Content uploaded by Marko Höyhtyä
Author content
All content in this area was uploaded by Marko Höyhtyä on Mar 12, 2018
Content may be subject to copyright.
Field Trial of the 3.5 GHz Citizens Broadband Radio Service
Governed by a Spectrum Access System (SAS)
Marko
Palola
1
, Marko
Höyh tyä
1
, Pekka Aho
1
,
Miia Mustonen
1
,
Tero Kippola
2
,
Marjo Heikkilä
2
,
Seppo Yrjölä3, Vesa Hartikainen3, Lucia Tudose3, Arto Kivinen4, Reijo Ekman4, Juhani Hallio4,
Jarkko Paavola4, Marko Mäkeläinen5, and Tuomo Hänninen5
1
VTT Technical Research Centre of Finland
Ltd
, Oulu, Finland
.
2
Centria University of Applied Sciences, Ylivieska, Finland
3Nokia Networks, Oulu, Finland
4Turku University of Applied Sciences, Turku, Finland
5Centre for Wireless Communications, University of Oulu, Finland
Abstract—In this paper, we describe a spectrum access system
(SAS) based Citizens Broadband Radio Service (CBRS) field trial
using a live LTE network in the 3.5 GHz band. The latest
WInnForum specification guided the implementation of the
relevant protocols for SAS operation. Here, we evaluate the
performance of a CBRS field trial by using one of the most
important performance indicators in a spectrum sharing scenario
– the evacuation time. It indicates how rapidly the secondary user
relinquishes the shared spectrum band to the primary user.
Following the applied protocols, we measure and analyze the time
scales for the evacuation and frequency change procedures in a
field trial environment. Our work shows that the set time limits
for the protection of primary users against interference are
realistic when using commercially available mobile networks and
equipment. Finally, utilizing knowledge of the latest base station
models, we propose ways to reduce the evacuation and
reconfiguration time by up to 70%.
I. INTRODUCTION
Spectrum sharing technologies have significantly advanced
since Mitola introduced the concept of cogn itive radio in [1].
Dynamic spectrum access technologies ar e already included in
multiple standards in different frequency bands, and
regulatory processes are being updated to include new forms
of licensing [2]. Coordinated spectrum sharing approaches are
being developed and proposed between different technologies
used in the industrial, scientific and medical (ISM) bands [3].
The idea behind coordinated and licensed sharing access
approaches is the ability to provide interference free operation
that leads to a better, or even guaranteed, quality of service
(QoS) in sharing applications.
One example is the European Licensed Shared Access
(LSA) concept, which allows spectrum sharing between
incumbent users and LSA licensees, both having exclusive
access to a portion of the spectrum at a given location and
time [4]. The first application of the concept concerned mobile
operators allocatin g mobile broadband services to new
spectrum bands, which already had other types of (incumbent)
use. For example, shared use of the 2.3–2.4 GHz band
between a mobile network and wireless cameras has been
demonstrated in Finnish LSA field trials [5].
In the United States, the prevailing approach to spectrum
sharing is the Citizens Broadband Radio Service (CBRS)
governed by SAS in the 3550–3700 MHz band [6]. In both
sharing concepts, incumbent users have the highest priority in
terms of spectrum access and protection against interference
from other users at any location and time. While LSA is a two-
tier model, SAS has three tiers, including the general
authorized access (GAA) tier, to facilitate opportunistic
spectrum use. Priority access (PA) users are allocated to
exclusive channels and protected from other PA and GAA
users. In the GAA tier, multiple users can use a given channel,
and thus there is no interference protection.
Compared to LSA, SAS allows a more dynamic and
complex sharing model, which is likely to promote
competition and foster innovation [7]. SAS is also more likely
to provide more efficient spectrum utilization and better
support for the deployment of small cells. Small cells with
low-power communication enable smaller exclusion zones [8],
thus providing more spectrum optimization opportunities than
macro sites. A major difference to the LSA concept is the use
of spectrum sensing in obtaining information about the current
spectrum use. To meet the mission critical requirements of
military incumbent users, it is required that sensing is used in
and adjacent to the 3.5 GHz band to detect incumbent radar
activity in coastal areas and near inland military bases.
Confidentiality of the sensitive military incumbent user
information is ensured through strict operational security
requirements and corresponding certification of the sensing
elements, as well as with operator authorization.
Similarly as in LSA, at the core of the SAS concept is a
database system. Incumbent users may provide spectrum
usage information, such as duration of the use and operational
parameters such as transmitter identity, location, antenna
height, transmission power, interference tolerance capabilit y
and protection contour, to be in cluded in the database [6]. SAS
can use either a database or a database-plus-sensing approach
to identify the available spectrum opportunities.
Previous studies on the SAS development have focused on
the technical and theoretical aspects of the research work. In
order to start practical testing and trials of the SAS concept,
design and implementation are also important. Recent
publications discussing architectural considerations of SAS
include [6], [11] and [12]. In addition, a messaging protocol
for the SAS operation has been proposed in [12]. Currently,
several member companies and research organizations of the
Wireless Innovation Forum (WInnForum) are jointly
developing interfaces, protocols and messaging formats for
SAS. Some of the specified requests enable spectrum
inquiries, granting permissions to use spectrum, and spectrum
relinquishment [13] between SAS and CBRS devices
(CBSDs).
What is currently missing in the literature is analyses of
the dynamics and time domain performance of SAS. Previous
work has focused more on spatial and frequency dom ain
considerations. One of the most important performance
indicators is the evacuation time from the first indication of
incumbent use in the same band and location to the time th e
band is clear ed of any interfering secondary systems such as
LTE base stations. In the case of an informing incumbent, the
evacuation time determines how much in advance the
incumbent user needs to declare its intention to use the
spectrum at a certain location to avoid interference. In the case
of sensing, the evacuation time is directly linked to the
detection requirements.
In this paper, we describe a CBRS field trial environment
and the related performance measurement results. We use the
recent standard specifications from [13]–[15] to study the time
domain operation of SAS with a focus on the evacuation and
reconfiguration performance. The system is implemented and
evacuation and reconfiguration time values are measured and
analyzed in a live commercial network environment. It is
important to carry out field tests to prove that the relevant SAS
requirements can be met using commercial networks and
systems. Our measurement results have already contributed to
updating the time limits in [9] and [10]. For example, we have
found that the configuration command to clear the band can
vary in network management systems. Therefore, reserving
some additional response time is practical to prepare for
worst-case scenarios and large-scale network oper ations. On
the other hand, the industry may also find our measurements
on the network reconfiguration time to restore the mobile
network operation after the evacuation interesting.
The paper is organized as follows: Section II presents the
reference SAS architecture and defines the messaging
protocol. The trial environment an d the measurement setup are
introduced in Section III, and the results are presented in
Section IV. Finally, Section V concludes the paper.
II. HIGH-LEVEL SYSTEM MODEL
A. SAS reference architecture
Fig. 1 illustrates the high-level SAS architecture, which
has been mainly defined in the WInnForum [14] and [17]. The
reference architecture shows the main components and
interfaces needed in defining th e messaging protocols. The
main component of the reference architecture is SAS 1, which
determines the available frequencies and assigns them to
different CBSDs and determines the maximum transmission
power limits at given locations. It also enforces exclusion and
protection zones around incumbent users such as U.S.
Department of Defense (DoD) sh ipborne radars operating in
coastal areas and non-federal Fixed Satellite Service (FSS)
earth stations.
To protect FSS earth stations, the Federal Communications
Commission (FCC) has adopted a rule requiring satellite
operators to register their stations annually [9]. In the case of
DoD shipborn e radars, the SAS uses in formation from
Environment Sensing Capability (ESC) devices to ensure that
CBSDs operate in a manner that does not interfere with the
incumbent users but still facilitates information exchange
between multiple SAS servers.
The protection of federal DoD incumben t users is
implemented by utilizing the static exclusion zones (EZ)
scattered in a large area of the country. In the second
deployment phase, the ESC system enables the rest of the
country, including major coastal areas, to become available, as
the exclusion zones are converted into protection zones (PZ).
An ESC deployment near exclusion zones consists of one or
more commercially operated networks of sensin g devices that
can be used to detect signals from federal radar systems in the
vicinity of the exclusion zones. Additionally, a CBSD
infrastructure based sensing could be considered under strict
operational security requirements. Prospective ESC operators
must have their systems approved through a similar process as
SAS servers and SAS administrators. An SAS would obtain
FCC maintained information about registered or licensed
commercial users from FCC databases and exclusion zone
information maintained by the National Telecommunications
and Information Administration (NTIA).
Where CBSDs are centrally controlled, a Domain Proxy
(DP) can act as a managing intermediary for a number of
separate CBSDs so that the SAS communicates with the proxy
instead of each individual CBSD. The DP selects the channels
for use by a specific CBSD, or alternatively notifies the
available channels to the Network Management System
(NMS) for CBSD channel selection.
There are two types of CBSDs in the CBRS/SAS concept.
Category A devices correspond to lower power access points
and femtocells, whereas Category B devices correspond to
point-to-point and point-to-multipoint types of architecture.
Category A devices can operate by using database only or
with ESCs, which are dedicated devices to detect incumbent
radar activity. Category B operation always requires an ESC.
In this paper, we focus on the small cell operation, i.e.,
Category A CBSDs.
B. Prerequisite procedures for operation
A connection to an SAS is required for CBSDs to access
the spectrum. However, there are some prerequisites and a
Figure 1. SAS system architecture.
dedicated procedure for opening a connection between an SAS
and a CBSD [15]. First, the CBSD or DP initiates SAS
discovery to locate a potential SAS to connect to. Then, the
CBRS user needs to register with the SAS to obtain a CBRS
User ID. PA license (PAL) rights management and PAL ID
registration allows the SAS to authenticate the claim to a PA
license for privileged users. Device type parameters such as
spurious emission mask, entered into the CBSD database, help
avoiding corruptions. Installation parameters such as the
location of the CBSD and antenna parameters must be entered
to the database by certified profes sional installers to ensure
accurate interference and allocation calculations. Finally, a
security framework enables trust and identification between
the SAS, CBSD and DP components.
There are some time and space domain limits discussed
recently regarding the CBRS model. First, CBSDs must be
able to determine their geographic coordinates with an
accuracy of 50 meters horizontal and 3 meters vertical [6].
Even though there are some technical challenges in achieving
indoor location accuracy due to lack of the GPS signal, this is
an achievable requirement, particularly because CBSDs are
fixed devices [9]. The requirement can be achieved with
automatic geolocation algorithms or with input of a
professional installer.
C. Messaging protocol
The SAS to CBSD signaling protocol has been defined in
[13]–[15]. The protocol specifies the messages and their
content and sequences needed to register a device to an SAS,
obtain permission to transmit, and to stop using allocated
resources. The reconfiguration process includes the evacuation
of the granted channels and configuring the associated CBSDs
to another frequency band. The communication between ESC,
SAS and DP/CBSD components related to the reconfiguration
process is depicted in Fig. 2. The messages are defined as
follows:
An ESC alert informs the SAS of the appearance of an
incumbent user. There is a time limit from an ESC alert to the
SAS to confirm that the interfering CBSDs have vacated the
spectrum. The CBSDs must cease transmission and move to
another frequency range or change their power level within th e
time limit following instructions provided by the SAS [9], [10]
[14]. The initial 60-second evacuation time requirement has
been increased to 300 seconds, partly due to the work done in
our field trials. Our tests have shown that the 60-second limit
cannot be achieved for large networks, not even under ideal
circumstances.
A heartbeat request from a CBSD informs that the CBSD
begins or continues using the allocated spectrum. If the SAS
does not receive a heartbeat within a certain period, it will
assume that the CBSD is no longer operating in the granted
spectrum. Similarly, CBSDs require a heartbeat response to be
able to operate in the allocated spectrum.
A heartbeat response allows the SAS to confirm, modify,
suspend or terminate a grant and to change the heartbeat
interval. A CBSD is authorized to use the spectrum during the
time interval defined in the latest heartbeat response message.
A grant may be suspended if an incumbent user such as a
naval radar arrives in the neighborhood. If an incumbent user
such as an FSS station moves into neighborhood to stay, th e
grant may be permanently denied.
A CBSD can request spectrum from the SAS at any time
by sending a grant request. CBSDs may also initiate a
spectrum inquiry procedure to check from the SAS spectrum
availability for one or more frequency ranges. A spectrum
inquiry does not guarantee channel availability but provides a
good indication of that. However, this information is useful to
be included in a grant request in order to enhance the
resource allocation optimization. CBSDs may also request
Figure 3. Radar sensing system.
ESC Alert
SASESC CBSD/DP
SAS
performs
channel
optimisation
assessm ent
Band
evac uat ion a nd
reconfiguration
of base stations
SAS
continues
resource
optimisation
Heartbeat Request
Heartbeat Response
Hea rtbe at d urat ion
expires; Reset
Heartbeat duration timer
Heartbeat R equest Heartbeat duration
exp ires ; Res et
Hea rtbe at d uration timer
Heartbeat Response
(ne w operational
parameter s)
Rel inq uish Requ est
Rel inq uish R espo nse
Grant Req uest
Gr ant R espo nse
Operation continues
with new operational
parameters
Waiting for
Heartbeat
Spectrum inquiry request
Spectrum inquiry response
SAS
performs
channel
availability
assessm ent
Heartbeat Request
Figure 2. CBSD reconfiguration process after receiving an
ESC
Alert
.
Optional spectrum inquiry included.
access to a specific channel based on network planning. After
receivin g a grant request, the SAS then performs a channel
interference assessment to determine if the requested
frequency range is acceptable.
Arelinquish request can be sent to notify the SAS that
the CBSD no longer uses the allocated spectrum. The SAS
answers with a relinquish response, and the freed spectrum
can be reused.
D. The evacuation process
In this study, we measured the time intervals of the
operations needed to evacuate the channel when the SAS
denies an existing grant. The evacuation procedure fulfilling
the specifications is depicted in Fig. 2. First, the CBSD is in
the Authorized state, transmitting on the granted spectrum.
The SAS receives an ESC Alert, which means that the sensor
system has detected an incumbent use, and then determines
which channels are affected by the incumbent use. The SAS
then delivers all the CBSDs transmitting on those channels a
heartbeat response, denying the use of those channels. The
DP/CBSDs then switch off the radio transmissions
accordingly. The CBSDs go in a Registered state in which
they connected to the SAS but cannot use the radio without the
allocated spectrum. Each DP/CBSD may send a new grant
request to the SAS to gain access to an alternative frequency.
III. TRIAL ENVIRONMENT AND THE MEASUREMENT SETUP
The Finnish live CBRS trial environment is depicted in
Fig. 4. The key building blocks are developed and governed
by multiple partners of the CORE++ project [16]. The Finnish
CBRS trial environment has been demonstrated with new
features on multiple occasions [19], [20]. The building blocks
of the trial are as follows:
A. Radar sensing system
The used radar sensing system, depicted in Fig. 3, consists
of a radar signal simulator (RSS), a spectrum-sensing receiver
(SSR) and a sensing software algorithm called Sensor
Commander. To avoid the need to have real naval or maritime
radars operating in the area, the RSS is used to generate radar
signals based on R2-ESC-01 in [17]:
Pulse repetition frequency 1 kHz
Pulse width 0.9 µs
Antenna scan rate 15 RPM
Antenna beam width 1.8 degrees
The RFeye spectrum sensing device is used to sense the radar
signals. The power level values (dBm) are recorded with a
frequency resolution of 19.531 kHz over the 20 MHz
bandwidth in the SAS band. The occupancy scan is performed
every 45 seconds. The developed ESC software processes the
occupancy data by detecting and recording power levels
higher than -90 dBm. The bandwidth, power level and center
frequency of the findings are stored.
B. LTE 3.5 GHz network
The CBRS trial environment consists of three 3GPP
Release 10 LTE-Advanced compliant base stations, a radio
access network, a management system and a core network.
Commercially available Flexi Multiradio time-division (TD)-
LTE 3.5 GHz base station at 3GPP spectrum band 42 (3.4–
3.6 GHz) are used and equipped as Category A low-power
access points for indoor usage. The radios are located inside
an office building, as shown in Fig. 5. Two of the CBSDs are
connected to a commercial Network Management System
Figure 4. SAS field trial environment.
(NMS) and are managed by a DP. One CBSD is equipped as a
standalone CBSD, having the core network functionalities
required to operate (Lite-EPC) and control (SAS-
controller/BTS_tools) locally.
The authorities have granted an indoor trial license to use
the 3.51–3.59 GHz band for field trial purposes.
C. Spectrum Access System
The SAS combines multiple functions to provide the SAS
capability for the field trial. The SAS component is
implemented on a Java Spark server on Linux with HTTPS
REST API for DPs and CBSDs. The SAS algorithm, Spark
server, ESC, LTE base stations, LTE network and DP are
located around Finland. They communicate using CORE++
tools, i.e., in practice by using publish/subscribe message-
passing communication [16].
The most important functions the SAS provides in the field
trial are as follows:
1) SAS Repository
The SAS repository is a database that gathers data about the
spectrum use in the area of interest, including CBSDs’
operational parameters such as identification, location,
antenna parameters, transmission power, and used channels.
The SAS repository stores all the information required by
other key components for channel allocations and interference
management in the network.
2) Environmental Sensing Capability
The ESC consists of networks of sen sors that detect the
presence of signals from incumbent systems in the band and
communicate that information to the SAS to facilitate
protection of operations in the band. Th e ESC module used in
the field trial combines information from the sensing system
and sends ESC alerts to alert the SAS to start an evacuation
process.
3) CBSD manager
The CBSD manager follows the protocols defined in the
SAS-CBSD protocol [13] specification for DPs and CBSDs to
access the SAS. It handles SAS requests, creates responses
and updates the SAS repository. In the field trial, the SAS
implementation also supports the development of alternative
SAS algorithms in order to test different channel allocation
methods.
4) SAS algorithm: Intelligence of the SAS
Frequency channel allocation is a challenging task due to
the stringent channel usage priority requirements and a
potentially large number of CBSDs. The CBSD manager
initiates the SAS algorithm processing in several cases, for
example, when incumbent activity is detected or ceases, a
CBSD is requesting a grant or relinquishing an existing grant,
or a CBSD has not sent a heartbeat within the set interval. The
CBSD manager generates an activity report based on the data
collected on activities such as channel allocations and on the
current state of CBSDs, pending grant requests, and the sensed
arrival and departure of incumbent users in or from a specific
band. The SAS algorithm uses the activity report as an input to
generate series of commands to the SAS to deny, change or
allocate CBSD grants. In a single run, the SAS algorithm can
carry out at most one state change for each CBSD. For
example, spectrum can be granted for multiple CBSDs
requesting at the same time. The SAS algorithm was
specifically developed to support the research and
development of the field trial. Thus, the measured SAS
Figure 5. LTE CBSD’s in field trial network. Figure 6. Domain Proxy architecture components.
algorithm is not using any advan ced radio propagation or
interference minimization calculations that are possible to
perform by using radio, antenna, power and other information
stored in the SAS system.
The basic idea of the algorithm is to control interference,
minimize the number of channel changes and use the SAS
band efficiently. First, the algorithm checks the activity report
for existence of incumbent users. If incumbent use is detected,
the algorithm then denies the grants of any overlapping
CBSDs. Second, the algorithm allocates the available free
channels to CBSDs in the order they send their grant requests.
If all channels are occupied, a requesting GAA user is put on a
shared GAA channel or a PAL channel with an existing GAA
user. Any requesting PA user is put on a PAL channel that was
previously being used by one or more GAAs, and the GAAs’
grants are denied. Third, the algorithm checks the possibility
to distribute GAA users sharing a channel to other available
free channels.
D. Domain Proxy
In a typical Mobile Network Operator (MNO) deployment
scenario, a CBSD is a base station in a managed radio network
comprising other base stations (BSs), DPs, core network, and
the NMS functionality. The DP acts as a managing
intermediary component between the SAS and several
CBSDs, enabling SAS usage for base stations that do not have
interface to an SAS. The two main functions of DPs are: (1)
communicating directly with the SAS by using the SAS-
CBSD protocol [13] and (2) configuring a number of CBSDs
by instructing the NMS to change the frequency, bandwidth,
transmit power and operational state of the impacted CBSDs.
The DP is able to report certain CBSD measurement data to
the SAS based on the SAS’s instructions.
A standalone GAA CBSD can be controlled directly via
the CBSD manager. Wi-Fi was originally thought to be the
most potential individual GAA customer. For MNOs, a valid
option is an unlicensed LTE in SAS band. For example, a
MulteFire system can be installed and operated in the same
way as a Wi-Fi access point. Right now, CBRS Alliance is
establishing an effective product certification program for
LTE equipment in the U.S. 3.5 GHz band to ensure multi-
vendor interoperability.
In the trial, the DP was implemented as a self-organizing
network (SON) module on top of the Eden-NET SON
platform [18] and Eden-NET configuration interface towards
NetAct, having the following functions (see Fig. 6):
Decoding/encoding SAS protocol messages
Maintaining the CBSD state and channel assigned by
the SAS
Aggregation of CBSD information for the SAS
Management of spectrum usage of the CBSD through
the heartbeat and relinquishment procedures.
The DP may also contain advanced functionalities, such as
flexible self-control and interference optimization. In addition
to lar ger MNO operated networks, the DP enables, for
example, combining small cells of a shopping mall or sports
venue to a single virtual base station entity that covers the
entire venue.
E. Evacuation procedure:
We made a performance analysis of the CBSD evacuation
and subsequent frequency change procedure in a live LTE test
network. The purpose was to find out the total evacuation time
in the field trial and record the processing times in each
component, so we divided the overall procedur e into several
steps based on field trial key component boundaries. The steps
T13
T0
Radar s ensed to
appear
T1 T2
ESC combines
sensor information
and g enerates an
ESC Aler t
T3
SAS algorithm
optimizes the band
allocations
T4
Domai n proxy s ends
commands to NMS
according to SAS
algo rithm resp onses
T5
NMS begins
conf igurati on of
CBSD
Band is cleared,
CBSDs no more
operating in the band
T6 T7
Grant request is sent
to SAS
Eva cuation tim e Te
Frequency change time Tw
Td
CBSD Ma nager
sends relevan t
inf ormation t o SAS
algorithm
T8
SAS algorithm
deter mines suitable
band and radio
parameters for
opera tion
Trans mission Td
T9
Grant response is
sent to CBSDs
Te
T10
Heart beat NMS begins
configuration of CBSD
T11 T12
Tw
Operation con tinues
in a new band
Heart beat
Figure 7. Timeline of the evacuation and frequency change process
are shown in the timeline in Fig. 7, and the details of each step
are provided below.
(1) In the beginning, the network is operating on SAS band
under SAS control and CBSDs are transmitting using
allocated 10 MHz channels. The evacuation begin, when an
ESC communicates to the CBSD Manager that it has detected
a signal (possibly from a radar) at time instant T0, including
information such as sensor location and sensed received
power, bandwidth and frequency. Based on the received
sensor information and current spectrum usage, the CBSD
Manager generates an ESC alert at time instant T1, if the
spectrum state has been changed and spectrum man agement is
needed
(2) At time instant T2, the CBSD manager sends an
activity report to the SAS algorithm. The activity report
includes information about the SAS spectrum range and
channel division, ESC alert (incumbent use), and state of each
CBSD (Grant Request,Granted or Authorized), band usage,
radio information and priority of CBSDs (GAA/PA).
(3) The SAS algorithm optimizes the band allocations
based on the sensed incumbent use and existing GAA/PA
users, licenses and radio types at time instant T3. The changes
made to channel allocations are communicated back to the
SAS.
(4) The DP is responsible for configuring the base stations
according to SAS algorithm responses. At time instant T4, the
DP receives the changes to the existing grants in the response
of heartbeat request of the CBSD.
(5) Th e NMS begins the con figuration of the base station
at time instant T5 to switch off the radio of the base station to
change power or frequency.
(6) During the configuration, the actual band becomes
cleared during the configuration process as the radio is
switched off at time instant T6, which ends the evacuation.
F. Frequency change procedure
(7) After the configuration command has ended, the DP
sends new grant requests to the SAS to gain access to new
frequencies for the base stations that were switched off. This
starts the frequency change procedure at time instant T7.
(8) The SAS algorithm allocates bands and radio
parameters to CBSDs during the grant request at time instant
T8. At this point, the new PA user allocations can also trigger
switch ing off/reconfiguration of some GAA radios on the PAL
channels.
(9) At time instant T9, DP/CBSDs receive grant responses.
This changes their state to Granted. The configuration of base
stations to a new frequency may begin at time instant T9,
provided that the radios remain off-air.
(10) Each DP/CBSD sends a single heartbeat request to
check their grant status at time instant T10. The heartbeat
response changes the state to Authorized/Transmission.
(11) Th e configuration of base station to switch them back
on-air begins at time instant T11 based on the algorithm’s
decision.
(12) Similarly, the base station switch back on-air using a
new channel at some point during the configuration command.
This ends the frequency change procedure.
(13) Once the base stations are back on-air, their operation
with regular heartbeat requests continues as per usual.
IV. DELAY ANALYSIS AND MEASUREMENT RESULTS
A. Evacuation and frequency change times
We performed ten consecutive measurements using the
above-described setup to define the time needed to perform
each step in the process composed of the evacuation phase and
the frequency change procedure. The following results were
obtained with the SAS algorithm described in Section III D.
The heartbeat interval was set to 20 seconds. We had two base
stations online behind a DP, and we measured th e duration of
the evacuation and reconfiguration process using a single base
station. The evacuation and frequency change measurement
points are presented in Tables I and II, respectively.
Table I. Evacuation measurement points, time in seconds.
Event
Avg
Max
Min
T1 ESC Alert 0 0 0
T2 SAS Algorithm begin 0.054 0.074 0.051
T3 SAS Algorithm ready 10.1 10.1 10.1
T4 Heartbeat denies grant 22.3 29.2 12.0
T5 NMS conf. begin 22.5 29.3 12.3
T6 Channel is clear 88 103 78
NMS conf. complete 103 115 94
As can be seen from Table I on the evacuation procedure,
the most time-consuming part of the procedure is deactivating
the base station by the NMS. This step takes on average
around one minute and 20 seconds after the base station has
lost the grant. Until the channel is freed, the evacuation takes
88 seconds on average. Time instant T6 was monitored by
using a spectrum analyzer.
The total evacuation time is 103 seconds on average; this
also covers the completion of the NMS configuration
comman d with additional checks to validate the configuration
success or failure overall. At this point, the NMS can confim
the band is evacuated.
Table II. Reconfiguration, frequency change time in seconds.
Event
Avg
Max
Min
T7 Grant request 0 0 0
T8 SAS Algorithm - - -
T9 Grant response 4.7 5.0 4.0
T10 Heartbeat, T11 24.8 25.2 24.1
T12 On Air 98 103 93
T13 Heartbeat 105 112 98
Time instant T3 shows that the SAS algorithm delay is on
average 10 seconds, which is too large for the algorithm run.
This delay includes additional four seconds spent on
synchronization, networking and queuing delays due to a slow
message passing protocol, and the algorithm is actually run
once for both grant requests. The algorithm delay could be
optimized to 1–2 seconds by running the algorithm locally and
processing the grant requests in a single algorithm run.
Time instant T4 also includes at most the heartbeat interval
until the SAS can communicate the grant denial to the DP in
the next response. The SAS can alter the heartbeat interval,
but here it was fixed to 20 seconds.
The reconfiguration and frequency change of the base
station begins right after the evacuation once the DP sends th e
grant request. Time instant T9 includes running the SAS
algorithm to select a new frequency for the base station. T10
includes the first heartbeat request and response after a
successful grant response and before the CBSD can turn on
the radio. The T11 configuration command starts at the same
time and continues on average for 80 seconds after the
heartbeat response. Time instant T12 was inspected on-air by
using a channel analyzer at the site. T13 indicates the time
when the DP communicates to the SAS that the base station is
transmitting. Most time in reconfiguring the CBRS system
(Table II) to operate in new bands is spent on unlockin g the
base station. Based on our measurements, it takes from an
ESC alert until the frequency change process is completed on
average around three minutes and 30 seconds, and less than
four minutes in the slowest case.
It should be noted that actual NMS command delays
depend on the base station manufacturer and model, selected
evacuation type, and load of the NMS during the
measurement. Manufacturers have their own LTE access,
management and core network systems with different
characteristics. In this case, the NMS provides three potential
means to evacuate the channel based on th e urgency of the
request and the type of the used base stations, as depicted in
Fig. 8. The fast evacuation type was used in the present
measurements, and it is needed, e.g., in the case of ESC alerts.
In fast evacuation, the CBSD locks the affected cells, and the
terminals will automatically start the cell reselection
procedure. In th e case of informin g incumbents, graceful
shutdown can be used when the MNO knows well beforehand
the need for evacuation. Graceful shutdown lowers the RX
power in the base station in small steps, and the terminals
initiate handover to a neighbor cell based on the handover
trigger levels. The latest versions of base station s provide an
option to change frequency “on the fly” with radio on. In this
case the terminal will start the cell reselection automatically.
As can be seen from the results, the duration of the
configuration command time varies between 78 and
103 seconds (Table I). This is mainly due to the overall load of
the shared operational NMS used in the field trial.
B. Relation to FCC limits
How do these results relate to the FCC time limits? The
FCC rules allow up to 300 seconds after the ESC
communicates that it has detected a federal incumbent user for
the SAS to confirm suspension of any CBSD in the band. The
FCC has not specified how the 300 seconds can be divided for
the SAS to process messages, communicate with CBSDs, etc.
That is up to each implementation. Furthermore, according to
the standard, the SAS cannot start the communication with the
CBSD since the standard allows the CBSD to open an IP port
for communication only when it sends a heartbeat request.
Thus, the SAS must wait for the next heartbeat request to be
able to communicate back to the CBSD.
The current requirement is that CBSDs must comply with
SAS commands within 60 seconds. The SAS can notify the
CBSD of incumbent use or needs only in the next heartbeat
response. From that, the CBSD has 60 seconds to process the
message before responding back. This may increase the
response time of the SAS, which must be able to confirm
within five minutes that the CBSD has vacated the relevant
spectrum.
Based on our measurements, we can confirm that the field
trial operates well and fulfills the above requirements. The
achieved evacuation time totals around 90 seconds when using
the base station locking procedure, which means that the band
is cleared well before the required five-minute time limit. The
total reconfiguration time, including the frequency change and
continuing operation in a new channel, takes at most four
minutes based on our measurements.
Our measurements were performed using a small-scale
commercial LTE system but with NMS being in shared use.
For large-scale networks with densified small cell installation,
the SAS scalability could be an issue, for example, when the
entire network needs to evacuate. In MNO use today, NMS is
operating very large networks and is capable of performing
parallel operations on several base stations at the same time.
Anoth er possible solution for con trolling large-scale
evacuations is an emergency evacuation type of operation in
NMS, where the evacuation takes place according to a pre-
defined emergency plan, which shuts down mobile networks
in two to three minutes. By defining a similar emergency plan
for a network in an SAS band, the DP could utilize this feature
to evacuate all CBSDs, if needed.
C. Techniques to reduce latency
There are a couple of meth ods that could be used to reduce
delays in the evacuation and frequency change process
compared to the field trial, which fulfils the current
specification requirements [13]–[15]. Since the SAS must wait
for a heartbeat message befor e it can initiate the evacuation
process, there is an initial waiting time totaling on average half
of the heartbeat interval th. If the SAS could initiate the
procedure immediately after receiving an ESC alert,
th/2 seconds would be saved in the process. For example, if th
is 10 seconds, the time saved in the evacuation would be
5 seconds.
Anoth er, and more remarkable, improvemen t could be
achieved in the frequency change process by enabling
frequency chan ge “on the fly” without a need to turn off ba se
Figure 8. Evacuation types on base station.
stations when the band is cleared. This could be done if other
channels were available. That would mean that at time instant
T5, the CBSDs would immediately send their grant request.
Then, after receivin g a heartbeat response, the DP could
switch the frequency on-air to the base station. This would
shave off more than a minute from both the evacuation and the
frequency change procedure. The total time reduction would
roughly amount to two minutes, which means an almost 70%
reduction in the reconfiguration time.
V. CONCLUSIONS
A key to the success of 5G, potentially implemented across
multiple industry verticals, lies in forward-looking spectrum
policies and unlocking new spectrum assets. In this paper, we
have concentrated on the 3.5 GHz band by implementing a
CBRS concept governed by an SAS in a live network using
standardized messages and protocols. This is the first work to
analyze time domain operation in a practical SAS based CBRS
system, focusing especially on the evacuation time, which is
one of the most important indicators of the feasibility and
dynamicity of the sharing system.
The trial highlights certain concerns regarding the ability
to meet the FCC’s 60-second requirement concerning the SAS
power down directive, when driven through the NMS and
SON. We analyzed how much time is needed for each process
phase and what are the most time-consuming phases in the
evacuation and frequency change process. The results show
that most time is spent on configuring the base stations by the
network management system. Th e standard process could be
improved by enabling frequency change “on th e fly” with out a
need to lock the base stations during the evacuation process.
This would reduce the total reconfiguration time to less than
half of the original.
To enable faster and more dynamic frequency change, the
base stations should be optimized to support fast on-air
frequency changes. Time scales, aggregate interference studies
and allocation algorithms could be studied further in large-
scale trials including, for example, hundreds of base stations
and advanced allocation algorithms.
VI. ACKNOWLEDGMENTS
The CORE++ research project [4] is funded by the 5thGear
program of the Finnish Funding Agency for Technology and
Innovation (Tekes) and the project consortium: VTT Technical
Research Centre of Finland, University of Oulu, Centria
University of Applied Sciences, Fairspectrum, Turku
University of Applied Sciences, Nokia, PehuTec, Bittium,
KeySight, Finnish Defence Forces. The project is also
supported by the Finnish Communications Regulatory
Authority.
REFERENCES
[1] J. Mitola III and G. Q. Maguire, Jr., “Cognitive radio: Making software
radios more personal,” IEEE Personal Communications, vol. 6, pp. 13–
18, Aug. 1999.
[2] S. M. Dudley et al., “Practical issues for spectrum management with
cognitive radios,” Proceedings of the IEEE, vol. 102, pp. 242–264, Mar.
2014.
[3] S. Sagari, S. Baysting, D. Saha, I. Seskar, W. Trappe and D.
Raychaudhuri, “Coordinated dynamic spectrum management of LTE-U
and Wi-Fi networks,” in Proc. DySPAN, pp. 209–220, Sep.–Oct. 2015.
[4] ECC report 205, “Licensed shared access,” Feb. 2014.
[5] M. Palola et al., “Live field trial of licensed shared access (LSA) concept
using LTE network in 2.3 GHz band,” in Proc. DySPAN, pp. 38–47,
Apr. 2014.
[6] M. M. Sohul, Y. Miao, T. Yang and J. H. Reed, “Spectrum access
system for the citizen broadba nd radio service,” IEEE Communications
Magazine, vol. 53, pp. 18–25, Jul. 2015.
[7] M. Mustonen, M. Matinmikko, D. Roberson and S. Yrjölä, “Evaluation
of recent spectrum sharing models from the regulatory point of view,” in
Proc. 5GU, pp. 11–16, November 2014.
[8] M. Höyhtyä, “Sharing FSS satellite C band with se condary smal l cells
and D2D communications,” in Proc. ICC workshop, pp. 1606–1611,
June 2015.
[9] FCC GN Docket No. 12-354, “Amendment of the Commission’s rules
with regard to Commercial Operations in the 3550-3650 MHz band,”
May 2016.
[10] The Office of the Federal Register (OFR) and the Government
Publishing Office. Electronic Code of Federal Regulations. Title 47:
Telecommunication, Part 96—Citizens Broadband Radio Service. URL:
http://www.ecfr. gov/cgi-bin/text-idx?node=pt47.5.96&rgn=div5
[11] M. B. H. Weiss, W. H. Lehr, A. Acker and M. M. Gomez, “Socio-
technical considerations for spectrum access system (SAS) design,” in
Proc. DySPAN, pp. 35–46, Sep.–Oct. 2015.
[12] C. W. Kim, J. Ryoo and M. M. Buddhikot, “Design and implementation
of an end-to-end architecture for 3.5 GHz shared spectrum,” in Proc.
DySPAN, pp. 23–34, Sep.-Oct. 2015.
[13] Wireless Innovati on Forum, “SAS to CBSD protocol technical r eport-
B,” version. V0.5.0, February 2016.
[14] Wireless Innovation Forum, “Interim SAS to CBSD Protocol Technical
Report A,” Version 1.0.0, October 2015.
[15] Wireless Innovation Forum, “SAS to User Protocol Technical
Specification,” July 2016.
[16] CORE++ project web site. URL: http://cor e.willab.fi
[17] Wireless Innovation Forum, “CBRS Operational and Functional
Requirements,” WINNF-15-S-0112-V1.0.0 Sept 2016.
[18] Nokia Eden-NET SON solution. URL:
http://networks.nokia.com/portfoli o/solutions/eden-net
[19] Field Trial of Citizens Br oadband Radio Service (CBRS),
Demonstration in European Wireless, May 2016.
[20] Field Trial of Citizens Broadband Radio Service (CBRS)/Spectrum
Access System (SAS), Demonstration in CrownCom, May–June 2016.
