IEEE Circuits and Systems Magazine - Q2 2020 - 12

Besides the aforementioned approaches, a special
type of method for delivering time to nodes is through
Global Navigation Satellite System (GNSS, GPS in North
America). It can provide synchronization-in-time with
high accuracy. However, it is expensive and requires
line-of-sight of satellites. Moreover, the low signal
strength makes it vulnerable to unintentional and intentional interferences.
C. Establishing A Common Notion of Time
Although time is a common notion for scheduling activities in all kinds of organizations, an explicit notion of
time is not an absolute necessity. For example, it is not
needed for an isolated system. Clock pulse train from an
oscillator is sufficient to drive the circuit of the system
since the circuit's output is deterministic as long as its
internal states evolve step by step. The sequential nature
of the clock pulses is able to do the trick. In this case,
the quality of the pulses (i.e. jitter, wander and spectral
purity) does not affect the function of event-sequencing
as long as the circuit is constructed correctly [18]. For
the case of centralized system, synchronized time is
not required either since there is no time ambiguity. As
in a single-processor computer with OS, a process can
get time by simply issuing a system call to the kernel.
When another process tries to get time, it will get either
an equal or higher time value. There is a clear ordering
among events [23]. Therefore, it is not always necessary
for a hardware clock being referenced to something (e.g.
a golden standard).
In distributed system, the story is different. There is
no global clock or common memory. Each processor has
its own internal clock and its own notion of time. Those
clocks can easily drift seconds per day due to frequency
instability, accumulating significant error over time. Different clocks drift at different rates and thus these clocks
most likely will not remain synchronized although, at
start, they might be. This is unacceptable to applications
running on distributed system. For such applications, we
need to acknowledge time in following senses.
1. The relative ordering of events happened on different machines in the network.
2. The time interval between two events happened
on different machines in the network.
3. The time of the day at which an event happened
on a specific machine in the network.
Those desiderata lead to the need of a common notion of time for a distributed system. For applications in
which a notion of time can be substituted by causalitybased relationship, clocks are relatively synchronized
to each other because the requirement is only to provide an ordering of events, not the exact real-world
time at which each event occurred (1 and 2). For such
12

clocks providing only relative synchrony, phase synchronization is sufficient. For clocks that must not only
be synchronized with each other but also have to adhere to a reference (the 3rd one), synchronization must
be performed to timestamp each significant instant using an accurate real time standard like universal coordinated time (UTC). As said, one strategy is to give each
node a GPS receiver and use the time signal sent by satellites. It of course has concern of reliability and cost.
An alternative is software based: to adopt protocol and
then design algorithm for seeking a common notion of
time. The underlying foundation of such algorithms is
the transfer mechanisms discussed in section II.B.
D. Applications of Packet-Switched
Synchronization
Mobile Cellular Networks: Synchronization is important
for synchronizing base stations and monitoring network
performance. Traditionally, synchronization in mobile
cellular networks is supported by TDM technologies
since TDM can provide sufficient timing performance.
However, it has low efficiency in utilizing bandwidth. It
is envisioned that traditional TDM synchronization architecture will be replaced by packed-based synchronization scheme due to the concern on cost efficiency and
the convergence of packet-based services provided by
an all-IP backhaul network [24], [25].
Industrial Applications: Synchronization plays critical
role in industrial applications of automation and data acquisition, to improve precision, productivity, and quality. In the case of data acquisition of synchronized inputs
from sensors, for instance, events measured in different
areas of an object need to be precisely time-stamped so
that the whole picture of correlated events can be reconstructed [9]. As another example, navy shipboards
require precise time measurements in order to detect,
locate, and identify moving targets [6].
Smart Grid: Smart grid is an enhancement of legacy
power grid. It consists of certain level of intelligence.
One of its salient features is the integration of two-wayflow of energy and information between the premises
of generation and customer. It supports emerging features such as distributed renewable energy resources
and demand-response. It improves grid efficiency and
reliability. To fulfill those desiderata, reliable timing for
the emerging smart grids is a critical factor [7], [8]. For
example, one significant modification to the grid with a
stringent synchronization requirement is the wide area
monitoring system, supporting the use of synchrophasor measurements of time-stamped harmonic power
components for real time grid control. With the arrival
of new applications such as demand-response, distributed energy resources (e.g., wind and solar) and advanced

IEEE CIRCUITS AND SYSTEMS MAGAZINE

SECOND QUARTER 2020

IEEE Circuits and Systems Magazine - Q2 2020

Table of Contents for the Digital Edition of IEEE Circuits and Systems Magazine - Q2 2020