by Ronald G. Todd
The popularity of Synchronous Optical Network (SONET) technology is
growing rapidly as more products demand the high bandwidth, connectivity,
and scalability that SONET offers. This enabling technology is necessary
in many cases for products to reach "next-generation" performance levels.
SONET is thus not only well-established within the traditional telecommunications
industry, but equipment vendors outside this community are now - or soon
will be - offering SONET interfaces on their products.
SONET equipment requires precise clocking to function properly and to
be compliant with SONET standards. However, clocking is an area that is
often misunderstood, even by people specifying and deploying SONET equipment.
This article will cover the major issues related to understanding and designing
a clocking system for SONET equipment.
The need for synchronization
SONET networks were originally designed to transport voice traffic in
an efficiently scalable way. The legacy time-division multiplex (TDM) hierarchy,
consisting of digital signals DS-0, DS-1, DS-3, and so on, incurs an increasing
loss of efficiency as the hierarchy rate increases. This is because the
TDM system is asynchronous and thus requires additional overhead each time
one multiplexes up to a higher rate in the hierarchy, in order to rate
match each lower-rate asynchronous source into the new higher rate.
SONET overcomes this scalability limitation by using a synchronous hierarchy,
resulting in an overhead percentage that does not vary as the rate increases.
Each time one multiplexes to a higher rate, the lower-rate signals are
synchronously byte interleaved to produce the higher rate. Since the lower-rate
signals are synchronous to one another, no additional overhead is needed
(with stuffing bytes and associated signaling) to support rate matching.
SONET has a built-in overhead of 4.4%, which can decrease slightly in
the special case where one transports concatenated payloads (such as with
an OC-3c payload rate of 149.76 Mbits/sec versus an OC-3 payload rate of
148.61 Mbits/sec). The initial mapping of a payload into the SONET synchronous
payload envelope can result in additional losses. For instance, a frame
of a DS-3 (44.736 Mbits/sec) is usually mapped into a full SONET 52-Mbit/sec
STS-1 (Synchronous Transport Signal 1 - the fundamental SONET rate and
format) frame, yielding an effective loss to overhead of 13.7%. But when
this STS-1 carrying a DS-3 is multiplexed to higher rates - for instance,
to an STS-192 (corresponding to OC-192) at 10 Gbits/sec - no further losses
to overhead occur.
SONET standards emanate from two different organizations: the American
National Standards Institute (ANSI) and Bell Communications Research (Bellcore).
In general, the Bellcore requirements are derived from the ANSI standards.
Clocking requirements are distributed throughout numerous standards. However,
the key requirements common to all systems are contained within a manageable
subset that includes:
-
ANSI T1-101: Synchronization Interface Standard
-
Bellcore GR-253-core: Synchronous Optical Network (SONET) Transport Systems:
Common Generic Criteria
-
Bellcore GR-1244-core: Clocks for the Synchronized Network: Common Generic
Criteria
-
Bellcore GR-436-core: Digital Network Synchronization Plan
-
Bellcore GR-378-core: Generic Requirements for Timing Signal Generators
-
Bellcore GR-499-core: Transport Systems Generic Requirements (TSGR): Common
Requirements
Other clocking standards usually pertain to specific equipment types, so
you must search for those standards when dealing with a specific equipment
type.
Stratum levels
Network clocks are divided into stratum levels based on their accuracy,
stability, and other parameters, according to Bellcore GR-1244-core. Stratum
levels are expressed as a number, sometimes along with a letter. The better
the clock, the lower the stratum level. The primary references used in
the network meet the Stratum 1 requirements. As a clock is distributed
across a network and among equipment, impairments are introduced that reduce
the stability and result in the clock being classified at a higher stratum
level. SONET equipment must either be synchronized with a Stratum 3 or
better clock or, if the equipment is not a digital crossconnect, clocked
from an oscillator with a minimum accuracy of ±20 parts per
million.
The defined clock stratum - along with the basic accuracy requirement
for that stratum - for an oscillator when it is not locked to a higher
stratum level clock (free-run accuracy) is indicated in the table. The
table also shows the accuracy with which the last frequency produced must
be maintained if connection to the higher stratum level clock is lost (holdover
stability).
SONET network timing architecture
The SONET network in the United States is timed from a limited number
of Stratum 1 primary reference sources (PRSs) distributed across the country,
forming timing domains. Historically, synchronization was distributed as
an analog 2.048-MHz signal. With the advent of the digital network, this
distribution occurred via DS-1 (1.544-Mbit/sec) signals passed from the
PRSs to network elements (NEs), which then distributed timing to other
NEs in a hierarchical fashion. This architecture is evolving such that
the SONET network will distribute the network timing instead of using the
tdm network. In the new architecture, one point of the SONET network within
a timing domain will be synchronized with the prs; the SONET network will
then distribute timing to other nodes of the SONET network and other non-SONET
NEs from DS-1s timed from the SONET network.
The hierarchy allows a small number of expensive, high-accuracy clocks
to be used to time an expansive network, providing reference to lower-accuracy
clocks, which then provide reference to yet lower-accuracy clocks. An advantage
of this system is that a detected failure of a clock high in the hierarchy
will cause the clocks it feeds to enter holdover mode. This will still
maintain a clock for that level and lower levels that is more accurate
than the free-running clock of the level just below the failure.
Equipment timing sources
There are five different ways that SONET equipment can be timed. A subset
of these is applicable in a given application:
External timing - Building integrated timing supply (BITs) is
the name given to the single master clock within a telephone company central
office. The BITs clock is derived from a timing reference that feeds the
central office, typically a DS-1. DS-1s carried on SONET cannot be used
for network synchronization distribution because they fail to meet ANSI
T1.101 synchronization interface specifications due to jitter introduced
by SONET network pointer adjustments. When SONET is used to distribute
a timing reference, timing is derived from the SONET line rate, not from
within the payload. As distributed within an office, the BITs clock can
assume one of two formats: a composite clock (CC) signal or a DS-1.
The CC conveys both bit and byte synchronization (64 and 8 kHz, respectively)
used in DS-0 (64-kbit/sec) signals all on a single waveform. The CC signal
consists of a 64-kHz, 5/8 duty cycle, return-to-zero, bipolar signal with
a bipolar violation every eighth bit. Data is clocked out on the leading
edge (departure from 0V) of this signal and sampled on the trailing edge
(return to 0V).
DS-1 signals used for bits typically consist of a framed "all-ones"
sequence, with a bipolar return-to-zero line format, and can be either
in the super-frame or extended-super-frame format.
In either case, the bits clock is multiplied up using a phase-locked
loop (PLL) to produce the bit- and byte-rate frequencies for SONET transmission.
Line timing - This method is used, for instance, with an add/drop
multiplexer. This kind of NE has a bidirectional SONET interface on both
sides. Timing is extracted from one incoming side and used as a reference
for both outgoing sides.
Loop timing - This is a special case of line timing and is applicable
to line terminating equipment. Here the NE has only one bidirectional connection
to the SONET network. The transmit timing is derived from recovered receive
timing.
Through timing - This mode is mostly used for regenerators. Timing
recovered from the incoming signal in one direction is used to clock the
transmitted signal continuing in the same direction. The same holds for
the opposite direction.
Free running - If an oscillator not refer- enced back to a network
timing source is used, both the oscillator and the mode are referred to
as free running.
Of the various ways that SONET NEs can be synchronized, the preferred
order for clock selection is as follows: bits (external timing), received
timing (loop, line, or through), and local timing (free running).
Clock switching
If multiple clock sources, such as external timing, loop timing, and
an internal free-running clock, are available to an NE, then a mechanism
must be implemented to select the appropriate synchronization source. The
selection can be statically provisioned or can be under hardware or software
control, based on the real-time health of the various clock sources.
One must be conscious of preventing timing loops in a network (such
as when NE #1 uses NE #2 as its timing reference, but NE #2 is already
using NE #1 for its reference). This can be an issue if an NE can switch
between external and line timing, such as upon an external reference failure.
This is not a problem with network terminating equipment, since downstream
equipment within the network would not look to the terminating equipment
for a reference. Thus, terminating equipment can be designed to switch
between external and line-timing sources. Switching can be either revertive
or nonrevertive. Revertive switching implies that there is a preferred
clock source, and that if it has failed (whereupon one switches away from
it) and then recovers, one should switch back to the preferred source.
With nonrevertive switching, once you switch to a new clock source you
stay with that source until it fails, even if the clock you switched away
from recovers. Nonrevertive switching is recommended by Bellcore in most
circumstances.
SONET implements a messaging protocol, via synchronization status messages,
to identify to the receiving NE the stratum level traceability of the clock
that was used to create a given SONET signal. With this information, an
NE can select the best synchronization reference from a set of available
references. This can aid in automatic reconfiguration of line-timed rings
and in troubleshooting synchronization problems.
Ronald G. Todd is vice president of engineering and technology at Kalman
Saffran Associates, Inc. in Newton, MA.
About Kalman Saffran Associates - Established in 1978, Kalman Saffran Associates
(KSA) is a leading OEM product development firm that rapidly creates innovative
products for premier companies in the data communications, telecommunications
and interactive/CATV industries. To ensure success, KSA applies a proven,
advanced engineering methodology to all projects. KSA has a staff of more
than 100, including engineers specializing in today's "building block"
technologies such as SONET/SDH, ATM, xDSL, fiber optics, MPEG and PCI.
KSA's clients include industry leaders such as AT&T, Lucent, Nortel,
Cisco Systems, 3COM, Scientific-Atlanta, PictureTel, Hewlett-Packard,
Motorola among many others. For further information, visit KSA on the
World Wide Web at http://www.ksa1.com.
We will be pleased to assist your product development efforts. For more
information: Contact Us