Optics Communications 474 (2020) 126112

Contents lists available at ScienceDirect

Optics Communications
journal homepage: www.elsevier.com/locate/optcom

Implementation of a multiband 5G NR fiber-wireless system using analog

radio over fiber technology
L.A.M. Pereira a , C.H.S. Lopes a , R.M. Borges a,b , E.S. Lima a , A.C. Ferreira a , M. Abreu c ,
L.L. Mendes a , Arismar Cerqueira S. Jr. a ,∗
a
Lab. WOCA, National Institute of Telecommunications (Inatel), 37540-000, Santa Rita do Sapucaí, MG, Brazil
b
Federal University of Itajubá (UNIFEI), 37500-903, Itajubá, MG, Brazil
c Venturus Innovation and Technology, 13086-530, Campinas, SP, Brazil

ARTICLE INFO ABSTRACT

Keywords: This work reports the implementation of a multiband 5G new radio (NR) fiber-wireless (FiWi) system,
5G NR using analog radio over fiber (A-RoF) technology. Our approach employs a 12.5-km optical fronthaul (FH)
Centralized radio access network for transporting fifth-generation of mobile network (5G) signals up to a remote antenna unit, from which
Fiber-wireless system
10-m reach wireless transmissions are performed aiming indoor and outdoor enhanced mobile broadband
Fronthaul
(eMBB) applications. The aforementioned 5G signals refer to a 5G NR standard at 26 GHz with 400-MHz
Microwave photonics
bandwidth and DSP-based flexible-waveform signal at 3.5 GHz from our transceiver, which can be set
as orthogonal frequency division multiplexing (OFDM), filtered-OFDM (F-OFDM) or generalized frequency
division multiplexing (GFDM). Additionally, an investigation on the coexistence between 5G NR and Long
Term Evolution-Advanced (LTE-A) standards in FiWi systems using the 700 MHz band, aimed for long-
range applications, is reported. Experimental results demonstrate 1.65 Gbit/s total wireless throughput in
accordance with the 3GPP Release 15 root mean square error vector magnitude (EVMRMS ) digital performance
requirements.

1. Introduction prefix orthogonal frequency division multiplexing (CP-OFDM) as wave-

form. However, CP-OFDM suffers from some major drawbacks, such
The fifth generation of mobile network (5G) is considered a remark- as high peak to average power ratio (PAPR) and high out-of-band
able evolution over the current wireless systems, since it is expected emission (OOBE) [8,9]. These issues have severe effects on the energy
to offer substantial improvements in terms of capacity, throughput, efficiency and coexistence with other communication systems, respec-
flexibility, energy efficiency and end-to-end latency [1]. 5G claims tively. Therefore, a considerable amount of works have focused on new
for innovative technical solutions, e.g. additional spectrum, multiband and orthogonal frequency division multiplexing (OFDM)-inspired wave-
allocation and new waveforms [2–4], for addressing three main sce- forms, including generalized frequency division multiplexing (GFDM)
narios, namely: enhanced mobile broadband (eMBB); massive machine and filtered OFDM (F-OFDM) [10,11].
type communication (mMTC); ultra-reliable low-latency communica- The 5G networks must operate with multiple frequencies to ad-
tion (URLLC) [5]. Moreover, enhanced remote area communications dress all application scenarios. In this way, 3GPP has specified two
(eRAC), aiming long-reach applications, has received considerable at- frequency ranges in the 5G NR standard. Frequencies between 0.450
tention in the last years [6]. The mix of scenarios gives rise to several and 7.125 GHz correspond to the frequency range 1 (FR1), whereas
applications with distinct and tough requirements. Recently, the 3rd frequency range 2 (FR2) includes millimeter-waves (mm-waves), more
Generation Partnership Project (3GPP) has developed the 5G NR stan- specifically from 24.5 to 52.6 GHz [12]. As a consequence, the most
dard, focused on the high-throughput applications from the eMBB appropriate frequency band might be selected in accordance with the
scenario [7]. desired usage scenario. For instance, FR1 might be used for cellular,
One of the most important features of a mobile communication mMTC, URLLC and eRAC applications [13]. On the other hand, FR2
system refers to the applied waveform since it impacts on the frame presents a lot of unexploited spectrum, which can be used for broad-
structure used to multiplex blocks of control, data and pilot sym- band transmissions aiming indoor Gbit/s eMBB applications [14,15]. In
bols. Fourth-generation of mobile network (4G) has adopted cyclic
this way, 5G networks must offer an unprecedented level of flexibility

∗ Corresponding author.
E-mail address: arismar@inatel.br (A. Cerqueira S. Jr.).

https://doi.org/10.1016/j.optcom.2020.126112
Received 11 February 2020; Received in revised form 10 April 2020; Accepted 25 May 2020
Available online 27 May 2020
0030-4018/© 2020 Elsevier B.V. All rights reserved.
L.A.M. Pereira, C.H.S. Lopes, R.M. Borges et al. Optics Communications 474 (2020) 126112

Fig. 1. C-RAN architecture for 5G multi-application systems.

to cover dense and heterogeneous networks (HetNet). Furthermore, The paper main contributions over the state-of-the-art and our
the radio access network (RAN) should evolve and shows a promising previous publications are the following ones: the implementation of
direction to be primarily centralized due to the imminent increase in a 5G NR FiWi system composed of a 12.5-km fronthaul, followed by
the number of base stations (BS). a 10-m reach wireless transmissions for indoor and outdoor eMBB
In traditional distributed RAN (D-RAN), the remote radio unit (RRU) applications; the implementation of all specified 5G NR FR2 band-
and baseband unit (BBU) are physically located at BS. On the other widths, i.e. 50, 100, 200 and 400 MHz, achieving Gbit/s through-
hand, centralized RAN (C-RAN) displaces BBUs from remote cell sites put and meeting the 3GPP root mean square error vector magnitude
to a central office (CO), creating the BBU pool concept. A fronthaul (EVMRMS ) requirements; the coexistence analysis of LTE-A and 5G NR
(FH) link, which can be optical or wireless, is used for connecting in a RoF-based transport network; the integration of our DSP-based 5G
BBUs and RRUs. C-RAN allows significant cost reduction on the capital
transceiver [30] at 3.5 GHz with a 5G NR system using a RoF-based
and operational expenditure (CAPEX and OPEX), since the network
FH.
management is simplified and its scalability is enhanced [16].
The manuscript is structured in five sections. Section 2 presents the
In this context, microwave photonics (MWP) plays an important
role on the integration of fiber-wireless (FiWi) systems to 5G networks. theoretical background of 5G waveforms, whereas Section 3 introduces
MWP allows generating, processing and transmitting radiofrequency our multiband 5G FiWi concept. Section 4 reports the experimental
(RF) signals, taking advantage of wireless and optical communication results regarding the implementation of the A-RoF-based multiband
benefits in a single system [17]. Particularly, the radio over fiber FiWi system, as well as the coexistence between 5G NR and LTE-A
(RoF) solution is used to transport analog or digital signals in optical signals in the referred network infrastructure. Finally, the conclusions
FH using analog RoF (A-RoF) or digital RoF (D-RoF) techniques, re- and future works are outlined in Section 5.
spectively [18]. Common Public Radio Interface (CPRI), evolved CPRI
(eCPRI), Open RAN (O-RAN) and Open Base Station Architecture Initia- 2. Waveforms for 5G networks
tive (OBSAI) have been proposed for D-RoF-based FH implementations.
Currently, CPRI has been widely employed despite this approach suffers
from scalability issues. As a consequence, the A-RoF solution might be 5G networks must address a large set of contrasting requirements
considered attractive, embracing spectral efficiency improvement and and a single physical layer (PHY) solution is unlikely to fulfill all
remote radio unit simplification [19,20]. challenging scenarios foreseen for the next mobile network. The 3GPP
The state-of-the-art on RoF-based FH solutions for 5G includes Release 15 [31] has defined OFDM as the waveform to support the
the distribution of 5G candidate waveforms and mm-waves [21–25], eMBB scenario. 3GPP Release 16 [32] is discussing the inclusion of
RoF transmission assisted by digital signal processing (DSP) [26], and the URLLC scenario in the 5G ecosystem, focusing mainly on the
physical layer split options [19]. Particularly, our research group has Industry 4.0 [33] use case, while it is expected that 3GPP Release
proposed the integration of a 5G transceiver in a gigabit passive op- 17 shall consider the mMTC and eRAC scenarios. The waveform to
tical network (GPON) to take advantage of network capillarity for be used in the next two 3GPP Releases is still an open question. This
distributing 5G signals [27]. Moreover, we have recently reported section exploits some potential solutions for improving the 5G networks
the implementation of a dual-band wireless FH assisted by an optical flexibility.
midhaul (MH), which connects CO and distribution unit (DU) [28].
Additionally, we have implemented a DSP-based flexible-waveform and
multi-application 5G FiWi system [29]. 2.1. OFDM
This work presents the implementation of a multiband 5G NR FiWi
system, as a FH solution to simultaneously cover outdoor and indoor OFDM [34] is a multicarrier waveform largely used in wireless
eMBB scenarios at 3.5 and 26 GHz, respectively. In addition, an exper- and wired digital communication standards. The main principle of
imental performance investigation of a FiWi system based on 5G NR this waveform consists on splitting the complex data symbols 𝑑𝑘 , pro-
and Long-term Evolution Advanced (LTE-A) standards in the 700 MHz vided by 𝐽 -quadrature amplitude modulation (QAM) mappers, into 𝐾
band is reported as a function of the frequency offset. Fig. 1 presents our orthogonal subcarriers, as depicted by Fig. 2.
4G/5G FiWi system proposal based on the C-RAN concept. COs access Although each subcarrier might carry different number of bits per
the network core via optical backhaul (BH) and connect distribution symbol, i.e. using different modulation orders, the symbol rate per
units and remote sites, using optical MH and FH, respectively. Densely subcarrier is constant and given by
distributed remote sites provide then broadband wireless access using
small cells at 3.5 GHz and femtocells at 26 GHz. Long-range cells can 𝑅S
𝑅MC = , (1)
also be provided in the 700 MHz band, even coexisting with the legacy 𝐾
systems. where 𝑅𝑆 is the overall symbol rate of the OFDM system.

2
L.A.M. Pereira, C.H.S. Lopes, R.M. Borges et al. Optics Communications 474 (2020) 126112

Fig. 2. Simplified OFDM block diagram.

In order to keep the subcarriers orthogonal to each other, the 2.2. F-OFDM
minimum frequency spacing between two adjacent subcarriers is given
by OFDM has some degree of freedom, allowing different modulation
order per subcarrier. Moreover, the total number of subcarriers can be
1
𝛥𝑓 = = 𝑅MC , (2) parameterized according with the available bandwidth or data symbols
𝑇OFDM
to be transmitted. However, all subcarriers must have the same OFDM
where 𝑇OFDM is the OFDM symbol period. Therefore, the OFDM symbol symbol duration and the same CP.
can be defined as The plethora of use cases and applications scenarios expected for
∑
𝐾−1 5G networks will demand a higher degree of freedom, since Internet of
𝑘
𝑠[𝑛] = 𝑑[𝑘]𝑒−𝑗2𝜋 𝐾 𝑛 , (3) Things (IoT) devices operating with very restrictive battery-life cannot
𝑘=0 use the same time–frequency structure employed by a smarthphone
which can be represented in a matrix form as downloading data ate multiple Gbit/s. F-OFDM [35] is a direct deriva-
tive of OFDM with filter banks applied to a set of adjacent subcarrier,
𝐬 = 𝐅H
𝐾 𝐝, (4) as illustrated in Fig. 3.
Each data source might provide pieces of information from a differ-
where (⋅)H is the Hermitian matrix, 𝐝 is the 𝐾 × 1 symbol vector and ent service, e.g. mMTC, eMBB, URLLC or eRAC and all communication
𝐅𝐾 is a 𝐾 × 𝐾 normalized Fourier
√ matrix with the element in the 𝑖th chains are independent, meaning that F-OFDM can apply specific mod-
row and 𝑗th column given by 1∕ 𝐾𝑒−𝑗2𝜋𝑖𝑗∕𝐾 . ulation order, Fourier matrix size and CP length for each one. The
OFDM protects the signal from the inter block interference (IBI), by filter bank, composed by the prototype pulses 𝑝𝑚 [𝑛], filters the subbands
coping the last 𝑁CP samples to its beginning. The introduction of the corresponding to the bandwidth (BW) from each data source. At the
cyclic prefix (CP) makes the OFDM symbol become circular, meaning receiver side, a matched filter bank, composed by the prototype pulse
that the signal at the output of the channel, after CP removal, is given 𝑞𝑚 [𝑛], is employed to split the signals provided by each communication
by chain and the remaining receiving process is equivalent to OFDM.
A guard band of at least half of the largest subcarrier is necessary
𝐲 = 𝐇𝐬 + 𝐰, (5) between two adjacent subbands to avoid interference. Hence, although
F-OFDM significantly increases the system flexibility, the spectrum
where 𝐇 is the 𝐾 ×𝐾 circulant channel matrix and 𝐰 is the 𝐾 ×1 additive efficiency can be reduced if the number of independent subbands is
white Gaussian noise (AWGN) vector with zero mean and variance 𝜎w 2.
high. The Inset (i) from Fig. 3 depicts the filtering process in F-OFDM
At the receiver side, frequency-domain equalization can be applied at the transmitter side.
after the inverse Fourier transform, i.e.
2.3. GFDM
𝐅𝐾 𝐲 = 𝐅𝐾 𝐇𝐅H
𝐾 𝐝 + 𝐅𝐾 𝐰. (6)
⏟⏞⏟⏞⏟
 GFDM is a block-based waveform composed by 𝐾 subcarriers with
𝑀 subsymbols each, as depicted in Fig. 4.
Notice that , highlighted in (6), is a 𝐾 × 𝐾 diagonal matrix with
A prototype pulse 𝑔[𝑛] is circularly shifted in time and frequency
the channel frequency response. Hence, zero-forcing (ZF) equalization
domains to carry data symbols in the time–frequency grid, leading to
can be performed after the subcarrier decoupling by dividing the a set of filters given by
received samples at each subcarrier frequency by the inverse of the
[ ] 𝑘
corresponding channel gain, leading to 𝑔𝑘,𝑚 [𝑛] = 𝑔 ⟨𝑛 − 𝑚𝐾⟩𝑁 𝑒−𝑗2𝜋 𝐾 𝑛 , (9)

𝐝̂ =  −1 𝐅𝐾 𝐲 = 𝐝 +  −1 𝐅𝐾 𝐰. (7) where 𝑚 = 0, 1, … , 𝑀 − 1 is the subsymbol index, 𝑘 = 0, 1, … , 𝐾 − 1 is

the subcarrier index, 𝑁 = 𝐾𝑀 is the total number of samples or data
Severe frequency-selective channel (FSC) can introduce deep symbols in a GFDM block and ⟨(⋅)⟩𝑁 is the modulo-𝑁 operator. The
notches in the channel frequency response, leading to very low gains at GFDM signal can be written as
specific subcarrier frequencies. Therefore, the noise at these subcarriers ∑ 𝐾−1
𝑀−1 ∑
frequencies will be enhanced by the frequency-domain equalization 𝑠G [𝑛] = 𝑑[𝑘, 𝑚]𝑔𝑘,𝑚 [𝑛], (10)
𝑚=0 𝑘=0
(FDE), which might hinder the correct detection of the data transmitted
at these subcarriers. where 𝑑[𝑘, 𝑚] is the data symbol transmitted in the 𝑚th subsymbol of
EVMRMS is an useful metric for evaluating the received signal per- the 𝑘th subcarrier.
All shifted versions of the prototype filter can be arranged in a
formance, which represents the ratio between the total error of the
transmit matrix defined as
estimated symbols at the receiver side in comparison to the possible
[ ]
𝐽 -QAM symbols and the total energy of the transmitted vector. The 𝐀 = 𝐠0,0 𝐠1,0 ⋯ 𝐠𝐾−1,0 ⋯ 𝐠𝐾−1,𝑀−1 , (11)
EVMRMS of a given OFDM symbol is defined as
where 𝐠𝑘,𝑚 is a vector containing the 𝑔𝑘,𝑚 [𝑛] samples.
Using the transmit matrix 𝐀, (10) can be rewritten as
√
√ ∑𝐾−1 | |2
√
√ 𝑘=0 ||𝑑 [𝑘] − 𝑑̂ [𝑘]|| 𝐬G = 𝐀𝐝, (12)
EVMRMS = 100√ 2
, (8)
∑𝐾−1 2
𝑘=0 |𝑑 [𝑘]|2 where 𝐝 is the vector with the data symbols arranged as
̂
where |(⋅)|𝑝 is the 𝑝−norm operator and 𝑑̂𝑘 is the 𝑘th element of 𝐝. 𝐝 = [𝑑[0, 0] 𝑑[1, 0] ⋯ 𝑑[𝐾 − 1, 𝑀 − 1]]T . (13)

3
L.A.M. Pereira, C.H.S. Lopes, R.M. Borges et al. Optics Communications 474 (2020) 126112

Fig. 3. Simplified block diagram of an F-OFDM system.

Fig. 4. GFDM block diagram.

As OFDM, GFDM uses a CP to protect the signal from FSC. However,

in GFDM, a single CP protects 𝑀 subsymbols, leading to a higher
spectrum efficiency when compared with OFDM. At the receiver side,
the CP is removed before the demodulation process resulting in

𝐲G = 𝐇𝐬G + 𝐰. (14)

GFDM requires equalization before the demodulation to decouple

the subcarriers. As OFDM, FDE can be used, leading to

𝐲eq = 𝐅H −1
𝑁  𝐅𝑁 𝐲G . (15)

After equalization, the data symbols can be retireved by

𝐝̂ = 𝐁ZF 𝐲eq = 𝐀−1 𝐲eq , (16)

where 𝐁ZF is the zero-forcing receive matrix.

GFDM is a very flexible multicarrier waveform that can mimic
OFDM by setting 𝐾 = 𝑁, 𝑀 = 1 and using a rectangular pulse
shape as prototype filter. Other waveforms can also be achieved by
GFDM, making it a good candidate for the air interface in future mobile
networks.

3. The proposed 5G NR Fronthaul

Fig. 5. Multiband FiWi 5G architecture based on A-RoF: LD—laser diode generator;
CO—central office; OPC—optical polarization controller; VSG—vector signal generator;
The proposed multiband FiWi system performance investigation AWG—arbitrary waveform generator; DD-MZM—dual-drive Mach–Zehnder modulator;
has considered eMBB scenario for two reasons: meeting the demands PD—photodetector; DPX—diplexer; EA—electrical amplifier; RAU—remote antenna
for high data rates per user; investigating indoor and outdoor eMBB unit; VSA—vector signal analyzer.
environments, operating at 3.5 and 26 GHz, respectively, since the
Release 15 from 3GPP mainly focus on eMBB scenarios. Fig. 5 depicts a
block diagram of the proposed 5G NR system, which exploits an analog signal, comprising the LTE-A at 778 MHz (RF3 ), whereas VSG3 provides
RoF solution based on external modulation and direct detection, for the 5G NR standard from 768 to 770 MHz, giving rise to the bottom RF-
performing a multiband FH in accordance with the C-RAN principle. driven signal (RF4 ). The marks labeled as ‘‘A", ‘‘B", ‘‘C" and ‘‘D" from
At CO, a dual-drive Mach–Zehnder modulator (DD-MZM) modulates Fig. 5 have been used for the performance analysis, which are going to
a 1560 nm optical carrier, from a distributed-feedback (DFB) laser
be reported in the next Section.
(LD), with two distinct RF-driven signals. The latter ones correspond
to a 3.5-GHz flexible-waveform (GFDM, F-OFDM and OFDM) signal The DD-MZM use favors to mitigate interference between the two
(RF1 ) from the 5G transceiver and a 26-GHz 5G NR signal (RF2 ) from applied RF-driven signals for each case, as is going to be demonstrated
a vector signal generator (VSG1 ), respectively. It enables to initially in Section 4.1. This phenomenon is due to the fact the RF-driven
investigating the RF transmission at both FR1 and FR2 simultaneously. signals individually drive the modulator upper and bottom RF inputs,
Additionally, we aim to evaluate the 5G NR and LTE-A coexistence in with independent bias voltages (VBIAS1 and VBIAS2 ) [27]. Specifically,
the 700 MHz band. In this case, VSG2 provides the upper RF-driven VBIAS1 and VBIAS2 are experimentally managed to lead the modulator

4
L.A.M. Pereira, C.H.S. Lopes, R.M. Borges et al. Optics Communications 474 (2020) 126112

to operate in the quadrature point for reducing nonlinearities and

avoiding distortions.
The modulated field in the optical domain is then launched into
12.5-km single-mode fiber (SMF) toward the remote antenna unit
(RAU), in which the RF signals are recovered by using a photodetector
(PD) and submitted to the first electrical amplification stage (EA1 ). It is
worth mentioning the maximum fronthaul length is limited by latency
at distances up to 20 km [36]. Since our experimental evaluation does
not consider any latency measurement, extending the optical fronthaul
link to 20-km reach only impacts in optical attenuation, resulting
in 1.5-dB additional loss. Assuming the RF1 and RF2 simultaneous
transmission, a diplexer (DPX) separates the 3.5 and 26 GHz signals.
The first signal goes to the 5G transceiver to be analyzed, whereas
the second one is amplified by EA2 , radiated and received by 25-dBi
horn antennas and amplified by EA3 at reception, aiming a mm-waves
femtocell. The latter one performance is investigated by using a vector
signal analyzer (VSA1 ), with the purpose of fulfilling the 3GPP Release
15 requirements, namely: maximum EVMRMS of 12.5respectively. In the Fig. 6. Block diagram of the investigated A-RoF techniques: (a) SD-MZM use;
coexistence case (transmission of RF3 and RF4 ), the amplified photode- (b) DD-MZM use. VOA—variable optical attenuator; OPM—optical power monitor;
ESA—electrical spectrum analyzer.
tected signals feed a commercial 5-dBi log-periodic antenna and are
received by an identical antenna, after 10 m wireless propagation. VSA2
takes charge of measuring the digital performance for 5G NR and LTE-A
signals, as a function of the frequency offset.

4. Experimental results

This section reports the experimental results from three specific

investigations, namely:

– A-RoF techniques for 5G NR fronthauls;

– Multiband 5G NR fronthaul;
– 5G NR and 4G coexistence in a FiWi System.

4.1. A-Rof techniques for 5G NR fronthauls

We investigated two A-RoF techniques for implementing the pro-

posed multiband 5G NR fronthaul. The first one is based on a single-
drive Mach–Zehnder modulator (SD-MZM), as reported in Fig. 6(a).
Since SD-MZM has only one RF input port, a diplexer combines the
two RF-driven signals and delivers the resultant electrical field to the
modulator, which also receives the 1550 nm optical carrier from a
laser and a bias voltage equal to 𝑉𝜋 ∕2. The second A-RoF approach
consists of using a DD-MZM, instead of SD-MZM, for simultaneously
modulating the optical carrier with the distinct RF-driven signals. As
reported in Fig. 6(b), it takes the advantage of not requiring an external
RF combiner, besides allowing independent bias voltage control at
the DD-MZM upper and bottom arms. V BIAS1 and V BIAS2 are properly
set for biasing the modulator close to the quadrature point. In both
methodologies, the modulated optical signal presents optical carrier
and sidebands spaced by the RF-driven frequencies (inset (i) from
Fig. 6). This electrical field in the optical domain was launched into
3-km SMF, before reaching the photodetector. A variable optical at-
tenuator (VOA) and an optical power monitor (OPM) were used for
managing and monitoring the optical power at the photodetector input, Fig. 7. Comparison between the Mach–Zehnder modulators: (a) SD-MZM performance;
respectively. The recovered RF signals were evaluated in terms of (b) DD-MZM performance.

signal-to-noise ratio (SNR), by using an electrical spectrum analyzer

(ESA). Among the available RF-driven-signals, RF-driven signal 1 (RF1 )
and RF-driven signal 2 (RF2 ) were selected to the current investigation, negatively influenced the SNR of the 26 GHz RF signal. For instance,
since higher frequencies typically imply in critical cases. SNR achieved 24.2 dB at 26 GHz in case of ensuring optical power =
Fig. 7 presents the SNR at the photodetector output for both A-RoF -3 dBm and individually transmitting RF2 . However, a degradation of
techniques. This experimental analysis takes into consideration the 10.2 dB was observed in the referred parameter, when the 3.5 GHz
RF signals coexistence and the photodetector input power, which was RF-driven signal was turning on. One can also note the obtained SNR
varied from −9 to −2 dBm. As reported in Fig. 7(a), the SNR obtained at 26 GHz is lower than that achieved at 3.5 GHz for all considered
with the SD-MZM-based approach reached values from 25 to 27.2 dB, optical power levels. It occurs as a consequence of the modulator
when transmitting only the 3.5 GHz RF-driven signal and has not and photodetector frequency response, as well as the additional power
significantly changed by turning RF2 on. In contrast, RF1 activation losses imposed by the RF cables at higher frequencies.

5
L.A.M. Pereira, C.H.S. Lopes, R.M. Borges et al. Optics Communications 474 (2020) 126112

Fig. 8. Measured electrical spectrum at the EA1 output. Fig. 9. The 5G transceiver performance analysis at 3.5 GHz. The insets report the
F-OFDM constellations for -50, -39 and -24 dBm received power levels.

Table 1
From Fig. 7(b), one can observe the SNR results in case of employing
5G NR signals transmission parameters for 120 kHz subcarrier spacing.
DD-MZM. The SNR parameter of the 3.5 GHz RF signal achieved
Bandwidth (MHz) 𝑁𝑅𝐵 𝑅𝑏 16-QAM 𝑅𝑏 64-QAM
approximately 33 dB, due to the optimized bias voltages and has not
50 32 130.59 Mbit/s 195.89 Mbit/s
suffered significant degradation when RF2 was turned on. Furthermore, 100 66 269.35 Mbit/s 404.03 Mbit/s
the SNR degradation at 26 GHz due to the RF1 activation was signifi- 200 132 538.71 Mbit/s 808.06 Mbit/s
cantly reduced, e.g. assuming 1.9 dB in case of ensuring optical power 400 264 1.07 Gbit/s 1.61 Gbit/s
= -3 dBm. As a conclusion, the obtained results indicate a promising
direction on employing DD-MZMs to support multiband A-RoF systems.
29.5 dBm. The performance investigation of the proposed higher-band
4.2. Multiband 5G NR fronthaul
FiWi system has been divided into three phases: EVMRMS as a function
of EIRP (Marker ‘‘C" from Fig. 5); EVMRMS as a function of the PD
The performance of the multiband 5G NR fronthaul depicted in
optical input power (Marker ‘‘A" from Fig. 5); EVMRMS as a function
Fig. 5 was investigated as a function of optical and electrical param-
of the 5G NR signal modulation order and bandwidth.
eters. Fig. 8 reports the measured electrical spectrum obtained at the
The back-to-back (B2B) condition has been added to the 64-QAM
EA1 output, in which one can observe the presence of both 5G signals
system results of EVMRMS as a function of EIRP from Fig. 10. B2B
centered at 3.5 and 26 GHz. The power level differences between the
corresponds to directly connecting the RF transmitter and receiver by
signals are due to their bandwidth and the frequency response of the using a 1m-long coaxial cable with 7.7 dB loss at 26 GHz. We have
opto-electronic devices. In addition, it was attested the chosen bias set the laser power as 10 dBm in order to reach the photodetector
voltages (VBIAS1 and VBIAS2 ) ensure spectral purity, since there are no input at 2 dBm, after propagating in the entire RoF-based 12.5-km FH
significant spurious through the entire evaluated frequency range from system. The 3GPP requirement was satisfied for EIRPs > 23.5 dBm for
0 to 30 GHz. all evaluated bandwidths from 50 to 400 MHz. Particularly for 50 MHz,
We have utilized QAM with 6 bits/symbol (64-QAM) with 10 MHz only 15.5 dBm was required to accomplish the EVMRMS threshold.
bandwidth, resulting in 40 Mbit/s, and RF1 power (PRF1 ) has been Moreover, one can note the maximum power penalty, considering the
varied from −35 to −3 dBm. Fig. 9 presents the 5G transceiver per- B2B condition, was 4 dB for 400-MHz bandwidth case with 1.61 Gbit/s
formance analysis for GFDM, OFDM and F-OFDM as a function of throughput.
the received electrical power level (Marker ‘‘B" from Fig. 5). GFDM Fig. 11 displays the measured results of EVMRMS as a function of
and OFDM provided similar performance in accordance with the 3GPP optical power at the PD input, obtained using a VOA and an OPM.
EVMRMS requirement for power levels higher than -49 dBm. On the The 26 GHz-driven signal power has been set as 9 dBm at one of the
other hand, F-OFDM presented EVMRMS < 8.0% only for levels higher DD-MZM arms, whereas the evaluated optical power range at photode-
than -46 dBm, i.e., required 3 dB higher power compared to the other tector has been varied from −10 to 2 dBm for 16-QAM and 64-QAM.
waveforms. Such degradation was due to the nonlinear edge response For EVMRMS > 12.5% limit, one can indeed note higher EVMRMS values
of the filters applied to the subbands for reducing OOBE. By increasing for 16-QAM at lower optical powers, when compared to 64-QAM.
PRF1 , we could enhance SNR and, consequently, significantly reduce However, in this case, the system was operating at very low optical
EVMRMS from 21.5 to 1.33%. One can clearly infer this remarkable signal-to-noise ratio, which degrades the measurement accuracy. This
improvement, by comparing the insets from Fig. 9, which report the phenomenon is due to the fact the distance among the symbols of
F-OFDM constellations for -50, -39 and -24 dBm. As the power level 64-QAM be smaller than that of 16-QAM for the same constellation
increases, the constellations yield even more well-defined symbols. average energy. It means the symbols deviation from the reference
The eMBB scenario implementation has incorporated a 5G NR in- symbols is lower for 64-QAM, which results in lower EVMRMS . It is
door femtocell at 26 GHz based on the transmission parameters sum- worth clarifying two points: i) even at low optical power scenario, 16-
marized in Table 1, including signal bandwidth, number of allocated QAM works properly and according to its 3GPP requirement, thus the
resource blocks (𝑁𝑅𝐵 ) and throughput (Rb ). We have used numerol- EVMRMS measurement is accurate; ii) at high optical power levels, both
ogy (𝜇) equal to 3, implying in subcarriers spaced by 120 kHz. The 16-QAM and 64-QAM operate with approximately the same EVMRMS
attained throughput for 64-QAM with 50, 100, 200 and 400 MHz was and meet their correspondent 3GPP requirements, as expected.
190 Mbit/s, 400 Mbit/s, 800 Mbit/s and 1.61 Gbit/s, respectively. The last and most important phase in the eMBB scenario imple-
The effective isotropic radiated power (EIRP) ranged from 11.5 to mentation was the FiWi system performance analysis, by taking into

6
L.A.M. Pereira, C.H.S. Lopes, R.M. Borges et al. Optics Communications 474 (2020) 126112

Table 2
Optical and electrical power levels.
Setup stage Power
Laser output 11 dBm
DD-MZM output 4.5 dBm
Photodetector input 2 dBm
EIRP 29.5 dBm
RX antenna output −35 dBm
EA3 output 0 dBm
VSA1 input 0 dBm

Fig. 10. EVMRMS as a function of EIRP (Marker ‘‘C" from Fig. 5).

Fig. 12. 5G NR FiWi system performance at 26 GHz, as a function of bandwidth

(Marker ‘‘D" from Fig. 5).

(FDD). The first version of non-standalone (NSA) mode, launched in

December 2017, represented an important breakthrough for making
this network upgrade feasible. The 3GPP Release 15 also standardizes
the standalone (SA) 5G NR and NSA 5G NR/LTE dual-connectivity
modes [37]. Therefore, it becomes important to evaluate the 5G and
4G coexistence in FiWi systems, targeting to share the same Xhaul- BHs,
Fig. 11. EVMRMS as a function of optical power at the PD input (Marker ‘‘A" from MHs and FHs- infrastructure with both cellular network generations.
Fig. 5). Our coexistence analysis proposes the use of an optical FH with
5G NR and LTE-A signals in the 700 MHz band, since this frequency
spectrum has already been used for 4G in the last years and is potential
account the wireless propagation over the 5G NR indoor femtocell, for long-range 5G applications. The vector signal generator (VSG2 )
operating at 26 GHz. This phase has been carried out by using the setup and vector signal generator (VSG3 ) have been used to generate the
from Fig. 5 and the received signal was analyzed by VSA1 at the edge of LTE-A (RF-driven signal 3 (RF3 )) and 5G NR (RF-driven signal 4 (RF4 ))
a 10-m reach femtocell. Since the wireless propagation attenuation in signals, respectively. We have set 10 MHz bandwidth for both RF
the millimeter waves frequency range is very high, we have chosen the signals and realized the coexistence analysis assuming 16- and 64-QAM
best performance points of the two last analyses for implementing the modulation orders, which led to 61.5 and 92.0 Mbit/s total throughput,
FiWi system, which were: 2 dBm at the PD input from Fig. 11; 29.5 dBm respectively. RF3 has been kept at 778 MHz, whereas the RF4 frequency
EIRP from Fig. 10, as summarized in Table 2. carrier has been varied, with the purpose of changing the frequency
One more time, all bandwidths specified by the FR2 5G NR have offset between the RF signals and evaluating a peaceful coexistence.
been analyzed, namely: 50, 100, 200 and 400 MHz. The received signal Concerning the power levels, power of RF3 (PRF3 ) and power of RF4
power at the RX horn antenna output (-35 dBm) has been amplified by (PRF4 ) were -5 dBm, while the PD received optical power was 0 dBm.
EA3 , which provides 35 dB gain and 4.5 dB noise figure. As a result, Fig. 13 presents an experimental performance investigation of
we have obtained 0 dBm channel power at VSA1 . Fig. 12 presents the EVMRMS at the photodetector output, as a function of the frequency
measurements of EVMRMS at the femtocell edge with 10-m reach at offset between the 5G NR and LTE-A signals. Basically, there are two
0 dBm channel power. The 3GPP requirements for either 16- and 64-
very different achievements, i.e. for lower and higher 9 MHz offset,
QAM have been accomplished for all evaluated bandwidths. The Fig. 12
which is the minimum frequency offset for avoiding overlap between
insets display the obtained 5G NR constellations for 50 and 400 MHz
the two wireless communications signals. From 0 to 8 MHz offset, the
bandwidths, in which the received symbols are clearly distinguishable,
interference level is extremely high, given rise to measured EVMRMS
implying in EVMRMS as low as 3.07 and 3.43% for 50 MHz bandwidth,
around 20%, which is prohibitive either for 16- and 64-QAM. The high
and 7.18 and 7.72% for 400 MHz bandwidth. The use of 400 MHz
bandwidth enabled to attaining 1.61 Gbit/s throughput for 64-QAM. interference level is a consequence of the 5G NR and LTE-A signals
overlap in the spectrum, as demonstrated in Fig. 13 inset for 8 MHz
4.3. 5G NR and 4G coexistence in a FiWi System offset (RF3 at 778 MHz and RF4 at 770 MHz). On the other hand,
EVMRMS has been shown always below 4% for frequency offset from
Many research groups and companies have claimed the 5G deploy- 9 to 17 MHz. The occupied bandwidth of each signal is delimited in
ment on top of the existing LTE-A, by using frequency division duplex this case, as we can see in Fig. 13 insets for RF4 at 769 and 768 MHz.

7
L.A.M. Pereira, C.H.S. Lopes, R.M. Borges et al. Optics Communications 474 (2020) 126112

EVMRMS margins could be properly exploited to extend the wireless

and/or optical links, as well as to the robustness purpose. From Fig. 14
insets, one can also note well-defined symbols in the 64-QAM LTE-A
and 5G NR constellations, ensuring a 186 Mbit/s FiWi communication
in case of using 20 MHz bandwidth.

5. Conclusions

This work has reported the implementation of a multiband 5G NR

fiber-wireless system with bandwidth up to 400 MHz, which is the
maximum allowed by the 3GPP Release 15. The simultaneous use
of ultra-high frequency (UHF), microwave and mm-waves allowed to
employ optical backhauls for transporting 5G signals up to a remote
antenna unit in order to accomplish the demands of enhanced mobile
broadband and long-reach scenarios. Our DSP-based flexible-waveform
5G transceiver at 778 MHz has been efficiently applied as an addi-
Fig. 13. Coexistence analysis of the RoF system as function of the frequency offset tional transmitter of our 5G multiband system in conjunction with
between the 5G NR and LTE-A signals in the 700 MHz band.
two commercial vector signal generators that produced the 5G NR
signal at 3.5 and 26 GHz. The OFDM, GFDM and F-OFDM waveforms
have been evaluated aiming the eMBB 5G scenario. OFDM and GFDM
provided similar performance with EVMRMS below the 3GPP limit
for power levels higher than -49 dBm. On the other side, F-OFDM
presented EVMRMS < 8.0% only for power levels higher than -46 dBm,
i.e., required 3 dB higher power compared to the other two waveforms.
Such degradation was due to the filtering process of their subbands,
which reduces OOBE at the price of increasing the EVMRMS . Experi-
mental results at 26 GHz have demonstrated 1.65 Gbit/s total wireless
throughput in accordance with the 3GPP EVMRMS digital performance
requirements.
Furthermore, to the best of our knowledge, an unprecedented 5G
and 4G FiWi system has been proposed and implemented. An efficient
and harmonious coexistence between 5G NR and LTE-A with clean
and well-defined constellations and EVMRMS as low as 1.88% has
successfully demonstrated the applicability of the proposed architecture
for the next wireless communications generations. Future works regard
Fig. 14. Coexistence analysis between the 5G NR and LTE-A standards in the proposed the extension of cell reach of the proposed multiband 5G FiWi system,
FiWi system. by applying digital pre-distortion to the 5G NR signals.

CRediT authorship contribution statement

It justifies the abrupt transition from the prohibitive EVMRMS to the
attractive digital performance. A remarkable EVMRMS of 1.67% has
L.A.M. Pereira: Investigation. C.H.S. Lopes: Investigation. R.M.
been achieved for frequency offsets higher or equal to 12 MHz. As a
Borges: Methodology. E.S. Lima: Investigation. A.C. Ferreira: Soft-
conclusion, the 3GPP EVMRMS requirements of 12.5 and 8% for 16-
ware,Validation. M. Abreu: Funding acquisition. L.L. Mendes: Soft-
and 64-QAM, respectively, have been accomplished with clearance in
ware,Validation. Arismar Cerqueira S. Jr.: Conceptualization, Project
case of not overlapping LTE-A and 5G NR signals.
administration, Supervision.
As a proof of concept, we have implemented a 10-m reach cell
for both LTE-A and 5G NR technologies, coexisting in the same FiWi
system. The proposed coexistence experiment has been carried out by Declaration of competing interest
using the setup from Fig. 5, with the switch at the EA1 output in the
down position. In this way, the photodetected and electrically amplified The authors declare that they have no known competing finan-
RF3 and RF4 signals feed a commercial 5-dBi gain log-periodic antenna cial interests or personal relationships that could have appeared to
and are received by an identical antenna, after 10 m wireless propaga- influence the work reported in this paper.
tion. Since the LTE-A standard specifies bandwidth of up to 20 MHz,
the 4G and 5G coexistence in the FiWi system has been investigated
Acknowledgments
assuming 10 and 20 MHz bandwidths. For RF signals with BW equal
to 10 MHz, as in the previous analysis, the frequency offset between
them was set to 9 MHz. On the other hand, by taking RF3 and RF4 with This work was partially supported by RNP, with resources from MC-
20 MHz bandwidth, the frequency offset was set to 19 MHz for making TIC, Brazil, Grant No. 01250.075413/2018-04, under the Radiocom-
the RF signals closed each other, but not overlapped. Once again, 16- munication Reference Center (Centro de Referência em Radiocomuni-
and 64-QAM modulation orders were considered in the experiment. cações - CRR) project of the National Institute of Telecommunications
The collected results using the vector signal analyzer (VSA2 ) are (Instituto Nacional de Telecomunicações - Inatel), Brazil. We would like
shown in Fig. 14. One can observe all evaluated configurations effi- to thank the Project No. D0640 from FAPEMIG/CEMIG/ANEEL. The
ciently met the 3GPP EVMRMS requirements, with margins of approx- authors also thank the technical support of José Reis from Rohde &
imately 10% and 5% for 16-QAM and 64-QAM, respectively. These Schwarz and Renato Guadagnini from Venturus.

8
L.A.M. Pereira, C.H.S. Lopes, R.M. Borges et al. Optics Communications 474 (2020) 126112

References [20] G. Kalfas, C. Vagionas, A. Antonopoulos, E. Kartsakli, A. Mesodiakaki, S. Pa-

paioannou, P. Maniotis, J.S. Vardakas, C. Verikoukis, N. Pleros, Next generation
[1] ITU-R, IMT Vision–Framework and overall objectives of the future development fiber-wireless fronthaul for 5G mmwave networks, IEEE Commun. Mag. 57 (3)
of IMT for 2020 and beyond, 2015, Recommendation ITU-R M.2083-0, 1-19. (2019) 138–144.
[2] S. Sun, B. Rong, R.Q. Hu, Y. Qian, Spatial domain management and massive [21] A. Delmade, C. Browning, A. Farhang, N. Marchetti, L.E. Doyle, R.D. Koilpillai,
MIMO coordination in 5G SDN, IEEE Access (ISSN: 21693536) 3 (2015) L.P. Barry, D. Venkitesh, Performance analysis of analog IF over fiber fronthaul
2238–2251. link with 4G and 5G coexistence, IEEE/OSA J. Opt. Commun. Networking 10
[3] H. Gamage, N. Rajatheva, M. Latva-Aho, Channel coding for enhanced mobile (3) (2018) 174–182.
broadband communication in 5G systems, in: 2017 European Conference on [22] F. Tian, D. Guo, B. Liu, Q. Zhang, Q. Tian, R. Ullah, X. Xin, A novel concatenated
Networks and Communications (EuCNC), IEEE, 2017, pp. 1–6. coded modulation based on GFDM for access optical networks, IEEE Photonics
[4] B. Farhang-Boroujeny, H. Moradi, OFDM Inspired waveforms for 5G, IEEE J. 10 (2) (2018) 1–8.
Commun. Surv. Tutor. (ISSN: 1553877X) 18 (4) (2016) 2474–2492. [23] M. Bi, W. Jia, L. Li, X. Miao, W. Hu, Investigation of F-OFDM in 5G fronthaul
[5] A. Osseiran, F. Boccardi, V. Braun, K. Kusume, P. Marsch, M. Maternia, O. networks for seamless carrier-aggregation and asynchronous transmission, in:
Queseth, M. Schellmann, H. Schotten, H. Taoka, H. Tullberg, M.A. Uusitalo, 2017 Optical Fiber Communications Conference and Exhibition (OFC), 2017, pp.
B. Timus, M. Fallgren, Scenarios for 5G mobile and wireless communications: 1–3.
The vision of the METIS project, IEEE Commun. Mag. (ISSN: 01636804) 52 (5) [24] W. Liu, T. Shao, J. Yao, Ultra-wideband and 60-GHz generation and transmission
(2014) 26–35. over a wavelength division multiplexing-passive optical network, IEEE/OSA J.
[6] I. Philbeck, Connecting the unconnected: working together to achieve Connect Opt. Commun. Networking 5 (9) (2013) 1076–1082.
2020 Agenda targets, 2017, ITU White Paper. [25] M. Morant, A. Trinidad, E. Tangdiongga, T. Koonen, R. Llorente, Experimental
[7] 3GPP, 5g; nr; overall description; stage-2, 2018, TS 38.300 version 15.3.1 Release demonstration of mm-wave 5G NR photonic beamforming based on ORRs and
15. multicore fiber, IEEE Trans. Microw. Theory Tech. 67 (7) (2019) 2928–2935.
[8] Y. Jiang, New companding transform for PAPR reduction in OFDM, IEEE [26] M. Noweir, Q. Zhou, A. Kwan, R. Valivarthi, M. Helaoui, W. Tittel, F.M.
Commun. Lett. (ISSN: 10897798) 14 (4) (2010) 282–284. Ghannouchi, Digitally linearized radio-over fiber transmitter architecture for
[9] J. Van De Beek, F. Berggren, Out-of-band power suppression in OFDM, IEEE cloud radio access network’s downlink, IEEE Trans. Microw. Theory Tech. 66
Commun. Lett. (ISSN: 10897798) 12 (9) (2008) 609–611. (7) (2018) 3564–3574.
[10] N. Michailow, M. Matthé, I.S. Gaspar, A.N. Caldevilla, L.L. Mendes, A. Festag, G. [27] R.M. Borges, T.R.R. Marins, M.S.B. Cunha, H.R.D. Filgueiras, I.F. da Costa, R.N.
Fettweis, Generalized frequency division multiplexing for 5th generation cellular da Silva, D.H. Spadoti, L.L. Mendes, Arismar Cerqueira S. Jr., Integration of a
networks, IEEE Trans. Commun. 62 (9) (2014) 3045–3061. GFDM-based 5g transceiver in a GPON using radio over fiber technology, J.
[11] R. Gerzaguet, N. Bartzoudis, L.G. Baltar, V. Berg, J.-B. Doré, D. Kténas, O. Font- Lightwave Technol. 36 (19) (2018) 4468–4477.
Bach, X. Mestre, M. Payaró, M. Färber, et al., The 5G candidate waveform race: [28] H.R.D. Filgueiras, R.M. Borges, M. Caldano Melo, T.H. Brandão, Arismar
a comparison of complexity and performance, EURASIP J. Wireless Commun. Cerqueira S. Jr., Dual-band wireless fronthaul using a FSS-based focal-
Networking 2017 (1) (2017) 13. point/cassegrain antenna assisted by an optical midhaul, IEEE Access 7 (2019)
[12] 3GPP, User Equipment (UE) radio transmission and reception; Part 3: Range 1 112578–112587.
and Range 2 Interworking operation with other radios, 2019, TS 38.101-3 version [29] R.M. Borges, L.A.M. Pereira, H.R.D. Filgueiras, A.C. Ferreira, M.S.B. Seda, E.R.
16.0.0 Release 16. Neto, D.H. Spadoti, L.L. Mendes, Arismar Cerqueira S. Jr., DSP-Based flexible-
[13] C. Bockelmann, N.K. Pratas, G. Wunder, S. Saur, M. Navarro, D. Gregoratti, G. waveform and multi-application 5G fiber-wireless system, J. Lightwave Technol.
Vivier, E. De Carvalho, Y. Ji, C. Stefanovic, P. Popovski, Q. Wang, M. Schellmann, 38 (3) (2020) 642–653.
E. Kosmatos, P. Demestichas, M. Raceala-Motoc, P. Jung, S. Stanczak, A. Dekorsy, [30] W. Dias, D. Gaspar, L. Mendes, M. Chafii, M. Matthé, P. Neuhaus, G. Fettweis,
Towards massive connectivity support for scalable mMTC communications in 5G Performance analysis of a 5G transceiver implementation for remote areas
networks, IEEE Access (ISSN: 21693536) 6 (2018) 28969–28992. scenarios, in: 2018 European Conference on Networks and Communications
[14] S. Rangan, T.S. Rappaport, E. Erkip, Millimeter-wave cellular wireless networks: (EuCNC), 2018, pp. 363–367.
Potentials and challenges, Proc. IEEE (ISSN: 00189219) 102 (3) (2014) 366–385. [31] A. Hoglund, D.P. Van, T. Tirronen, O. Liberg, Y. Sui, E.A. Yavuz, 3GPP Release
[15] Y. Niu, Y. Li, D. Jin, L. Su, A.V. Vasilakos, A survey of millimeter wave 15 early data transmission, IEEE Commun. Stand. Mag. 2 (2) (2018) 90–96.
communications (mmwave) for 5G: opportunities and challenges, Wirel. Netw. [32] R. Ahmed, F. Tosato, M. Maso, Overhead Reduction of NR type II CSI for NR
21 (8) (2015) 2657–2676. Release 16, in: WSA 2019; 23rd International ITG Workshop on Smart Antennas,
[16] M.A. Habibi, M. Nasimi, B. Han, H.D. Schotten, A comprehensive survey of RAN vol. 1, 2019, pp. 1–5.
architectures toward 5G mobile communication system, IEEE Access 7 (2019) [33] R. Drath, A. Horch, Industrie 4.0: Hit or hype? [industry forum], IEEE Ind.
70371–70421. Electron. Mag. 8 (2) (2014) 56–58.
[17] J. Capmany, D. Novak, Microwave photonics combines two worlds, Nature [34] J.A.C. Bingham, Multicarrier modulation for data transmission: an idea whose
Photon. 1 (6) (2007) 319–330. time has come, IEEE Commun. Mag. 28 (5) (1990) 5–14.
[18] D. Wake, A. Nkansah, N.J. Gomes, Radio over fiber link design for next [35] P. Weitkemper, J. Bazzi, K. Kusume, A. Benjebbour, Y. Kishiyama, On regular
generation wireless systems, J. Lightwave Technol. 28 (16) (2010) 2456–2464. resource grid for filtered OFDM, IEEE Commun. Lett. 20 (12) (2016) 2486–2489.
[19] C. I, H. Li, J. Korhonen, J. Huang, L. Han, RAN Revolution with NGFI (xhaul) [36] S. Ahmadi, 5G NR: Architecture, Technology, Implementation, and Operation of
for 5G, J. Lightwave Technol. 36 (2) (2018) 541–550. 3GPP New Radio Standards, Academic Press, 2019.
[37] S. Parkvall, E. Dahlman, A. Furuskar, M. Frenne, NR: The new 5G radio access
technology, IEEE Commun. Stand. Mag. 1 (4) (2017) 24–30.