Background
for Telephone Switching
2nd Edition (Revised and Expanded)
Chapter 7
Physical Design
OUTLINE
OBJECTIVES:
This chapter will introduce the reader to some of the
practical details necessary to convert a concept into a physical
system that can actually serve the customer.
PREVIEW QUESTIONS
1. How is a telephone switching
system put together?
2. How does it expand to serve
more customers?
3. What sort of requirements does
a switching system impose on the space around it?
PHYSICAL
DESIGN
Knowledge of the concepts and
principles of switching, even when augmented with appreciation of
users' needs, is not enough to produce a telephone switch.
Ultimately, a physical structure must be designed and built, and
then transported to the site, installed, cut over, and operated for
20 years or more. The mechanical design involved is as important as
the electrical and program design. And, like the electrical and
program design, the mechanical design must consider the real world
if it is to be successful.
FRAMES AND
CABINETS
A telephone switch is composed of
thousands of circuits in a very compact package. All these circuits
are interconnected with miles of wire and/or optical fiber, and must
be easily accessible for maintenance and repair. Over the years,
frames and cabinets have been developed to facilitate such mounting.
In spite of the differences between LSI, transistors, relays and the
venerable clockwork of SXS, some general principles emerge.
Frames
Frames are, for all practical
purposes, parallel steel posts to which mounting plates of equipment
can be bolted. In crossbar systems, the posts are vertical, while in
SXS systems they are horizontal. One spoke of "shelves" of SXS
switches, but one frame supported a number of shelves. In SXS PBX
systems, a frame six feet long, three feet wide and seven feet high
contained a complete switching system for 200 lines. There were two
shelves of line finders, two of connectors, and two of first
selectors; there was space for one or two shelves of second
selectors, if and when they were needed. Each shelf was wired for
ten or twelve switches. There was even space for the 200 line
circuits (line and cut-off relays), 20 to a mounting plate.
In SXS systems, each switch contained
its own control relays mounted under a dust cover and was a separate
plug-in unit, easily removable for maintenance. Removal of any idle
switch in a shelf did little more to the system than degrade traffic
handling capacity somewhat.
In a SXS frame, shelves were usually
mounted back to back to maximize the equipment in a given area. Some
crossbar systems also used the back-to-back arrangement with
switches and other equipment accessible from the aisles. This made
some wiring difficult to reach, but wiring was seldom a problem
compared with the switches themselves.
Another commonly used approach was to
have electromechanical equipment mounted on one side of a frame
only, with narrow wiring aisles separating two lines of frames, and
wider access aisles to reach the relays and other circuitry. In
systems such as 5XBAR, a frame was, for all practical purposes, one
relay thick.
Frame height
Frame height is a factor of some
importance. Traditionally, most CO equipment was built in 9 or
11-ft. frames while PBX equipment was mounted in 7-ft. frames or
smaller. Obviously, with higher frames, less floor space was needed
for the same amount of equipment. However, to reach equipment 11
feet from the floor, rolling ladders were necessary. These ladders
had to have power sockets for soldering irons and trouble lamps;
their support rail, at the top, sometimes blocked cable runs, and
the safety of those working at such heights had to be considered.
Finally, a high frame tended to be heavy, and floor loading become a
problem.
Cabinets
Dust, along with cigarette smoke and
other dirt, is one of the traditional enemies of exposed contacts in
both electromechanical switching equipment and the connectors used
by electronic switching. Thus dust covers protected the relays (but
usually not the line banks and wipers) in SXS, wire-spring relays
had built-in plastic dust covers over their contacts, and various
kinds of covers were available for individual frame-mounted
circuits. Covers which kept out dust also kept in heat, and
sometimes one problem was exchanged for another.
With the advent of electronic
switching, open frames gave way to closed cabinets. Unlike frames,
cabinets provide overall dust protection as well as sound-proofing
(both acoustic and electrical), additional structural possibilities,
and new opportunities for system wiring. Cabinets have two major
difficulties, however: by closing off the circulation of room air,
they require some form of internal air circulation and/or cooling,
and they tend to encourage very high equipment densities.
Flow and quality of air within the
cabinet is easier for the designer to control than room air;
however, air intake filters must be easy to clean or replace, and
fans and cooling equipment must be designed with the same
reliability and maintainability as anything else in the system.
European standards forbid forced air cooling, making power
conservation in combination with careful plans for convection air
currents necessary design ingredients.
High equipment density often leads to
excessive weight, making floor loading a factor of some importance.
With increased structural knowledge available to architects and
civil engineers, less safety margin is allowed in modern building
design. Thus older buildings are sometimes able to carry more weight
on their floors than new buildings rated at the same loading. In any
event, cabinet weight should be limited to something like 150 pounds
per square foot wherever possible.
Size and weight enter the picture in
other ways. Equipment frames and cabinets must be moved into place
through doors and elevators by ordinary people, often without
special handling gear. This suggests that equipment should be
designed with shipping and installation in mind, with the
limitations of buildings well understood, and with the capabilities
of the work force carefully considered.
A final factor becoming ever more
important in hardware design is disposal of equipment after its
useful life is over. Dismantling and removal must be possible with
minimum impact on power and communication wiring, and materials must
be selected with recycling in mind. Although recovery of precious
metals on switch and connector contacts has long been standard, such
practices will undoubtedly be expanded to include most other
materials in the future. Refurbishing equipment for resale on the
"secondary market" has implications not only for design but also for
documentation and long-term support.
Frame line-ups
In many CO installations reflecting
earlier thinking, frames or cabinets were arranged in straight,
parallel lines, spaced far enough apart to permit access by
maintenance and cleaning personnel with their equipment. Further,
the main distributing frame (to be discussed) paralleled the
equipment. In later CO layouts, the MDF was run at right angles to
equipment line-ups so that wiring could go from the MDF straight
down cable troughs on top of the frames or cabinets.
Cabinets, usually limited to 7 foot
heights and somewhat wider than frames, are often designed to
support special cable troughs with separate compartments for
different kinds of wiring. This allows power to be kept away from
communication wires, and control pairs and co-axial cables carrying
high speed pulses to be separated from tips and rings carrying
analog voice. With the coming of optical fiber, separate slots are
provided for it; although optical fiber does not radiate and is not
bothered by the radiation of other circuits, it has mechanical
constraints that recommend separate physical channels.
Even "permanent" wiring serving a
large switch must be changed every now and then to accommodate
addition and/or removal of individual frames or cabinets. Cable
troughs must be designed to allow for both possibilities to be
implemented quickly and safely.
Another approach to wiring uses a grid
above the equipment; a cable can then be run in any direction via
the grid, dropping down to the cabinet for which it is destined.
Such overhead grids permit great wiring flexibility, but must be
carefully coordinated with lighting and air conditioning ducts and
may pose problems when cables have to be removed.
Frames gave better access to equipment
than cabinets, important when individual relays required
adjustments. Cabinets need space for swinging doors, and often the
equipment within must either swing or slide out. This extra
complexity increases costs to some extent but increases the packing
density of the circuitry, reduces intercabinet wiring, and provides
the environment control needed by electronics. Floor-space reduction
is important, but, as has been indicated, it must not be achieved at
the expense of exceeding the floor's ability to support weight.
Some cabinets require access, not only
from the front and rear, but also from the sides, necessitating free
space all around. Although this may help with floor loading when the
building can average out concentrated loads separated from one
another, it reduces the space saving that modern electronics is
supposed to bring about. Today, cabinets requiring front access only
are widely used, and can be mounted side by side, with two lines of
cabinets back to back. Some cabinet designs have the "backplane" in
the center, becoming, in effect, two lines of cabinets in one, back
to back, with "front" access both front and rear. Many smaller PBXs
are packaged in wall-mounted cabinets requiring front access only.
One of the stranger requirements
voiced by PBX customers is "stackable" cabinets. One would suppose a
6 or 7 foot cabinet would be quite acceptable for a small PBX, even
if it only needed one shelf for implementation. Although many
systems come with a half-height cabinet for 2 shelves and a full
height cabinet for 4 shelves, customers have expressed strong
preference for each shelf to come complete in its own cabinet, with
provision for cabinets (shelves) to be stacked one on top of another
as the system grows. Why unused space should be left unenclosed is
not clear, but the message from customers is.
Interconnecting devices
Prior to about 1975, labor was less
expensive than hardware. Inter-frame wiring, and even wiring from
one shelf to another within a frame, was done in the field by
installation crews who ran multi-pair cables from one terminal block
to another, usually attaching the ends with "wire wrap guns" which
had replaced earlier hand-soldering techniques.
As the cost of labor rose, it became
apparent that connectorized cables, in spite of their higher cost,
speeded up installation and saved more in labor than the cost of the
connectors. Once it became economical to build whole systems,
including their interconnecting cables, in the factory rather than
in the field, it was possible to also test those systems before
taking them apart for shipping. At the site, the frames or cabinets
could then be put in place, the interconnecting cables attached to
their connectors, and the job was done. Small PBXs in single
cabinets, assembled to order to meet customer requirements,
pioneered the factory system-testing approach.
During the 60s and 70s, wiring of
individual circuits was often installed by computer controlled
automatic wiring machines with a great improvement in speed and
accuracy over manual factory wiring. However, the whole technology
of building assemblies on printed circuit boards, used also in the
computer industry, matured during this interval to become the
standard means of construction. Today, printed circuit boards
(PCBs), often with several layers of wiring laid down
photographically (and thus without error once the correct layout is
completed) support and interconnect a variety of circuit elements
from relays to resistors and capacitors to LSI chips.
In electronic switching systems, the
hardware revolution is probably better characterized by connectors
(not to be confused with SXS connector switches) than by the
electronic devices themselves. Because it is impossible to repair
LSI in the field, almost all equipment is built on printed circuit
boards which can easily be replaced so that faulty boards can be
sent to a repair depot. The connectors which receive these boards
are built into much larger printed circuit boards, backplanes
similar to those used in computers. Typically, each shelf has its
own backplane, and wiring from one backplane to another, in the same
or a different cabinet, is done with connectorized cables. This
allows shelves to be added as needed, and circuit boards to be
plugged into shelves.
Obviously, connectors must be highly
reliable so that they will permit insertion and removal of matching
plugs over the life of the equipment. This may require 500
insertions, and mechanical wear must be considered, to say nothing
of electrical erosion when cards are to be removed without cutting
off power. Further, electrical contact must be nearly perfect over
the same period of time; with relatively small signals used by
electronic equipment, great care must go into designing the plug and
connector interface.
Mechanical accuracy of a very high
degree is required, particularly when 100 or more signal paths are
involved. Further, tension on each contact must be great enough to
insure a good connection, but not so great that the plug-in unit
needs a hydraulic jack for removal. Considering the number of
contacts involved, this is no small feat; many circuit boards are
designed with small pivoted levers (called extractor clips) top and
bottom to assist in removal. To insure good electrical connections,
special precious metal coatings are often used on contacts. Note
that silver is NOT used because it tends to migrate when current
flows through it, shorting out nearby connections.
For any given system, PCBs look pretty
much alike. Thus when they differ, it is important to key or at
least color code them so that a card will not be plugged into the
wrong slot and connector. It is also a good idea to be sure that
power and ground connections are the same on all cards, and pin
lengths are selected so that power will not be applied until other
connections are established. Devising a foolproof system requires
careful planning.
It must be remembered that while
plug-in units can easily be replaced for maintenance, connectors
themselves are not so readily dealt with. Any connector failure, or
any failure in backplane or cabinet wiring, can become a major
disaster. Much of the design process must focus on making connectors
and wiring truly reliable.
One of the most interesting aspects of
connectorization comes in the design process. For many years,
various design techniques, including a modified form of Boolean
Algebra, have been employed to "minimize" circuits. In relay
equipment, minimization of contacts on relays as well as the number
of relays was vital. In early electronic logic, diodes in gates were
minimized, but transistors, usually required for gain and logical
inversion, were the main subjects of minimization. As medium and
then large scale integration were perfected, all this turned out to
be of little importance; the main trick now is to minimize the
number of wires connecting things together. The complexity of the
circuitry within a chip is far less important in determining price
than the total quantity of that chip to be produced.
With LSI, one tries to maximize what
can be done on one chip with a small and standardized number of
leads to the printed circuit board on which it is mounted. Then,
PCBs are designed to have as much related circuitry as possible on a
given board so that connections to the back-plane can be made via
standard connectors with limited numbers of contacts. Finally,
shelves and cabinets are organized to be as complete as possible,
again to minimize the wires that must run through connectorized
cables to other cabinets and to simplify modular expansion.
Electronic circuitry, grown in lavish profusion in almost
microscopic sizes, is readily traded off against wires and
connectors. Irrelevant minimization techniques have gone the way of
wire wrap guns and soldering irons.
PCBs, shelves and cabinets
In electromechanical systems, the
switching matrix dominated floor-space and cost. A line circuit
consisted of nothing more than a line relay and cut-off contacts,
sometimes on a separate relay and sometimes associated with the hold
magnet of a crossbar switch. Trunk and service circuits were
appreciably more complex in systems like 5XBAR, but when most of
their work could be done within the system program, as in 1ESS, they
became small, even when built with traditional relays. Thus SXS,
crossbar switches and reed relay devices, suitably packaged,
occupied most of space in the switch room.
Electronic switching turned the whole
thing around. Because line circuits, providing a buffer between
internal electronics and outside plant, were usually not switched
out but remained connected to the customer's line at all times, they
had to contain toll-grade transmission components, battery feed and
supervision, ringing and test access, and, for digital systems using
1950s telephones, codecs to convert between analog speech and
digital signals for switching. PBXs now consist mostly of line and
trunk cards augmented with perhaps one or two PCBs for the common
control; the switching matrix, often incorporated into the port or
control cards, has vanished for all practical purposes. CO switches,
because of their vastly greater size, still have physical equipment
dedicated to the matrix function, but it tends to be quite small
compared to earlier systems.
As a result, major design effort
centers on circuit boards and custom LSI for lines and, to a lesser
extent, on trunks (because of their smaller numbers). In PBXs, it
quickly became evident that "universal card slots," those that could
accept line cards for electronic or conventional telephones and
trunks of all denominations, were vital to user acceptance; having
separate shelves or dedicated card slots for trunks vs. lines, where
the customer, wishing to add trunks, might find only slot space for
line cards, was unacceptable.
Control information went past all port
cards on a standard bus and, when time-division switching became
dominant, matrix connections also took advantage of a bus-like
access on which a time slot could be assigned on a per-port or
per-call basis. Thus port cards interfaced the rest of the system
via a fixed number of connector terminals, independent of the number
of ports on the PCB. This gave PBX designers motivation to put as
many ports on a PCB as possible to make full use of the time-shared
control and matrix leads as well as shared control circuitry on the
board.
In PBXs today, it is common to have
large line cards, often 20 inches on a side, which can support 8 or
16 telephones. Three- or 4-bit address can then be used to identify
the particular port, with additional address bits to select the card
slot and shelf. Clearly, it is desirable to have PCBs with the same
number of trunks so as not to waste equipment numbers.
Unfortunately, trunk components often require more space than
components for lines, so it is not uncommon to have PCBs for CO
trunks with half as many circuits as line cards.
Although addressing and control may be
standard, the number of wires from each PCB to lines and trunks is
highly variable. CO trunks and 2500 type telephones need only one
pair per port, as do some electronic phones, but other electronic
phones require two pairs. E&M trunks usually require at least four
pairs to the outside world, so their port-density per PCB may be
only a quarter that of a card for analog phones.
T-carrier cards for digital trunks or
subscriber loop carrier (SLC) introduce new problems in address
numerology. Twenty-four-trunk digroups in North America and Japan,
like 30-trunk groups used elsewhere, require 5 bits for addressing
individual circuits but only 2 pairs to the outside world, one for
each direction. Multiplexed digital circuits can be handled
efficiently by LSI; thus some PBXs terminate two T-spans on a single
PCB. Thus occasionally the PBX "uniform card slot" rule is violated
to take advantage of the simpler wiring and larger addresses needed
by digital trunks, and to leave more "universal" card slots for
lines and analog trunks.
Shelves for PCBs, sometimes called
"card cages," include in their design slots to guide cards into
their backplane connectors, matching backplanes, and the careful
design required to control both the physical and electromagnetic
environment. Cabinets have to be structurally strong to support
shelves of equipment, and have the physical rigidity to insure
continuing fit that is not upset by the addition or removal of cards
or other shelves. Further, by enclosing the shelves they support,
cabinets provide additional requirements and opportunities for
environment control. All this is more expensive than the simple
frames of earlier systems, and such cost must be pro-rated over the
circuits supported.
Central office switches such as those
in the DMS series from Northern Telecom use small line cards, about
3" on a side, each housing a single line circuit, mounted on shelves
in large sliding drawers. Trunks, usually digital, interface as
T-spans with different hardware. The 5ESS from AT&T uses a similar
approach for its digital trunks and ISDN telephone sets, but serves
analog sets via a space division line concentrator using electronic
crosspoints and providing the traditional line and cut-off functions
per line with new technology. Clearly, the uniform card slot rule is
not as important in CO equipment where quantities of port circuits
are vastly larger and technical support is more readily available.
Most digital central offices, because
of their size, are built on the line-group/group-selector pattern,
with optical fiber connecting line groups to the group selector. A
few strands of glass, carrying something like 500 simultaneous
conversations, greatly simplify system wiring. Further, because
optical fiber neither radiates nor picks up electrical noise, it
simplifies the design of the rest of the system which otherwise
would have to be protected against the noise produced by high speed
pulses on copper, even when co-ax is used. CO line groups typically
support up to 5000 lines, taking advantage of concentration, but
trunks are interfaced in much smaller numbers to omit concentration.
Electromechanical equipment is, in
certain ways, more rugged than electronic equipment, particularly
with regard to temperature and extraneous voltages which may enter
the switch via lines or trunks exposed to lightning, power crosses,
etc. However, electronic components have so many advantages that it
has been necessary to find ways to allow their use.
Air conditioning
All generations of switching equipment
have needed clean air, as has been discussed, but electronic
equipment is also quite sensitive to both humidity and temperature.
All three of these factors must be combined to condition the air in
which the system is to survive. American design often uses cabinets
with air filtering at the intake and duplicated fans, with external
air conditioning for humidity and temperature control. Air
conditioning to fit at the cabinet air intake, allowing the cabinet
air to be controlled independent of the switch room, is also
available, particularly for PBXs.
Enclosed cabinets may lead to "hot
spots" at certain points in their interior; thus forced air flow is
often required although careful design, following European practice,
can make fans unnecessary. When fans or air conditioning are used,
they must, of course, not generate vibration, power surges, or
electrical noise harmful to the circuitry they support, and their
reliability and maintainability must be equal to that of the
switching system itself. Needless to say, design efforts should
minimize power dissipation because both reliable power and cooling
will become increasingly expensive in the future.
In the final analysis, the reliability
of the electronic switching system may depend on that of the air
conditioner serving the switch room. Further, air conditioning often
increases power consumption to such an extent that power savings
predicted through use of electronic circuitry are wiped out,
particularly when air conditioning must be run from the same
reliable power source used by the telephone equipment. In PBX
installations where regular office air conditioning is assumed,
conditions after hours and on weekends must be investigated. These
are times when building management typically turns off the central
air conditioning to save money. Similarly, if air conditioning ducts
are used for heating in winter rather than the removal of
system-generated heat, the PBX may hibernate until spring.
Most switching equipment
specifications give both an upper and a lower bound for temperature
and humidity if equipment is to function properly, with somewhat
wider limits for equipment storage. It should be noted that very low
humidity facilitates the build-up of electrostatic charges which can
destroy electronic equipment; high humidity, on the other hand,
leads to condensation on circuit boards and connectors which can
short out signals and otherwise harm the wiring itself. People
handling circuit boards should always ground themselves carefully to
prevent static discharges.
Illumination
With more and more switching equipment
operating unattended, there is no need for switchroom illumination
most of the time. However, when maintenance is required, lighting
levels must be suitable to the task at hand. With high packing
densities used for equipment, labels must be readable; color coding,
when used, must not be altered or rendered ineffective by lighting
variations. One advantage of going to cabinets, usually 7 feet high
or less, is that overhead lights can be lower than for 11 foot
frames, allowing less lighting power to produce the desired level of
illumination at floor level.
At switchboards, consoles, test panels
and the like, proper illumination is even more important. LEDs must
be very bright if they are to be used in bright areas, and LCDs must
be avoided for areas normally dim. Glare must be minimized and
cathode ray tube displays must be suitably positioned; otherwise
eyestrain and poor operations will result.
Noise and vibration
Older switching systems were notorious
for the noise they generated. SXS, particularly when used in PBX
applications, had to have a special sound-proofed room, or else
cabinets lined with sound-insulating material. Indeed, one of the
early arguments for electronic switching was its quietness. A modern
electronic PBX can be located right in the reception room, next to
the console; it makes no more noise than a coffee table.
Perhaps the grand champion noise-maker
was the card translator used in the 4XBAR toll switch. The card
translator was the first commercial application of transistors
(1952), and for its day it was a wonderful machine. It had a metal
card for each route, and when a route was requested, all the other
cards would be dropped, permitting light to shine through a pattern
of holes in the remaining card. A room full of these machines, all
going full tilt, was spectacular. Fortunately, computers have made
such ingenuity unnecessary, and have also reduced the acoustic noise
level almost to zero.
Just as a switching system could, in
the old days, generate noise, unpleasant for switchboard employees
in adjacent rooms, so vibration from outside could impact the
switches. Subway trains and traffic can shake an entire building and
wind sway in gusty weather may cause trouble, particularly for
magnetically latched relays and reed switches which can be released
mechanically under adverse circumstances. Fans and blowers internal
to cabinets, elevators and other external equipment, and personnel
themselves can all interact mechanically.
Electronic equipment is not
particularly sensitive to vibration; indeed, its ability to
withstand shock and vibration is one of its many advantages.
However, contacts in connectors are a potential source of trouble,
as are high speed rotating machines such as disk drives.
Electrical noise
Elevator motors, motors in floor
waxers, fluorescent lights, radio transmitters and various other
electrical and electronic appliances may produce signals that
interfere with the operation of electronic switching. Metallic
cabinets can provide shielding from such noise, but cables between
cabinets can act like receiving antennas.
The inverse problem, radiation from
the switching system to the detriment of surrounding radios, TVs,
computers and other devices, became painfully evident with the
earliest ESS installations. In large systems, where data buses using
unshielded twisted pairs were run to many cabinets, noise radiation
turned out to be considerable. Relatively sharp pulses, repeated at
regular intervals, generated signals at many harmonics of the basic
system clock rate.
Over the years, clock speeds have gone
up several orders of magnitude and high speed pulses have
proliferated, not only for control purposes but as digitized voice.
As a result, chips, PCBs, backplane wiring and cabling to other
cabinets, digital telephones and trunks have all been extensively
studied with an eye to minimizing electromagnetic radiation and
keeping it within the limits specified by Part 15 of the
Communication Act. Although cabinets can be designed to contain or
exclude electromagnetic noise by brute force, it is far more
satisfactory and often less expensive to design equipment from the
beginning to minimize the impact of radiation.
For wiring between cabinets, coaxial
cable has been used but once again, careful design can allow twisted
pairs to function at a considerable saving in cost. Optical fiber is
fast becoming the preferred approach, however, making most earlier
approaches obsolete.
For external wiring to digital
customer telephones, whether PBX proprietary or ISDN BRI, twisted
copper pairs can be expected to continue in use for some time. With
phones near PCs and data terminals, to say noting of electric
typewriters and fluorescent lights, it is important to design the
system not only to limit radiation which might be harmful to others,
but to protect itself from radiation from others.
In small systems, proper grounding and
shielding can minimize noise and facilitate the transmission of
digital signals internally. In larger systems, spread out on one or
more floors of a building that may occupy a city block, the meaning
of "ground" is not obvious. Ground for dc battery and for protection
of personnel is relatively simple to consider, but ground for high
speed digital signals is something else again. Often, transformer
coupling is used to digital paths on copper to keep individual
cabinet grounds independent. It is unlikely that the last word has
been said by anyone on the principles of grounding in large
electronic digital systems, or even in small ones. However, before
it is, optical fiber and optical circuit components may make the
whole question moot.
With the exception of relatively small
switches such as PBXs in a single cabinet where cards can be
inserted as needed, the ability to expand (or contract) gracefully
is a requirement which has to be planned carefully.
SXS modularity and implications
Perhaps the greatest single advantage
of SXS switching was the way it could grow. A 701B PBX could start
at 60 lines or so and expand to over 10,000 while remaining
economical every step of the way (actually, there was no theoretical
limit; the original Pentagon SXS PBX had over 30,000 lines). For
smaller systems, however, the 740 was used; it could handle up to 80
lines, but topped out at that point.
The difference between the two
demonstrates the basis of maximum size in almost all modular
systems. The 740 consisted of line-finders back to back with
connectors; the connectors had extension lines on their lower
levels, reached with two digits, while their upper levels (typically
9 and 0 for outside and switchboard) were each selected with a
single digit and then hunted for an idle trunk. For systems as small
as 20 or 30 lines, there was no need to plug in the full complement
of switches.
In the 701, line finders were tied
directly to first selectors rather than connectors; the first
selectors could connect directly to connectors in a 3-digit system,
or through second selectors to connectors in a larger 4-digit
system; each stage of selectors permitted the switch to expand by a
factor of 10. The 740, by connecting directly to connectors,
sacrificed this flexibility to provide a more economical system (no
selector switches at all) in small sizes. All modern switches have a
fixed upper limit to their matrices; in particular, small electronic
systems, guided by the 740, are optimized by limiting the ability to
grow beyond a certain size.
It happens that most PBXs are
relatively small. Below about 50 lines, electronic systems are often
called "hybrids" or electronic key systems although, as has been
discussed, they are PBXs. A separate size range, from about 40 to
400 lines, has also become common, usually having one shelf for the
control and a few port cards, and three or four additional shelves,
often stackable, for port cards only. The next size break is in the
1000 to 2000 line range, while large PBXs usually top out at around
5000 lines, roughly the size of a small central office. Below each
break point, a different switching matrix format is used to keep
costs down.
For very large customers such as
universities where telephone service is provided to students,
central office switches are sometimes employed as Centrex when
rented from the telephone company or as PBXs when owned by the
institution. Until ISDN phones are available, however, most of these
systems are optimized to serve conventional analog single-line
telephones, leading to the installation of 1A2 key or small PBXs
behind the large switch to provide key-system features for staff and
administration.
Modularity in link type switches
The great majority of central office
lines in the United States are switched on large machines. Prior to
the coming of digital switches, the matrices in these large switches
were "link type systems," illustrated in Figs. 8 and 9 in Chapter 1,
and exemplified by AT&T's 5XBAR and 1ESS. 5XBAR could grow to about
30,000 lines, while 1ESS could go over 100,000. 5XBAR was a
four-stage switch, composed of two-stage line link frames (LLF) with
degrees of concentration which could be selected as needed, and
two-stage trunk link frames (TLF). 1ESS was built of Line Frames and
Trunk Frames, each of 4 stages, although local calls generally used
only the Line Frames for interconnection.
Because of its relative simplicity,
5XBAR will be used to illustrate modular switching-matrix growth in
link type systems. LLFs, as shown in Fig. 7-1, were made up of two
bays, one of which, Bay 0, held ten crossbar switches with 20
verticals and 10 horizontals. The horizontals in each switch were
split in the middle so that, effectively, there were twenty 10 x 10
switches, with customer lines coming in on one group, and "junctors"
going off to TLFs on the other group. The two groups were cross
connected by "line links" so that one level of each line switch went
to a different junctor switch, and each level on a junctor switch
could be traced back to a different line switch.
Bay 0, by itself, connected any line
to any junctor; full access existed. However, to provide
concentration, one or more bays such as Bay 1 could be added by
extending the line switch horizontals in a process called "building
out." For very high traffic situations, Bay 1 contained ten 10-input
switches. More typical was a bay of ten 20-input switches, building
the line appearances out to 300, 10 of which were used for no-test
access. For very quiet residential areas, additional bays could be
added so that, with light traffic, 590 lines were given access to
the same 100 junctors. An LLF, composed of two or more bays,
depending on traffic, was a basic equipment unit.
TLFs, shown in Fig. 7-2, were similar;
they consisted of a two-stage array that connected 200 junctors to
160 trunks. Actually, the design was quite subtle: ten 20x10
switches were split to produce twenty 10x10 equivalent switches to
terminate junctors. Cross-connection internally went to ten 20 x 10
crossbar switches which accessed trunks. Here "trunk links" cross
connected the verticals rather than the horizontals as in the LLFs.
But there was one final trick. The trunk access switches were 6-wire
and handle two trunks per output. Two horizontal levels were used to
steer the input to the right hand or left hand group of three leads
on the outputs, so eight levels per switch were left, serving two
trunks each. With 16 outputs per switch, each frame handled 160
trunks.
Given LLFs that could connect from 190
to 590 lines to 100 junctors, and TLFs capable of extending 200
junctors to 160 trunks, how was a matrix assembled? With 10 LLFs and
10 TLFs, the answer is easy. Each of ten bundles of ten junctors
from each LLF were connected to one of the 10 TLFs as shown in Fig.
7-3. One junctor in each bundle came from each of the LLF junctor
switches, and went to each of the TLF junctor switches. With 10
junctors between every LLF-TLF pair, there were ten possible paths
from any line to any trunk.
A switch of this size handled between
7000 and 8000 lines (20 LLF times 390 lines per LLF), and 1600
trunks and service circuits such as originating registers and
circuits for call progress tones. For smaller systems, there were
fewer frames, so more junctors could be used between the frames that
remained. With larger systems, one could not simply provide fewer
junctors between frames or traffic-handling capacity would suffer.
Thus a different approach was required.
Just as the line switches in a LLF
were built out by adding additional crossbar switches, so the TLFs
were built out. This permitted doubling or tripling the number of
line link frames that could be connected to a given trunk frame by
simply increasing the TLF inputs. However, more trunks had to be
added to meet the needs of the added lines. Thus two or three of
these extended TLFs were put in parallel on each group of junctors.
Each TLF added 160 trunks, and the line-trunk ratio remained the
same. The line-junctor ratio also remained the same because each LLF
had 100 junctors. Thus the system was expanded while holding the
grade of service constant.
The matrix in 1ESS, using four-stage
line-frames serving from 1000 to 4000 lines, depending on
concentration, and four stage trunk frames, grew to much larger
sizes, but used similar approaches. In either system, the apparent
complexity of design was justified by lower cost and better traffic
performance.
Distributing frames
In large space-division systems with
metallic matrices, the equipment frames for the matrix took up a lot
of space, had to be connected together properly, and had to be
re-configured for growth. Thus an orderly means was required to
associate matrix frames with customer lines, with trunks, and with
each other. Distributing frames of various sorts performed this
valuable function. Equipment frames were positioned and cabled to
distributing frames on a "mass production" basis, culminating in
modern connectorized cables. Interconnection could then be carried
out by running jumpers as needed between terminals on the
distributing frames, making the actual operation relatively simple
and isolating the fragile switches themselves from contact with the
not inconsiderable activity at the distributing frames.
The Main Distributing Frame or MDF is
the major cross-connect frame in any central office. It terminates
cables coming in from the local customers. These cables normally
enter the building through a cable vault in the basement and come up
through the floor where they are terminated on vertically mounted
terminal strips on the MDF.
In CO MDFs, the vertical terminals
have, in the past, been associated with protection equipment:
over-voltage protection which connects tip and/or ring to ground
when lightning strikes, power lines are crossed to telephone lines,
etc., and excess current protection which opens the tip or ring when
too much current flows, typically from crosses to 120 volt AC power.
For many years, carbon blocks and heat coils where used for these
two functions; a carefully placed electrode would arc to a grounded
carbon block when lightning struck, and a heat coil would open, not
unlike a fuse, on excessive current. Although carbon blocks were
self-healing for transitory lightning hits, both carbon blocks and
heat coils were designed to also place a permanent ground on the tip
and ring from the outside world when current continued to flow; this
prevented dangerous external voltages from going beyond the MDF to
harm either personnel or equipment.
The mountings for the heat coils and
carbon blocks have also been used for manual test access, the
connection of traffic measuring and service observing equipment,
etc. They were a convenient place to open the path to the outside
world for manual testing, allowing independent looks toward the
customer and the switch.
Today, complex electronic line
circuits, more expensive and far more fragile than metallic matrices
and line and cut-off relays, are in direct contact with pairs to the
outside world. As a result, carbon blocks have been pretty much
replaced by specially designed gas tubes or solid-state break-down
devices which can provide faster operation and more accurate
limitation of over-voltage. Gas tubes usually have one electrode
from each wire which can ionize gas to provide a self-healing short
to ground. An advantage of gas-tube protectors is that once the gas
ionizes from excess voltage on either tip or ring, both are grounded
until trouble-current ceases. A perceived disadvantage is that, over
time, gas may leak out of its glass bottle, leaving the line
unprotected. As a result, solid state devices are becoming the
standard.
Aerial and underground cable have
replaced open wire lines, minimizing the exposure to power crosses;
as a result, heat coils are often omitted in the CO. PBXs, which
have most of their wiring inside the building they serve, usually
omit both over-voltage and over-current protection on everything
except trunks and OPXs.
The switching matrix, via customer
line circuits, is cabled to the other side of the MDF. Here the
terminal strips for the inside plant cables are arranged in
horizontal rows. Connections from vertical to horizontal are made by
means of jumpers which are run through cable troughs between the "V"
and "H" sides of the frame. When a telephone customer moves to
another location served by the same CO, the same matrix port on the
H side can be connected to the pair to the new location from the V
side by simply running another jumper, allowing the customer to
retain the familiar number.
This jumper change at the MDF is a
variation of the translation function discussed in Chapter 2.
Because the software in many stored program controlled PBXs and CO
switches associates features, class of service, etc., with a
specific matrix port, and requires not only the telephone number but
all the supporting data to be entered at the new port (and deleted
from the old one) when a move is made, it often turns out to be
quicker and easier to change jumpers than to use a VDT to modify the
system's data base. When the data base is changed, the new
information must be backed up on tape or disk; this can lead to
further complexities when others are contending for access to the
same resources for similar changes.
When different kinds of PBX line card
are in use, a translation change requires the line circuit serving
the pair to the new location to be the same type as the line circuit
serving the old; a program change alone cannot allow a card for a
proprietary electronic phone to serve a 2500 set and vice versa.
Thus the system manager, even when equipped with VDT access, also
has to change the line card, awkward when it supports 8 or 16
telephones. As ISDN phones come into general use, CO switches will
have to pay particular attention to this same problem. Although
digital CO switches usually employ single-line line cards, making
line card changes easier, pulling jumpers on the MDF will continue
to be a viable alternative for years.
Some PBX manufacturers have added
inexpensive "single line" digital sets to their product line, priced
competitively with analog sets, so that a single type of line card
can be used and changes made via software only. A "set relocation"
feature then allows the user to unplug a phone from one jack and
plug it into another. Prior to unplugging, a MOVE code is keyed into
the system to warn it; after plugging in at the new location, a
"here I am" code is keyed in and the system makes the change. A
serial number stored in ROM is used in some of the sets for newer
systems. When the set is plugged in at its new location, the switch
reads the serial number, relates it to its assigned directory number
and moves both to the new switch position; the phone, complete with
all its features, is ready to go without the user having to key in
any codes at all. When there is no possibility of plugging a digital
phone into a 2500 set's jack or vice versa, this is a powerful tool.
Users have been known to take their phones with them to the lunch
room, conference room, etc., so as not to miss calls, anticipating
the functions of radio-based personal communication services (PCS).
In central offices, MDFs are
functionally part of the outside plant and are seldom changed when a
new switching system replaces an old one. In the PBX world, however,
the station wiring is the property and responsibility of the
customer, and if not changed when a new PBX is installed may
invalidate the PBX warranty. This has led to rapid development of
wiring approaches, and today the connection between the PBX and MDF
is almost completely connectorized at both the PBX and MDF ends.
The need for "universal card slots,"
discussed previously, leads to some interesting considerations when
a port card may have 4, 8, 16 or some similar number of circuits on
it, requiring anywhere from 8 to 32 pairs to the MDF. Fifty pin (25
pair) connectors have become a de-facto standard; for analog phones
and proprietary electronic phones which, like analog CO trunks,
require a single pair, one connector can serve three 8-pair PCBs.
When the number of circuits doubles to 16, or when each circuit
needs two pairs, two connectors can serve three 16-pair PCBs; to
double again, four connectors can serve three 32-pair PCBs. On the
PCB, the printed wiring from each circuit to its pins on the
connector must be able to carry the fairly heavy line and signaling
currents involved, withstand voltage and current surges below the
limits of the protection devices, and, along with the cables
connected to them, not produce cross-talk.
Cables are put in place at
installation and, with luck, are not touched until the PBX is
replaced. Thus the MDF must show clearly which card slot in which
shelf and cabinet is being served, the kind of card in place, and
relate pairs to a particular port on the card. Obviously, replacing
one port card with a different type requires a change in indication
at the MDF.
The intermediate distributing frame,
or IDF, was another common cross-connect frame found in
electromechanical offices. Initially, the IDF was used in SXS COs
between line finders and first selectors. Traffic balancing was
achieved by mixing line finders from busy line groups with those
from groups with less activity so that all groups of first selectors
would be offered essentially equal traffic. Note that this was at
variance with PBX practice where line finders and first selectors
were usually hard-wired tail to tail. In non-SXS offices, the IDF
was sort of a catch-all cross-connect frame.
The junctor grouping frame or JGF is
used between line frames and trunk frames in link-type systems such
as 5XBAR and 1ESS. These frames have their junctor sides cabled to
the JGF, where line-frame to trunk-frame jumper connections are
made. In 1ESS, where both line-frames and trunk-frames handle 1000
junctors rather than the 100 or 200 as in 5XBAR, junctor
rearrangements are made in groups rather than individually.
Plugended cords for 16 junctors can be arranged as required to give
all frames access to each other for the type of service to be
rendered. The 1ESS patterns are further complicated in that line
frames can be connected to line frames and trunk frames to trunk
frames, all in addition to the line-trunk pattern of 5XBAR; the 1ESS
control has a program to generate optimum junctor grouping patterns
when a system is expanded.
A trunk distributing frame or TDF
follows the trunk side of a metallic matrix, and permits trunk and
service circuits to be cross-connected to the matrix itself. With a
TDF, trunk and service circuits can be mounted in special frames or
cabinets of their own; circuits of a particular type may all be
mounted together for convenience, but may be accessed by a number of
different trunk frames for traffic balance. The far side of the
trunk circuits may be cabled to an intermediate distributing frame (IDF)
for cross-connection to transmission facilities, signaling sets,
tone distribution terminals and other similar points. Fig. 4,
patterned loosely after 1ESS, shows the location of the various
distributing frames relative to switching equipment; it emphasizes
just how much hardware and human activity can be eliminated when
line and trunk circuits interface the matrix directly, and when
multiplexed digital trunks can go directly into a digital switch.
As a practical matter, in many central
office and PBX installations, there is no need for an IDF, JGF or
TDF, and the MDF alone provides all the cross connect functions
which remain. Because not all cross-connections require protection,
it is now common for central offices to have a protection frame
separate from the MDF. All outside lines terminate on the protection
frame, while the outputs of the protection frame, along with pairs
carrying trunk transmission and signaling to nearby carrier
equipment and signaling sets, tone distribution, etc., are brought
to the MDF without carbon blocks or heat coils. This can produce a
considerable saving in some situations.
The Change to Digital
The development of electronic
switching and T-carrier started in the mid '50s at Bell Labs;
T-carrier was first introduced in 1962, and 1ESS three years later.
Because 1ESS, and the smaller 2ESS and 3ESS which followed, were
intended for local switching, dealing primarily with residential
lines terminated in 500 and 2500 type telephones, they took full
advantage of a 2-wire space division matrix, using a metallic
connection to extend the customer loop to trunks and service
circuits for regular telephone processing, and to line insulation
test equipment and test desks for maintenance of outside plant and
station apparatus (see Chapter 8).
T-carrier was intended for short-haul
use, primarily on copper pairs already in place between nearby
switches in metropolitan areas. Obtaining 24 channels on two pairs
increased by an order of magnitude the number of channels available
without having to dig up city streets, but even when cable
installation was easy, T-carrier's remarkable economics made it the
preferred approach for distances as short as 6 miles.
Because carrier systems have to be
4-wire while local lines are 2-wire, and lines outnumber trunks by a
wide margin, it was logical, in the days of metallic switching, to
put the hybrid circuit in the transmission equipment and switch
lines and trunks on a two-wire basis. As a result, the tradition
established by analog carrier systems was continued and T-carrier
trunks were made to look to local central offices like conventional
2-wire trunks on copper pairs using standard supervision such as
loop/reverse-battery or E&M.
But with the change from copper pairs
to digital exchange carrier, it no longer made sense for tandem and
toll switches, which switched nothing but trunks, to continue to
pretend that trunks were simple copper pairs; rather, it quickly
became clear that digital trunks should enter such switches in their
T-carrier multiplexed format so that digital signaling could be
accessed directly, digitized voice channels could be switched
without converting back to analog, and channels carrying non-voice
digital signals could be switched directly. AT&T's 4ESS, introduced
in 1976, used the T-carrier 24 channel digroup as its basic input,
and multiplexed five of these DS-1 channels together in a digital
digroup terminal to make a DS-120 channel for application to its
switching matrix. Although a voice interface terminal was also
available to support 120 conventional analog trunks typical of those
on long-haul microwave, the replacement of microwave with optical
fiber quickly made digital digroup terminals dominant.
4ESS called trunks that ran between
the digroup terminals on two switches "digital trunks." However,
because a T-spans could also terminate in conventional D-type
channel banks to interface analog switches such as 1ESS, a trunk
between a digroup terminal and a channel bank was called a
"combination trunk," not to be confused with the combination trunk
used with PBXs for dial 9 outgoing and attendant handled incoming
calls. With channel banks on only one end (if used at all),
T-carrier became even more economical and encouraged the use of
digital interoffice trunks to a greater extent. Although local
offices prior to about 1980 were almost exclusively analog,
"combination trunks" moved the inputs to digital tandem and toll
switches right to the switchrooms of local CO switches, even when
they were 20 miles or more away.
Prior to the coming of 4ESS, Western
Union had observed that a digital transmission system should be able
to handle digital data directly without modems; several independent
manufacturers made T-carrier voice/data systems for WU to use within
metropolitan areas, subdividing 64 kBps channels to handle
economically the lower bit rates then standard. Although digital
PBXs started coming on the market in 1975, it took some years for
them to consider handling digital data like Western Union, or
digital voice trunks like AT&T.
Northern Telecom introduced its
DMS-100 local CO switch in 1979, and quickly followed it with DMS
switches for various kinds of toll service. AT&T's 5ESS, a digital
switch designed to replace the 1ESS, 4ESS, and TSPS to provide
local, toll and operator service, followed in 1982, and Ericsson,
Siemens and NEC were quick to join the party. All of these switches
were at their best accepting digital trunks in multiplexed form, and
also emphasized their ability to use digital facilities toward
customers to meet PBXs and remote switching units, and to interface
directly with digital subscriber loop carrier (SLC). Although at
present most of these digital local switches are actually serving
conventional analog lines, they make it possible for individual
customers to convert to digital lines when ISDN or some other
approach is accepted.
Digital CO Switches.
When digital switching finally reached
the local CO, A/D conversion had to be moved to the line side, along
with the hybrid to interface four-wire digital switching to analog
pairs. AT&T, as has been described, chose to develop a special
electronic crosspoint for use in line-group concentration in the
5ESS, following an approach that had been tried by NEC in the NEAX
22 PBX. This reduces by a factor of 4 or 5 to 1 the need for hybrids
and codecs, and provides a certain amount of reliability in that a
line is switched to a codec/hybrid combination on a per call basis;
if the electronics fails, it can be detected and taken out of
service, leaving the remaining codec/hybrids to carry the load with
only a slight degradation in traffic handling capacity.
This approach was quite economical as
long as telephone sets and other customer equipment, based on 1950s
standards, were used. PBXs, however, were able to introduce
proprietary telephone sets, allowing them to move the codec to the
set itself, preserving digital and 4-wire integrity end to end. When
ISDN telephones become available, and if customers choose to buy
them, a similar extension of digital technology from central office
switches will become more generally available.
Northern Telecom and others chose to
develop CO switches that were digital right to the line card, taking
advantage of the falling price of codecs and other electronic
circuitry. Line groups are designed to be different from trunks
which enter in multiplexed groups. Each line has its own 3"x3" line
card; line cards are mounted 64 to a drawer, five drawers to a
shelf, and two shelves to a line group of 640 lines. With only one
line per line card, and each line card containing its own relay for
test access, maintenance is simplified, and a change from analog to
digital telephone sets can be made by simply replacing the line card
along with the phone on a per-line basis. AT&T's 5ESS has similar
line circuits and drawers for ISDN phones (but 4 drawers per line
group for 512 lines). Line cards which can support 2500 sets are
also available; they are particularly important in digital Centrex
system when customers insist on using modems and fax machines.
Digital CO switches and many of the
larger PBXs follow the line-group/group-selector pattern suggested
in Fig. 3 of Chapter 1. The individual line groups provide
concentration and expansion for lines, while somewhat different port
groups are designed to interface trunks without concentration. The
group selector must approach non-blocking, and often follows the
design approach of a link-type system (Figures 8 and 9 of Chapter
1).
In space division, links and junctors
were used to interconnect stages of switching; links were hard wired
in a fixed pattern between stages within a given equipment unit such
as a line or trunk link frame, while junctors, which ran between
such equipment units, made use of the JGF to facilitate
rearrangements and growth. In digital systems, the connections
between port groups and the group selectors are often called
"links," or "multiplexed links" because they take advantage of the
simplicity of digital multiplexing to carry from 30 to 500 or more
channels on a single medium. Port groups can be added as needed,
each with its own links to connect it to the group selector.
The group selector in digital systems,
although a single entity, usually has to be able to expand from a
small size for just a few port groups to a size large enough to
accommodate 100,000 or more lines and trunks. Thus the term "junctor"
is often applied to cross-connections within a group selector which
allow modules to be added in an orderly way. In general, both links
and junctors in digital systems run directly from one shelf or
cabinet to another, omitting separate cross-connect facilities such
as a JGF. Because multiplexing greatly reduces the number of
physical channels required, such cross-connects are much easier to
manage than in analog systems.
In digital CO switches, adding port
groups is straightforward, with system management simplified by
families of single-line line cards (analog line, ISDN line, coin
line, ground start, etc.). Expanding group selectors, particularly
in tandem and toll switches when several cabinets containing
high-speed circuitry are required, is a highly specialized
operation, and differs considerably from one system to another.
Digital (and Other Electronic) PBXs
Because PBXs are smaller and cover a
wider size range (roughly 30 to 5000 lines) than central office
switches, they introduce wider variety into architecture and
modularity than do CO switches. Up to about 400 lines, the simple
bus-like approach of Fig. 7a, Chapter 1, works well for both space
and time division. The Mitel SX-200 and the one-shelf version, the
SX-100, allowed up to 31 simultaneous conversations on individual
wires across the back-plane; lines and trunks had equal access to
each of the 31 space-division paths. Similarly, AT&T's Dimension 400
had a serial bus subdivided into 64 two-way time slots for PAM
samples. In a later version, a second bus was added to double
capacity and improve reliability. The Rolm CBX used a 16 bit
parallel bus for its 12-bit linear PCM coding, assigning two time
slots, one for each direction, to each conversation, and serving up
to 192 simultaneous calls.
In such systems, modularity at its
lowest level was represented by port cards; additions were made by
simply sliding cards into empty slots and activating them via the
program. With 8 or 16 circuits per card, a certain amount of
"granularity" was introduced, but options such as loop-start and
ground-start for CO trunks, program controlled on a per-port basis,
reduced the number of card types required. Some manufacturers
developed additional cards with only one or two circuits on them,
less expensive than fully implemented cards, to reduce initial cost
in competitive bidding; savings to customers were often wiped out
when growth demanded additional circuits.
Shelves might have slots for 10 to 20
port cards, and systems could be expanded by simply extending the
bus to additional shelves (as many as five per cabinet), and then to
additional cabinets; however, the fixed maximum number of
simultaneous connections limited the number of ports which could be
handled, and the transit-time of high speed pulses limited the
system's physical size. To make larger systems, another level of
modularity was required. Use of small PBXs as modules of a larger
PBX was one idea that had much to recommend it, particularly from a
manufacturing and training point of view.
The Dimension 2000 and the Rolm VL
chose to provide direct "trunk groups" between each pair of modules,
where a module was quite similar to a Dimension 400 or Rolm CBX.
Although ten or more such modules could be assembled into a large
system, the number of trunk groups between modules (N*[N-1]/2) along
with their support circuitry could become unwieldy. Further, the
problem of control had to be addressed. Should a single powerful
control be designed for the group, or should each group have its own
autonomous control with special provisions for calls to ports on
other modules? The latter turned out to be far more difficult,
particularly when hunt and pick-up groups were scattered across
modules for reliability; however, it was successfully used by a
number of PBXs and by AT&T in its 5ESS CO switch, although the 5ESS
simplified its inter-module trunking by introducing a central
switching stage.
The use of a central switch (group
selector), of course, permits PBXs to add port groups the same way
that CO switches do. AT&T's System 85, similar to 5ESS and using
many of its components, consisted of modules which could be
stand-alone PBXs (up to about 1500 ports); each such unit had a time
slot interchanger to make intra-module connections. When more than
one module was needed, the group selector had to be added to make
connections between ports on different modules. The group selector
was actually a space-division switch; the TSI in each module had to
select the same time slot (to and from the group selector) for a
given connection, and the group selector would make different
connections in each time slot.
Northern Telecom's Meridian 1 (best
known as the SL-1) used a different approach. All its calls,
including those between ports on the same line group, went through
the group selector. The group selector's inputs were 30 channel
multiplexed links (called multiplexed loops by Northern). In one
version, a small group selector supported 32 of these links, with
Time-Space switching to move an incoming time slot on one link to
any outgoing time slot on the same or a different link. Originally,
one multiplexed link supported a line group of one or two shelves,
each with 10 slots for port cards serving 4 analog lines or 2
trunks. This gave a maximum of 80 ports access to the 30 time slots,
which were assigned on a per-call basis. Port density was doubled
twice over the years, eventually supporting 16 phones and 32 ports
per PCB, the number of multiplexed links per line group was
increased, and other improvements were made. But the basic
architecture continues to include concentration/expansion between
ports and multiplexed links, and distribution among time slots on
multiplexed links.
When LSI development made possible a
switch quite similar to the Northern Telecom 32x32 group selector on
a single chip (see Chapter 1), and several chips could be grouped to
make a bigger switching matrix in a very small space, some systems
were designed to assign time slots to each port permanently and use
the group selector alone as a single-stage rectangular time-space
matrix. A number of PBX families, including the Siemens Saturn and
the Harris 20-20, chose to follow this approach to provide
non-blocking switching among many ports. Individual port modules can
be added until all multiplexed links are in use.
A matrix of this sort establishes
connections at 64 kBps in each direction; however, proponents of
"bandwidth on demand" suggest that this may not be suitable for the
future where both wider and narrower channels may be necessary. In
the parallel bus type of matrix, where it is possible, at least in
principle, to assign two or more time slots in each direction to a
given connection (super-multiplexing), or different connections on
each of the bus wires in the same time slot (sub-multiplexing),
bandwidth on demand takes on more meaning.
The Rolm CBX has offered
sub-multiplexing, and has the potential for super-multiplexing; with
direct channels rather than a group selector in multi-group systems,
these features could function both within and between modules.
AT&T's System 75, a digital PBX which can be thought of as a
replacement for the Dimension 400, uses two 8-bit parallel busses
for companded PCM, and has similar single-module potential. When
upgraded into Definity Generic 3, individual line groups, each of
which looks very much like a System 75, are interconnected by a
group selector used only for connections between ports on different
modules. This group selector is also designed on a parallel bus
basis, and suggests that both sub- and super-multiplexing could be
provided on a switched basis between modules. A system providing
modular physical growth and modular variations of the bandwidth to
be switched should have considerable utility in the future.
Digital Cross-Connect Systems.
Digital CO switches and many digital
PBXs support a variety of remote switching units to serve large
groups of distant telephones, and station carrier for smaller
groups. These remote units come to the group selector either via
optical fiber or T-carrier on copper; as a result, the lines they
serve bypass the MDF at the CO. PBXs, which are, in reality,
autonomous remote switching units, can also arrive at the CO on
T-spans, again bypassing the MDF when the PBX homes on a digital CO.
Business customers often have private
lines for data, and tie-trunks for switched voice and data
connections. Further, direct connections are often required to long
distance carriers for WATS lines and similar services. When
T-carrier extends all the way to the customer's premises, it makes
sense to include these circuits as well as CO trunks in the same
carrier system, even though anything other than CO trunks will not
require connection to the local telephone company's switch. It would
be silly and expensive to demultiplex such circuits and bring them
back to analog just to cross-connect them at the MDF; such a
procedure would be a disaster if the digital properties of these
customer circuits were the reason for their existence in the first
place.
What has been done is create a digital
switch to replace the MDF for digital circuits. Called DACS or DCS
for digital (access and) cross connect system, such switches accept
T-spans and make "nailed up" connections as required at the DS0 (64
kBps channel) level. Typically, a DACS is a non-blocking
single-stage time slot interchanger; because it is actually a
switch, it can be controlled electronically, and remote control is
usually used. Similar cross-connect systems are used at the
locations of long-distance hubs, and allow private networks to be
set up, modified and taken down quite quickly from a central point.
Although systems similar to DACS found
their way into military specifications in the early 1970s, it was
almost a decade later, when digital local switching began to be used
in commercial telephony, that they made their appearance in civilian
life. One might have supposed that building DACS capability directly
into the local and tandem switches would have been more economical,
even if such switches had to become non-blocking to handle both call
switching and the cross connect function. Unfortunately, design
philosophy, rooted in the high cost of metallic crosspoints in a
century of analog switches, has limited the consideration of such a
possibility.
For better or for worse, DACS are now
widely deployed. As suggested earlier, they are particularly useful
in handling circuits from PBXs, where "grooming" may be needed to
obtain proper fill for local, long distance and data connections.
This appears to be an almost classic example of sophistication
replacing copper to provide better service.
Optical Cross-connects.
Optical fiber was installed initially
for short-haul trunks, but when single-mode fiber became available,
it took over long distance trunking very quickly. However, because
so many individual circuits could be multiplexed on each fiber, the
number of strands needed was actually quite small. Further, the
experimental nature of many early installations led to ad hoc
solutions for running fiber within telco buildings, often extending
it directly to the specific device needed to provide the proper
termination.
As the trend toward fiber in the loop,
to the curb and to the home accelerates, such techniques will give
way to a more organized approach. Already, many manufacturers of
wiring and cross-connect frames are developing similar technologies
for optical fiber (such as the FDF or fiber distributing frame), and
are providing the flexibility of cross-connect jumpers and other
amenities which the past shows the future will be able to use. The
whole field is brand new, however, and catalogs and trade journals
must be consulted as a standard approach evolves.
Power plants for telephone switching
systems are highly specialized and well beyond the scope of this
book. However, there are a few points that relate to switch design
and operation that are important here.
In the first place, a telephone system
is expected to be reliable. When our lights go out, the first thing
we do is reach for the telephone and call the power company. We
never question the ability of the telephone system to perform this
minor miracle, but the miracle doesn't just happen. It has been
designed in.
Traditionally, telephone power plants
have consisted of large batteries floating across the output of a
battery charger (rectifier) fed by the commercial ac mains. The
charger supplies power to the CO equipment, including the switch,
trunk equipment and other communication devices. The battery acts
very much like a large filter capacitor smoothing and regulating the
output voltage. When the commercial power fails, the battery takes
over; usually it is designed to handle about four hours of service
before it discharges to the point where permanent damage will take
place. During this time, a local generator is expected to start up,
either automatically or under manual control, and replace the
commercial ac power at the system input.
Batteries are particularly good in
switching systems because of their ability to supply very large
overloads for short (millisecond) intervals. Thus average drains
rather than peak surges can be considered, a factor of no small
importance in the days of SXS when a switch magnet could draw
several amperes during each dial pulse. Electronic power supplies
are limited by the peak load they can deliver; momentary surges can
be fatal.
Battery plants work very well with
electromechanical equipment, but ac power supplies, with the ac to
dc conversion taking place in each frame or each shelf have
advantages for electronic equipment. With a local power supply,
frame grounds can be handled easily, a great advantage with pulse
circuits containing high frequency components. Further, distribution
of power to the many frames in a switch is often easier in terms of
transformer-coupled ac than one large, common dc feeder system.
However, the development of inexpensive dc-to-dc converters has lead
to continued distribution of the reliable -50 dc battery voltage to
individual frames in the CO, with local power supplies in each
frame, cabinet or shelf developing the several voltages needed by
electronic circuits.
Per-shelf electronic power supplies,
particularly in line groups serving conventional 2500 type
telephones, are often designed to supply limited power, taking
advantage of traffic fluctuations; that is, they do not assume that
all phones in a group will be off hook (or ringing) at the same
time. Thus a PBX may have a non-blocking matrix suitable for
advertising, but not be able to power all the phones which might be
in use under heavy traffic conditions. One way to relieve this
problem is to have shelf power supplies duplicated in a load-sharing
configuration for both reliability and higher peak power if the
customer needs such service.
Small PBXs are usually ac powered,
designed to plug into the wall; unless some form of power (or
system) failure transfer is available, loss of ac power takes them
out of service and isolates their customers. To add power
reliability, a "UPS" or uninterruptible power supply is often
inserted between the wall and the cabinet. A UPS contains a
rectifier which converts the ac to dc, a battery kept under constant
charge by the rectifier to store emergency energy, and an inverter,
to convert the battery dc voltage back to ac for the PBX.
Larger PBXs, like CO switches, are
often designed to be dc powered, taking advantage of the well
understood battery art. However, commercial power is quite reliable,
and much of the rest of the equipment needed for a business to
operate, such as computers, runs from ac only. Further, air
conditioning, necessary to keep both computers and electronic
telephone equipment working properly, is also usually ac operated,
and many modern office buildings have a large number of windowless
inside offices which require ac-operated electric lights. Thus
people cannot easily continue working when power fails.
As a result, it is not uncommon to
find even large PBXs operating from commercial ac. When this is the
case, non-volatile memory for programs and system parameters, as
well as CDR data, traffic records, etc., is vital. PBXs should be
designed to update system parameters automatically whenever changes
are made; it is more than embarrassing to have hunt groups, speed
calling numbers and a variety of other pieces of idiosyncratic data
vanish when the lights go out for a few seconds. Hard disks make
such storage and reloading easy. Again, note the need for power (or
system) failure transfer of a few analog trunks to strategically
placed 2500 type telephones to allow calling for help.
Voltage variations, power failure and
switchover
Most modern telephone systems operate
on 50 volts dc, with the positive terminal grounded. Under normal
conditions, the voltage is regulated to about ±1 volt around the
nominal. The batteries are being charged at a relatively light rate;
thus the terminal voltage of the dc power supply is the open circuit
of the batteries, plus their IR drop due to the small charging
current.
When commercial power fails, two
things happen to reduce the terminal voltage of the battery. First,
the battery goes from the charge condition to the discharge
condition, and second, the magnitude of the current through the
battery goes from small in the charging direction to large in the
direction of discharge. Thus there is a four or five volt drop
almost immediately; this jolt can often be felt by electronic
components in spite of the filters in each frame unless great care
is taken in the design of dc-to-dc converters. For the remainder of
the trouble interval, or until the diesel or gasoline driven
generator is started, there is a slow drop as the internal
resistance of the battery builds up.
To reduce the impact of the voltage
change when commercial power fails, end-cells have been used in the
past. Special switch-gear inserted these cells in series with the
main battery during the switchover process. The end-cells were
shorted out momentarily, inserted in the main power bus, and then
unshorted, as indicated by the make-before-break transfer contact in
Fig. 7-5. Although the short during transfer drew a brief surge of
current from the end cell, there was no interruption of current to
the system and the voltage drop on transfer was minimized.
In smaller offices, CEMF cells were
sometimes used. These cells were not batteries, but had a voltage
drop about equal to the battery change at transfer, relatively
independent of the amount of current drawn. When change-over took
place, the CEMF cells were shorted out, and their
reference-diode-like voltage drop vanished, canceling the drop in
the main battery.
Upon restoration of outside power or
activation of the stand-by generator, the end cells had to be
removed or the CEMF cells unshorted to prevent the 50 volt supply
from rising too far above nominal. By and large, improvements in
electronic battery feed to customer lines and stable outputs of
derived voltages for electronics have made CEMF and end cells less
necessary today.
Although central offices are usually
supplied with gasoline or diesel generators to handle prolonged
outages of commercial power, this approach is seldom practical for
PBXs except at hospitals or other emergency sites. Rather, the
reserve batteries in a UPS are designed to maintain operations over
relatively short outages, and to provide time for an orderly
shut-down of operations if it is evident the outage will persist. In
those situations where non-emergency organizations must maintain
some form of telecommunication, power failure transfer is often
sufficient.
Power failure transfer relates analog
CO trunks to compatible PBX telephones; when commercial power goes
off, relays held operated by that power release to disconnect the
phones and trunks from the PBX and connect them to one another,
making them stations served directly by the central office.
Obviously, this procedure works best with loop-start trunks. Where
ground-start trunks are used, the associated phones are equipped
with ground buttons which, when depressed momentarily, draw CO dial
tone. More elaborate power failure transfer equipment eliminates the
ground buttons, making emergency calling much easier for users who
have forgotten the purpose of the ground button by the time it is
needed.
PBX system failure can block phone
calls as surely as power failure. As a result, power failure
transfer can be generalized to system failure transfer so that the
transfer circuitry can activate the emergency phones even when power
is still available. Another emergency approach, particularly when CO
access is via a PRI or other digital trunking, is simply to have
permanent phones homing directly on the CO at strategic locations in
an organization, making sure, of course, that such phones are
powered from the CO. How ISDN and other electronic phones, data
terminals, and complex systems such as voice mail and automated
attendant can best handle power failure remains to be seen.
Fuses and filters
In central offices and very large PBX
installations, battery power is delivered via main fuses at a power
distribution board to the various frame line-ups. Each frame usually
has its own frame filter (several millifarads) and fuse panel to
distribute current to its own circuits. When a system is first put
into service, the frame filters are first charged up manually so
that their cumulative effect will not short out the main power
feeders and blow the main fuses. Fuse alarms, which also provide an
indication to the system that something has gone wrong, will be
discussed in Chapter 8.
Ringing and tone supplies
Traditionally, the ringing and tone
plant has been part of the power system. This comes from the old
pre-vacuum tube days when the only way to generate such signals was
to use a rotating electrical machine. A typical system used a dc
motor powered by the battery to drive a 20 Hz 86 volt ac ringing
generator. The generator would have a few extra windings on its
shaft so that it could also generate 600 Hz, 420 Hz, and the various
other frequencies then used for dial tone, busy and overflow tone,
audible ringing, etc., as well as power at harmonic ringing
frequencies when used. Via a gear train, cams would be operated to
activate interrupters that produced the 2-second-on 4-second-off
regular ringing signal (in three phases, dividing the ringing load
three ways), all the party-line ringing signals, and the 30, 60 and
120 IPM interruption patterns for the audible tones.
Although it is still not unreasonable
to generate 20 Hz power ringing with a rotating machine,
particularly in a large central office, many other techniques are
available including various electronic circuits and sub-harmonic
converters (60 Hz to 20 Hz) using saturable reactors. Call progress
tones, if they are to follow the North American precise tone plan
(Table 1, Chapter 3), require electronic generation; when delivered
by a space division switch, tones from a high quality generator can
be monitored for frequency and amplitude, and duplicated for
reliability.
In digital systems, tones are easily
generated using digital techniques as described in Chapter 3.
Digital coding controls the amplitude, and the frequency is locked
to the system clock. Although space division matrices are designed
to prevent multiple connections, digital matrices can easily provide
listen-only connections from one tone source to as many lines and
trunks as desired. When the switching matrix takes the form of a bus
subdivided into time slots, certain time slots may be dedicated to
specific call progress tones to which all ports can listen.
Ringing and coin control, being power
signals, can operate relays and blow fuses directly for checking and
protection. However, because they are power signals, they can only
be returned through the matrix when metallic or other high voltage
space division crosspoints are available. Otherwise, they must be
delivered to each line at its line circuit, distributed by a bus
subject to opens and shorts as it wends its way past line cards
serving analog phones. Shorts can blow alarm fuses to alert the
system, but finding opens usually requires some means of reading the
flow of ringing or coin-control current. Note that if coin control
failures are not picked up immediately, user frustration may produce
additional component failures.
In digital systems, it will be
necessary for decades to supply power ringing to 2500 type
telephones, answering machines, modems and a variety of other
devices. Often, ringing is generated in each frame or cabinet
supporting digital line cards, and is applied via per-line relays.
Even this relatively short ringing distribution system must be
designed to be reliable, must facilitate automatic testing on a
per-line basis, and must notify the maintenance force if it goes
down. When ringing is generated locally by electronic circuits, the
number of ports that can be rung at one time is often severely
limited; as has been discussed, this may make claims for
non-blocking matrix properties irrelevant.
-
Frame
-
Cabinet
-
Shelf
-
Module
-
Protection
-
MDF
-
DACS
-
Battery
Click Here for
Answers
1. What is the difference in a frame
and a cabinet?
2. Which was most commonly used with
electromechanical equipment?
3. What are some advantages of 7 foot
frames or cabinets as opposed to 9 or 11 foot frames?
4. Give some advantages of cabinets
over frames.
5. What is an advantage of the MDF
running at right angles to the frame line-ups rather than parallel?
6. When is a "shelf" a "cabinet"?
7. How does modern electronic switch
construction differ from that used with SXS and Crossbar systems?
8. What are some major environmental
considerations for switching systems?
9. What size ranges are typical of
PBXs?
10. How is growth provided in modern
CO switches?
11. Why are distributing frames used?
12. What is the principal
cross-connect frame?
13. Where are protection devices
usually located?
14. What changes have been made in
protection practice in the last 30 years or so?
15. Is a jumper change at the MDF
sometimes easier than using VDT in an SPC system?
16. When is it not possible to make a
translation change using only SPC techniques?
17. How is a Junctor Grouping Frame
different from the other distributing frames mentioned in the text?
18. Why can we expect digital switches
to greatly reduce the number of TDFs and IDFs in COs?
19. What is a DACS or DCS and how does
it compare with an MDF?
20. Why do CO switches operate on DC
rather than AC?
21. When commercial power fails, does
the battery voltage stay constant?
22. What are the requirements on a PBX
telephone for power or system failure transfer to work?
23. What is an alarm fuse?
24. Do ringing and tone sources and
distribution systems need protection?
[ Top ] [
Next Chapter ] [
Table of
Contents ] |
