. If the normal-
tions, and massive machine-type communications. To achieve ized timing offset is larger than the cyclic prefix (CP) length,
these goals, 3GPP has introduced a unified network archi- then a misalignment of the fast Fourier transform (FFT)
tecture, with a new physical layer design that supports very window can be observed, which results in ISI and ICI.
high carrier frequencies (mmWaves), large frequency band- Otherwise, only a phase offset can be observed [4], [6]. The
widths, and new techniques such as massive multiple-input estimation and correction of the time offset should be done
and multiple-output (MIMO), and beamforming [1]–[3]. in the pre-FFT synchronization stage, by using specific syn-
Those major modifications increase the synchronization chronization algorithms. In the literature, the auto-correlation
procedure challenges. and cross-correlation algorithms are well known and used
In fact, the very high defined carrier frequencies result in wireless communication systems [10], [11]. In the first
in large values of frequency and time offsets, which need algorithm, the received signal is correlated with a delayed
an accurate and expensive oscillator to align a transmitter version of the same signal. However, in the second algorithm,
and a receiver for interference free communications. The the received signal is correlated with a stored pattern known
sources of interference are mainly related to the imperfec- to the receiver to estimate the time offset [4], [6]. Beside
tions of orthogonal frequency-division multiplexing (OFDM) the time offset, the errors in the transmitter and the receiver
systems, which suffer from the time and frequency oscillators result in frequency offset, which is a linear phase
offsets that result in inter-carrier interference (ICI), and over the time domain samples and it causes ICI over the sub-
carriers [10], [12], [13]. In contrast to the time offset, this
The associate editor coordinating the review of this manuscript and phase offset increases in time as it is directly proportional
approving it for publication was Walid Al-Hussaibi. to the discrete time index. The frequency offset in OFDM
2169-3536
2019 IEEE. Translations and content mining are permitted for academic research only.
41286 Personal use is also permitted, but republication/redistribution requires IEEE permission. VOLUME 7, 2019
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
A. Omri et al.: Synchronization Procedure in 5G NR Systems

is usually normalized to the sub-carrier spacing as the ratio

between the frequency error and the sub-carrier spacing.
In addition, a mismatch between the sampling frequency at
both transmitter and receiver presents another source for the
time offset.
In addition, the 3GPP has introduced a new high-
dimensional phased arrays based mechanism to establish
highly directional transmission links between the next gener-
ation node base station (gNB) and the user equipments (UEs).
This mechanism requires fine alignment of the transmitter
and the receiver beams, achieved through a set of operations
known as beam management. The beam management needs FIGURE 1. 5G NR physical channels and signals.
complex algorithms and high level processing at gNBs and
UEs to perform a variety of control tasks, including initial
access, and beam tracking [14], which increase the synchro-
nization procedure challenges.
In the reminder of this paper, the 5G NR physical layer
is presented and detailed in Section II. In Section III,
the required synchronization procedure and steps for 5G NR
are described. Then, the main challenges and issues within
the 5G NR synchronization are investigated in Section IV.
Finally, conclusions are drown in Section V.

II. OVERVIEW OF 5G NR PHYSICAL LAYER

In this section, we present an overview of the 5G NR physical
layer. The 3GPP references that detail all the specifications
of the 5G NR as well as the used versions in this paper are
presented in Table 1.

TABLE 1. 3GPP 5G NR specifications [15].

FIGURE 2. 5G NR frame structure.

Based on these 3GPP specifications, we present and detail

in the following the 5G NR physical channels and signals as flexible number of slots, which depends on the used numerol-
well as the frame structure. ogy (µ). As presented in Fig. 2, this number is given by
Subframe,µ µ
NSlot = 2 . Consequently, the time slot period is equal
µ
A. PHYSICAL CHANNELS AND SIGNALS to 1 ms / 2 . And, for each slot, there are 14/12 OFDM
Slot
The different 5G NR physical channels and signals are pre- symbols, denoted by NSym , for normal/extended cyclic prefix,
sented in Fig. 1 [1]. To figure out the differences between CP. Based on that, each symbol duration is equal to 1 ms
µ µ
the 5G NR and the LTE in terms of physical channels and / [2 × 14] (1 ms / [2 × 12]) for normal/extended CP. This
signals, we present in Table 2, a comparison between the two is the frame structure in time domain.
standards for each physical channel and each physical signal, In frequency domain, and similar to LTE standard,
where SHCCH denotes the shared and control channels. a resource block (RB) is defined as 12 consecutive sub-
carriers. Let 1f denotes the subcarrier spacing. In 5G
B. 5G NR FRAME STRUCTURE NR, 1f is flexible and depends on the used numerology,
Based on the 3GPP specifications in [1], the 5G NR frame e.g., 1f = 2µ × 15 [kHz].
structure is described in this subsection. The 5G NR frame In the following, we focus on the time-frequency
frame
structure consists of Nsubframe = 10 subframes each of structure of the 5G synchronization block (SS/PBCH),
1 ms. Different than LTE frame, each 5G NR subframe has which is denoted by SSB. Overall description on the

VOLUME 7, 2019 41287

 A. Omri et al.: Synchronization Procedure in 5G NR Systems

TABLE 2. Channels and signals: NR vs. LTE [1], [2].

TABLE 3. Resource allocation for SS/PBCH block [1, Tab. 7.4.3.1-1], l and
k are relative to the beginning of the SS/PBCH block.

FIGURE 3. Resource allocation for SSB.

resource allocation for SSB is described in Table 3 • Case B - 30 kHz subcarrier spacing: the first symbols
Cell Cell
[1, Tab. 7.4.3.1-1], where v = NID mod 4, and NID is the of the candidate SSBs have indexes {4, 8, 16, 20}+28×
corresponding physical cell ID. n. For carrier frequencies smaller than or equal to 3 GHz,
Based on Table 3, the summary of the SSB specification n = 0. For carrier frequencies larger than 3 GHz and
are as follows: smaller than or equal to 6 GHz, n = 0, and 1.
• SS (PSS and SSS) and PBCH in NR is transmitted in the • Case C - 60 kHz subcarrier spacing: the first symbols
same 4 symbol block. of the candidate SSBs have indexes {2, 8} + 14 × n. For
• SSB consists of 240 contiguous subcarriers (20 RBs). carrier frequencies smaller than or equal to 3 GHz, n =
• The subcarriers are numbered in increasing order from 0, and 1. For carrier frequencies larger than 3 GHz and
0 to 239 within the SSB. smaller than or equal to 6 GHz, n = 0, 1, 2, and 3.
• Case D - 120 kHz subcarrier spacing: the first symbols
Table 3 can be represented in resource grid as shown of the candidate SSBs have indexes {4, 8, 16, 20} +
in Fig. 3. This SSB structure is general and defined for all the 28 × n. For carrier frequencies larger than 6 GHz, n =
numerology in 5G NR. However, for a half frame, the number 0, 1, 2, 3, 5, 6, 7, 8, 10, 11, 12, 13, 15, 16, 17, and 18.
and first symbol indexes for candidate SSBs are different • Case E - 240 kHz subcarrier spacing: the first symbols
and determined according to the corresponding subcarrier of the candidate SSBs have indexes {8, 12, 16, 20, 32,
spacing as follows [3]: 36, 40, 44} + 56 × n. For carrier frequencies larger than
• Case A - 15 kHz Subcarrier Spacing: The first symbols 6 GHz, n = 0, 1, 2, 3, 5, 6, 7, and 8.
of the candidate SSBs have indexes of {2, 8} + 14 × n.
For carrier frequencies smaller than or equal to 3 GHz, C. SYNCHRONIZATION SIGNALS
n = 0, and 1. For carrier frequencies larger than 3 GHz Similar to 4G networks, the synchronization signals PSS and
and smaller than or equal to 6 GHz, n = 0, 1, 2, and 3. SSS are used in 5G NR systems to help the UEs to get radio

41288 VOLUME 7, 2019

A. Omri et al.: Synchronization Procedure in 5G NR Systems

frame boundary and to detect the cell identity (ID). In 5G NR follows [1]
systems, each radio cell is identified by a cell ID from 1008
IDs that are arranged into 336 different groups. Each group dSSS (n) = [1 − 2x0 ([n + m0 ] mod 127)]
× [1 − 2x1 ([n + m1 ] mod 127)] , (5)
(1)
is identified by the cell ID group, NID ∈ {0, . . . , 335}, and
consists of three different sectors, which are specified by the
(2) where,
cell ID sector, NID ∈ {0, 1, 2}. The UE can detect the value
(2) (1) " (1) #
of NID from the PSS, and the value of NID from the SSS. NID (2)
Based on that, the UE computes the serving cell ID as follows: m0 = 15 + 5 NID , (6)
cell (1) (2)
112
NID = 3 NID + NID . The 5G NR cell ID concept is the same (1)
as in 4G networks. However, the number and the generation m1 = NID mod 112, 0 ≤ n < 127, (7)
of the corresponding synchronization signal sequences are x0 (i + 7) = [x0 (i + 4) + x0 (i)] mod 2, (8)
different. In the following, we present these differences, and x1 (i + 7) = [x1 (i + 4) + x1 (i)] mod 2, (9)
the details of the 5G NR signal generations.
[x0 (6) x0 (5) x0 (4) x0 (3) x0 (2) x0 (1) x0 (0)]
1) PSS = [0 0 0 0 0 0 1], (10)
In 4G networks, each PSS consists of one of three 62-symbols and,
Zadoff-Chu sequences and is mapped to the central 72 sub-
carriers with a guard band of 10 subcarriers. For FDD frame, [x1 (6) x1 (5) x1 (4) x1 (3) x1 (2) x1 (1) x1 (0)]=[0 0 0 0 0 0 1],
PSS is allocated on symbol # 6 (last symbol) of slot # 0 (11)
(subframe 0) and slot # 10 (subframe 5) of each radio frame.
And, for TDD frame, PSS is allocated on symbol # 2 of slot III. SYNCHRONIZATION PROCEDURE IN 5G NR SYSTEMS
# 2 (subframe 1) and slot # 12 (subframe 6) of each radio Due to the wide range of the new defined frequencies
frame. in 5G NR systems, the corresponding synchronization pro-
Similar to 4G networks, the 5G-NR PSS is a physical layer cedure becomes more critical and presents many challenges.
specific signal and helps UEs to get radio frame boundary, Recently, the synchronization procedure in 5G NR systems is
(2)
and to detect the cell ID sector, e.g., NID . However, the 5G presented in [3]. In this section, the main steps of the 5G NR
NR PSS consists of one of three 127-symbols m-sequences synchronization procedure are detailed first. Then, numerical
and is allocated on the first symbol of each SSB, and on results are presented to evaluate this procedure.
127 subcarriers. The 3 possible m-sequences for the PSS are
defined as follows [1] A. MAIN STEPS OF 5G NR SYNCHRONIZATION
dPSS (n) = 1 − 2 x(m), (1) The 5G NR synchronization procedure is based on the beam
management operations, including (i) initial access for idle
where, users, which allows a mobile UE to establish a physical link
h i connection with a gNB, and (ii) beam tracking, for connected
(2)
m = n+43 NID mod 127, 0 ≤ n < 127, (2) users, which enable handover, beam adaptation, and radio
x(i + 7) = [x(i + 4) + x(i)] mod 2, (3) link failure recovery procedures. The beam management
is based on four different operations; (i) Beam sweeping,
and, (ii) Beam measurement, (iii) Beam determination, and (iv)
Beam reporting. Based on theses four operations, we present
[x(6) x(5) x(4) x(3) x(2) x(1) x(0)] = [1 1 1 0 1 1 0]. (4) in Fig. 4, the general 5G NR synchronization procedure,
which can be summarized as follows:
2) SSS By using a beam sweeping, the gNB periodically transmits
In 4G networks, each SSS consists of one of 168 62-symbols SS burst that carries multiple SSBs, with periodicity TSS =
m-sequences. Similar to PSS, the SSS is mapped to the central {5, 10, 20, 40, 80, or 160} ms [14]. As shown in Fig. 5, each
72 subcarriers with a guard band of 10 subcarriers. For FDD SSB is transmitted via a specific beam with pre-specified
frame, SSS is allocated on symbol # 5 of slot # 0 (subframe interval and direction. The SSBs in a half frame are indexed
0) and slot # 10 (subframe 5) of each radio frame. And, for in an ascending order in time from 0 to L − 1.
TDD frame, SSS is allocated on symbol # 6 (last symbol) of For initial access case (idle mode), the UE should start
slot # 1 (subframe 0) and slot # 11 (subframe 5) of each radio by the cell search procedure. By using the PSS, SSS, and
frame. specific synchronization algorithms, the UE can estimate and
Similar to 4G networks, the 5G-NR SSS is used to detect correct the frequency and time offsets. In the next subsection,
(1)
the cell ID group, e.g., NID . However, the 5G NR PSS consists we present and details some well know and used synchroniza-
of one of 336 127-symbols gold sequences and is allocated tion algorithms in the literature.
on the third symbol of each SSB, and on 127 subcarriers. After decoding the PSS, the UE can detect the cell ID sec-
(2) (2)
The 336 possible gold sequences for the SSS are defined as tor (NID ). Then, by using the detected NID , the UE can decode

VOLUME 7, 2019 41289

 A. Omri et al.: Synchronization Procedure in 5G NR Systems

FIGURE 4. 5G NR synchronization procedure.

should identify the corresponding SSB index (0, 1, 2, .. or

L − 1). Based on the received PBCH, the UE shall determine
the 2 least significant bits (LSBs), for L = 4, or the 3 LSBs,
for L > 4, of the candidate SSB index per half frame from
a one-to-one mapping with an index of the DMRS sequence
transmitted in the PBCH. The rest of the SSB index is carried
by the PBCH data.
For example, let us consider the case of 15 kHz subcarrier,
a carrier frequency less than 3 GHz, and L = 4. Based on that,
the SBBs are allocated at symbol indexes 2, 8, 16, and 22.
Each of them has a unique SSB index, iSSB , (2 bits in this case)
and is corresponding to unique sample timing, which refers
FIGURE 5. SS burst transmission. to the sample accurate timing, e.g., the sample corresponds
to the first sample, of a given symbol, in a given slot per a
(1)
frame. Therefore, the UE can detect the iSSB from decoding
the SSS and detect the cell ID group NID , and hence it can the corresponding PBCH, where the corresponding DMRS
cell (1) (2)
compute the serving cell ID as follows: NID = 3 NID + NID . is a function of iSSB . For example, if the UE successfully
After detecting the cell ID, the UE should identify the decodes the PBCH with a DMRS carries iSSB = 0, then the
candidate SSB within the SS burst (set of L SSBs). As we corresponding SSB is mapped to symbol 2 of the half-frame,
have detailed in Subsection II. B, for each subcarrier spacing, and then the UE can detect the sample timing within the 5 ms.
there is a set of candidate SSBs with different first symbol After that, the UE has to know which 5 ms is this in a radio
indexes per a half-frame. Each SSB is corresponding to a frame (the first 5 ms or the second 5 ms). This extra bit is
specific beam. So, to detect the serving beam index, the UE provided by the master information block (MIB), and also by

41290 VOLUME 7, 2019

A. Omri et al.: Synchronization Procedure in 5G NR Systems

the DMRS in case of L = 4. So after decoding the PBCH, are neither provided nor presented in the 3GPP specifications.
the UE knows the sample timing within the full frame. In the literature, the cross-correlation and auto-correlation
In addition to the sample timing, the UE should detect the methods are well known and used in wireless communication
system frame number (SFN). The SFN is divided into two systems [10], [11]. In the following, we details these two
parts (i) 4 LSBs that are used within the channel encoding of correlation based synchronization algorithms that can be used
the MIB message itself, and (2) 6 most significant bit (MSBs) in 5G NR Networks.
that are part of the MIB data. During the MIB decode, there
will be blind decode trials since UE does not know the 4 LSBs 1) CROSS-CORRELATION SYNCHRONIZATION ALGORITHM
of the MIB that are used during the channel encoding. So, USING PSS
it will try different combinations while decoding the MIB, In this cross-correlation process, the received signal is corre-
each corresponds to a unique 4 LSBs. One of them should lated with the PSS stored pattern known to the receiver. The
get into success, and then the UE will know the 4 LSBs cross-correlation based algorithm can be summarized as fol-
of the SFN. Hence, once the MIB is decoded successfully, lows: Let r(n) be the perfect received signal in the discrete
the corresponding SFN will be known in full. time domain, which is expressed as follows:
The decoded DMRS (associated with the PBCH) can be r(n) = s(n) ⊗ h(n) + w(n), (12)
used to measure the reference signal received power (RSRP)
of the candidate SSB, which can be used for the beam mea- where, s(n) is the transmitted signal, h(n) is the channel
surement, with the other measurements of the RSRP, and the impulse response (CIR), w(n) is the noise term, and ⊗ is
reference signal reported quality (RSRQ) given by SSS. the linear convolution operator. In the presence of carrier fre-
Based on the beam measurement, the beam determination quency offset (CFO), the received signal expression becomes
can be done and the UE can decide which is the best beam. h i  j2πn 
r (n) = s(n) ⊗ h(n) + w(n) exp  , (13)
During the beam reporting, and based on the necessary NFFT
parameters that are including in the MIB, the UE can decode where,  is the normalized frequency offset to the sub-carrier
the system information block (SIB) that is transmitted by the spacing, which can be separated into integer frequency offset
gNB over PDSCH and includes the RACH resources. Based (IFO: I ), and fractional frequency offset (FFO: F ). Let
on that, the UE can transmit Message 1 (RACH preamble) on θ̂ denotes the estimated timing synchronization position of
the configured RACH resources via the selected beam. PSS, and î ∈ {0, 1, 2} denotes the estimate of cell ID
After receiving Message 1, the gNB responds with Mes- sector. By using the time-domain received signal, a joint
sage 2 (random access response (RAR)). Then, the UE trans- estimation of θ̂ and î can be done based on the following
mits Message 3 (RRC connection request), and the gNB criterion [10], [11]
sends Message 4 (RRC connection setup) to the UE. Finally,
the UE sends to the gNB Message 5 (RRC connection setup (θ̂ , î) = arg max(C(θ, i)), (14)
θ,i
complete).
where, C(θ, i) is the cross-correlation, which is expressed
Once random access procedure is completed, dedicated
as [11]
connection is established between UE and gNB with dedi-  
NFFT −1
cated connection ID.  P ∗

r(θ + k)pi (k)

These procedures are used to update the optimal transmit- 
 k=0 
ter and receiver beam pair over time. C(θ, i) = . (15)
NFFT −1
For beam tracking case (connected mode), the beam man- P
|r(θ + k)|2
agement operations use CSI-RS (in DL) and SRS (in UL) k=0
to transfer the beam/CSI report and the DL control informa- Here, r(θ + k) presents the delayed received signal, pi (k)
tion (DCI) between the gNB and the UE, and to update the denotes the PSS sequence time-domain waveform of cell ID
connection via the strongest selected beam. sector i, and NFFT is the FFT number without CP. Based on
For initial access case (idle mode), as we detailed before, that, the FFO estimation can be evaluated as follows [11]
the beam management operations use PSS, SSS and PBCH  NFFT /2−1 ∗
DMRS, where the cell search procedure is required to detect 16 X
ˆF = r(θ̂ + k)pî (k)
∗
the cell ID. The cell search procedure is a crucial step in π
k=0
the 5G NR synchronization as it needs accurate time and NFFT −1
 
frequency offsets detection and correction. In the following X
we focus on the details of this procedure. × r(θ̂ + k)p∗î (k) , (16)
k=NFFT /2
To access to the network or during a handover opera-
tion, the UE has to acquire correct timing and frequency where, 6 (.) and [.]∗ are the argument and complex conjugate
synchronization to recover the cell ID of the serving gNB. operators, respectively.
This first operation is known as initial cell-search, which The PSS based cross-correlation synchronization
can use different algorithms of estimation and correction of algorithm is efficient for small values of CFO, as ˆF in
frequency and time offsets. These synchronization algorithms equation (16) presents the estimate FFO only.

VOLUME 7, 2019 41291

 A. Omri et al.: Synchronization Procedure in 5G NR Systems

2) AUTO-CORRELATION SYNCHRONIZATION ALGORITHM TABLE 4. Simulation parameters.

USING CP
In this auto-correlation method, a part of the received sig-
nal is correlated with the corresponding CP part. The auto-
correlation based algorithm can be summarized as follows:
Let γ (θ) and α(θ ) denote the correlation and energy terms,
which are expressed as
θ+L−1
X
γ (θ ) = r(k)r ∗ (k + NFFT ), (17)
k=θ
and,
θ +L−1
X
α(θ) = |r(k)|2 + |r(k + NFFT )|2 , (18)
k=θ
respectively, where L is the CP length. Based on that, the max- TABLE 5. Power-delay profile of TDL-A channel [16].
imum likelihood (ML) can be used to estimate θ and  as
follows [11]
θ̂ = max (2|γ (θ)| − ρ 6 (θ)), (19)
θ
and,
1 6
ˆ = − (γ (θ̂)), (20)
2π
where, ρ is the magnitude of the correlation coefficient
between r(k) and r(k + N ), which is expressed as
σS2
ρ= , (21)
σS2 + σN2
with σS2 and σN2 are the signal and the noise powers, respec-
tively.
This method’s accuracy can be improved by averaging the
estimates of time and frequency offsets over several OFDM
symbols. In a large CFO scenario, i.e.,  > 1, the FFO can
be estimated by using the auto-correlation algorithm, and the
IFO can be recovered in the frequency domain, by evaluating
the shift of the received PSS. This method is insensitive to
IFO and can offer an accurate estimation of FFO before
the estimation of IFO, which is not the case for the cross- specifications to evaluate the average BER for different time-
correlation method. However, the auto-correlation method frequency synchronization scenarios.
cannot provide the cell ID. Fig. 6 presents a performance comparison between cross-
As presented in Fig. 4, the estimation and correction of correlation and auto-correlation algorithms in terms of time
time-frequency offset is a crucial task in the overall 5G offset estimation error. It is clear here that both algorithms
NR synchronization procedure. In the following sub-section, offer good performances in terms of time offset estimation
we present numerical results to investigate the 5G NR syn- error, with a slightly advantage of the auto-correlation method
chronization procedure. for θ = 5. These results confirm that both methods are
efficient and able to offer a good estimation of time-frequency
B. NUMERICAL RESULTS AND INTERPRETATIONS offset.
The conducted simulation parameters are provided in Table 4 The impacts of time and frequency offsets on BER, with
and represent typical values representative of numerology different values of , θ = 4, and θ̂ = 4 are presented in Fig. 7.
3 of 5G NR network. As shown in this figure, for small values of , the two methods
The used 5G NR channel model is the tapped delay line are able to estimate and correct the frequency offset to offer
(TDL)-A, where the corresponding channel parameters are an acceptable BER. By increasing the values of , the auto-
presented in Table 5. correlation method outperforms the cross-correlation method
The conducted numerical results is based on our devel- as it can offer a low CFO estimation error. However, for
oped 5G NR simulator, which uses the corresponding 3GPP high CFO values, e.g.,  > 0.5, both method suffer from

41292 VOLUME 7, 2019

A. Omri et al.: Synchronization Procedure in 5G NR Systems

FIGURE 6. Comparison of the time-offset estimation for cross-correlation FIGURE 8. Trade-off between the crystal oscillator accuracy and the
and auto-correlation algorithms. carrier frequency; α = 21f
F
, for numerology µ = 3.
C

Based on that, we present in the following sub-section,

the main synchronization challenges in 5G NR systems.

IV. SYNCHRONIZATION CHALLENGES

IN 5G NR SYSTEMS
The main challenge in 5G NR systems is the time-frequency
synchronization in the presence of the wide range of the
new defined frequencies. The new defined frequency band
increases the imperfections of the used OFDM systems,
which results in serious interference problems. To over-
come this problems, new efficient synchronization procedure
should be proposed, and a more accurate and expensive oscil-
lator must be used. In fact, the frequency error is mainly
related to the accuracy of the used oscillator, and the envi-
ronment factors (mainly temperature) may deviate the used
FIGURE 7. Impact of time and frequency offsets on BER, with different
carrier frequencies. The frequency deviation is measured in
values of , θ = 4 and θ̂ = 4. parts per million (ppm), which is a value that represents the
part of a whole number in units of 1/1000000, i.e., one ppm
the high estimation error, which results in significant BER is equal to 1/1000000 of the whole:
performance degradation. 1ppm = 1 × 10−6 = 0.0001%. (22)
In Fig. 8, we present the required accuracy (maximum
allowed error) of a given oscillator that can be used in 5G In wireless communication, the ppm indicates how much the
NR to offer a maximum CFO equal to 0.2 and 0.4. This used oscillator’s frequency may deviate from the nominal
figure shows that the oscillator accuracy is decreasing with value. For example, in LTE, a 10 ppm crystal oscillator is used
the increased values of the carrier frequency (FC ). This is for the frequencies range of 3GHs and below. It means, that
because, for a fixed value of , FC and the oscillator accu- for a given LTE equipment that is using a carrier frequency
racy are inversely proportional; α = 21f FC . In addition, of 3 GHz, its frequency error is equal to +/ − 10 ppm ×
it is clear that the CFO increases with the increased values 3 GHz= 30 kHz. Consequently, the real carrier frequency
of the maximum allowed error of the oscillator, which is will be between (3 GHz-30 kHz) and (3 GHz+30 kHz), which
expected. presents the accuracy of the used oscillator. In LTE, this
As shown in the simulation results, to offer an acceptable range is acceptable for frequency tolerance, where the cor-
BER, the synchronization methods are able to manage only responding frequency offset can be estimated and corrected
a specific range of CFO, e.g.,  < 0.4. And to offer such during the synchronization procedure between the eNBs and
values of CFO, an accurate oscillator is needed, where its the UEs. However, for 5G NR, a very wide range of fre-
accuracy is increasing with the increased values of the used quencies is defined, which needs a very accurate oscillator,
carrier frequency, which presents a crucial issue for the 5G note that the oscillator’s cost is inversely proportional to
NR networks. its ppm.

VOLUME 7, 2019 41293

 A. Omri et al.: Synchronization Procedure in 5G NR Systems

In addition, the used beamforming technique is based on [6] K. Shamaei and M. Sabbaghian, ‘‘Analytical performance evaluation of
precise alignment of the transmitter and the receiver beams, SC-FDMA systems in the presence of frequency and time offset,’’ IEEE
Trans. Wireless Commun., vol. 14, no. 11, pp. 6230–6239, Nov. 2015.
which needs complex algorithms and high level processing [7] D. Li, Y. Li, H. Zhang, L. J. Cimini, and Y. Fang, ‘‘Integer frequency offset
at gNBs and UEs. These algorithms and processing may estimation for OFDM systems with residual timing offset over frequency
increase the latency of establishing a link, and have impor- selective fading channels,’’ IEEE Trans. Veh. Technol., vol. 61, no. 6,
pp. 2848–2853, Jul. 2012.
tant implications on control layer procedures, such as initial [8] V.-T.-D. Huynh, N. Noels, and H. Steendam, ‘‘Effect of offset mismatch in
access, handover, and beam tracking. time-interleaved ADC circuits on OFDM-BER performance,’’ IEEE Trans.
In this context, many challenges for the 5G NR should be Circuits Syst. I, Reg. Papers, vol. 64, no. 8, pp. 2195–2206, Aug. 2017.
[9] A. M. Hamza and J. W. Mark, ‘‘Closed-form expressions for the BER/SER
addressed: of OFDM systems with an integer time offset,’’ IEEE Trans. Commun.,
• The current 10 ppm crystal oscillator that is used in vol. 63, no. 11, pp. 4461–4473, Nov. 2015.
[10] R. Zeng, H. Huang, L. Yang, and Z. Zhang, ‘‘Joint estimation of frequency
LTE presents a high frequency offset, e.g., 60 kHz for offset and Doppler shift in high mobility environments based on orthogonal
6 GHz. Based on that a more expensive oscillator should angle domain subspace projection,’’ IEEE Trans. Veh. Technol., vol. 67,
be used to offer less ppm, and hence less frequency no. 3, pp. 2254–2266, Mar. 2018.
[11] S. Huang, Y. Su, Y. He, and S. Tang, ‘‘Joint time and frequency offset
offset. estimation in LTE downlink,’’ in Proc. 7th Int. Conf. Commun. Netw.
• What is the maximum acceptable frequency error in 5G China, Aug. 2012, pp. 394–398.
NR? [12] O. H. Salim, A. A. Nasir, H. Mehrpouyan, and W. Xiang, ‘‘Multi-relay
communications in the presence of phase noise and carrier frequency
• How would the current commercial oscillators behave in offsets,’’ IEEE Trans. Commun., vol. 65, no. 1, pp. 79–94, Jan. 2017.
the 5G NR? [13] C. Chen, Y. Chen, Y. Han, H.-Q. Lai, and K. J. R. Liu, ‘‘High resolu-
• What is the expected accuracy for the current commer- tion carrier frequency offset estimation in time-reversal wideband com-
munications,’’ IEEE Trans. Commun., vol. 66, no. 5, pp. 2191–2205,
cial oscillators?, and is it possible to offer an acceptable May 2018.
frequency error for the very wide range of the new [14] M. Giordani et al., ‘‘A tutorial on beam management for 3GPP NR at
frequencies? mmWave frequencies,’’ IEEE Commun. Surveys Tuts., vol. 21, no. 1,
pp. 173–196, Sep. 2018.
• During the signaling procedure, are the gNBs and the [15] 3GPP. The 3rd Generation Partnership Project. Accessed: Mar. 17, 2019.
users able to estimate and correct the frequency offset of [Online]. Available: http://www.3gpp.org/
the current commercial oscillators? [16] Study on Channel Model for Frequency Spectrum Above 6 GHz, document
TR 138.900, V.14.2.0, Jun. 2017.
• Are the gNBs and the UEs able to use the proposed
beamforming technique to establish a precise alignment
of the transmitter and the receiver beams, especially for
communication scenarios with mobility?
AYMEN OMRI received the degree in telecom-
munication engineering from The Academy of
V. CONCLUSION Aviation (EABA), Tunisia, in 2007, and the
In this paper, we have described and investigated the syn- M.Res. and Ph.D. degrees in telecommunications
chronization procedure in 5G NR systems. First, we have from the Engineering National School of Tunis
detailed the 5G NR physical layer. Then, we have described (ENIT)/Tunis El Manar University, in 2009 and
2012, respectively.
the required synchronization procedure, and different syn- From 2012 to 2017, he was a Postdoctoral
chronization algorithms for 5G NR. Based on that, numerical Researcher with the Electrical Engineering (EE)
results are conducted and investigated to evaluate the syn- Department, Qatar University. He is currently a
chronization procedure, and to present the main challenges, Postdoctoral Researcher with the Department of Electrical and Computer
Engineering, Texas A&M University at Qatar. His research interests include
and issues within the 5G NR synchronization. As expected, modeling, design, and performance analysis of wireless communication sys-
the main challenge for the 5G NR systems is the time- tems. His current specific research interests include device-to-device com-
frequency synchronization in the presence of the wide range munications, UAV-based networks, and the fifth-generation (5G) wireless
of the new defined frequencies, and the performance limita- communication networks.
tion of the current commercial crystal oscillator.

REFERENCES
[1] NR; Physical Channels and Modulation Release 15, document TS 38.211,
MOHAMMED SHAQFEH received the B.Sc.
V.15.4.0, 3GPP, 2018.
degree in electrical engineering (communications
[2] NR; Multiplexing Channel Coding Release 15, document TS 38.212,
V.15.4.0, 3GPP, 2018.
stream) from United Arab Emirates University,
in 2003, the M.Sc. degree in communications tech-
[3] NR; Physical Layer Procedures for Control Release 15, document
TS 38.213, V.15.4.0, 3GPP, 2018. nology from Ulm University, Germany, in 2005,
[4] M. A. Alvarez and U. Spagnolini, ‘‘Distributed time and carrier frequency
and the Ph.D. degree from The University of
synchronization for dense wireless networks,’’ IEEE Trans. Signal Inf. Edinburgh, Edinburgh, U.K., in 2009. In 2009,
Process. Netw, vol. 4, no. 4, pp. 683–696, Dec. 2018. he joined the Department of Electrical and Com-
[5] H. Abdzadeh-Ziabari, W.-P. Zhu, and M. N. S. Swamy, ‘‘Joint maximum puter Engineering, Texas A&M University at
likelihood timing, frequency offset, and doubly selective channel esti- Qatar, where he is currently an Associate Research
mation for OFDM systems,’’ IEEE Trans. Veh. Technol., vol. 67, no. 3, Scientist. His research interests include wireless communication networks,
pp. 2787–2791, Mar. 2018. information theory, and smart transportation systems.

41294 VOLUME 7, 2019

A. Omri et al.: Synchronization Procedure in 5G NR Systems

ABDELMOHSEN ALI was born in Giza, Egypt, HUSSEIN ALNUWEIRI (S’81–M’83–SM’17)

in 1982. He received the B.Sc. and M.A.Sc. received the Ph.D. degree in electrical and com-
degrees in electronics and communication engi- puter engineering from the University of Southern
neering from Cairo University, Giza, in 2004 and California, Los Angeles, in 1989. From 1991 to
2008, respectively, and the Ph.D. degree in elec- 2007, he was a Professor with the Department
trical and computer engineering from Concordia of Electrical and Computer Engineering, The
University, Montreal, QC, Canada, in 2016. From University of British Columbia. From 1996 to
2008 to 2012, he was with Wasiela, where he was 1998, he represented The University of British
involved in digital communications modem design Columbia, Vancouver, BC, Canada, at the ATM
including DVB-C/T2, and LTE. He is currently Forum. From 2000 to 2006, he served as a
with Riot Micro as a Senior Lead for the IoT physical-layer solutions. Canadian Delegate to the ISO/IEC JTC1/SC29 Standards Committee
He also serves as an affiliated Assistant Professor with Concordia Uni- (MPEG-4 Multimedia Delivery), where he worked within the MPEG4 stan-
versity. His research interests include cognitive radio networks, multi-user dardization JTC1-SC29WG11 Group to develop the first client–server
communications, communication signal processing, VLSI architectures for MPEG4 video streaming reference software. He has a long record of indus-
communication systems, machine-to-machine communications, the Internet trial collaborations with several major companies worldwide. He is currently
of Things, and 5G. He has received many awards, including the Best Paper a Professor with the Department of Electrical and Computer Engineer-
Award of the WCNC 2016 and the Canadian Governor General’s Academic ing, Texas A&M University at Qatar, Doha, Qatar. He is also an inventor
Gold Medal 2017. and holds four U.S. patents. He has authored or coauthored more than
200 refereed journal and conference papers in various areas of computer
and communications research. In particular, his research interests include
the mobile Internet technologies, cyber security and cyber systems, mobile
cloud computing, wireless communications, routing and information dissem-
ination algorithms in mobile networking, and quality-of-service provisioning
and resource allocation in wireless networks.

VOLUME 7, 2019 41295