1

A Tutorial on Beam Management for

3GPP NR at mmWave Frequencies
Marco Giordani, Student Member, IEEE, Michele Polese, Student Member, IEEE, Arnab Roy, Member, IEEE,
Douglas Castor, Member, IEEE, Michele Zorzi, Fellow, IEEE

Abstract—The millimeter wave (mmWave) frequencies offer In this context, the millimeter wave (mmWave) spectrum –
the availability of huge bandwidths to provide unprecedented roughly above 10 GHz1 – has been considered as an enabler
data rates to next-generation cellular mobile terminals. However, of the 5G performance requirements in micro and picocellular
arXiv:1804.01908v2 [cs.NI] 4 Nov 2019

mmWave links are highly susceptible to rapid channel variations

and suffer from severe free-space pathloss and atmospheric networks [6], [7]. These frequencies offer much more band-
absorption. To address these challenges, the base stations and the width than current cellular systems in the congested bands
mobile terminals will use highly directional antennas to achieve below 6 GHz, and initial capacity estimates have suggested
sufficient link budget in wide area networks. The consequence that networks operating at mmWaves can offer orders of
is the need for precise alignment of the transmitter and the magnitude higher bit-rates than 4G systems [8]. Nonetheless,
receiver beams, an operation which may increase the latency of
establishing a link, and has important implications for control the higher carrier frequency makes the propagation conditions
layer procedures, such as initial access, handover and beam harsher than at the lower frequencies traditionally used for
tracking. This tutorial provides an overview of recently proposed wireless services, especially in terms of robustness [9]. Sig-
measurement techniques for beam and mobility management in nals propagating in the mmWave band suffer from increased
mmWave cellular networks, and gives insights into the design pathloss and severe channel intermittency, and are blocked by
of accurate, reactive and robust control schemes suitable for
a 3GPP NR cellular network. We will illustrate that the best many common materials such as brick or mortar [10], and
strategy depends on the specific environment in which the nodes even the changing position of the body relative to the mobile
are deployed, and give guidelines to inform the optimal choice device can lead to rapid drops in signal strength.
as a function of the system parameters. To deal with these impairments, next-generation cellular
Index Terms—5G, NR, mmWave, 3GPP, beam management. networks must provide a set of mechanisms by which User
Equipments (UEs) and mmWave Next Generation Node Base
I. I NTRODUCTION (gNB) stations2 establish highly directional transmission links,
typically using high-dimensional phased arrays, to benefit from
From analog through Long Term Evolution (LTE), each the resulting beamforming gain and sustain an acceptable
generation of mobile technology has been motivated by the communication quality. Directional links, however, require fine
need to address the challenges not overcome by its pre- alignment of the transmitter and receiver beams, achieved
decessor. The 5th generation (5G) of mobile technology is through a set of operations known as beam management. They
positioned to address the demands and business contexts of are fundamental to perform a variety of control tasks including
2020 and beyond. It is expected to enable a fully mobile (i) Initial Access (IA) [11], [12] for idle users, which allows a
and connected society, related to the tremendous growth in mobile UE to establish a physical link connection with a gNB,
connectivity and density/volume of traffic that will be required and (ii) beam tracking, for connected users, which enable beam
in the near future [2], to provide and guarantee: (i) very high adaptation schemes, or handover, path selection and radio link
throughput (1 Gbps or more), to support ultra-high definition failure recovery procedures [13], [14]. In current LTE systems,
video and virtual reality applications; (ii) very low latency these control procedures are performed using omnidirectional
(even less than 1 ms in some cases), to support real-time signals, and beamforming or other directional transmissions
mobile control and Device-to-Device (D2D) applications and can only be performed after a physical link is established, for
communications; (iii) ultra high reliability; (iv) low energy data plane transmissions. On the other hand, in the mmWave
consumption; and (v) ultra high connectivity resilience and bands, it may be essential to exploit the antenna gains even
robustness [3] to support advanced safety applications and during initial access and, in general, for control operations.
services. In order to meet these complex and sometimes con- Omnidirectional control signaling at such high frequencies,
tradictory requirements, 5G will encompass both an evolution indeed, may generate a mismatch between the relatively short
of traditional 4G-LTE networks and the addition of a new radio range at which a cell can be detected or the control signals
access technology, globally standardized by the 3rd Generation can be received (control-plane range), and the much longer
Partnership Project (3GPP) as NR [4], [5]. range at which a user could send and receive data when
Marco Giordani, Michele Polese and Michele Zorzi are with the Department using beamforming (data-plane range). However, directionality
of Information Engineering (DEI), University of Padova, Italy, and Consorzio
1 Although strictly speaking mmWave bands include frequencies between
Futuro in Ricerca (CFR), Italy. Email:{giordani,polesemi,zorzi}@dei.unipd.it.
Arnab Roy and Douglas Castor are with InterDigital Communications, Inc., 30 and 300 GHz, industry has loosely defined it to include any frequency
USA. Email:{arnab.roy,douglas.castor}@interdigital.com. above 10 GHz.
Part of this work has been submitted for publication at Med-Hoc-Net 2018 [1]. 2 Notice that gNB is the NR term for a base station.
 2

can significantly delay the access procedures and make the advanced beamforming capabilities (e.g., digital or hybrid
performance more sensitive to the beam alignment. These are beamforming), which allow the transceiver to sweep
particularly important issues in 5G networks, and motivate the multiple directions at any given time.
need to extend current LTE control procedures with innovative • Illustrate some of the complex and interesting trade-
mmWave-aware beam management algorithms and methods. offs to be considered when designing solutions for next-
generation cellular networks by examining a wide set
of parameters based on 3GPP NR considerations and
A. Contributions
agreements (e.g., the frame structure and other relevant
This paper is a tutorial on the design and dimensioning physical-layer aspects).
of beam management frameworks for mmWave cellular net- In general, the results prove that the optimal design choices
works. In particular, we consider the parameters of interest for for implementing efficient and fast initial access and reactive
3GPP NR networks, which will support carrier frequencies tracking of the mobile user strictly depend on the specific envi-
up to 52.6 GHz [5]. We also report an evaluation of beam ronment in which the users are deployed, and must account for
management techniques, including initial access and tracking several specific features such as the base stations density, the
strategies, for cellular networks operating at mmWaves under antenna geometry, the beamforming configuration and the level
realistic NR settings and channel configurations, and describe of integration and harmonization of different technologies.
how to optimally design fast, accurate and robust control-
plane management schemes through measurement reports in
different scenarios. More specifically, in this tutorial we: B. Organization
• Provide an overview of the most effective measure-
The sections of this tutorial are organized as follows. Sec. II
ment collection frameworks for 5G systems operating reports the related work on beam management at mmWave
at mmWaves. We focus on Downlink (DL) and Uplink frequencies. Sec. III provide basic information on the 3GPP
(UL) frameworks, according to whether the reference Release 15 frame structure for NR, and presents the candidate
signals are sent from the gNBs to the UEs or vice DL and UL measurement signals that can be collected by
versa, respectively, and on Non-Standalone (NSA) and the NR nodes for the beam management operations. Sec. IV
standalone (SA) architectures, according to whether the describes the beam management frameworks whose perfor-
control plane is managed with the support of an LTE mance will be analyzed, simulated and compared in the
overlay or not, respectively. A DL configuration is in remainder of the work. Sec. V defines the parameters that
line with the 3GPP specifications for NR and reduces the affect the performance of beam management in NR. Sec. VI
energy consumption at the UE side, but it may be lead to reports a performance evaluation and some considerations on
a worse beam management performance than in the UL. the trade-offs and on which are the best configurations for
Moreover, when considering stable and dense scenarios beam management frameworks. Additional considerations and
which are marginally affected by the variability of the final remarks, aiming at providing guidelines for selecting the
mmWave channel, an SA architecture is preferable for optimal IA and tracking configuration settings as a function of
the design of fast IA procedures, while an NSA scheme the system parameters, are stated in Sec. VII. Finally, Sec. VIII
may be preferable for reducing the impact of the overhead concludes the paper.
on the system performance and enable more robust and
stable communication capabilities. II. R ELATED W ORK
• Simulate the performance of the presented measurement Measurement reporting is quite straightforward in LTE [41]:
frameworks in terms of signal detection accuracy, using the DL channel quality is estimated from an omnidirectional
a realistic mmWave channel model based on real-world signal called the Cell Reference Signal (CRS), which is
measurements conducted in a dense, urban scenario in regularly monitored by each UE in connected state to cre-
which environmental obstructions (i.e., urban buildings) ate a wideband channel estimate that can be used both for
can occlude the path between the transmitter and the demodulating downlink transmissions and for estimating the
receiver. The tutorial shows that accurate beam man- channel quality [42]. However, when considering mmWave
agement operations can be guaranteed when configuring networks, in addition to the rapid variations of the channel,
narrow beams for the transmissions, small subcarrier CRS-based estimation is challenging due to the directional
spacings, denser network deployments and by adopting nature of the communication, thus requiring the network and
frequency diversity schemes. the UE to constantly monitor the direction of transmission of
• Analyze the reactiveness (i.e., how quickly a mobile user each potential link. Tracking changing directions can decrease
gets access to the network and how quickly the frame- the rate at which the network can adapt, and can be a major
work is able to detect an updated channel condition), obstacle in providing robust and ubiquitous service in the face
and the overhead (i.e., how many time and frequency of variable link quality. In addition, the UE and the gNB may
resources should be allocated for the measurement op- only be able to listen to one direction at a time, thus making it
erations). In general, fast initial access and tracking hard to receive the control signaling necessary to switch paths.
schemes are ensured by allocating a large number of To overcome these limitations, several approaches in the
time/frequency resources to the users in the system, at literature, as summarized in Table I, have proposed directional-
the expense of an increased overhead, and by using based schemes to enable efficient control procedures for both
 3

Topic Relevant References

IEEE 802.11ad [15] [16], [17], [18]. Not suitable for long-range, dynamic and outdoor scenarios.
[20], [21], [22] Exhaustive search.
[23], [24], [25] More advanced searching schemes.
Initial Access [11], [12], [19]
[26], [27], [28], [29] Context-aware initial access.
[30], [31] Performance comparison.
[32], [33], [34] Mobility-aware strategies.
Beam Management [14]
[35], [36], [37], [38], [39], [40] Multi-connectivity solutions.

TABLE I: Relevant literature on measurement reporting, initial access and beam management strategies for mmWave networks.

the idle and the connected mobile terminals, as surveyed in and maintenance under node mobility. In [34], the authors
the following paragraphs. proposed the use of an extended Kalman filter to enable a static
Papers on IA3 and tracking in 5G mmWave cellular systems base station, equipped with a digital beamformer, to effectively
are very recent. Most literature refers to challenges that have track a mobile node equipped with an analog beamformer
been analyzed in the past at lower frequencies in ad hoc after initial channel acquisition, with the goal of reducing the
wireless network scenarios or, more recently, referred to the 60 alignment error and guarantee a more durable connectivity. Re-
GHz IEEE 802.11ad WLAN and WPAN scenarios (e.g., [15], cently, robust IA and tracking schemes have been designed by
[16], [17]). However, most of the proposed solutions are un- leveraging out-of-band information to estimate the mmWave
suitable for next-generation cellular network requirements and channel. In [14], [35], [36], [37] an approach where 5G
present many limitations (e.g., they are appropriate for short- cells operating at mmWaves (offering much higher rates) and
range, static and indoor scenarios, which do not match well traditional 4G cells below 6 GHz (providing much more robust
the requirements of 5G systems). Therefore, new specifically operation) are employed in parallel have been proved to enable
designed solutions for cellular networks need to be found. fast and resilient tracking operations. In [38], a framework
In [20], [21], the authors propose an exhaustive method which integrates both LTE and 5G interfaces is proposed as
that performs directional communication over mmWave fre- a solution for mobility-related link failures and throughput
quencies by periodically transmitting synchronization signals degradation of cell-edge users, relying on coordinated trans-
to scan the angular space. The result of this approach is missions from cooperating cells are coordinated for both data
that the growth of the number of antenna elements at either and control signals. In [39], a novel approach for analyzing
the transmitter or the receiver provides a large performance and managing mobility in joint sub-6GHz–mmWave networks
gain compared to the case of an omnidirectional antenna. is proposed by leveraging on device caching along with the
However, this solution leads to a long duration of the IA capabilities of dual-mode base stations to minimize handover
with respect to LTE, and poorly reactive tracking. Similarly, failures, reduce inter-frequency measurement, reduce energy
in [22], measurement reporting design options are compared, consumption, and provide seamless mobility in emerging
considering different scanning and signaling procedures, to dense heterogeneous networks. Moreover, the authors in [40]
evaluate access delay and system overhead. The channel illustrate how to exploit spatial congruence between signals
structure and multiple access issues are also considered. The in different frequency bands and extract mmWave channel
analysis demonstrates significant benefits of low-resolution parameters from side information obtained in another band.
fully digital architectures in comparison to single stream ana- Despite some advantages, the use of out-of-band information
log beamforming. Additionally, more sophisticated discovery for the 5G control plane management poses new challenges
techniques (e.g., [23], [24]) alleviate the exhaustive search that remain unsolved and which deserve further investigation.
delay through the implementation of a multi-phase hierarchical Context information can also be exploited to improve the
procedure based on the access signals being initially sent in cell discovery procedure and minimize the delay [26], [27],
few directions over wide beams, which are iteratively refined while capturing the effects of position inaccuracy in the pres-
until the communication is sufficiently directional. In [25] a ence of obstacles. In the scheme proposed in [28], booster cells
low-complexity beam selection method by low-cost analog (operating at mmWave) are deployed under the coverage of an
beamforming is derived by exploiting a certain sparsity of anchor cell (operating at LTE frequencies). The anchor base
mmWave channels. It is shown that beam selection can be station gets control over IA informing the booster cell about
carried out without explicit channel estimation, using the user locations, in order to enable mmWave gNB to directly
notion of compressive sensing. steer towards the user position. Finally, in [29], the authors
The issue of designing efficient beam management solutions studied how the performance of analog beamforming degrades
for mmWave networks is addressed in [32], in which the author in the presence of angular errors in the available Context
designs a mobility-aware user association strategy to overcome Information during the initial access or tracking procedures,
the limitations of the conventional power-based association according to the status of the UE (connected or non-connected,
schemes in a mobile 5G scenario. Other relevant papers on respectively).
this topic include [33], in which the authors propose smart The performance of the association techniques also de-
beam tracking strategies for fast mmWave link establishment pends on the beamforming architecture implemented in the
3 We refer to works [11], [12], [19] for a detailed taxonomy of recent IA transceivers. Preliminary works aiming at finding the optimal
strategies. beamforming strategy refer to WLAN scenarios. For example,
 4

the algorithm proposed in [18] takes into account the spatial PBCH
239
distribution of nodes to allocate the beamwidth of each antenna 192

pattern in an adaptive fashion and satisfy the required link 182

budget criterion. Since the proposed algorithm minimizes the

PBCH (with DMRS)

PBCH (with DMRS)

collisions, it also minimizes the average time required to

subcarriers
transmit a data packet from the source to the destination

SSS
PSS
through a specific direction. In 5G scenarios, papers [20],
[21], [23] give some insights on trade-offs among different
beamforming architectures in terms of users’ communication 56
quality. More recently, articles [30], [31] evaluate the mmWave 47
cellular network performance while accounting for the beam 0
PBCH

training, association overhead and beamforming architecture. 0 1 2 3

OFDM symbols
The results show that, although employing wide beams, initial
beam training with full pilot reuse is nearly as good as perfect Fig. 1: SS block structure [60].
beam alignment. However, they lack considerations on the
latest 3GPP specifications for NR. Finally, paper [56] provides A. NR Frame Structure
an overview of the main features of NR with respect to initial The 3GPP technical specification in [43] and the report
access and multi-beam operations, and article [57] reports the in [5] provide the specifications for the PHY layer. Both
details on the collection of channel state information in NR. Frequency Division Duplexing (FDD) and Time Division
However, both these papers only present a high level overview, Duplexing (TDD) will be supported.
and do not include a comprehensive performance evaluation of The waveform is Orthogonal Frequency Division Multiplex-
NR beam management frameworks at mmWave frequencies. ing (OFDM) with a cyclic prefix. Different numerologies4
The above discussion makes it apparent how next-generation will be used, in order to address the different use cases of
mmWave cellular networks should support a mechanism by 5G [59]. The frame structure follows a time and frequency
which the users and the infrastructure can quickly determine grid similar to that of LTE, with a higher number of con-
the best directions to establish the mmWave links, an op- figurable parameters. The subcarrier spacing is 15 × 2n kHz,
eration which may increase the latency and the overhead n ∈ Z, n ≤ 4. In Release 15, there will be at most 3300
of the communication and have a substantial impact on the subcarriers, for a maximum bandwidth of 400 MHz. A frame
overall network performance. In the remainder of this paper lasts 10 ms, with 10 subframes of 1 ms. It will be possible to
we will provide guidelines to characterize the optimal beam multiplex different numerologies for a given carrier frequency,
management strategies as a function of a variety of realistic and the whole communication must be aligned on a subframe
system parameters. basis. A slot is composed of 14 OFDM symbols. There are
multiple slots in a subframe, and their number is given by
the numerology used, since the symbol duration is inversely
proportional to the subcarrier spacing [4]. Mini-slots are also
III. F RAME S TRUCTURE AND S IGNALS FOR supported: they can be as small as 2 OFDM symbol and have
3GPP NR AT MM WAVE F REQUENCIES variable length, and can be positioned asynchronously with
respect to the beginning of the slot (so that low-latency data
can be sent without waiting for the whole slot duration).
Given that NR will support communication at mmWave
frequencies, it is necessary to account for beamforming and
B. NR Measurements for Beam Management
directionality in the design of its Physical (PHY) and Medium
Access Control (MAC) layers. The NR specifications will thus Regular beam management operations are based on the con-
include a set of parameters for the frame structure dedicated trol messages which are periodically exchanged between the
to high carrier frequencies, as well as synchronization and ref- transmitter and the receiver nodes. In the following paragraphs
erence signals that enable beam management procedures [5]. we will review the most relevant DL and UL measurement sig-
In this regard, in Sec. III-A and Sec. III-B we introduce the nals supported by 3GPP NR for beam management purposes,
3GPP frame structure and measurement signals proposed for as summarized in Table II.
NR, respectively, which will provide the necessary background 4 The term numerology refers to a set of parameters for the waveform, such
for the remainder of this tutorial. as subcarrier spacing and cyclic prefix duration for OFDM [58].

Initial Access (Idle UE) Tracking (Connected UE)

SS blocks (carrying the PSS, the SSS, and the PBCH). CSI-RSs and SS blocks.
Downlink
See references [5], [43], [44], [45], [46], [47]. See references [5], [43], [48], [49], [50], [51], [52], [53].
3GPP does not use uplink signals for initial access, but the
Uplink SRSs. See references [5], [43], [54], [55].
usage of SRSs has been proposed in [36], [35], [14]
TABLE II: Reference signals for beam management operations, for users in idle and connected states, in downlink or uplink.
 5

frequency frequency

NCSI NCSI

SS Burst SS Burst SS Burst SS Burst

CSI-RS

CSI-RS

CSI-RS

CSI-RS
CSI-RS
TCSI TCSI TCSI TCSI
SS Block

SS Block
SS Block

SS Block

SS Block
SS Block
OCSI OCSI

Δf

Δf
time time
TSS TSS

(a) Option 1: the first CSI-RS is sent TCSI ms after an SS burst. (b) Option 2: the first CSI-RC is sent OCSI ms after an SS burst.

Fig. 2: Examples of CSI-RS measurement window and periodicity configurations. SS blocks are sent every TSS ms, and they embed time and frequency
offsets indicating the time and frequency allocation of CSI-RS signals within the frame structure.

Downlink Measurements: SS Blocks. In the most recent which represents the time interval between the end of the
versions of the 3GPP specifications [43], the concept of SS SS burst and the first CSI-RS. The CSI-RSs, which may not
block and burst emerged for periodic synchronization signal necessarily be broadcast through all the available frequency
transmission from the gNBs. An SS block is a group of 4 resources [49], may span N =1, 2 or 4 OFDM symbols [64].
OFDM symbols [43, Sec. 7.4.3] in time and 240 subcarriers For periodic CSI-RS transmissions, the supported periodicities
in frequency (i.e., 20 resource blocks) [44], as shown in Fig. 1. are TCSI,slot ∈ {5, 10, 20, 40, 80, 160, 320, 640} slots [43],
It carries the PSS, the SSS and the PBCH. The DeModulation thus the actual periodicity in time depends on the slot duration.
Reference Signal (DMRS) associated with the PBCH can be As we assessed in the previous sections of this work, when
used to estimate the Reference Signal Received Power (RSRP) considering directional communications, the best directions
of the SS block. In a slot of 14 symbols, there are two possible for the beams of the transceiver need to be periodically
locations for SS blocks: symbols 2-5 and symbols 8-11. identified (e.g., through beam search operations), in order to
The SS blocks are grouped into the first 5 ms of an SS maintain the alignment between the communicating nodes.
burst [45], which can have different periodicities TSS . At the For this purpose, SS- and CSI-based measurement results can
time of writing, the value of TSS is still under discussion in be jointly used to reflect the different coverage which can be
3GPP, and the candidates are TSS ∈ {5, 10, 20, 40, 80, 160} achieved through different beamforming architectures [51].
ms [61]. When accessing the network for the first time, the As far as CSI signals are concerned, the communication
UE should assume a periodicity TSS = 20 ms [62]. quality can be derived by averaging the signal quality from
the NCSI,RX best beams among all the available ones, where
The maximum number L of SS blocks in a burst is
the value of NCSI,RX can be configured to 1 or more than
frequency-dependent [45], and above 6 GHz there could be
1 [48]5 . Nevertheless, to avoid the high overhead associated
up to 64 blocks per burst. When considering frequencies
with wide spatial domain coverage with a huge number of
for which beam operations are required [63], each SS block
very narrow beams, on which CSI-RSs are transmitted, it
can be mapped to a certain angular direction. To reduce
is reasonable to consider transmitting only subsets of those
the impact of SS transmissions, SS can be sent through
beams, based on the locations of the active UEs. This is also
wide beams, while data transmission for the active UE is
important for UE power consumption considerations [53]. For
usually performed through narrow beams, to increase the gain
example, the measurement results based on SS blocks (and
produced by beamforming [47].
referred to a subset of transmitting directions) can be used
to narrow down the CSI-RS resource sets based on which a
Downlink Measurements: CSI-RS. It has been agreed that UE performs measurements for beam management, thereby
CSI-RSs can be used for Radio Resource Management (RRM) increasing the energy efficiency.
measurements for mobility management purposes in connected
mode [5]. As in LTE, it shall be possible to configure multiple Uplink Measurements: SRS The SRSs are used to monitor
CSI-RS to the same SS burst, in such a way that the UE the uplink channel quality, and are transmitted by the UE and
can first obtain synchronization with a given cell using the received by the gNBs. According to [54], their transmission
SS bursts, and then use that as a reference to search for is scheduled by the gNB to which the UE is attached, which
CSI-RS resources [48]. Therefore, the CSI-RS measurement also signals to the UE the resource and direction to use for
window configuration should contain at least the periodicity the transmission of the SRS. The UE may be configured with
and time/frequency offsets relative to the associated SS burst.
Fig. 2 shows the two options we consider for the time offset of 5 The maximum value for N
CSI,RX has not been standardized yet. In [61]
the CSI-RS transmissions. The first option, shown in Fig. 2a, it is specified that, for the derivation of the quality of a cell, the UEs should
allows the transmission of the first CSI-RS TCSI ms after consider an absolute threshold, and average the beams with quality above the
threshold, up to NCSI,RX beams. If there are no beams above threshold, then
the end of an SS burst. The second one, shown in Fig. 2b, the best one (regardless of its absolute quality) should be selected for the cell
has an additional parameter, i.e., an offset in time OCSI , quality derivation.
 6

multiple SRSs for beam management. Each resource may be

gNB UE
periodic (i.e., configured at the slot level), semi-persistent (also SS Burst
at the slot level, but it can be activated or deactivated with Beam sweep and
messages from the gNB) and a-periodic (the SRS transmission measurement

is triggered by the gNB) [55]. The SRSs can span 1 to SS Blocks to get
4 OFDM symbols, and a portion of the entire bandwidth RACH resources UE decides which is
the best beam
Beam determination
available at the UE [54].

IV. B EAM M ANAGEMENT F RAMEWORKS FOR

UE receives RACH
5G C ELLULAR S YSTEMS resource allocation
Beam reporting
In this section, we present three measurement frameworks RACH preamble
for both initial access and tracking purposes, whose perfor-
mance will be investigated and compared in Sec. VI.
As we introduced in the above sections of this tutorial,
the NR specifications include a set of basic beam-related
procedures [5] for the control of multiple beams at frequencies Fig. 3: Signals and messages exchanged during the SA-DL beam management
above 6 GHz and the related terminologies, which are based procedure (with the beam reporting step of the IA). Notice that the duration
of the three phases is not in scale, since it depends on the actual configuration
on the reference signals described in Sec. III. The different of the network parameters.
operations are categorized under the term beam management,
which is composed of four different operations:
• Beam sweeping, i.e., covering a spatial area with a set of the surrounding UEs, while in the second case the measure-
beams transmitted and received according to pre-specified ments are based on SRSs forwarded by the mobile terminal
intervals and directions. instead. Notice that the increasing heterogeneity in cellular
• Beam measurement, i.e., the evaluation of the quality of networks is dramatically changing our traditional notion of
the received signal at the gNB or at the UE. Different a communication cell [3], making the role of the uplink
metrics could be used [66]. In this paper, we consider important [69] and calling for the design of innovative UL-
the Signal to Noise Ratio (SNR), which is the average driven solutions for both the data and the control planes.
of the received power on synchronization signals divided In the following, we will describe in detail the three con-
by the noise power. sidered measurement schemes6 . Table III provides a summary
• Beam determination, i.e., the selection of the suitable of the main features of each framework.
beam or beams either at the gNB or at the UE, according
to the measurements obtained with the beam measure- A. Standalone-Downlink (SA-DL) Scheme
ment procedure. The SA-DL configuration scheme is shown in Fig. 3. No
• Beam reporting, i.e., the procedure used by the UE to support from the LTE overlay is provided in this configura-
send beam quality and beam decision information to the tion. The beam management procedure is composed of the
Radio Access Network (RAN). following phases:
These procedures are periodically repeated to update the (i) Beam sweeping. The measurement process is carried out
optimal transmitter and receiver beam pair over time. with an exhaustive search, i.e., both users and base sta-
We consider a NSA or a standalone (SA) architecture. tions have a predefined codebook of directions (each iden-
Non-standalone is a deployment configuration in which a NR tified by a beamforming vector) that cover the whole an-
gNB uses an LTE cell as support for the control plane man- gular space and are used sequentially to transmit/receive
agement [67] and mobile terminals exploit multi-connectivity synchronization and reference signals [20].
to maintain multiple possible connections (e.g., 4G and 5G (ii) Beam measurements. The mmWave-based measurements
overlays) to different cells so that drops in one link can be for IA are based on the SS blocks. The tracking is done
overcome by switching data paths [36], [35], [14], [38], [37], using both the measurements collected with the SS bursts
[68]. Mobiles in a NSA deployment can benefit from both and the CSI-RSs. These last elements cover a set of
the high bit-rates that can be provided by the mmWave links directions which may or may not cover the entire set
and the more robust, but lower- rate, legacy channels, thereby of available directions according to the users’ needs, as
opening up new ways of solving capacity issues, as well as explained in Sec. III. No support from the LTE overlay
new ways of providing good mobile network performance and is provided in this configuration.
robustness. Conversely, with the standalone option, there is no (iii) Beam determination. The mobile terminal selects the
LTE control plane, therefore the integration between LTE and beam through which it experienced the maximum SNR,
NR is not supported. if above a predefined threshold. The corresponding sector
The measurement frameworks can be also based on a
6 Notice that we do not consider the SA-UL configuration for both IA and
downlink or an uplink beam management architecture. In the
tracking applications. In fact, we believe that uplink-based architectures will
first case, the gNBs transmit synchronization and reference likely necessitate the support of the LTE overlay for the management of the
signals (i.e., SS blocks and CSI-RSs) which are collected by control plane and the implementation of efficient measurement operations.
 7

SA-DL NSA-DL NSA-UL

LTE overlay available for robust control operations and
Multi-RAT connectivity Not available
quick data fallback [38], [36], [37].
Reference signal transmission Downlink Downlink Uplink
Network coordination Not available Possibility of using a centralized controller [14].

Beam management phase SA-DL NSA-DL NSA-UL

Beam sweep Exhaustive search based on SS blocks [20]. Based on SRS [35].
Beam measurement UE-side UE-side gNB-side
Each gNB sends infor-
mation on the received
Beam determination The UE selects the optimal communication direction. beams to a central con-
troller, which selects the
best beam pair [36].
The UE signals the best The gNB signals the best
Exhaustive search at the beam pair using LTE, a beam pair using LTE, a
Beam reporting
gNB side [65]. RACH opportunity in that RACH opportunity in that
direction is then scheduled. direction is then scheduled.

TABLE III: Comparison of the beam management frameworks.

will be chosen for the subsequent transmissions and gNB UE

SS Burst
receptions and benefit from the resulting antenna gain.
Beam sweep and
(iv) Beam reporting. For IA, as proposed by 3GPP, after beam measurement
determination the mobile terminal has to wait for the
gNB to schedule the RACH opportunity towards the best
UE decides which is
direction that the UE just determined, for performing the best beam
Beam determination
random access and implicitly informing the selected
Feedback on LTE
serving infrastructure of the optimal direction (or set of Beam reporting
directions) through which it has to steer its beam, in gNB schedules directional
order to be properly aligned. It has been agreed that RACH resource

for each SS block the gNB will specify one or more RACH preamble
RACH opportunities with a certain time and frequency
offset and direction, so that the UE knows when to
transmit the RACH preamble [65]. This may require
an additional complete directional scan of the gNB,
thus further increasing the time it takes to access the Fig. 4: Signals and messages exchanged during the NSA-DL beam manage-
network. For the tracking in connected mode, the UE ment procedure (with the beam reporting step of the IA). Notice that the
duration of the three phases is not in scale, since it depends on the actual
can provide feedback using the mmWave control channel configuration of the network parameters.
it has already established, unless there is a link failure
and no directions can be recovered using CSI-RS. In this
case the UE must repeat the IA procedure or try to recover direction with the full beamforming gain. Moreover, the LTE
the link using the SS bursts while the user experiences a link can be also used to immediately report a link failure, and
service unavailability. allow a quick data-plane fallback to the sub-6-GHz connection,
while the UE recovers the mmWave link.
B. Non-Standalone-Downlink (NSA-DL) Scheme
The sub-6-GHz overlay can be used with different levels of C. Non-Standalone-Uplink (NSA-UL) Scheme
integration. As shown in Fig. 4, the first three procedures are Unlike in traditional LTE schemes, this framework (first
as in the SA-DL scheme. However, non-standalone enables proposed in [36] and then used in [14]) is based on the
an improvement in the beam reporting phase. Thanks to the channel quality of the UL rather than that of the DL signals
control-plane integration with the overlay, the LTE connection and, with the joint support of a central coordinator (i.e.,
can be used to report the optimal set of directions to the an LTE evolved Node Base (eNB) operating at sub-6 GHz
gNBs, so that the UE does not need to wait for an additional frequencies), it enables efficient measurement operations. In
beam sweep from the gNB to perform the beam reporting or this framework, a user searches for synchronization signals
the IA procedures. Thanks to this signaling, a random access from conventional 4G cells. This detection is fast since it
opportunity can therefore be immediately scheduled for that can be performed omnidirectionally and there is no need
 8

gNB with gNB UE thereby removing a possible point of failure in the control
central SRS
coordinator Beam sweep and signaling path. Moreover, since path switches and cell
measurement
additions in the mmWave regime are common due to link
Best beam and SNR failures, the control link to the serving mmWave cell may
not be available either. Finally, the coordinator notifies the
Coordinator Beam determination
decides which is designated gNB, through a backhaul high-capacity link,
the best beam
with the reports about the optimal direction in which to steer the beam
from all gNBs
for serving each UE.
Feedback on LTE
Beam reporting
Feedback to gNB
gNB schedules directional V. P ERFORMANCE M ETRICS AND 3GPP F RAMEWORKS
RACH resource
PARAMETERS
RACH preamble
In this section we define the metrics that will be used to
compare and characterize the performance of the different
Fig. 5: Signals and messages exchanged during the NSA-UL beam manage-
ment procedure (with the beam reporting step of the IA). Notice that the
beam management frameworks. Moreover, we will list the rel-
duration of the three phases is not in scale, since it depends on the actual evant parameters that affect the performance of the frameworks
configuration of the network parameters. in 3GPP NR.

A. Performance Metrics
for directional scanning. Under the assumption that the 5G
mmWave eNBs are roughly time synchronized to the 4G cell, The performance of the different architectures and beam
and the round trip propagation times are not large, an uplink management procedures for IA and tracking will be assessed
transmission from the UE will be roughly time aligned at using three different metrics. The detection accuracy is mea-
any closeby mmWave cell7 [35]. The NSA-UL procedure8 sured in terms of probability of misdetection PMD , defined
is shown Fig. 5 with a detailed breakout of the messages as the probability that the UE is not detected by the base
exchanged by the different parties. In detail, it is composed station (i.e., the Signal to Noise Ratio (SNR) is below a
of: threshold Γ) in an uplink scenario, or, vice versa, the base
station is not detected by the UE in a downlink scenario.
(i-ii) Beam sweeping and beam measurements. Each UE direc-
The reactiveness differs according to the purpose of the
tionally broadcasts SRSs in the mmWave bands in time-
measurement framework. For non-connected users, i.e., for IA,
varying directions that continuously sweep the angular
it is represented by the average time to find the best beam pair.
space. Each potential serving gNB scans all its angular
For connected users, i.e., for tracking, it is the time required
directions as well, monitoring the strength of the received
to receive the first CSI-RS after an SS burst, and thus react
SRSs and building a report table based on the channel
to channel variations or mobility in order to eventually switch
quality of each receiving direction, to capture the dynam-
beams, or declare a Radio Link Failure (RLF). Moreover, we
ics of the channel.
also consider the time it takes to react to the RLF. Finally,
(iii) Beam determination. Once the report table of each
the overhead is the amount of time and frequency resources
mmWave gNB has been filled for each UE, each
allocated to the framework with respect to the total amount of
mmWave cell sends this information to the LTE eNB
available resources, taking into account both the IA (i.e., SS
which, due to the knowledge gathered on the signal
blocks or SRSs and the RACH) and the tracking (i.e., CSI-
quality in each angular direction for each gNB-UE pair,
RSs).
obtains complete directional knowledge over the cell
it controls. Hence, it is able to match the beams of
the transmitters and the receivers to provide maximum B. 3GPP Framework Parameters
performance. In this section, we list the parameters that affect the per-
(iv) Beam reporting. The coordinator reports to the UE, on formance of the measurement architectures, as summarized in
a legacy LTE connection, which gNB yields the best Table IV. Moreover, we provide insights on the impact of each
performance, together with the optimal direction in which parameter on the different metrics.
the UE should steer its beam, to reach the candidate Frame Structure – As depicted in Fig. 6, we consider the
serving cell in the optimal way. The choice of using frame structure of 3GPP NR, with different subcarrier spacings
the LTE control link during the tracking is motivated ∆f . Given that in [45] the only subcarrier spacings considered
by the fact that the UE may not be able to receive for IA at frequencies above 6 GHz are ∆f = 120 and 240 kHz,
from the optimal mmWave link if not properly aligned, i.e., 15 × 2n kHz, with n ∈ [3, 4], we will only consider these
cases. The slot duration in ms is given by [4]
7 For example, if the cell radius is 150 m (a typical mmWave cell), the
round trip delay is only 1 µs.
1
Tslot = , (1)
8 Unlike the conventional DL-based measurement configuration, the uplink 2n
scheme has not been considered by 3GPP. Nevertheless, we will freely adapt
the same NR frame structure proposed for the downlink case to the NSA-UL
while the duration of a symbol in µs is [4]
scheme, using for the uplink SRSs the resources that would be allocated to 71.35
SS blocks in a downlink framework. Tsymb = . (2)
2n
 9

Parameter ∆f D NSS TSS CSI NCSI,RX KBF M , Nθ and Nφ Nuser λb

∗
Accuracy x x x x x
∗ ∗
Reactiveness x x
Overhead x x
TABLE IV: Relation among performance metrics and parameters.
∗
This depends on the tracking strategy.

frequency the SS blocks will be transmitted in the first NSS oppor-

(B = 400 MHz)
tunities. The actual maximum duration of an SS burst is
17.84 µs 17.84 µs 35.68 µs 35.68 µs Dmax,SS = 2.5 ms for ∆f = 240 kHz and Dmax,SS = 5 ms
BSS = 57.6 MHz BSS = 28.8 MHz
for ∆f = 120 kHz. We will also investigate all the possible
values for the SS burst periodicity Tss , as defined in [46], i.e.,
D D TSS ∈ {5, 10, 20, 40, 80, 160} ms.
A A
T T CSI-RS Configuration – As for the tracking, there
A A
are different options for the configuration of the CSI-RS
structure. These options include (i) the number NCSI of
CSI-RS per SS burst period, (ii) the CSI-RS periodicity
(a) (b) (c) (d) TCSI,slot ∈ {5, 10, 20, 40, 80, 160, 320, 640} slots, and (iii) the
Fig. 6: SS block structure. For configurations (a) and (b), each blue rectangle offset OCSI with respect to the end of an SS burst. In the
is an SS block (with 4 OFDM symbols) of duration 17.84 µs (i.e., ∆f = 240 analysis in Sec. VI we will also refer to TCSI = TCSI,slot Tslot ,
kHz) and bandwidth BSS = 57.6 MHz. For configurations (c) and (d) (for which represents the absolute CSI-RS periodicity in ms. These
which ∆f = 120 kHz), instead, the blocks last 35.68 µs and have bandwidth
BSS = 28.8 MHz. Cases (a) and (c) implement a frequency repetition scheme settings will be specified by the system information carried
(with Nrep = 5 and 11, respectively) while, for cases (b) and (d), a data by the SS blocks of each burst. Other CSI-related parameters
solution (i.e., Nrep = 1) is preferred. are the number of symbols of each CSI-RS transmission,
i.e., Nsymb,CSI ∈ {1, 2, 4}, and the portion of bandwidth ρB
allocated to the CSI-RSs. Moreover, the user will listen to
Therefore, for n = 3 and 4 the slot duration is 125 µs NCSI,RX CSI-RSs through an equivalent number of directions,
or 62.5 µs, respectively. Moreover, according to the 3GPP when in connected state. We will consider NCSI,RX ∈ {1, 4}.
specifications [43], the maximum number of subcarriers allo- Array Geometry – As shown in Fig. 7 and Table V,
cated to the SS blocks is 240, thus the bandwidth reserved another fundamental parameter is the array geometry, i.e., the
for the SS blocks would be respectively 28.8 and 57.6 MHz. number of antenna elements M at the gNB and UE and the
As mentioned in Sec. III, we consider a maximum channel number of directions that need to be covered, both in azimuth
bandwidth B = 400 MHz per carrier [5]. Nθ and in elevation Nφ . At the gNB we consider a single
Frequency Diversity – It is possible to configure the sector in a three sector site, i.e., the azimuth θ varies from
system to exploit frequency diversity, D. Given that 240 −60 to 60 degrees, for a total of ∆θ = 120 degrees. The
subcarriers are allocated in frequency to an SS, the remaining
bandwidth in the symbols which contain an SS block is
B − 240∆f . Therefore, it is possible to adopt two different
20
strategies: (i) data (as represented in Figs. 6(b) and (d)),
Directivity (dBi)

i.e., the remaining bandwidth B − 240∆f is used for data

0
transmission towards users which are in the same direction
in which the SS block is transmitted, or (ii) repetition (as
−20
displayed in Figs. 6(a) and (c)), i.e., the information in the
first 240 subcarriers is repeated in the remaining subcarriers to M =4 M = 16 M = 64
−40
increase the robustness against noise and enhance the detection −60 −40 −20 0 20 40 60
capabilities. The number of repetitions is therefore Nrep = 1 Azimuth Angle (degrees)
if frequency diversity is not used (i.e., D = 0, and a single Fig. 7: Relationship between beamwidth and antenna array size.
chunk of the available bandwidth is used for the SS block), and
Nrep = 11 or Nrep = 5 when repetition is used (i.e., D = 1) M θ [deg] Nθ gNB Nθ UE
with ∆f = 120 kHz or ∆f = 240 kHz, respectively. There is 4 60 2 6
a guard interval in frequency among the different repetitions of 16 26 5 14
64 13 10 28
the SS blocks, to provide a good trade-off between frequency
diversity and coherent combining [21]. TABLE V: Relationship between M , θ and Nθ , for the azimuth case. Each
SS Block Configuration – We consider different configu- gNB sector sweeps through ∆θ,gNB = 120◦ , while the UE scans over
∆θ,UE = 360◦ . In our evaluation, we consider a single antenna array at the
rations of the SS blocks and bursts. The maximum number UE modeled as a uniform rectangular array with isotropic antenna elements,
NSS of SS blocks in a burst for our frame structure and following the approach of the literature [70]. Real handheld devices will be
carrier frequencies is L = 64. We assume that, if NSS < L, equipped with multiple patch antennas able to cover the whole angular space.
 10

elevation φ varies between −30 and 30 degrees, for a total Parameter Value Description
of ∆φ = 60 degrees, and also includes a fixed mechanical B 400 MHz Total bandwidth of each mmWave gNB
tilt of the array pointing towards the ground. There exists fc 28 GHz mmWave carrier frequency
a strong correlation among beamwidth, number of antenna PTX 30 dBm Transmission power
elements and BF gain. The more antenna elements in the Γ −5 dB SNR threshold
system, the narrower the beams, the higher the gain that TABLE VI: Main simulation parameters.
can be achieved by beamforming, and the more precise and
directional the transmission. Thus, given the array geometry, Symbol Meaning
we compute the beamwidth ∆beam at 3 dB of the main lobe ∆f Subcarrier spacing
of the beamforming vector, and then Nθ = ∆θ /∆beam and Tslot Duration of a slot
Nφ = ∆φ /∆beam . Tsymb Duration of a symbol
B Bandwidth
Beamforming Architecture – Different beamforming D Usage of frequency diversity
architectures, i.e., analog, hybrid or digital, can be used both Nrep Number of repetitions in frequency of an SS block
at the UE and at the gNB. Analog beamforming shapes the PMD Probability of misdetection
beam through a single Radio Frequency (RF) chain for all the Γ SNR threshold for the misdetection
λb gNB density
antenna elements, therefore the processing is performed in the
analog domain and it is possible to transmit/receive in only one NSS Number of SS blocks per burst
L Maximum number of SS blocks per burst
direction at any given time. This model saves power by using Dmax,SS Maximum duration of an SS burst
only a single pair of Analog to Digital Converters (ADCs), TSS SS burst periodicity
but has a little flexibility since the transceiver can only SD Number of SS blocks for a complete sweep
TIA Time required to perform IA
beamform in one direction. Hybrid beamforming uses KBF RF Tlast Time to transmit the SS blocks in the last (or only) burst
chains (with KBF ≤ M ), thus is equivalent to KBF parallel TBR Time to perform beam reporting during IA
analog beams and enables the transceiver to transmit/receive NCSI Number of CSI-RSs per SS burst periodicity
in KBF directions simultaneously. Nevertheless, when hybrid TCSI CSI-RS periodicity
beamforming is used for transmission, the power available TCSI,slot CSI-RS periodicity in slot
OCSI Time offset between the end of the SS burst and the first
at each transmitting beam is the total node power constraint CSI-RS
divided by KBF , thus potentially reducing the received power. Nsymb,CSI Number of OFDM symbols for a CSI-RS
Digital beamforming requires a separate RF chain and data ρ Portion of bandwidth B for CSI-RSs
NCSI,RX Number of directions that a UE monitors
converters for each antenna element and therefore allows ZCSI Number of CSI-RSs to be transmitted
the processing of the received signals in the digital domain, Ttot,CSI Time available for the CSI-RS transmission between two SS
potentially enabling the transceiver to direct beams at infinitely bursts
NCSI Number of CSI-RS that can be transmitted between two bursts
many directions. Indeed, the availability of a sample for each Ttr Average time needed to receive the first CSI-RS
antenna allows the transceiver to apply arbitrary weights to NCSI,⊥ Number of orthogonal CSI-RSs between two SS bursts
the received signals, and perform a more powerful and flexible Nmax,neigh Number of neighbors that can be supported with orthogonal
CSI-RSs
processing than that in the analog domain. As in the hybrid TRLF RLF recovery delay
case, the use of digital beamforming to transmit multiple
M Number of antenna elements at the transceiver
beams simultaneously leads to a reduced transmit power being θ Azimuth angle
available to each (i.e., the total power constraint applies to the φ Elevation angle
sum of all beams, not to each of them individually). Moreover, ∆θ Angular range for the azimuth
∆φ Angular range for the elevation
the digital transceiver can process at most M simultaneous and Nθ Number of directions to cover in azimuth
orthogonal beams without any inter-beam interference (i.e., Nφ Number of directions to cover in elevation
through a zero-forcing beamforming structure [71]). For this ∆beam Beamwidth at 3 dB
KBF Number of beams that the transceiver can handle simultane-
reason, we limit the number of parallel beams that can be ously
generated to M . Furthermore, as previously mentioned, we Nuser Number of users
implement a digital beamforming scheme only at the receiver RSS Time and frequency resources occupied by SS blocks
side to avoid higher energy consumption in tranmsission. For Ω5ms SS blocks overhead in 5 ms
ΩTSS SS blocks overhead in TSS
the sake of completeness, we also consider an omnidirectional ΩCSI CSI-RS overhead in TSS
strategy at the UE i.e., without any beamforming gain but Ωtot Total overhead in TSS
allowing the reception through the whole angular space at any U [a, b] Uniform random variable in the interval [a, b]
given time.
Network Deployment – Finally, the last parameters are the TABLE VII: Notation.
number of users Nuser ∈ {5, 10, 20} per sector of the gNBs
and the density of base stations λb , expressed in gNB/km2 .
Our results are derived through a Monte Carlo approach, where
multiple independent simulations are repeated, to get different
C. Channel Model statistical quantities of interest. The channel model is based
The simulations for the detection accuracy performance on recent real-world measurements at 28 GHz in New York
evaluation are based on realistic system design configurations. City, to provide a realistic assessment of mmWave micro
 11

and picocellular networks in a dense urban deployment. A MgNB = 4, MUE = 4 MgNB = 64, MUE = 4
complete description of the channel parameters can be found λb = 10 gNB/km2 λb = 30 gNB/km2
in [72], while the main simulation parameters for this paper 1
are reported in Table VI.
0.8
10 gNB/km2
0.6

CDF
VI. R ESULTS AND D ISCUSSION
0.4
In this section, we present some simulation results aiming at
(i) evaluating the performance of the presented initial access 0.2
gNB/km2
MgNB(i.e.,
schemes in terms of detection accuracy =probability
4, MUEof = 4 0 MgNB = 3016, MUE = 4
misdetection), as reported in Sec. VI-A; (ii) describing the −40 −20 0 20 40 60

analysis and the results related M = 64, M the = 4 M SNR [dB] = 64, M
to gNB
the performance of UE gNB UE = 16
Fig. 8: CDF of the SNR, for different antenna configurations. ∆ = 120
measurement frameworks for the reactiveness and the over- kHz, Nrep = 0. The red dashed line represents the SNR thresholdf Γ = −5
M gNB = 64, M UE = 1 (omni)
head, respectively in Sec. VI-B-VI-C and Sec. VI-D. Table VII dB that has been considered throughout this work.
reports the notation used in this 0
section.
10 MgNB = 4, MUE = 4 MgNB = 16, MUE = 4
MgNB = 64, MUE = 4 MgNB = 64, MUE = 16
Misdetection probability

A. Detection Accuracy Results MgNB = 64, MUE = 1 (omni)

100
Array size and gNB density – Fig. 8 shows the Cumulative

Misdetection probability
Distribution Function (CDF) 10of−1 the SNR between the mobile
10−1
terminal and the gNB it is associated to, for different antenna
configurations and considering two density values. Notice
that the curves are not smooth because of the progressive 10−2
transitions of the SNR among the different path loss regimes,
i.e., Line of Sight (LOS),10 −2Line of Sight (NLOS) and
Non
10−3
outage. We see that better detection accuracy performance 10 20 30 40 50 60
λb [gNB/km2 ]
can be achieved when densifying the network and when using
larger arrays. In the first case, the endpoints are progressively Fig. 9: PMD as a function of λb , for different antenna configurations.
closer, thus ensuring better signal quality and, in general,
stronger received power. In 10the−3second case, narrower beams ∆ = 120 kHz, D = 0
can be steered thus guaranteeing 10 20 by
higher gains produced 30 40
∆ = 240 kHz, D = 1
50∆∆ f

f
f

f
= 120 kHz, D = 1
60
= 240 kHz, D = 0
2
beamforming. We also notice that, for good SNR regimes, the λb [gNB/km
10 ] 0

MgNB = 4, MUE = 4 and MgNB = 64, MUE = 4 configura- MgNB = 4, MUE = 4

Misdetection probability

tions present good enough SNR values: in these regions, the

10−1
channel conditions are sufficiently good to ensure satisfactory
signal quality (and, consequently, acceptable misdetection)
even when considering small antenna factors. Finally, the red 10−2 MgNB = 64, MUE = 4
line represents the SNR threshold Γ = −5 dB that we will
consider in this work. 10−3
10 20 30 40 50 60
Similar considerations can be deduced from Fig. 9, which λb [BS/km2 ]
illustrates how the misdetection probability monotonically
Fig. 10: PMD as a function of λb , for different subcarrier spacings ∆f
decreases when the gNB density λb progressively increases and repetition strategies and for different antenna configurations. MgNB =
or when the transceiver is equipped with a larger number of 4, MUE = 4, Γ = −5 dB.
antenna elements, since more focused beams can be gener-
ated in this case. Moreover, we notice that the beamforming Subcarrier spacing and frequency diversity – Fig. 10
strategy in which the UE transmits or receives omnidirec- reports the misdetection probability related to λb , for different
tionally, although guaranteeing fast access operations, does subcarrier spacings ∆f and repetition strategies D. First, we
not ensure accurate IA performance and leads to degraded see that, if no repetitions are used (i.e., D = 0), lower
detection capabilities. More specifically, the gap with a fully detection accuracy performance is associated with the ∆f =
directional architecture (e.g., MgNB = 64, MUE = 16) is 240 kHz configuration, due to the resulting larger impact
quite remarkable for very dense scenarios, and increases as of the thermal noise and the consequent SNR degradation.
the gNB density increases. For example, the configuration Furthermore, the detection efficiency can be enhanced by
with 16 antennas (i.e., MUE = 16) and that with a single repeating the SS block information embedded in the first 240
omnidirectional antenna at the UE reach the same PMD , but subcarriers in the remaining subcarriers (i.e., D = 1), to
at different values of gNB density λb , respectively 30 and 35 increase the robustness of the communication and mitigate
gNB/km2 : the omnidirectional configuration requires a higher the effect of the noise in the detection process. In fact, if a
density (i.e., 5 gNB/km2 more) to compensate for the smaller frequency diversity approach is preferred, the UE (in the DL
beamforming gain. measurement technique) or the gNB (in the UL measurement
 12

technique) has Nrep > 1 attempts to properly collect the Then, Tlast is
synchronization signals exchanged during the beam sweeping ( NSS,left
phase, compared to the single opportunity the nodes would 2 Tslot − 2Tsymb if NSS,left mod 2 = 0
have had if they had not implemented any repetition strategy. Tlast = j NSS,left k
2 Tslot + 6Tsymb otherwise,
We also observe that the ∆f = 120 kHz with no frequency
(6)
diversity configuration and the ∆f = 240 kHz scheme with
The two different options account for an even or odd remaining
Nrep = 5 produce the same detection accuracy results, thus
number of SS blocks. In the first case, the SS blocks are sent
showing how the effect of increasing the subcarrier spacing
in NSS,left /2 slots, with total duration NSS,left /2Tslot , but the
and the number of repetitions of the SS block information in
last one is actually received in the 12th symbol of the last slot,
multiple frequency subbands is similar in terms of misdetec-
i.e., 2 symbols before the end of that slot, given the positions
tion capabilities. Finally, we observe that the impact of the
of the SS blocks in each slot described in [45]. If instead
frequency diversity D and the subcarrier spacing ∆f is less
NSS,left is odd, six symbols of slot bNSS,left /2c + 1 are also
significant when increasing the array factor, as can be seen
used.
from the reduced gap between the curves plotted in Fig. 10
A selection of results is presented in the next paragraphs.
for the MgNB = 4, MUE = 4 and MgNB = 64, MUE = 4
configurations. The reason is that, when considering larger Number of SS blocks per burst and beamforming tech-
arrays, even the configuration with ∆f = 240 kHz and no nology – In Fig. 11 we consider first the impact of the number
repetitions has an average SNR which is high enough to reach of SS blocks in a burst, with a fixed SS burst periodicity
small misdetection probability values. TSS = 20 ms and for different beamforming strategies and
antenna configurations. In particular in Fig. 11a, in which
both the UE and the gNB use analog beamforming, the initial
B. Reactiveness Results for IA access delay heavily depends on the number of antennas at the
transceivers since all the available directions must be scanned
Analysis – For initial access, reactiveness is defined
one by one. It may take from 0.6 s (with NSS = 64) to 5.2
as the delay required to perform a full iterative search in
s (with NSS = 8) to transmit and receive all the possible
all the possible combinations of the directions. The gNB
beams, which makes the scheme infeasible for practical usage.
and the UE need to scan respectively Nθ,gNB Nφ,gNB and
A reduction in the sweeping time can be achieved either by
Nθ,UE Nφ,UE directions to cover the whole horizontal and
using an omnidirectional antenna at the UE or by decreasing
vertical space. Moreover, they can transmit or receive respec-
the number of antennas both at the UE and at the gNB. In this
tively KBF,gNB and KBF,UE beams simultaneously. Notice
case, the only configurations that manage to complete a scan
that, as mentioned in Sec. V-B, for digital and omnidirectional
in a single SS burst are those with 4 antennas at both sides
architectures KBF = min{Nθ Nφ , M }, for hybrid KBF =
and NSS ≥ 16, or that with MgNB = 64, an omnidirectional
min{Nθ Nφ , M }/ν, where ν is a factor that limits the number
UE and NSS = 64.
of directions in which it is possible to transmit or receive at
the same time, and for analog KBF = 1 [73]. Another option is the usage of hybrid or digital beamform-
Then the total number of SS blocks needed is9 ing at the UE in a downlink-based scheme, or at the eNB
   in an uplink-based one. Fig. 11b shows TIA when the UE
Nθ,gNB Nφ,gNB Nθ,UE Nφ,UE uses hybrid beamforming to receive from half of the available
SD = . (3)
KBF,gNB KBF,UE directions at any given time (i.e., L = 2), while in Fig. 11c the
UE receives from all available directions at any given time.
Given that there are NSS blocks in a burst, the total delay
This leads to an increased number of configurations which are
from the beginning of an SS burst transmission in a gNB to
able to complete a sweep in an SS block, even with a large
the completion of the sweep in all the possible directions is
number of antennas at the gNB and the UE.
  
SD Finally, Fig. 11d shows the performance of an uplink-based
TIA = TSS − 1 + Tlast , (4)
NSS scheme, in which the SRSs are sent in the same time and
frequency resource in which the SS blocks would be sent, and
where Tlast is the time required to transmit the remaining SS the gNB uses digital beamforming. It can be seen that there is
blocks in the last burst (notice that there may be just one burst, a gain in performance for most of the configurations, because
thus the first term in Eq. (4) would be 0). This term depends the gNB has to sweep more directions than the UE (since it
on the subcarrier spacing and on the number of remaining SS uses narrower beams), thus using digital beamforming at the
blocks which is given by gNB-side makes it possible to reduce TIA even more than

SD
  when it is used at the UE-side.
NSS,left = SD − NSS −1 . (5) SS burst periodicity – For the setup with hybrid beam-
NSS
forming at the UE, that generally requires more than one SS
9 We recall that hybrid or digital architectures consume more power than burst periodicity, we show in Fig. 12 the dependency of TIA
analog ones, if the same number of bits in the ADCs is used, and thus and TSS . It can be seen that the highest periodicities are not
are more likely to be implemented only at the receiver side. Nevertheless, suited for a mmWave deployment, and that in general it is
some ADC configurations enable energy efficient digital beamforming (e.g.,
3 bits ADC [74]), with a power consumption comparable to that of an analog better to increase the number of SS blocks per burst in order
implementation. to try to complete the sweep in a single burst.
 13

MgNB = 4, MUE = 4 MgNB = 16, MUE = 4 MgNB = 4, MUE = 4

MgNB = 64, MUE = 4 MgNB = 16, MUE = 16 MgNB = 16, MUE = 4
MgNB = 64, MUE = 16 MgNB = 64, MUE = 1 (omni) MgNB = 64, MUE = 4
600 MgNB = 16, MUE = 16
MgNB = 64, MUE = 16
MgNB = 64, MUE = 1 (omni)

TIA [ms]
4,000 400
TIA [ms]

200
2,000

0
50 100 150
0
8 16 32 64 TSS [ms]
NSS
(a) gNB Analog, UE Analog Fig. 12: TIA as a function of TSS for the downlink configuration with analog
gNB and hybrid UE. NSS = 64

50 directions to sweep.
600 ∆f = 120 kHz
3
∆f = 240 kHz
TIA [ms]

TIA [ms]
400 2

200 1 12 directions to sweep.

0 0
8 16 32 64 MgN B = 4, MU E = 4 MgN B = 64, MU E = 1
NSS
(b) gNB Analog, UE Hybrid (DL-based configuration) Antenna configuration

Fig. 13: TIA for different antenna configurations and subcarrier spacing ∆f ,
with gNB Analog, UE Analog.
300
TIA [ms]

200 1
3GPP advocates the implicit reporting of the chosen direction,
e.g., the strongest SS block index, through contention-based
0.5
100 random access messages, agreeing that the network should
0
32 allocate multiple RACH transmissions and preambles to the
0
8 16 32 64
UE for conveying the optimal SS block index to the gNB
NSS [75]. When considering an SA configuration, beam reporting
(c) gNB Analog, UE Digital (DL-based configuration) might require an additional sweep at the gNB side while,
if an NSA architecture is preferred, the beam decision is
100 forwarded through the LTE interface (and requires just a
single RACH opportunity) which makes the beam reporting
reactiveness equal to the latency of a legacy LTE connection.
TIA [ms]

50
1 Assuming a 0% BLER data channel, the uplink latency in
0.5 legacy LTE, including scheduling delay, ranges from 10.5 ms
0
to 0.8 ms, according to the latency reduction techniques being
32 implemented [76].
0
8 16 32 64 In Table VIII, we analyze the impact of the number of SS
NSS
blocks (and, consequently, of RACH opportunities) in a burst,
(d) gNB Digital, UE Analog (UL-based configuration)
with a fixed burst periodicity TSS = 20 ms and for a subcarrier
Fig. 11: TIA as a function of NSS with TSS = 20 ms. spacing of ∆f = 120 KHz. The results are independent of the
antenna configuration at the UE side, since the mobile terminal
Subcarrier spacing – Another parameter that has an impact steers its beam through the previously determined optimal
on TIA is the subcarrier spacing ∆f . As shown in Fig. 13, direction and does not require a beam sweeping operation to be
when the larger spacing is used the OFDM symbols have a performed. It appears clear that the SA scheme presents very
shorter duration and the transmission of the SS blocks in the good reactiveness for most of the investigated configurations
directions of interest can be completed earlier. and, most importantly, outperforms the NSA solution even
Impact of Beam Reporting – For initial access, in addition when the LTE latency is reduced to 0.8 ms. The reason is
to the time required for directional sweeping, there is also a that, if the network is able to allocate the needed RACH
delay related to the allocation of the resources in which it resources within a single SS burst, then it is possible to
is possible to perform initial access, which differs according limit the impact of beam reporting operations on the overall
to the architecture being used. As introduced in Sec. IV, initial access reactiveness, which is instead dominated by the
 14

TBR,SA [ms]
NSS = 8 NSS = 64 Opt. 1, TCSI = 10 ms Opt. 2, TCSI = 10 ms
MgN B Analog Digital Analog Digital
Opt. 1, TCSI = 0.625 ms Opt. 2, TCSI = 0.625 ms
4 0.0625 0.0625 0.0625 0.0625
16 0.5 0.0625 0.5 0.0625 500
64 40.56 0.0625 1.562 0.0625
400

Ttr [ms]
TBR,NSA ∈ {10, 4, 0.8} ms, according to [76].
300
TABLE VIII: Reactiveness performance for beam reporting operations con- 200
sidering an SA or an NSA architecture. Analog or digital beamforming is
100
implemented at the gNB side, while the UE configures its optimal beamformed
direction. TSS = 20 ms, ∆f = 120 KHz. 0
20
1
2
10 3
beam sweeping phase. In particular, when considering small 5 4
Nuser NCSI,RX
antenna factors and when digital beamforming is employed,
beam reporting can be successfully completed through a single (a) MgNB = 64, analog beamforming, TSS = 20 ms
RACH allocation, thus guaranteeing very small delays.
Opt. 1, TSS = 10 ms
C. Reactiveness Results for Beam Tracking Opt. 1, TSS = 40 ms
Analysis – For tracking, we define the reactiveness as the

Ttr [ms]
average time needed to receive the first CSI-RS after the end 20
of each SS burst.
We assume that the Nuser UEs are uniformly distributed in
the space covered by the k = Nθ,gNB Nφ,gNB beams available 0
at the gNB. Moreover, each UE has to monitor NCSI,RX 20
1
directions. Given that a UE may or may not be in LOS, it is not 2
10 3
obvious that these directions will be associated to the closest Nuser
5 4
NCSI,RX
beams with respect to the one selected during the initial access.
Therefore, we also assume that this scenario is equivalent to a (b) MgNB = 64, analog beamforming, TCSI = 0.625 ms
scenario with n = Nuser NCSI,RX uniformly distributed UEs, Fig. 14: Performance of tracking using CSI-RSs for Option 1 and Option 2,
each of them monitoring a single direction. We will refer to as described in Fig. 2, as a function of different parameters (e.g., TCSI , TSS ),
for ∆f = 120 kHz.
n as the number of measures.
Consequently, on average there are n/k measurements for
the area belonging to each beam, if the beams divide the space The last sum accounts for the case ZCSI < NCSI and for the
into equally sized regions. Therefore, if n ≥ k, a CSI-RS is CSI-RS in the last SS burst periodicity when ZCSI > NCSI .
needed in each beam, otherwise it is sufficient to send at least The sum over p, instead, accounts for ZCSI ≥ NCSI .
n CSI-RSs, and thus the total number of CSI-RS that need to Option 2: thanks to the additional parameter OCSI it is
be transmitted is on average ZCSI = min{n, k}. possible to transmit NCSI = dTtot,CSI /TCSI e, as shown in
Depending on the combination of TSS , TCSI = TCSI,slot Tslot Fig. 2b. The offset is computed as
and ZCSI , it may not be possible to allocate all the CSI-RS
transmissions between two consecutive SS bursts. Notice that Ttot,CSI − (NCSI − 1)TCSI
OCSI = . (9)
after the end of an SS burst, there are Ttot,CSI = TSS − 2
Dmax,SS ms available for the CSI-RS transmission. Then, the The metric Ttr,opt2 is computed as for option 1, but taking
number NCSI of CSI-RS that can be allocated between two SS into account also OCSI , in Eq. (8).
bursts may depend on which of the options shown in Fig. 2 Notice that if ZCSI > NCSI , a signal in a certain direction
is chosen. could be either received as SS block in the next burst, or as
Option 1: the first CSI-RS is transmitted TCSI ms after CSI-RS, depending on how the transmission of SS blocks and
the transmission of the SS burst. In this case, NCSI = CSI-RSs is scheduled.
bTtot,CSI /TCSI c, and single periodicity is not enough if Scheduling options, number of users and CSI-RS pe-
ZCSI > NCSI . For option 1, the metric Ttr,opt1 is given by (7). riodicity – Fig. 14a shows the value of Ttr for different

j k
ZCSI
NCSI −1 P
NCSI
 P
ZCSI mod NCSI
j k 
P ZCSI
p=0 i=1 (pTSS + iTCSI ) + i=1 NCSI TSS + iTCSI
Ttr,opt1 = (7)
ZCSI
j k
ZCSI
NCSI −1 P
NCSI −1
 P
ZCSI mod NCSI −1
j k 
P ZCSI
p=0 i=0 (pTSS + iTCSI + OCSI ) + i=0 NCSI TSS + iTCSI + OCSI
Ttr,opt2 = (8)
ZCSI
 15

Opt. 1, TCSI = 1.25 ms Opt. 2, TCSI = 1.25 ms

Opt. 1, TCSI = 10 ms
Opt. 1, TCSI = 2.5 ms Opt. 2, TCSI = 2.5 ms
Opt. 1, TCSI = 5 ms Opt. 2, TCSI = 5 ms 800 Opt. 1, TCSI = 0.625 ms
Opt. 1, TCSI = 10 ms Opt. 2, TCSI = 10 ms
Same bandwidth as SS

150 600 Same bandwidth as SS burst for ∆f = 120 kHz.

Nmax,neigh
20 burst for ∆f = 240 kHz.

15
400
100
10
NCSI

5
200
0
50 5 10 20

0
4
0 2 0.5 0.2 0.1 0
0 20 40 60 80 100 120 140 160 1 1
TSS [ms] Nsymb,CSI ρ

Fig. 15: NCSI as a function of the TSS and TCSI periodicities. Fig. 16: Nmax,neigh as a function of Nsymb,CSI and ρ for different TCSI
periodicities, with TSS = 20 ms and ∆f = 120 kHz.

parameters, such as the different scheduling option 1 or 2,

in which CSI-RSs can be scheduled, and the second and third
the number of users per gNB Nuser and of directions of
express the number of CSI-RSs per slot (there are at most 4
interest NCSI,RX , for SS burst periodicity TSS = 20 ms and
OFDM symbols per slot for CSI-RSs). Then, the maximum
64 antennas at the gNB. The fundamental parameter is the
number of neighbors that a gNB can support is
periodicity of the CSI-RS transmission: only a small CSI-RS  
periodicity makes it possible to sweep all the directions to NCSI,⊥
Nmax,neigh = − 1, (11)
be covered during a relatively short interval, and to avoid the NCSI
dependency on TSS . Moreover, if the periodicity is small (i.e., with NCSI computed as in the previous paragraphs.
TCSI = 0.625 ms, or 5 slots with ∆f = 120 kHz), then there Fig. 16 reports the value of Nmax,neigh for a different
is no difference between the two scheduling options, while number of OFDM symbols for the CSI-RSs and bandwidth
this becomes notable for TCSI = 10 ms, as expected. scaling factor ρ, which ranges from 0.1 to 1, and represents
SS burst periodicity – Fig. 14b compares two different TSS also the bandwidth values corresponding to 240 subcarriers
periodicities, i.e., 10 and 40 ms, using the smallest TCSI,slot with ∆f ∈ {120, 240} kHz, i.e., the bandwidth occupied by
available (i.e., 5 slots, or 0.625 ms at ∆f = 120 kHz). It can an SS burst. Notice that for the frequencies in the mmWave
be seen that using a higher TSS would allow a decreased Ttr , spectrum it is advisable not to use the entire bandwidth for
since more CSI-RSs can be scheduled between two SS bursts CSI-RSs [49], and the number of neighbors of a mmWave
and consequently a larger number of directions can be swept. gNB will be limited, given the short propagation distance
For the sake of completeness, Fig. 15 shows the number of typical of these frequencies. If TCSI = 10 ms, then even
CSI-RSs that can be scheduled in between two SS bursts as when using 4 OFDM symbols and the whole bandwidth
a function of TSS and of the different scheduling options and it is possible to support only 14 neighbors. Instead, when
periodicities. Since in a mmWave scenario there may be a need TCSI = 0.625 ms it is not feasible to use the whole bandwidth
to scan a large number of CSI-RSs, it is advisable to either and 4 symbols, but more conservative configurations should be
use an adaptive scheme for the scheduling of CSI-RSs, which adopted. For example, with ρ = 0.072 (i.e., 240 subcarriers
adapts the periodicity according to the number of users in the with ∆f = 120 kHz) it is possible to support 15 or 31
different directions, or adopt a conservative approach and use neighbors, respectively with 2 or 1 OFDM symbols.
a short TCSI interval. Standalone vs non-standalone – Notice that when the
Limits on the CSI-RS periodicity – Since the CSI-RSs that standalone scheme is used and the UE experiences a link
a user receives from multiple base stations should not overlap failure on all the NCSI,RX directions it is monitoring, then
in time and frequency (otherwise the RSRP value would be the UE has no choice but using the SS blocks in the SS
over-estimated), there is a maximum number of neighboring burst to perform either a link recovery or a new initial access,
cells that a gNB can support. According to [64], there are 4 and meanwhile it is not able to transmit or receive data or
symbols per slot in which a CSI-RS can be sent (additional control information [14]. When a non-standalone architecture
symbols are under discussion), and a CSI-RS can last 1, 2 is used, instead, the UE could signal this event to the RAN
or 4 symbols, each with bandwidth ρB. Assuming a common on the lower-frequency control link, and the data plane can be
configuration for the gNBs deployed in a certain area, the total switched to the sub-6-GHz RAT, and faster recovery options
number of orthogonal CSI-RS transmission opportunities is could be designed, for example, by instructing the UE to
TSS − Dmax, SS 4
 
1 monitor additional CSI-RSs.
NCSI,⊥ = , (10) Downlink vs uplink and beamforming architecture –
Tslot Nsymb,CSI ρ
Finally, we observe that, when a digital architecture is chosen,
where the first ratio is the number of slots in the time interval there exist some specific configurations in which a UL-based
 16

Antenna TRLF,SA [ms]

NSS = 8, TSS = 20 NSS = 64, TSS = 40 NSS = 64, TSS = 80
MgN B MU E
gN B ABF, UE ABF gN B DBF, UE ABF gN B DBF, UE ABF
4 4 30.2322 20.3572 40.3572
64 1 130.1072 20.0535 40.0535
64 16 5250 22.6072 42.6072
TRLF,NSA ∈ {10, 4, 0.8} ms, according to the considerations in [76].
TABLE IX: RLF recovery delay considering the SA or the NSA measurement frameworks, for different values of NSS , TSS and for different beamforming
configurations. ∆f = 120 kHz. ABF stands for Analog Beamforming, and DBF for Digital.

measurement framework can ensure more efficient tracking • Much more responsive RLF recovery operations may be
operations than its DL counterpart. In fact, due to the gNB’s prompted if the failure notification is forwarded through
less demanding space constraints with respect to a mobile the LTE overlay (i.e., by implementing an NSA-based
terminal, a larger number of antenna elements can usually be measurement framework), which may also serve the UE’s
packed at the base station side, resulting in a larger number traffic requests until the mmWave directional commu-
of directions that can potentially be scanned simultaneously nication is successfully restored. If an NSA-DL frame-
through a digital beamforming scheme. Moreover, hybrid or work is designed, the RLF recovery delay TRLF,NSA
fully digital receivers are more costly in terms of power is equal to the latency of a traditional LTE connection
consumption, and hence are more likely to be implemented (which depends on the implemented latency reduction
in a gNB rather than in a UE. technique, as assessed in [76]). Alternatively, the gNB
RLF recovery – Another important factor that affects the can autonomously declare an RLF event (without the
reactiveness of beam management schemes is the time it takes user’s notification) and react accordingly by monitoring
to recover from an RLF. As assumed by 3GPP [77], RLF the SRS messages. Without loss of generality, assuming
occurs when the quality of an associated control channel falls that SRSs are uniformly allocated within two SS bursts
below a certain threshold. As soon as the failure is detected, with periodicity TSRS , an RLF is detected as soon as the
mechanisms to recover acceptable communication capabilities gNB is not able to correctly receive NSRS consecutive
(e.g., by determining an alternative suitable direction of trans- SRSs from its reference user. In this case, the reactiveness
mission or possibly handing over to a stronger and more robust of the RLF recovery operation depends on the periodicity
gNB) need to be quickly triggered upon notifying the network. of the sounding signals and is equal to
Natural candidates for monitoring the link quality and detect TSRS
the link failure are the SS blocks in a burst [78]. Assume that TRLF,NSA = + (NSRS − 1) TSRS . (12)
2
an object blocks the propagation path of the transceiver at time
T ∼ U[t, t + TSS ], i.e., on average at time T̄ = TSS /2 within Analogously, if an NSA-UL framework is designed, the
two consecutive SS bursts. recovery may be immediately triggered by the gNB by
switching the traffic to the LTE eNB in TRLF,NSA sec-
• When implementing an SA architecture, as soon as an onds, as given by Eq. (12). From the results in Table IX,
impairment is detected, the UE may no longer be able it appears that fast and efficient RLF recovery operations
to communicate with its serving gNB since the optimal can be guaranteed if an NSA solution is preferred over an
directional path connecting the endpoints is affected by SA one for all the investigated network configurations.
the failure. The recovery phase is most likely triggered at
the beginning of the subsequent SS burst (i.e., on average
after TSS − T̄ = TSS /2 seconds) and at least after the D. Overhead Results
completion of an IA operation of duration TIA seconds.10
In this section, we characterize the overhead for IA and
Table IX reports the RLF recovery delay TRLF,SA for
tracking in terms of the ratio between the time and frequency
some network configurations when an SA architecture
resources that are allocated to SS bursts and the maximum
is implemented. We observe that the latency is quite
duration of the SS burst (i.e., 5 ms), or the entire TSS interval.
high for all the investigated settings and is dominated
Analysis – The total number of time and frequency re-
by the IA delay, as illustrated in Fig. 11. Moreover,
sources RSS scheduled for the transmission of NSS SS blocks,
in some circumstances (e.g., NSS = 8, TSS = 20 ms,
each spanning 4 OFDM symbols and 240 (or multiple of 240)
MgN B = 64, NgN B = 16 and when analog beamform-
subcarriers, is given by
ing is implemented), the RLF recovery delay assumes
unacceptably high values. RSS = NSS 4Tsymb 240Nrep ∆f , (13)

10 In some circumstances, the UE can autonomously react to an RLF event where Tsymb is expressed in ms and ∆f in kHz. The overhead
by selecting an alternative direction of communication, as a sort of backup for the 5 ms time interval with the SS burst transmission and
solution before the transceiver fully recovers the optimal beam configuration total bandwidth B (in Hz) is then given by
[35]. Although having a second available link, when the primary path is
obstructed, adds diversity and robustness to the communication, it may not NSS 4Tsymb 240Nrep ∆f
always guarantee sufficiently good communication performance. Ω5ms = , (14)
5B
 17

∆f = 120 kHz, D = 0 0.4 ∆f = 120 kHz, Nrep = 0

0.4
∆f = 240 kHz, D = 0 ∆f = 240 kHz, Nrep = 0
∆f = 120 kHz, D = 1 ∆f = 120 kHz, Nrep = 11
∆f = 240 kHz, D = 1 ∆f = 240 kHz, Nrep = 5

ΩTSS
Ω5ms

0.2 0.2

0 0
8 16 32 64 0 20 40 60 80 100 120 140 160
NSS TSS [ms]

(a) Ω5ms as a function of NSS , for different subcarrier spacings ∆f and repetition (b) ΩTSS as a function of TSS , for different subcarrier spacings ∆f and repetition
strategies. strategies. NSS is set to the maximum value, i.e., 64.

Fig. 17: Overhead for initial access, introduced by the transmission of the SS blocks. Notice that the number of repetitions for the different subcarrier spacings
∆f is chosen to send as many repetitions of the SS blocks as possible.

and the overhead considering the total burst periodicity TSS is TCSI = 0.625 ms
0.06
NSS 4Tsymb 240Nrep ∆f TCSI = 5 ms
ΩTSS = . (15)
TSS B
0.04

ΩCSI
Moreover, additional overhead is introduced by the trans-
mission of CSI-RSs after the SS burst. The value of the
0.02
overhead ΩCSI depends on the number of symbols Nsymb,CSI
and the bandwidth ρB for each CSI-RS, as well as on the
0
number of CSI-RSs NCSI computed as in Sec. VI-C for the 4
0
two CSI-RS scheduling options: 0.2 0.1
2 0.5
1 1
NCSI Nsymb,CSI Tsymb ρB NCSI Nsymb,CSI Tsymb ρ Nsymb,CSI ρ
ΩCSI = = .
(TSS − Dmax,SS )B (TSS − Dmax,SS ) (a) Overhead ΩCSI as a function of Nsymb,CSI and ρ, for different TCSI periodicities,
(16) with TSS = 20 ms.
Finally, the total overhead Ω takes into account both the SS
bursts and the CSI-RSs in TSS :
∆f = 240 kHz, D = 0 ∆f = 120 kHz, D = 0
NCSI Nsymb,CSI Tsymb ρB + RSS
Ωtot = . (17) ∆f = 240 kHz, D = 1 ∆f = 120 kHz, D = 1
TSS B
Subcarrier spacing and frequency diversity – Fig. 17 re-
ports the overhead related to the maximum duration of the SS 0.1
burst (i.e., 5 ms) for different subcarrier spacings and repetition
Ωtot

strategies. It can be seen that if no repetitions are used (i.e.,

0.05
D = 0) then the overheads for the configurations with ∆f =
120 kHz and ∆f = 240 kHz are equivalent. In fact, when
configuring large subcarrier spacings (i.e., ∆f = 240 kHz), 0
4
0
the OFDM symbols used for the SS blocks have half the 0.2 0.1
2 0.5
duration, but they occupy twice the bandwidth of the systems 1 1
Nsymb,CSI ρ
with narrower subcarrier spacings (i.e., ∆f = 120 kHz), given
that the same number of subcarriers are used. Instead, when (b) Overhead Ωtot as a function of Nsymb,CSI and ρ, for different subcarrier spacings
∆f and repetition strategies. NSS is set to the maximum value, i.e., 64, and TCSI,slot =
a repetition strategy is used (i.e., D = 1), the overhead is 5 slot.
different. As mentioned in Sec. V-B, we consider 5 repetitions
Fig. 18: Overhead for the CSI-RS transmission and total overhead, with
for ∆f = 240 kHz and 11 for ∆f = 120 kHz. Therefore, the TSS = 20 ms. Notice that the number of repetitions for the different subcarrier
actual amount of bandwidth that is used is comparable, but spacings ∆f is chosen to send as many repetitions of the SS blocks as
since the OFDM symbols with ∆f = 120 kHz last twice as possible.
long as those with the larger subcarrier spacing, the overhead
in terms of resources used for the SS burst is higher with
∆f = 120 kHz. SS blocks. For TSS = 20 ms or higher, instead, the overhead
SS burst periodicity – Fig. 17b shows the dependency of is always below 10%.
the overhead for initial access on TSS , which follows an inverse CSI-RS periodicity – The overhead due to the transmission
proportionality law. In particular, for very small TSS (i.e., 5 of CSI-RSs is shown in Fig. 18a for different TCSI periodicities
ms) the impact of the SS bursts with repetitions in frequency and time and frequency resource allocation to the CSI-RSs. It
is massive, with up to 43% of the resources allocated to the is always below 0.008 with TCSI = 5 ms, and below 0.06
 18

ΩBR,SA ·10−3 of an increase in the overhead in the order of 2 times, while the
∆f,RACH = 60 kHz ∆f,RACH = 120 kHz accuracy gain for the configuration with λ = 30 gNB/km2 and
MgN B Analog Digital Analog Digital
the 4 × 4 antenna arrays is in the order of 23%, according to
4 0.0894 0.0894 0.0894 0.0894 Fig. 10. A smaller subcarrier spacing has also a negative effect
16 0.7149 0.0894 0.7149 0.0894
on the reactiveness, as shown in Fig. 13, since the OFDM
64 2.2341 0.0894 2.2341 0.0894
symbols last longer and the SS blocks sweep takes more time.
TABLE X: Overhead for beam reporting operations considering an SA b) Frequency diversity: The repetition in frequency of
architecture. Analog or digital beamforming is implemented at the gNB side, multiple SS signals for the same OFDM symbol results in an
for different antenna array structures.
increased accuracy (e.g., up to 45%, when λ = 60 gNB/km2
and considering the 4 × 4 array configuration). The overhead
for TCSI = 0.625 ms. However, for practical values of the is, however, from 5 to 11 times higher in our setup (according
configuration of the CSI-RSs, in which the bandwidth for the to the ∆f used), thus there is a trade-off between the amount
reference signal is smaller than half of the entire bandwidth, of resources to allocate to the users that are already connected
then also for TCSI = 0.625 ms the overhead reaches very small (which is higher if frequency diversity is not used) and
values, i.e., below 0.028. the opportunity to discover new users (which increases with
Impact of IA and tracking – The trend of Ωtot is shown frequency diversity for the SS blocks). However, notice that
in Fig. 18b, where it can be immediately seen that the largest the accuracy gain reduces when increasing the array dimension
impact is given by the term RSS at the numerator and not (e.g., when λ = 60 gNB/km2 and considering the 64 × 4 array
by the CSI-RS-related overhead. The parameters on the x and configuration, a gain of just 15% is achieved, as seen from
y axes have indeed a limited effect on the gradient of the Fig. 10). In those circumstances, it may not be desirable to
surfaces, which are almost horizontal. The main difference is adopt a frequency diversity scheme which would inevitably
introduced by the different subcarrier spacings and repetition increase the overhead while only providing marginal accuracy
strategies. Notice that, contrary to what is shown in Fig. 17a, gain.
there is a difference between the two different subcarrier c) Number of SS blocks in a burst NSS : This parameter
spacings for the total overhead Ωtot and for the CSI-RS- has a fundamental impact on the reactiveness, since a higher
related overhead ΩCSI , because we consider a different Tsymb number of SS blocks per burst increases the probability of
in Eq. (16), but the same ρ factor, thus a different number of completing the sweep in a single burst and thus prevents TIA
subcarriers for the different values of ∆f . from being dependent on TSS . The number of SS blocks per
Impact of beam reporting – For the SA case, as reported burst, however, increases also the overhead linearly. NSS has
in Table X, the completion of the beam reporting procedure a strict relationship with the number of directions to be swept,
for initial access may require an additional overhead, due to i.e., with both the beamforming architecture and the number
the need for the system to allocate possibly multiple RACH of antennas: if, for example, hybrid or digital beamforming is
resources11 for the reporting operations. Conversely, for the used at the receiver, a larger number of antennas (i.e., narrower
NSA case, the beam decision is forwarded through the LTE beams) can be supported even with a smaller NSS , as shown
overlay and requires a single RACH opportunity, with a total in Fig. 11
overhead of 0.0894 · 10−3 . Nevertheless, from Table X, we d) SS burst periodicity TSS : The periodicity of a burst
notice that the SA additional reporting overhead is quite has an impact on the reactiveness for initial access, since a
limited due to the relatively small number of directions that smaller TSS enables a larger number of opportunities in which
need to be investigated at this stage, especially when designing a UE can receive synchronization signals. However, if the
digital beamforming solutions. beam sweeping procedure is completed in a single burst, TSS
does not impact TIA as previously defined. The overhead is
inversely proportional to TSS , which has a major impact also
VII. F INAL C ONSIDERATIONS
on the reactiveness related to the tracking and the transmission
In the following paragraphs we will provide some insights of CSI-RSs, as shown in Fig. 14. Overall, if the sweep can be
on the trade-offs related to the different parameters we investi- completed in a single burst, a higher TSS would decrease the
gated and the three metrics considered, and some suggestions overhead and increase the reactiveness for the CSI-RSs.
and guidelines to optimally design a measurement framework e) CSI-RS periodicity TCSI : A short TCSI allows an
for NR at mmWave frequencies. improved reactiveness for the beam tracking of connected
a) Subcarrier spacing ∆f : When using a smaller sub- users. In particular, when the number of users per gNB is
carrier spacing (i.e., ∆f = 120 kHz) it is possible to achieve high then a short CSI-RSs periodicity enables a much shorter
a higher accuracy (i.e., smaller misdetection probability), Ttr . On the other hand, the overhead related to the CSI-RSs
either because the impact of the noise is less relevant, when is small if compared with that of the SS bursts.
frequency diversity is not used, or because it is possible f) Number of CSI-RSs to be monitored at the UE side
to allocate a larger number of repetitions, when frequency NCSI,RX : The impact of this parameter on the reactiveness
diversity is used. This last option comes however at the price is related to both the number of users per gNB and the total
11 According to the 3GPP agreements [79], a bandwidth of 10 MHz (for
number of directions to be swept with the reference signals.
∆f,RACH = 60 kHz) or a bandwidth of 20 MHz (for ∆f,RACH = 120 If there is a limited number of directions and a large number
kHz) is reserved for the RACH resources. of users, uniformly distributed in the available directions, then
 19

Reactiveness i) Antenna Arrays MgN B , MU E : The antenna array is

(1/TIA ) one of the parameters that has the largest impact on the
accuracy. A larger number of antennas enable narrower beams
and higher accuracy, since the received power at the UE (in
downlink) or at the gNB (in uplink) increases. The width of the
beam has, however, an inverse relationship with the number
of directions to scan, thus configurations that provide a higher
Accuracy accuracy perform worse in terms of reactiveness and overhead.
(1/PMD ) Notice that the choice of the antenna array and of the beam
design is strictly tied to the beamforming architecture (if dig-
ital or hybrid beamforming is used then narrower beams can
be supported without penalizing reactiveness and overhead)
and the configuration of the SS bursts (a large number of
directions to be swept with a limited number of SS blocks per
bursts has a negative impact on the reactiveness). In Fig. 19
Overhead (Ωtot )
a direct comparison among three different schemes is shown.
It can be seen that the MgN B = 4 × MU E = 4 configuration
presents a smaller overhead and an improved reactiveness with
respect to the MgN B = 64 × MU E = 16 configuration.
MgN B = 4 × MU E = 4, analog beamforming, NSS = 8 Moreover, both configurations with 64 antennas at the gNB
MgN B = 64 × MU E = 1 (omni), analog beamforming, NSS = 64 have the same overhead, but there is a trade-off between the
MgN B = 64 × MU E = 16, digital beamforming at the UE, NSS = 64
reactiveness (the configuration with the omnidirectional UE
Fig. 19: Comparison of three different configurations for accuracy, reactive- has the best reactiveness) and accuracy (using 16 antennas at
ness and overhead. The common parameters are ∆f = 120 kHz, Nuser = 10, the UE provides the best accuracy, at the cost of a higher
Nsymb,CSI = 2, ρ = 0.072, λb = 30 gNB/km2 , NCSI,RX = 3, energy consumption because of digital beamforming).
TSS = 20 ms, TCSI = 0.625 ms.
j) Measurement Framework: As far as initial access
is concerned, the implementation of a standalone scheme
generally guarantees more reactive access capabilities. The
the monitoring of additional CSI-RSs does not impact Ttr or reason is that faster beam reporting operations are ensured if
the overhead at the network side. The UE may, however, be multiple SS blocks and RACH opportunities can be allocated
impacted by the energy consumption related to the monitoring within a single SS burst. On the other hand, a non-standalone
of too many directions, i.e., by a needlessly high NCSI,RX . framework may be preferable to: (i) reduce the impact of
the overhead in the beam reporting phase; (ii) in connected
g) gNB density λb : As the network density increases,
mode, implement efficient and reactive recovery operations as
the accuracy and the average received power increase, and
soon as a radio link failure event is detected; (iii) guarantee a
this allows a larger number of users to be served by a
more robust control signaling exchange (e.g., when forwarding
mmWave network. Besides the cost in terms of equipment
the beam reporting messages). Moreover, a non-standalone
and energy, a higher density has also a negative effect on
architecture is also better than an SA one when it is not
the interference [80]. Moreover, the number of neighbors
possible to allocate in the same SS burst the SS blocks
of each single gNB increases, and this limits the available
for the first sweep and the subsequent RACH opportunities,
configurations for the CSI-RSs.
because for example there are too many directions to monitor
h) Beamforming architecture KBF : A digital beam- at the gNB. Finally, NSA enables a centralized beam decision:
forming architecture at the receiver side would improve the re- unlike in traditional attachment policies based on pathloss
activeness of the measurement scheme and decrease the over- measurements, by leveraging on the presence of an eNB
head, without penalizing the accuracy. The same improvement operating at sub-6 GHz frequencies, an NSA-based beam
in terms of reactiveness and accuracy can be achieved with an association can be performed by taking into account the
omnidirectional receiver, but the accuracy would decrease with instantaneous load conditions of the surrounding cells, thereby
a loss of around 30% (when λ = 30 gNB/km2 ) with respect promoting fairness in the whole cellular network [35].
to the MgN B = 64 configuration, as displayed in Fig. 9. The Overall, it is possible to identify some guidelines for the
complexity of the transceiver implementation and the energy configuration of the measurement framework and the de-
consumption [81] are, however, two important parameters that ployment of a NR network at mmWave frequencies. First, a
must be taken into account. A hybrid configuration could setup of NSS , the RACH resources, the beamforming and the
represent a trade-off between an improved reactiveness and a antenna array architectures that allows the completion of the
simpler and less consuming transceiver design. Finally, notice beam sweeping and reporting procedures in a single burst is
that a digital architecture allows a higher gain with respect preferable, so that it is possible to increase TSS (e.g., to 20 or
to the reactiveness if used at the gNB in an uplink-based 40 ms), and consequently allocate a larger number of CSI-RSs
framework, since the directions to be swept at the gNB are for the connected users (to guarantee more reactive tracking
usually more than at the UE. operations) and reduce the overhead of the SS blocks.
 20

Second, the adoption of a frequency diversity scheme for the 17th Annual Mediterranean Ad Hoc Networking Workshop (Med-Hoc-
SS blocks depends on the load of the gNBs: if many users are Net), 2018.
[2] Cisco, “Cisco Visual Networking Index: Global Mobile Data Traffic
connected to a certain gNB, this could disable the frequency Forecast Update, 2016–2021,” White Paper, March 2017.
diversity to both reduce the overhead and avoid discovering [3] F. Boccardi, R. W. Heath, A. Lozano, T. L. Marzetta, and P. Popovski,
new users. Third, with low network density, larger antenna “Five disruptive technology directions for 5G,” IEEE Communications
Magazine, vol. 52, no. 2, pp. 74–80, February 2014.
arrays make it possible to detect farther users, and provide
[4] Ericsson, “5G New Radio: designing for the future,” Ericsson Technol-
a wider coverage but, as λb increases, it is possible to use ogy Review, 2017.
a configuration with wide beams for SS bursts (so that it is [5] 3GPP, “Study on New Radio (NR) Access Technology - Physical Layer
more likely to complete a sweep in a single burst) and narrow Aspects - Release 14,” TR 38.802, 2017.
[6] S. Rangan, T. S. Rappaport, and E. Erkip, “Millimeter-Wave Cellular
ones for CSI-RS, to refine the pointing directions for the data Wireless Networks: Potentials and Challenges,” Proceedings of the
transmission and achieve higher gains. IEEE, vol. 102, no. 3, pp. 366–385, March 2014.
Finally, when considering stable and dense scenarios which [7] T. S. Rappaport, S. Sun, R. Mayzus, H. Zhao, Y. Azar, K. Wang, G. N.
Wong, J. K. Schulz, M. Samimi, and F. Gutierrez, “Millimeter wave
are marginally affected by the variability of the mmWave mobile communications for 5G cellular: It will work!” IEEE access,
channel, a standalone architecture is preferable for the design vol. 1, pp. 335–349, 2013.
of fast initial access procedures, since it enables rapid beam [8] S. Akoum, O. El Ayach, and R. W. Heath, “Coverage and capacity
reporting operations. Conversely, an NSA configuration should in mmwave cellular systems,” in Signals, Systems and Computers
(ASILOMAR), 2012 Conference Record of the Forty Sixth Asilomar
be employed by users in connected mode to guarantee higher Conference on. IEEE, 2012, pp. 688–692.
resilience and an improved reactiveness in case a radio link [9] Z. Pi and F. Khan, “An introduction to millimeter-wave mobile broad-
failure occurs. A downlink configuration is in line with the band systems,” IEEE communications magazine, vol. 49, no. 6, 2011.
[10] J. S. Lu, D. Steinbach, P. Cabrol, and P. Pietraski, “Modeling human
3GPP design for NR and reduces the energy consumption at blockers in millimeter wave radio links,” ZTE Communications, vol. 10,
the UE side (since it has just to receive the synchronization no. 4, pp. 23–28, 2012.
or reference signals), but is less reactive because the gNBs [11] M. Giordani, M. Mezzavilla, C. N. Barati, S. Rangan, and M. Zorzi,
have a larger number of directions to sweep with downlink “Comparative analysis of initial access techniques in 5G mmWave
cellular networks,” in Information Science and Systems (CISS), 2016
SS blocks or CSI-RSs. Annual Conference on. IEEE, 2016, pp. 268–273.
[12] M. Giordani, M. Mezzavilla, and M. Zorzi, “Initial access in 5G
mmWave cellular networks,” IEEE Communications Magazine, vol. 54,
VIII. C ONCLUSIONS no. 11, pp. 40–47, 2016.
In this paper, we have presented a tutorial on beam man- [13] M. Giordani and M. Zorzi, “Improved user tracking in 5G millimeter
wave mobile networks via refinement operations,” in 16th Annual
agement frameworks for mmWave communications in 3GPP Mediterranean Ad Hoc Networking Workshop (Med-Hoc-Net), June
NR. The harsh propagation at mmWave frequencies requires 2017.
the implementation of directional transmissions supported by [14] M. Polese, M. Giordani, M. Mezzavilla, S. Rangan, and M. Zorzi, “Im-
beamforming techniques to increase the link budget. There- proved Handover Through Dual Connectivity in 5G mmWave Mobile
Networks,” IEEE Journal on Selected Areas in Communications, vol. 35,
fore, control procedures such as initial access must be up- no. 9, pp. 2069–2084, Sept 2017.
dated to account for the lack of an omnidirectional broadcast [15] T. Nitsche, C. Cordeiro, A. Flores, E. Knightly, E. Perahia, and
channel, and the optimal beam pair with which a base station J. Widmer, “IEEE 802.11ad: directional 60 GHz communication for
multi-Gigabit-per-second Wi-Fi [Invited Paper],” IEEE Communications
and a UE communicate should be tracked when needed. Magazine, vol. 52, no. 12, pp. 132–141, December 2014.
Consequently, the design and configuration of efficient IA [16] J. Wang, “Beam codebook based beamforming protocol for multi-Gbps
and tracking procedures is of extreme importance in cellular millimeter-wave WPAN systems,” IEEE Journal on Selected Areas in
Communications, vol. 27, no. 8, pp. 1390–1399, October 2009.
networks operating at mmWaves.
[17] R. Santosa, B.-S. Lee, C. K. Yeo, and T. M. Lim, “Distributed Neighbor
After a brief overview of the literature on beam man- Discovery in Ad Hoc Networks Using Directional Antennas,” in The
agement at mmWave frequencies, we described the frame Sixth IEEE International Conference on Computer and Information
structure and reference signals in 3GPP NR, focusing on Technology, Sept 2006, pp. 97–97.
[18] K. Chandra, R. V. Prasad, I. G. Niemegeers, and A. R. Biswas,
the settings for communication at frequencies above 6 GHz. “Adaptive beamwidth selection for contention based access periods in
Then, we described several beam management procedures millimeter wave WLANs,” in IEEE 11th Consumer Communications
according to different network architectures (standalone and and Networking Conference (CCNC). IEEE, 2014, pp. 458–464.
[19] D. Liu, L. Wang, Y. Chen, M. Elkashlan, K.-K. Wong, R. Schober, and
non-standalone) and signal transmission directions (downlink L. Hanzo, “User association in 5G networks: A survey and an outlook,”
or uplink). We also evaluated the impact of several parameters IEEE Communications Surveys & Tutorials, vol. 18, no. 2, pp. 1018–
(specified by 3GPP for NR) on their performance. We showed 1044, 2016.
[20] C. Jeong, J. Park, and H. Yu, “Random access in millimeter-wave
that there exist trade-offs among better detection accuracy, beamforming cellular networks: issues and approaches,” IEEE Commu-
improved reactiveness and reduced overhead. Finally, we nications Magazine, vol. 53, no. 1, pp. 180–185, January 2015.
provide insights and guidelines for determining the optimal [21] C. N. Barati, S. A. Hosseini, S. Rangan, P. Liu, T. Korakis, S. S. Panwar,
initial access and tracking strategies in different network and T. S. Rappaport, “Directional cell discovery in millimeter wave
cellular networks,” IEEE Transactions on Wireless Communications,
deployments, according to the need of the network operator vol. 14, no. 12, pp. 6664–6678, December 2015.
and the specific environment in which the nodes are deployed. [22] C. N. Barati, S. A. Hosseini, M. Mezzavilla, P. Amiri-Eliasi, S. Rangan,
T. Korakis, S. S. Panwar, and M. Zorzi, “Directional initial access
for millimeter wave cellular systems,” in 49th Asilomar Conference on
Signals, Systems and Computers. IEEE, 2015, pp. 307–311.
R EFERENCES
[23] V. Desai, L. Krzymien, P. Sartori, W. Xiao, A. Soong, and A. Alkhateeb,
[1] M. Giordani, M. Polese, A. Roy, D. Castor, and M. Zorzi, “Initial Access “Initial beamforming for mmWave communications,” in 48th Asilomar
Frameworks for 3GPP NR at mmWave Frequencies,” submitted to the Conference on Signals, Systems and Computers, 2014, pp. 1926–1930.
 21

[24] L. Wei, Q. Li, and G. Wu, “Exhaustive, Iterative and Hybrid Initial Ac- [50] ——, “Summary of offline discussions on CSI-RS,” Huawei, HiSilicom
cess Techniques in mmWave Communications,” in 2017 IEEE Wireless - Tdoc R1-1718947, 2017.
Communications and Networking Conference (WCNC). IEEE, 2017. [51] ——, “Measurement reporting for NR SS and CSI-RS ,” Huawei - Tdoc
[25] J. Choi, “Beam selection in mm-Wave multiuser MIMO systems using R2-1708703, 2017.
compressive sensing,” IEEE Transactions on Communications, vol. 63, [52] ——, “Details of cell quality derivation,” Ericsson - Tdoc R2-1704101,
no. 8, pp. 2936–2947, August 2015. 2017.
[26] A. Capone, I. Filippini, and V. Sciancalepore, “Context information for [53] ——, “Consideration on CSI RS for beam management,” ZTE Corpo-
fast cell discovery in mm-wave 5G networks,” in 21th European Wireless ration - Tdoc R2-1708123, 2017.
Conference; Proceedings of European Wireless, 2015. [54] ——, “Remaining issues on SRS,” InterDigital, Inc. - Tdoc R1-1716472,
[27] A. Capone, I. Filippini, V. Sciancalepore, and D. Tremolada, “Obstacle 2017.
avoidance cell discovery using mm-waves directive antennas in 5G [55] ——, “Discussion on remaining issues on SRS design,” CATT - Tdoc
networks,” in IEEE 26th Annual International Symposium on Personal, R1-1712386, 2017.
Indoor, and Mobile Radio Communications (PIMRC). IEEE, 2015, pp. [56] J. Liu, K. Au, A. Maaref, J. Luo, H. Baligh, H. Tong, A. Chassaigne,
2349–2353. and J. Lorca, “Initial Access, Mobility, and User-Centric Multi-Beam
[28] Q. C. Li, H. Niu, G. Wu, and R. Q. Hu, “Anchor-booster based Operation in 5G New Radio,” IEEE Communications Magazine, vol. 56,
heterogeneous networks with mmWave capable booster cells,” in IEEE no. 3, pp. 35–41, March 2018.
Globecom Workshops. IEEE, 2013, pp. 93–98. [57] E. Onggosanusi, M. S. Rahman, L. Guo, Y. Kwak, H. Noh, Y. Kim,
[29] W. B. Abbas and M. Zorzi, “Context information based initial cell search S. Faxer, M. Harrison, M. Frenne, S. Grant, R. Chen, R. Tamrakar, and
for millimeter wave 5G cellular networks,” in European Conference on a. Q. Gao, “Modular and High-Resolution Channel State Information
Networks and Communications (EuCNC). IEEE, 2016, pp. 111–116. and Beam Management for 5G New Radio,” IEEE Communications
[30] A. Alkhateeb, Y. H. Nam, M. S. Rahman, J. Zhang, and R. W. Magazine, vol. 56, no. 3, pp. 48–55, March 2018.
Heath, “Initial Beam Association in Millimeter Wave Cellular Systems: [58] A. A. Zaidi, R. Baldemair, H. Tullberg, H. Bjorkegren, L. Sundstrom,
Analysis and Design Insights,” IEEE Transactions on Wireless Commu- J. Medbo, C. Kilinc, and I. D. Silva, “Waveform and Numerology
nications, vol. 16, no. 5, pp. 2807–2821, May 2017. to Support 5G Services and Requirements,” IEEE Communications
[31] Y. Li, J. Luo, M. Castaneda, R. Stirling-Gallacher, W. Xu, and G. Caire, Magazine, vol. 54, no. 11, pp. 90–98, November 2016.
“On the Beamformed Broadcast Signaling for Millimeter Wave Cell [59] A. Osseiran, F. Boccardi, V. Braun, K. Kusume, P. Marsch, M. Maternia,
Discovery: Performance Analysis and Design Insight,” arXiv preprint O. Queseth, M. Schellmann, H. Schotten, H. Taoka, H. Tullberg, M. A.
arXiv:1709.08483, 2017. Uusitalo, B. Timus, and M. Fallgren, “Scenarios for 5G mobile and
[32] A. S. Cacciapuoti, “Mobility-Aware User Association for 5G mmWave wireless communications: the vision of the METIS project,” IEEE
Networks,” IEEE Access, vol. 5, pp. 21 497–21 507, 2017. Communications Magazine, vol. 52, no. 5, pp. 26–35, May 2014.
[33] J. Palacios, D. De Donno, and J. Widmer, “Tracking mm-Wave chan- [60] 3GPP, “NR and NG-RAN Overall Description - Rel. 15,” TS 38.300,
nel dynamics: Fast beam training strategies under mobility,” in IEEE 2018.
Conference on Computer Communications (INFOCOM). IEEE, 2017. [61] 3GPP, “NR - Radio Resource Control (RRC) protocol specification -
[34] S. Jayaprakasam, X. Ma, J. W. Choi, and S. Kim, “Robust Beam- Release 15,” TS 38.331, 2017.
Tracking for mmWave Mobile Communications,” IEEE Communications [62] ——, “NR - Physical layer procedures for control - Release 15,” TS
Letters, vol. 21, no. 12, pp. 2654–2657, Dec 2017. 38.213, 2018.
[35] M. Giordani, M. Mezzavilla, S. Rangan, and M. Zorzi, “An Efficient [63] ——, “Framework for beamformed access,” Samsung - Tdoc R1-
Uplink Multi-Connectivity Scheme for 5G mmWave Control Plane 164013, 2016.
Applications,” submitted to IEEE TWC, 2017. [Online]. Available: [64] ——, “Discussion on CSI-RS Design,” Qualcomm - Tdoc R1-1718546,
https://arxiv.org/abs/1610.04836 2017.
[36] ——, “Multi-Connectivity in 5G mmwave cellular networks,” in Ad Hoc [65] ——, “NR PRACH preamble resource allocation,” Ericsson - Tdoc R1-
Mediterranean Networking Workshop (Med-Hoc-Net). IEEE, 2016. 1611905, 2016.
[37] M. Polese, M. Mezzavilla, and M. Zorzi, “Performance Comparison of [66] ——, “NR - Physical layer measurements - Rel. 15,” TS 38.215, 2017.
Dual Connectivity and Hard Handover for LTE-5G Tight Integration,” [67] ——, “Study on new radio access technology: Radio access architecture
in Proceedings of the 9th EAI International Conference on Simulation and interfaces,” TR 38.801, 2017.
Tools and Techniques. ICST (Institute for Computer Sciences, Social- [68] ——, “NR - Multi-connectivity - Overall description (Stage 2),” TS
Informatics and Telecommunications Engineering), 2016, pp. 118–123. 37.340, 2018.
[38] F. B. Tesema, A. Awada, I. Viering, M. Simsek, and G. P. Fettweis, [69] J. Oueis and E. C. Strinati, “Uplink traffic in future mobile networks:
“Mobility modeling and performance evaluation of multi-connectivity in Pulling the alarm,” in International Conference on Cognitive Radio
5G intra-frequency networks,” in IEEE Globecom Workshops. IEEE, Oriented Wireless Networks. Springer, 2016, pp. 583–593.
2015. [70] T. Bai and R. W. Heath, “Coverage and rate analysis for millimeter-wave
[39] O. Semiari, W. Saad, and M. Bennis, “Caching Meets Millimeter Wave cellular networks,” IEEE Transactions on Wireless Communications,
Communications for Enhanced Mobility Management in 5G Networks,” vol. 14, no. 2, pp. 1100–1114, February 2015.
arXiv preprint arXiv:1701.05125, 2017. [71] T. Yoo and A. Goldsmith, “Optimality of zero-forcing beamforming with
[40] N. Gonzalez-Prelcic, A. Ali, V. Va, and R. W. Heath, “Millimeter-Wave multiuser diversity,” in IEEE International Conference on Communica-
Communication with Out-of-Band Information,” IEEE Communications tions (ICC), vol. 1, May 2005, pp. 542–546.
Magazine, vol. 55, no. 12, pp. 140–146, Dec. 2017. [72] M. R. Akdeniz, Y. Liu, M. K. Samimi, S. Sun, S. Rangan, T. S.
[41] S. Schwarz, C. Mehlführer, and M. Rupp, “Calculation of the spatial Rappaport, and E. Erkip, “Millimeter wave channel modeling and
preprocessing and link adaption feedback for 3GPP UMTS/LTE,” in 6th cellular capacity evaluation,” IEEE Journal on Selected Areas in Com-
conference on Wireless advanced (WiAD). IEEE, 2010. munications, vol. 32, no. 6, pp. 1164–1179, June 2014.
[42] M. Giordani, M. Mezzavilla, A. Dhananjay, S. Rangan, and M. Zorzi, [73] S. Sun, T. S. Rappaport, R. W. Heath, A. Nix, and S. Rangan,
“Channel dynamics and SNR tracking in millimeter wave cellular “MIMO for millimeter-wave wireless communications: beamforming,
systems,” in 22th European Wireless Conference. VDE, 2016. spatial multiplexing, or both?” IEEE Communications Magazine, vol. 52,
[43] 3GPP, “NR - Physical channels and modulation - Release 15,” TS no. 12, pp. 110–121, December 2014.
38.211, V15.0.0, 2018. [74] S. Dutta, C. N. Barati, A. Dhananjay, and S. Rangan, “5G Millimeter
[44] ——, “Remaining details on sync signals,” Samsung - Tdoc R1- Wave Cellular System Capacity with Fully Digital Beamforming,” in
1721434, 2017. 51st Asilomar Conference on Signals, Systems and Computers, Nov
[45] ——, “Discussion on remaining issues of SS block and SS burst set,” 2017.
Motorola Mobility, Lenovo - Tdoc R1-1714212, 2017. [75] 3GPP, “Discussion on NR 4-Step Random Access Procedure,” Ericsson
[46] ——, “SS burst periodicity for initial cell selection in NR,” Nokia, - Tdoc R1-1718052, 2017.
Alcatel-Lucent Shanghai Bell - Tdoc R4-1705123, 2017. [76] K. Takeda, L. H. Wang, and S. Nagata, “Latency Reduction toward 5G,”
[47] ——, “Measurement configuration and reporting considering additional IEEE Wireless Communications, vol. 24, no. 3, pp. 2–4, 2017.
RS,” Huawei - Tdoc R2-1703387, 2017. [77] 3GPP, “Relation between radio link failure and beam failure,” Ericsson
[48] ——, “Measurement configuration for CSI-RS,” Ericsson - Tdoc R2- - Tdoc R1-1705917, 2017.
1704103, 2017. [78] ——, “Beam failure detection and beam recovery actions,” Ericsson -
[49] ——, “NR CSI-RS configuration for RRM measurement,” Samsung - Tdoc R1-1705893, 2017.
Tdoc R2-1709593, 2017. [79] ——, “LS on NR PRACH BW Aspects,” Tdoc R1-1716814, 2017.
 22

[80] M. Rebato, M. Mezzavilla, S. Rangan, F. Boccardi, and M. Zorzi,

“Understanding Noise and Interference Regimes in 5G Millimeter-Wave
Cellular Networks,” in 22th European Wireless Conference, 2016.
[81] W. B. Abbas, F. Gomez-Cuba, and M. Zorzi, “Millimeter Wave Receiver
Efficiency: A Comprehensive Comparison of Beamforming Schemes
With Low Resolution ADCs,” IEEE Transactions on Wireless Commu-
nications, vol. 16, no. 12, pp. 8131–8146, Dec 2017.