This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2016.2600759, IEEE
Transactions on Power Delivery
1

Design of a Time Synchronization System based on

GPS and IEEE 1588 for Transmission Substations
Hao Guo, Student Member, IEEE, and Peter Crossley, Member, IEEE

 Abstract—Accurate timing is often required for the Intelligent time signal to the IEDs, in the format of one-pulse-per-second
Electronic Devices (IEDs) used in transmission substations. A (1-PPS) and/or the IRIG-B time code, via dedicated fibre optic
common method of achieving this is direct connection of a device cables [2]–[3]. Considering the significant number of IEDs
to a local GPS receiver and the use of its 1-PPS synchronizing requiring a time source and that a single fibre optic cable may
signal and the IRIG-B coded message. However, concerns about only carry time information to one device, it is inevitable that
GPS reliability are encouraging the use of timing systems less dozens of distributed GPS receivers are deployed within a
dependent on the direct use of local GPS receivers. IEEE 1588 power substation [4]. Many utilities and timing experts have
protocol is a network based time synchronization technique raised concern about the reliability of local, inexpensive GPS
designed to co-exist with IEC 61850 applications and deliver sub
receivers and a number of protective relay mal-operation
micro-second timing accuracy. Many utilities are now considering
caused by incorrect timing data obtained from GPS receivers
the adoption of IEEE 1588, but they need confidence in the
reliability of this technology before it can be rolled out to real
have been reported [4]-[5]. In addition, GPS jamming [6] also
substations. Hence, comprehensive tests were undertaken on an poses a threat to the continuous availability of GPS in a
IEEE 1588 timing system, to help gain insight into the limitations substation and this severely affects the use of synchronized data
of the system. for automation control, especially when information from
This paper presents a procedure to assess the performance of multiple IEDs, each connected to its own GPS receiver, is
a timing system based on distributed GPS receivers and one based required [7]. Natural interference (e.g. Solar Flares) is also
on a mixture of GPS receivers and IEEE 1588 devices. Test results considered to have negative effects on the GPS signal reception
indicate whichever system is selected, high quality devices and [8] and a GPS satellite anomaly [9] interrupted the 1-PPS
systems, with appropriate installation and engineering, are output on almost all the GPS receivers used in several
essential to satisfy the stringent ±1 µs accuracy requirements substations in Japan. For obvious reasons, there is no legitimate
needed by critical IED applications. commercially available GPS spoofer, although researchers
have built their own and successfully mislead critical devices
Index Terms—Time Synchronization, Global Positioning that rely on a GPS signal, such as a Phasor Measurement Unit
System, IEEE 1588, Performance Evaluation, Ethernet Networks, (PMU) [10]. As a consequence of these fears and greater
IEC 61850, Protection and Control, Power Transmission,
awareness of threats to critical infrastructure, a time
Substation Automation
synchronization system that uses fewer, but more reliable, GPS
receivers is preferred by utilities.
I. INTRODUCTION
Synchronization based on the IEEE 1588-2008 standard

A substation is a critical part of a power transmission

network and its associated Protection and Control (P&C)
systems or intelligent electronic devices (IEDs) limits the
[11] can be implemented over Ethernet and realizes sub
micro-second accuracy when 1588 compliant hardware is used.
A substation now requires two or three 1588 master clocks in
damage caused by a fault, allows switching operations to occur conjunction with a data network, instead of using a local GPS
and in extreme situations prevents the power grid from receiver with each IED [12]. The 1588 traffic shares the
collapsing. In general, an IED measures the local voltage and Ethernet with IEC 61850 Sampled Value (SV) [13] and Generic
current signals, and when multiple IEDs are synchronized to a Object Oriented Substation Event (GOOSE) [14] applications,
common time reference, they can be used to reflect the system which allows the use of a unified network consisting of 1588
state in real time [1]. This requirement for time synchronization clocks, 61850 Merging Units (MUs) and IEDs [12] [15]. In the
is becoming increasingly important, especially when the future, substation timing or synchronizing systems will fall into
network operators require knowledge of the actual system state two main categories [12] as shown in Fig. 1(a) and 1(b). The
in real time, or feeder protection is based on differential former consists of a network of point-to-point connections that
protection. carry 1-PPS / IRIG-B information to a small number of 61850
Local Global Positioning System (GPS) receivers are devices. Whilst the latter involves a large number of 61850
widely used for accurate timing in the power industry due to devices and a network of 1588 compliant Ethernet switches
their low cost and accessibility. Having received and decoded shared by IEDs, MUs and 1588 slaves that convert 1588 to
the time information from satellites, a GPS receiver feeds the 1-PPS / IRIG-B.
The performance of the IEEE 1588 technology, in terms of
1588 device characteristics, were assessed in [16] and the
This work was supported by The University of Manchester, National Grid
and the Engineering and Physical Sciences Research Council (EPSRC) Centre
impact of different data networks conditions analyzed in [3],
for Doctoral Training (CDT) in Power Networks project. [17]-[18]. When 1588 timing is needed within a harsh
The authors are with the School of Electrical and Electronic Engineering, substation environment, it must be both reliable and sufficiently
The University of Manchester, Manchester, M60 1QD, UK (e-mail: robust so that critical automation devices relying on
[email protected]; [email protected]).

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2016.2600759, IEEE
Transactions on Power Delivery
2

Wiring Legend Section III presents and discusses the test results, and Section
(a) 1-PPS / IRIG-B IV describes the conclusions.
over Fibre Optic

Control / II. EXPERIMENT SETUP AND TEST METHODS

Communication Room Outdoor Cubicle
To assess the timing accuracy, the 1-PPS output from
GPS Receiver 1 MU 1 different timing devices are directly compared, as discussed
previously in [2]-[3], [20]. However, instead of using an
MU 2
oscilloscope to measure the time difference, a Linux server
with 22 peripheral 1-PPS input channels is employed as shown
in Fig. 2. The server can simultaneously compare the time of
each 1-PPS signal with the reference 1-PPS; the pulse delay
Bay
values are plotted in real time and then recorded in a text file for
IED 1 later analysis. A 24 hour measurement file that contains 86,400
records for each input channel, consumes 1.7 MB of disk space;
hence, the long term accuracy of a timing device can be easily
GPS Receiver 2 IED N
monitored and analyzed. The tests are performed in a
laboratory without air conditioning and the ambient
temperature varies between 10oC and 20oC over the test period.
Wiring Legend
(b) Devices related to timing are equipped with high quality
Ethernet with 1588
internal oscillator such as Over Controlled Crystal Oscillator
1-PPS / IRIG-B and Temperature Controlled Crystal Oscillator. Hence, the
over Fibre Optic
temperature variation should not affect the timing accuracy.
Control /
Communication Room Outdoor Cubicle A. Accuracy Validation of Measurement Server
GPS Receiver 1 1588 Switch 1588 Delay measurements are made by the server; which requires
Compliant MU
(1588 Master 1) the validation of the measurement accuracy. Hence, the 1-PPS
output of a GPS receiver is split using a T type BNC connector.
1588 Slave non-1588 MU As the BNC connector is a passive component, it can be
regarded as a short electrical cable accepting the input 1-PPS
1588 Switch and routing out the 1-PPS via two copper cores within itself.
Therefore, the delay between the two output 1-PPS is
Bay
negligible. The output 1-PPS is then connected to the reference
1588 Switch 1588 port and the input port using BNC cables of identical length.
Compliant IED
GPS Receiver 2 The validation procedure for each channel is run for 1000
(1588 Master 2) seconds, i.e. 1000 pulse delay values are recorded. Ideally, the
1588 Slave non-1588 IED pulse delay should be 0 ns since the same 1-PPS is compared
and the propagation delay from the GPS receiver to the server is
identical (all 1-PPS wiring is carried out using BNC connectors
Fig. 1. Timing System based on (a) local GPS receivers and (b) a mixture of
GPS receivers and IEEE 1588 devices.
and cables and there is no electrical signal conversion on the
path between the GPS receiver and the measurement server).
synchronization will not mal-operate. The IEEE 1588 Power However, measurement error is inevitable and as illustrated in
Profile is defined by the IEEE C37.238-2011 standard [19] for Table I, the error caused by the server is < ±7 ns, which is
the power system applications and specifies a number of fixed negligible as compared to the pulse delay accuracy
features so that interoperability and predictable performance requirements of a substation.
can be obtained. Ingram et al. conducted comprehensive tests B. Requirement for Synchronization Accuracy
on a network using the 1588 Power Profile in conjunction with
61850 SV and GOOSE applications; the results were used to When designing a timing system, applications requiring a
assess the timing performance in terms of the 1-PPS delay, time reference have to be identified and the associated accuracy
message compensation accuracy and system reliability [2] requirements determined. The de facto implementation of SV is
[20]-[21]. specified by the 61850-9-2 Light Edition [24] and the required
Synchronization test results on the GPS receivers used in accuracy of a synchronization source (i.e. devices providing
substations are rarely published, despite the usual “quoted 1-PPS) is < ± 1 µs. Authors in [22] [25] also suggest ±1 µs
accuracy” of ±100 ns. This paper, will evaluate the long term timing accuracy for Phasor Measurement Units (PMUs).
accuracy of various GPS receivers and their transient behavior Consequently, ± 1 µs is the performance baseline used in this
during satellite signal loss and restoration, and extend the work paper.
described in [2] and [20], by considering the effect of:- network C. Long Term Synchronization Accuracy of GPS Receivers
topologies, excessive 1588 traffic, 1588 packet loss,
The long term timing accuracy of three different GPS
communication link loss and complete GPS signal loss. The
receivers “B, C and D” is assessed by measuring the 1-PPS
test setups and methods are described in Section II. Whilst
delay over a 24 hour period for each of the receivers, using the

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2016.2600759, IEEE
Transactions on Power Delivery
3

Wiring Legend Wiring Legend

Measurement Measurement Server Link Loss
Server Copper 1-PPS (reference)
Copper 1-PPS (Reference)
Copper 1-PPS (input)
Copper 1-PPS (Input) Ethernet with 1588
GPS Receiver B Background Traffic

GPS Receiver A
(Reference Clock) GPS Receiver C
GPS Receiver A
(Primary 1588 Master) 1588 Slave X

1588 Switch 1 1588 Switch 2 1588 Switch 3

GPS Receiver D

GPS Receiver B
Fig. 2. Experiment Setup for Synchronization Assessment of GPS Receivers. 1588 Slave Y
(Backup 1588 Master)
Traffic Generator
TABLE I 1010101010101010
ERROR OF MEASUREMENT SERVER
Fig. 3. Synchronization Assessment of 1588 Slaves using Star Topology.
Average Standard Deviation Maximum Minimum
1.392 ns 2.1 ns 7 ns -6 ns 1588
GPS Receiver A 1588 Switch 1 1588 Switch 2 1588 Switch 3
Slave X
(Primary 1588 Master)
experimental setup in Fig. 2. The mask angle in the
configuration of Receiver B can be raised to ensure RSTP Ring
1588
low-elevation satellites seen by the antenna can be excluded GPS Receiver B 1588 Switch 4 1588 Switch 5
Slave Y
(Backup 1588 Master)
[24]; this reduces the timing error due to signal multipath
propagation. To investigate this effect, the 24 hour pulse delay (a)
measurement of GPS Receiver B is repeated at different mask 1588
1588 Switch 1
angle configurations (5o, 15o and 25o). GPS Receiver A Slave X
(Primary 1588 Master)
D. Transient Behavior of GPS Receivers PRP RedBox 1 PRP LAN A PRP RedBox 2

When using GPS receivers, it is possible for the GPS signal PRP LAN B
GPS Receiver B
1588 Switch 2 1588 Switch 3
to disappear and recover later. Previous research has indicated (Backup 1588 Master) 1588
Slave Y
the 1-PPS of certain GPS receivers were mistimed during GPS
re-synchronization [2] [4]. This emphasizes the importance of (b)
assessing transient behavior of receivers during GPS loss and 1588
1588 Switch 1
restoration. This involves disconnecting and re-connecting the GPS Receiver A Slave X
(Primary 1588 Master)
GPS antenna and monitoring the effect on the pulse delay using PRP RedBox 1 PRP LAN A PRP RedBox 2
the setup in Fig. 2.
GPS Receiver B PRP LAN B
E. Long Term Synchronization Accuracy of 1588 Slaves (Backup 1588 Master) 1588
1588 Switch 2 1588 Switch 3 1588 Switch 4
Since Receiver A has a better internal oscillator than Slave Y

Receiver B, the IEEE 1588 Best Master Clock Algorithm RSTP Ring
automatically selects Receiver A as the primary Master when 1588 Switch 5 1588 Switch 6 1588 Switch 7

all configurable parameters are set to the same default values

specified by the IEEE 1588 Power Profile. The long term (c)
timing accuracy of two 1588 slaves were assessed by
1588
comparing their 1-PPS to the reference 1-PPS (from GPS GPS Receiver A
HSR RedBox 2
Slave X
Receiver A) over a 24 hour monitoring period, as indicated in (Primary 1588 Master)

Fig. 3. HSR RedBox 1

HSR Ring
HSR RedBox 3

To provide a benchmark for the 1588 synchronization GPS Receiver B

assessment, Receiver A was directly connected to each of the (Backup 1588 Master)
HSR RedBox 4 HSR RedBox 5 1588
1588 Slaves. In real applications, various communication Slave Y

architectures are deployed between the GPS receivers and the (d)
1588 slaves. The five topologies commonly used in substation Fig. 4. Network Topologies used between GPS Receivers and 1588 Slaves.
automation systems are assessed in this paper, i.e. Star
connection, Rapid Spanning Tree Protocol (RSTP) [27] ring, manufacturer and each is configured as a Transparent Clock
Parallel Redundancy Protocol (PRP) [28] network, (TC) with the Peer-to-Peer mechanism as specified by the 1588
High-available Seamless Ring (HSR) [28] and combined Power Profile. Receiver B, with a 25o mask angle, is configured
PRP/RSTP; see Fig. 3 and Fig. 4. The link between Switch 3 as the backup 1588 master clock if Receiver A degrades or
and 5 in Fig. 4 (a) and the link between Switch 5 and 6 in Fig. 4 fails. The rate of all IEEE 1588 messages including Announce,
(c), denoted by dotted lines, is configured as the backup link Sync, Follow_Up and Pdelay_Req is set as one packet per
and will be activated if there is a network fault. second according to the Power Profile. In addition, the Power
Each 1588 device uses a two-step operation mode where the Profile also specifies the worst-case timing error, this should
1588 Sync and Follow_Up messages appear together. not exceed ±1 µs when the network load occupies up to 80% of
Furthermore, all the Ethernet switches are from the same the total bandwidth [21]. To investigate the impact of

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2016.2600759, IEEE
Transactions on Power Delivery
4

background traffic, the delay measurement is conducted over HSR RedBox 1 HSR 1588
GPS Receiver B
24 hours without and with traffic injection. Since the 1588 (Primary 1588 Master) Ring Slave X
synchronization shares the data network with 61850 SV and
GOOSE applications, a significant portion of the network load 1588
will be multicast traffic. The 61850 standards propose the use GPS Receiver A Slave Y
(1588 Traffic Generator)
1588 Traffic
of IEEE 802.1Q tagging for traffic prioritization, with a default
value of 4 [14] [15]. Hence, a traffic generator is utilized to Fig. 5. Injection of 1588 Traffic in HSR Ring Topology.
inject 80% multicast non-1588 traffic with priority 4 at the
Original Impaired
leftmost Ethernet switch in all the topologies and the 1588 Sync/Follow_Up Sync/Follow_Up
traffic uses priority 4 as defined in the 1588 Power Profile. All Stream Stream 1588
Traffic 1588 Switch Slave X
Ethernet switches in the testbed can be configured as RSTP GPS Receiver A Impairment
(1588 Master) Emulator
switch, PRP RedBox or HSR RedBox. When the hardware 1588
testbed is constructed, the PRP/HSR RedBox can only support Slave Y
data rate up to 100 Mb/s. Therefore, the bandwidth of each Fig. 6. Emulation of 1588 Message Loss.
Ethernet switch is 100 Mb/s. It is reasonable to conduct
experiments on a 100 Mb/s testbed as it has smaller traffic using the traffic impairment emulator, connected between
throughput capability, which means IEEE 15888 traffic may Receiver A and the slaves; see Fig. 6. The 1-PPS of both 1588
experience longer latency and more contingencies. This is slaves is compared with the 1-PPS of Receiver A.
useful to identify the limit of IEEE 1588 timing. H. 1588 Synchronization during Communication Link Loss
F. 1588 Synchronization under Excessive Network Load Communication link loss often occurs in substations, which
If a 1588 timing network is not carefully designed and requires P&C systems to use communication links with an
engineered, excessive traffic could appear which may adversely appropriate level of redundancy. Following a link loss, the
affect synchronization accuracy and result in mal-operation of RSTP ring topology, which is widely applied in substations,
critical protection and control applications, such as feeder can restore communications within 50 to 2000 ms [3] [27]. In
differential protection or PMU based wide area control. comparison, a PRP LAN or a HSR ring delivers seamless
Consequently, it is essential to overload the network to verify redundancy (0 s recovery time), but at present, these are not
the effect on the timing performance. Two types of network widely used in substations. Prior to the deployment of 1588
load are injected into the HSR topology during the tests, namely timing in substations, it is crucial to assess the impact of a link
non-1588 traffic and 1588 traffic. For non-1588 traffic, the loss on synchronization, and to ensure the network
traffic generator injects into the leftmost Ethernet switch:- architecture(s) can build a robust 1588 system. Link loss, as
100%, 160% and 200% multicast traffic with priority 4 for denoted by the red cross, is introduced to all the topologies
1000 seconds. described in Fig. 3 and Fig. 4, and 80% background traffic is
With regard to 1588 traffic, the setup is shown in Fig. 5, injected during the link loss to produce more realistic results.
where Slaves X and Y are synchronized by Receiver B, whilst During this test, the 1-PPS of Slave X and Y is compared with
Receiver A with two 1588 ports operates as the traffic generator the pulse signal of Receiver A using the measurement server.
and provides up to 128 Sync and Follow_Up messages per I. Impact of Complete GPS Signal Loss on 1588 Slaves
second. The Best Master Clock Algorithm (BMCA) defined in
For the system shown in Fig. 7, degradation of the time
1588 can select the best clock in the network as the primary
quality of the Primary 1588 Master, will force the Slaves to
master, whilst the other potential master(s) enters the “Passive”
track the Backup Master which now has a higher time quality
state and does not send 1588 Sync and Follow_Up messages.
than the Primary Master. Assuming, there is minimal offset
To force Receiver A to inject 1588 traffic when it is not the
between the internal oscillators of both masters, the timing
primary master, it is configured to accept only 1588 Announce
accuracy of the Slaves will not be significantly affected [21].
packets with 802.1Q VLAN tag. In the HSR RedBox 1, the
However, in an extreme situation where GPS is lost for all 1588
VLAN tags of all the 1588 packets transmitted to Receiver A
masters, the impact on the 1588 slaves needs to be investigated.
are set to be stripped. Consequently, Receiver A does not know
A star topology with 80% non-1588 background traffic, as
Receiver B is a better clock and regards itself as the primary
shown in Fig. 7 is used and the GPS antennas is first
master, sending out Sync and Follow_Up packets. The
disconnected from Receiver A and then Receiver B. Note:
capability of Slave X and Y in recovering from excessive 1588
Receiver A loses the GPS signal during this test and thus
traffic is also tested by stopping the output from Receiver A.
Receiver C which is locked to GPS is used in this test as the
Note: 1-PPS of Slave X and Y is compared to Receiver A
reference for pulse delay measurement. The 1-PPS of Receiver
during non-1588 traffic injection and to Receiver B when
A, Receiver B, Slave X and Slave Y is now compared to
injecting 1588 traffic.
Receiver C even though it does not operate as the IEEE 1588
G. Impact of 1588 Message Loss on 1588 Slaves master clock.
1588 slaves rely on the receipt of Sync and Follow_Up
messages for time synchronization. However, Ethernet frames III. TEST RESULTS AND DISCUSSIONS
can occasionally be discarded due to network contingencies. A. Long Term Synchronization Accuracy of GPS Receivers
Hence, it is important to assess how 1588 slaves will respond to
The 24 hour 1-PPS delay measurement for GPS Receivers
the loss of Sync and Follow_Up messages, which is achieved
B, C and D are plotted in Fig. 8 and Fig. 9; with detailed

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2016.2600759, IEEE
Transactions on Power Delivery
5

Wiring Legend
High Wall to the South
Copper 1-PPS (reference) Copper 1-PPS (input) GPS Antenna
Disconnection of GPS Antennas
Copper 1-PPS (input) Background Traffic

Measurement Server

Fig. 10. Installation and Position of GPS Antennas.

GPS Receiver C
delay spikes and helps maintain the accuracy of Receiver B
within ±420 ns. Without a configurable mask angle, Receiver D
1
can only deliver timing accuracy in the range of ±880 ns, with a
GPS Receiver A
(Primary 1588 Master) 1588 Slave X standard deviation of 127 ns. When Receivers B and D are
1588 Switch 1 1588 Switch 2 1588 Switch 3
compared with Receiver C, they experience many delay spikes,
2 which are probably caused by GPS multipath propagation
GPS Receiver B resulting from a high wall to the South of the GPS antennas, as
1588 Slave Y
(Backup 1588 Master)
Traffic Generator
well as the close proximity of the antennas [28], as shown in
1010101010101010 Fig. 10. Further work is required to verify this observation; this
Fig. 7. Test Setup to Investigate Impact of Complete GPS Signal Loss. involves moving the GPS antennas to a location where they can
be installed far apart and with no obstruction to the sky.
In conclusion, Receiver C is the best for SV and phasor
15o Mask Angle 5o Mask Angle
measurements because of its high accuracy and immunity to
25o Mask Angle
multipath problems. Receiver B can also satisfy the ± 1 µs
requirement with an appropriate mask angle configuration, but
it may not be advisable to use Receiver D for accurate timing.
B. Transient Behavior of GPS Receivers
Test results for the transient behavior of GPS receivers are
shown in Fig. 11 and 12 and in all case the GPS signal loss
occurs at t=100 s. Note: all plots in terms of drifting rate and
step change are based on the worst case results.
(i) Loss of GPS Signal
Fig. 8. Long Term Synchronization Accuracy of GPS Receiver B. Fig. 11 shows Receiver C stops outputting 1-PPS once it
detects loss of the GPS signal, as illustrated by the blue line at
GPS Receiver D the zero axis. However, Receiver B and D continue to generate
1-PPS after GPS loss and exhibit drifting rates with diverse
signs and magnitudes. Receiver B usually drifts positively at an
approximate rate 2.22 ns/s as indicated by the orange curve.
Although rarer, Receiver B can also drift in the negative
direction at 1.61 ns/s, according to the amber plot. Hence, with
a worst case initial delay value of 622 ns, Receiver B reaches
the ±1 µs limit after 170 s. Similarly, Receiver D always drifts
GPS Receiver C negatively, as shown by the purple line, and its slope varies
between -1.25 ns/s and 0.57 ns/s. Receiver D breaks the ±1 µs
limit after 193 s, assuming a worst case initial delay value of
-758 ns. When considering these results, it is important to
Fig. 9. Long Term Synchronization Accuracy of GPS Receiver C and D.
recognize that most MUs and IEDs can detect the loss of the
TABLE II synchronization input and in turn inhibit operation [29] [30].
LONG TERM ACCURACY OF GPS RECEIVERS (VS. RECEIVER A) Therefore, Receiver C, which stops outputting 1-PPS when
GPS Receiver ̅
𝜶 𝝈 Range GPS is lost, is better than Receiver B and D, since relay
B with 5o Mask Angle 3.448 ns 30.9 ns -390 ns to 493 ns mal-operation can be avoided.
B with 15o Mask Angle 1.287 ns 26.2 ns -255 ns to 622 ns
B with 25o Mask Angle -8.036 ns 19.6 ns -129 ns to 411 ns (ii) Restoration of GPS Signal
C -10.532 ns 23.4 ns -128 ns to 121 ns Fig. 12 shows the effect of GPS restoration on receivers B,
D 80.906 ns 126.8 ns -758 ns to 873 ns C and D. When B regains GPS, the 1-PPS delay suddenly
increases from 200 ns to -3,800 ns, which is not acceptable for
statistics of timing accuracy summarized in Table II. The most types of IEDs. In comparison, the delay change is
average delay value is denoted as 𝛼̅ whilst the standard negligible when C re-synchronizes to GPS, and small (<400 ns)
deviation is denoted as 𝜎 in Table II. when D re-synchronizes. Consequently, Receivers C and D can
Test results indicate Receiver C has the most stable 1-PPS guarantee correct operation of MUs, IEDs and PMUs.
output and can deliver timing accuracy better than ±150 ns. Fig.
8 suggests increased mask angle reduces the occurrence of

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2016.2600759, IEEE
Transactions on Power Delivery
6

TABLE III
LONG TERM SYNCHRONIZATION ACCURACY OF SLAVE X (VS. RECEIVER A)
GPS Signal Loss Receiver B Topology and Traffic ̅
𝒙 𝝈 Range
from t = 100 s Drift 1
Direct -8.688 ns 8.1 ns -32 ns to 22 ns
Receiver D
Drift 2 Star without Traffic -11.482 ns 29.6 ns -107 ns to 73 ns
Star with 80% Traffic -9.401 ns 38.1 ns -125 ns to 113 ns
Receiver C RSTP without Traffic 7.069 ns 32.0 ns -87 ns to 113 ns
RSTP with 80% Traffic 5.46 ns 29.6 ns -83 ns to 88 ns
Receiver D
PRP without Traffic 11.998 ns 23.6 ns -63 ns to 90 ns
Drift 2 PRP with 80% Traffic 11.32 ns 27.9 ns -67 ns to 98 ns
Receiver D
Drift 1 Combined PRP/RSTP
14.891 ns 22.6 ns -60 ns to 83 ns
without Traffic
Combined PRP/RSTP
15.497 ns 23.8 ns -60 ns to 93 ns
with 80% Traffic
Fig. 11. Impact of GPS Loss on GPS Receiver B, C and D.
HSR without Traffic 29.412 ns 18.9 ns -38 ns to 95 ns
HSR with 80% Traffic 30.024 ns 18.4 ns -39 ns to 82 ns

TABLE IV
LONG TERM SYNCHRONIZATION ACCURACY OF SLAVE Y (VS. RECEIVER A)
Topology and Traffic ̅
𝒚 𝝈 Range
Direct -13.047 ns 9.5 ns -43 ns to 16 ns
Receiver D Receiver C Star without Traffic -14.959 ns 18.0 ns -66 ns to 46 ns
Star with 80% Traffic -10.324 ns 27.8 ns -93 ns to 74 ns
RSTP without Traffic 4.761 ns 21.1 ns -60 ns to 68 ns
GPS Signal Restoration
from t = 100 s Receiver B RSTP with 80% Traffic 0.501 ns 18.6 ns -56 ns to 71 ns
PRP without Traffic 9.769 ns 15.6 ns -46 ns to 63 ns
PRP with 80% Traffic 11.191 ns 18.5 ns -53 ns to 74 ns
Combined PRP/RSTP
13.939 ns 15.8 ns -42 ns to 69 ns
without Traffic
Fig. 12. Impact of GPS Restoration on GPS Receiver B, C and D. Combined PRP/RSTP
15.744 ns 17.7 ns -38 ns to 71 ns
with 80% Traffic
The transient behavior of Receiver B indicates it may HSR without Traffic 62.149 ns 15.2 ns 0 ns to 122 ns
severely break the ±1 µs requirement during GPS restoration HSR with 80% Traffic 63.166 ns 14.8 ns 14 ns to 117 ns
and is only able to maintain micro-second accuracy for about
500 s after the loss of GPS. Similarly, D can only maintain ±1 flow; this involves synchronizing to the upstream Master and
µs accuracy for  700 s following loss of GPS, but it does work disciplining the downstream Slaves. Multicast traffic does not
correctly during GPS resynchronization. Receiver C provides greatly affect the timing performance but the primary 1588
accurate timing during GPS restoration and stops outputting master (Receiver A) will stop working when > 25 Mb/s traffic
1-PPS when it detects loss of GPS. appears on its port. Hence, multicast traffic not needed by the
master should be filtered out using the MAC address and/or the
C. Long Term Synchronization Accuracy of 1588 Slaves VLAN setting in the Ethernet switch.
Timing accuracy varies from slave to slave and may be
affected by different network topology and load conditions. D. 1588 Synchronization under Excess Network Load
Statistics of 1-PPS delay of Slave X compared to Receiver A is (i) Non-1588 Traffic
summarized in Table III whilst details of Slave Y are covered in Timing accuracy of Slave X and Y during excessive level of
Table IV. non-1588 multicast traffic is listed in Table V. Although the
In general, Slave X and Slave Y can both deliver accuracy network is extremely congested, both Slave X and Y obtain
much better than ±1 µs (i.e. Slave X = ±125 ns; Slave Y = ±122 accuracy < ±120 ns. Theoretically, when a large amount of
ns) even when 80% of the bandwidth is occupied by other traffic with the same priority competes for transmission, 1588
equal priority traffic. The introduction of 1588 Ethernet packets will experience additional delay. If the delay is not
switches will increase the standard deviation of the 1-PPS properly measured by 1588 switches, the accuracy could be
delay; they introduce errors when calculating the path delay and severely affected. Furthermore, excessive network traffic can
residence time, but the increase is negligible (<35 ns). The Star fill the switch buffer and 1588 packets may be discarded when
topology has the least impact on the timing accuracy, whilst the buffer is full, leading to unexpected mistiming. In fact, the
HSR topology has the greatest. This is because the HSR 1588 switches used during the experiment are able to preserve
RedBox connected to the 1588 master acts as a Boundary Clock the transmission of 1588 packets even when non-1588 traffic
(BC), even if it is configured as a TC. This can be illustrated by with higher priority is injected. Results in Table V also indicate
the fact that the source MAC addresses of all IEEE 1588 Ethernet switches accurately measure the delay experienced by
messages including Announce, Sync and Follow_Up originated 1588 packets even if the non-1588 load is excessive.
from Receiver A are the addresses of switches. In general, a
1588 TC will not change the source MAC address of 1588 (ii) 1588 Traffic
messages. Increased errors are present during the process where Fig. 13 shows Slave X and Y correctly synchronize to
a BC terminates the 1588 packets and regenerates the 1588 Receiver B within 100 s when Receiver A only injects 96 Sync

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2016.2600759, IEEE
Transactions on Power Delivery
7

TABLE V
ACCURACY OF SLAVES UNDER EXCESSIVE NON-1588 TRAFFIC (VS. RECEIVER A)
Slave and Traffic ̅
𝜶 𝝈 Range Sync Packet Loss Slave X; 99%
from t = 100 s Sync Loss
X, 100% Traffic 34.74 ns 18.8 ns -35 ns to 85 ns
X, 160% Traffic 30.985 ns 18.4 ns -35 ns to 79 ns Slave Y; 95%
Sync Loss
X, 200% Traffic 34.895 ns 19.1 ns -25 ns to 91 ns
Y, 100% Traffic 64.148 ns 16.1 ns 17 ns to 113 ns
Y, 160% Traffic 61.435 ns 14.9 ns 12 ns to 102 ns
Slave Y with New
Y, 200% Traffic 63.576 ns 16.1 ns 17 ns to 117 ns Firmware; 96% Sync Loss Slave X; 100%
Sync Loss
Slave Y; 96% Slave Y; 99%
Sync Loss Sync Loss
Excess 1588 Traffic
Stops from t = 820 s

Fig. 14. Impact of 1588 Sync Packet Loss, commencing from 100 s.
Excess 1588 Traffic
Injection from t = 100
s quickly even if it receives only one Sync message. When the
Slave Y Sync messages are completely not available, Slave X drifts at a
rate -0.368 ns/s. As mentioned in previous section, the worst
case initial error of Slave X can be ±125 ns, giving a margin of
-875 ns. Hence, Slave X will reach the ±1 µs after 2377 s. If
Slave X
there is a Sync packet available during this period (i.e. Sync
available rate = 1/2377 = 0.04%), Slave X recovers
synchronization again, so theoretically it can withstand a
Fig. 13. Impact of Excess 1588 Traffic on 1588 Timing. 99.96% Sync loss rate.
and Follow_Up messages per second. After 100 s, 128 Sync Slave Y, can maintain synchronization when 95% Sync
and Follow_Up packets are injected into the network every packets are missing but with an error up to -800 ns. When 96%
second and the slaves start to drift away. Sync packets are discarded; Slave Y loses synchronization
The Wireshark capture indicates the 1588 switches cannot within 100 s and shows an unusual drifting behavior, probably
process this many 1588 Sync packets and most of the Sync and due to its synchronization algorithm. After updating the
Follow_Up originated from Receiver B are discarded. Hence, firmware of Slave Y, the unusual drifting behavior upon 96%
the slaves keep losing synchronization and the clocks drift. Sync loss is resolved. More specifically, the pulse delay of
However, when some 1588 packets from Receiver B are not Slave Y fluctuates and this is because Slave Y occasionally
dropped, the slaves use these packets to adjust their internal receives Sync messages and adjusts its internal time. If 99%
oscillators, leading to the sudden decreases in pulse delay as Sync packets are dropped, Slave Y drifts at a faster rate (-0.947
shown in Fig. 13. Instead of using the VLAN setting to force ns/s) than Slave X; this is because Slave X includes a more
Receiver A to inject 1588 traffic when it is not the best clock, stable Oven Controlled Crystal Oscillator while Slave Y uses a
then if the Ethernet switch can forward 1588 packets with Temperature Controlled Crystal Oscillator. This also suggests
different domain numbers, Receiver A can also inject 1588 the excess 1588 traffic injected in the previous experiment can
packets and stress the network. Consequently, the 1588 cause at least 99% Sync packet loss.
network should be carefully configured to ensure it is not (ii) Loss of 1588 Follow_Up Packets
breached by excessive 1588 traffic resulting from
Similar to the scenario where Sync messages are dropped,
misconfiguration or a cyber-attack.
Slave X is not affected by a Follow_Up packet loss of up to
When excessive 1588 traffic stops flooding the network,
Slave X recovers in less than 10 s and there is a step change in 99%, as shown in Fig. 15. If the Follow_Up packets are not
the pulse delay from 200 ns to -200 ns, after which the pulse presenting at all, Slave X will start drifting at a rate similar to
that observed when the Sync packets are completely lost.
delay settles in the range between 0 ns to 100 ns within 20 s. In
Slave Y is very sensitive to the loss of Follow_Up packets
terms of Slave Y, it takes about 40 s to react and the pulse delay
drifts during this period. After that, the pulse delay will and the pulse delay will dramatically become larger than 80 s
decrease from more than 1000 ns to within ±100 ns and settles even when a single Follow_Up message is missing. Once the
in 30 s. Experimental results suggest both Slave X and Y can anomaly occurs, it takes about 200 s for Slave Y to recover,
recover synchronization without obvious timing spikes when which would increase the likelihood of relay mal-operation. If
excess 1588 traffic is removed, which can ensure the associated new firmware is used on Slave Y, its ability to tolerate
IEDs do not mal-operate. Follow_Up loss increases from a single packet loss to 62% loss.
However, when the loss rate exceeds 62%, an anomaly with a
E. Impact of 1588 Message Loss on 1588 Slaves smaller timing spike will occur and it will last for a much longer
(i) Loss of 1588 Sync Packets period 600 s. This can also easily result in relay mal-operation.
Experimental results demonstrate a loss rate lower than 99%
Worst case test results (from the perspective of drifting rate) will lead to the anomaly for the old firmware, which might be
with various Sync loss rates are plotted in Fig. 14 and the Sync caused by a bug in the synchronization algorithm. Using new
loss begins at t=100 s. Slave X can tolerate up to 99% Sync firmware, Slave Y can handle Follow_Up loss rate up to 62%,
message loss and can synchronize itself to the master very but an anomaly occurs when the loss rate is between 62% and

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2016.2600759, IEEE
Transactions on Power Delivery
8

Follow_Up Packet
Loss from t = 250 s
Slave Y; Star
Link Loss

Communication Link Slave X; PRP

Loss from t = 100 s Link Loss
Slave Y with New Firmware;
63% Follow_Up Loss
Slave Y;
1 Follow_Up Loss Slave X; 99% Follow_Up Loss
or
Slave Y with New Firmware; 62% Follow_Up Loss
Slave X; Star Link Loss

Fig. 15. Impact of 1588 Follow_Up Packet Loss.

Fig. 16. Impact of Communication Link Loss on 1588 Timing.
99%. When the loss rate is ≥ 99%, Slave Y (with old or new
firmware) exhibits a drifting rate similar to that observed when
≥ 99% Sync packets are absent. Hence, further firmware Receiver B
GPS Antenna Disconnected
updates are required for Slave Y to resolve the synchronization from Receiver B from t = 500 s
issue related to considerable loss (≥ 63%) of Follow_Up
messages. Receiver A

F. 1588 Synchronization during Communication Link Loss

The link loss in Fig. 16 occurs at t=100 s, and when Star Slave X
topology is used Slave X and Y drift away because no 1588 GPS Antenna Disconnected
packets from the master are available. When RSTP, HSR and from Receiver A from t = 100 s Slave Y

combined PRP/RSTP topologies are employed, link loss does

not affect the timing accuracy, which indicates the 1588
switches and slaves can automatically adjust themselves when Fig. 17. Pulse Delay between Reference Receiver C and Slaves upon Complete
swapping to a different communication path. In comparison, GPS Signal Loss.
link loss in PRP topology introduces considerable fluctuations
in pulse delay with a peak value of 50 ns, as indicated by the IV. CONCLUSION
green curve. However, in general, Slave X and Y are able to Experimental results described in this paper indicate that
achieve an accuracy significantly less than ±1 µs during the link not all GPS receivers can deliver ±100 ns accuracy, but
loss if communication redundancy is in place. increasing mask angle could further reduce the timing error. It
G. Impact of Complete GPS Signal Loss on 1588 Slaves is advisable that a GPS receiver does not generate any timing
output when it loses the GPS signal, this is necessary to ensure
Fig. 17 illustrates what happens when Slave X and Y are the associated IEDs can block operation to avoid
initially synchronized to Receiver A and the GPS antenna is mal-operation. Consequently, system operators do not need to
disconnected from Receiver A at t=200 s. After about 40 s, worry about the drifting rate of distributed GPS receivers. One
Receiver B notices that the performance of Receiver A has GPS receiver introduced unacceptable synchronization errors,
deteriorated and it takes the master role for the whole network. > 3 µs, upon GPS signal restoration and this has to be resolved
As a result, Slave X and Y start to follow Receiver B with a 200 to guarantee correct operation of secondary devices.
ns increase in their pulse delay. At t=500 s, the GPS antenna is A data network consisting of 1588 clocks and 1588 Ethernet
also disconnected from Receiver B; it continues to act as the switches was used to achieve synchronization accuracy better
master until t=540 s when Receiver A discovers it is a better than ±150 ns even when the network is heavily loaded. But this
clock and recovers the master role again. Sequentially, Slave X requires the use of 1588 Ethernet switches to accurately
again synchronizes to Receiver A, but Slave Y drifts away. measure the delay experienced by the 1588 packets and ensure
When Slave Y drifts, its port is in “Slave” state and according the 1588 messages are not dropped when they need to compete
to 1588, it should synchronize to Receiver A once the same pair for transmission. This supports the use of a unified IEC 61850
of Sync and Follow_Up packets is received. The reason the and IEEE 1588 network for future substation automation
algorithm of Slave Y refuses to synchronize with Receiver A is systems. However, such a network must be carefully designed
because the clock accuracy value of Receiver A is “unknown” and configured because considerable non-1588 traffic could
and it does not satisfy Slave Y’s requirement. “shut down” certain 1588 clocks, whilst excessive 1588 traffic
Receiver A is more stable than Receiver B when the GPS can exceed the processing ability of a 1588 switch, causing loss
signal is completely lost and a 1588 slave (e.g. Slave X) follows of synchronization. Communication redundancy techniques
a stable clock when there is no GPS signal, ensuring the correct can ensure 1588 timing is not affected during communication
operation of the IEDs. However, if a less stable clock becomes link loss. However, one 1588 slave could not properly handle
the master after a complete GPS loss, it may be advisable that a the loss of Follow_Up message and introduced timing errors >
1588 slave, i.e. Slave Y, does not follow the master and issues
80 s for about 200 s (when old firmware is used) or timing
an alarm so that the associated IEDs can block operation to
errors > 10 µs for 600 s (when new firmware is used); this must
avoid mal-operation.

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2016.2600759, IEEE
Transactions on Power Delivery
9

be fixed by a further firmware upgrade. 1588 slaves from IEC 61850 Process Bus,” IEEE Trans. Power Del., vol. 29, no. 3, pp.
1382-1389, Jun. 2014.
different vendors use diverse algorithms to deal with the
[16] L. Cosart, “Characterizing grandmaster, transparent and boundary clocks
situation where the master does not lock to the GPS – some still with a precision packet probe and packet metrics,” in Proc. 2011 IEEE
synchronize to the master whilst others will not. Therefore, the Int. Symp. Precis. Clock Synchr. Meas. Control Commun. (ISPCS),
stability of the 1588 master(s) and the behavior of 1588 slaves Munich, Germany, 12-16 Sep. 2011, pp. 56-61.
[17] R. Harada, A. Abdul and P. Wang, “Best Practice of Transporting PTPv2
during complete GPS loss must be investigated and understood
over RSTP networks,” in Proc. 2012 IEEE Int. Symp. Precis. Clock
to avoid mal-operation caused by timing drift. Synchr. Meas. Control Commun. (ISPCS), San Francisco, USA, 24-28
Time synchronization with stringent ±1 µs accuracy Sep. 2012, pp. 1-6.
requirement is important for digital substation applications and [18] R. Zarick, M. Hagen and R. Bartos, “The Impact of Network Latency on
the Synchronization of Real-World IEEE 1588-2008 Devices,” in Proc.
solutions based on direct use of local GPS receivers have been
2010 IEEE Int. Symp. Precis. Clock Synchr. Meas. Control Commun.
widely adopted. However, sophisticated tests must be carried (ISPCS), Portsmouth, USA, 27 Sep. – 01 Oct. 2010, pp. 135-140.
out to assess the long term accuracy and transient behavior of [19] IEEE Standard Profile for Use of IEEE 1588TM Precision Time Protocol
GPS receivers during incidents, before they can be considered in Power System Applications, IEEE Std C37.238-2011, Jul. 2011.
[20] D. M. E. Ingram, P. Schaub, D. A. Campbell and R. R. Taylor,
adequately reliable for transmission substations. The described
“Performance Analysis of PTP Components for IEC 61850 Process Bus
solution, based on GPS receivers and 1588 devices, is Applications,” IEEE Trans. Instrum. Meas., vol. 62, no. 4, pp. 710-719,
promising as it can be easily integrated into the IEC 61850 Apr. 2013.
process and station buses, and delivers accurate and reliable [21] D. M. E. Ingram, P. Schaub, D. A. Campbell and R. R. Taylor,
“Quantitative Assessment of Fault Tolerant Precision Timing for
time synchronization.
Electricity Substations,” IEEE Trans. Instrum. Meas., vol. 62, no. 10, pp.
2694-2703, Oct. 2013.
REFERENCES [22] UCA International Users Group. (2004, July 7). Implementation
Guideline for Digital Interface to Instrument Transformers using IEC
[1] E. Southern, “GPS Synchronized Current Differential Protection,” Ph.D.
61850-9-2. (R2-1) [Online]. Available:
dissertation, School of EEE, UMIST, 1998.
http://iec61850.ucaiug.org/Implementation%20Guidelines/DigIF_spec_
[2] D. M. E. Ingram, P. Schaub and D. A. Campbell, “Use of Precision Time
9-2LE_R2-1_040707-CB.pdf
Protocol to Synchronize Sampled-Value Process Bus,” IEEE Trans.
[23] F. Steinhauser, C. Riesch and M. Rudigier, “IEEE 1588 for Time
Instrum. Meas., vol. 61, no. 5, pp. 1173 – 1180, May 2012.
Synchronization of Devices in the Electric Power Industry,” in Proc.
[3] C. M. De Dominicis, P. Ferrari, A. Flammini, S. Rinaldi and M.
2010 IEEE Int. Symp. Precis. Clock Synchr. Meas. Control Commun.
Quarantelli, “On the Use of IEEE 1588 in Existing IEC 61850-Based
(ISPCS), Portsmouth, USA, 27 Sep. – 01 Oct. 2010, pp. 1-6.
SASs: Current Behavior and Future Challenges,” IEEE Trans. Instrum.
[24] L. Heng, T. Walter, P. Enge and G. X. Gao, “GNSS Multipath and
Meas., vol. 60, no. 9, pp. 3070-3081, Sep. 2011.
Jamming Mitigation using High-Mask-Angle Antennas and Multiple
[4] W. An, T. Nick, D. Barron, M. Bingham and A. Hackett, “A transmission
Constellations,” IEEE Trans. Intell. Transp. Syst., vol. 16, no. 2, pp.
utility’s experience to date with feeder unit protection systems,” in Proc.
741-750, Mar. 2015.
11th Int. Conf. on Develop. Power Syst. Protection (DPSP), Birmingham,
[25] IEEE Standard for Local and metropolitan area networks: Media Access
UK, 23-26 Apr. 2012, pp. 1-6.
Control (MAC) Bridges, IEEE Std 802.1D-2004, Jun. 2004.
[5] C. Xu, H. Xiong, L. He, Z. Li and J. Yang, “The Research of Intelligent
[26] Industrial communication networks – High availability automation
Substation Time Synchronization System and the Influence of its Fault to
networks Part 3: Parallel Redundancy Protocol (PRP) and
Relay Protection,” Energy and Power Eng., vol. 5, no. 4B, pp. 468-473,
High-availability Seamless Redundancy (HSR), IEC 62439-3:2012, Sep.
Jul. 2013.
2012.
[6] C. Curry. (2014, April). SENTINEL Project – Report on GNSS
[27] G. Prytz, “Network Recovery Time Measurements of RSTP in an
Vulnerabilities. Chronos Technology, UK. [Online]. Available:
Ethernet Ring Topology”, in Proc. 2007 IEEE Conf. Emerging
www.chronos.co.uk/files/pdfs/gps/SENTINEL_Project_Report.pdf
Technologies and Factory Automation (ETFA), Patras, Greece, 25-28
[7] B. Baumgartner, C. Riesch and W. Schenk, “GPS Receiver
Sep. 2007, pp. 1427-1253.
Vulnerabilities Urban Legends or Sad, Hard Truth?” in Proc. PAC World
[28] Chronos Technology. (2013, July). GPS Antenna Installations Best
Amer. Conf. 2014, Raleigh, USA, 22–25 Sep. 2014, pp. 1–20.
Practice. (r1.1) [Online]. Available:
[8] A. P. Cerruti, P. M. Kintner Jr, D. E. Gary, A. J. Mannucci, R. F. Meyer,
www.chronos.co.uk/files/pdfs/cs-an/GPS-Installation-Best-Practice.pdf
P. Doherty and A. J. Coster, “Effect of December 2006 solar radio bursts
[29] PCS-221G Merging Unit Instruction Manual, 1.00 ver., NR Electric Co.,
on GPS receivers” Space Weather, vol. 6, no. 10, pp.1-10, Oct. 2008.
Nanjing, Jiangsu, 2012, pp. 11-12.
[9] D. Itagaki, K. Ohashi, I. Shuto and H. Ito, “Field Experience and
[30] Technical Manual Current Differential Protection Relays,
Assessment of GPS Signal Receiving and Distribution System for
P544&6/EN/TM/M, Alstom Grid, Stafford, pp.6-10.
Synchronizing Power System Protection, Control and Monitoring,” in
Proc. 2006 IEEE Power India Conf., New Delhi, India, 10-12 Apr. 2006,
pp. 1-8. BIOGRAPHIES
[10] D. P. Shepard, T. E. Humphreys and A. A. Fansler, “Evaluation of the
Hao Guo received the BEng. Degree from the University
vulnerability of phasor measurement units to GPS spoofing attacks,” Int.
of Manchester, Manchester, UK in 2012. He is currently a
J. of Critical Infrastructure Protection, vol. 5, no. 3-4, pp 146-153, Dec.
PhD student at The University of Manchester. His research
2012.
interests include time synchronization, Ethernet
[11] IEEE Standard for a Precision Clock Synchronization Protocol for
redundancy and engineering within power substations.
Networked Measurement and Control Systems, IEEE Std 1588-2008, 24
Jul. 2008.
[12] Communication networks and systems for power utility automation Part
90-4: Network engineering guidelines, IEC 61850-90-4:2013, Aug. 2013. Peter Crossley received the B.Sc. degree from UMIST,
[13] Communication networks and systems for power utility automation Part Manchester, U.K., in 1977 and the Ph.D. degree from the
9-2: Specific communication service mapping (SCSM) – Sampled values University of Cambridge, U.K., in 1983. He is a Professor
over ISO/IEC 8802-3, IEC 61850-9-2:2011, Dec. 2011. of Electrical Power Systems Engineering at The University
[14] Communication networks and systems for power utility automation Part of Manchester, U.K. and Director of Doctoral Training
8-1: Specific communication service mapping (SCSM) – Mappings to Centre. He has published over 250 technical papers on
MMS (ISO 9506-1 and ISO 9506-2) and to ISO/IEC 8802-3, IEC protection and is an active member of various CIGRE,
61850-8-1:2011, Sep. 2011. IEEE, and IET committees.
[15] D. M. E. Ingram, P. Schaub, R. R. Taylor and D. A. Campbell,
“System-Level Tests of Transformer Differential Protection Using an

0885-8977 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.