Background
for Telephone Switching
2nd Edition (Revised and Expanded)
Chapter 3
Interfacing Subscriber Lines
OUTLINE
OBJECTIVES: The objective of this chapter is to describe
-
The properties of telephone lines
-
The classification of telephone lines
-
And the functions of telephone lines.
PREVIEW QUESTIONS:
As you read, watch for he answers to the
following important questions:
1. What do telephone lines do?
2. How to they interact with the rest of the telephone system?
INTERFACING SUBSCRIBER LINES
The
terms "line" and "trunk" are deceptive. In a simple world, lines
connect customers to central office switches, and trunks connect
central office switches to each other. We will start off with this
definition, observe the way it has been modified almost beyond
understanding by common usage, and summarize the terminology in
Chapter 4.
For
a hundred years, lines connected telephone sets to "line-side" ports
on a switch. Thus the sets, wires and ports were designed to more or
less work together. Because there were so many lines, every effort
was made to minimize cost, not only of the hardware, but also of
installation, maintenance, record keeping, etc. For instance,
traditional telephone sets are powered from the central office. As
suggested in Fig. 1, they contain a "hook switch" which is operated
by the "switch hook," and when the phone is "hung up," power to the
set is disconnected. This power flow is monitored by the switch in a
process called "supervision" to tell when the phone is on-hook
(idle) or off-hook (in use). The ringer, which alerts the subscriber
to an incoming call, is "outside" the switch-hook so that it can be
operated when the phone is on hook.
The
last Bell System analog telephone, the 500 set, came into use in
1950. In the late 60s, the DTMF pad for tone signaling replaced the
rotary dial; the designation for sets using DTMF is 2500. Whether
500 or 2500, these sets have the same transmission package
(transmitter, receiver, coil, etc.) which was designed to produce a
uniform audio level at the switch independent of distance from the
CO, as will be discussed below. This not only improved transmission,
but also eliminated the need for transmission field adjustments at
the time of installation.
As a
result of the deregulation of the telephone industry, customers were
no longer permitted to rent telephones from local telephone
companies as part of their service; rather, they were required to
buy phones to create a market for outside equipment vendors.
Although many of these vendors have shown considerable ingenuity in
using new chips and other modern technology in the construction of
their competitive sets, and have added a variety of features that
will be discussed in Chapter 5, their basic designs are still
directly related to that of the 500 set so that such phones can
replace the older phones without changing the wires they use or the
CO on which they home. Indeed, now that customers must own their
phones, continuing innovation in both terminals and the telephone
network will be much more difficult.
A
major innovation, as opposed to cosmetic variations more typical of
the automotive field, would be to extend digital technology to the
customer's telephone set. ISDN is an attempt to establish standards
for such an attempt, and will, of course, require a completely
different telephone set and matching port circuit (and digital
switching matrix) at the CO, although existing copper wires will, at
least in the near term, suffice. Note that conventional matrix ports
will not work with ISDN phones, and 2500 sets will not work with
ISDN matrix ports. Thus careful coordination will be necessary at
both ends of the line when (and if) customers wish to upgrade to
modern technology and leave the 2500 set behind.
ISDN
phones will differ from 2500 sets in several important ways. First,
they will probably not be powered by the CO switch; local power
(perhaps with battery back-up for reliability) will then be
required. Second, the sets will be powered at all times, requiring
other means for the switch to supervise them for on-hook/off-hook
status. Third, signaling will use a separate, out-of-band channel,
not only for supervision but also for telephone numbers, feature
requests, calling party identification, incoming call notification,
answer supervision from the called party and certain kinds of data
transmission. Finally, analog to digital and D to A conversion for
voice will take place in the set, requiring 4-wire transmission even
though a single pair will still be used.
The
copper web connecting customers to central offices represents a
slightly larger investment to a telephone company than the CO switch
itself and, although a pair of wires seems simple, the design
sophistication required to maximize performance while minimizing
overall system cost is considerable. Trading electronics for copper
in the last few years has widened both the engineering and economic
process, and today many options are available.
In
particular, modern digital CO switches can interface directly with
subscriber loop carrier (SLC, usually one fixed channel per phone),
remote concentrators (flexible assignment of many phones to a
smaller number of channels), and remote switching units (RSUs, like
concentrators but also able to complete local calls internally).
These sub-systems usually connect to the CO via multiplexed links in
the T-carrier DS1 (24 channel) format, but interface the 2500 sets
of their subscribers over individual copper pairs.
Optical fiber for much bigger multiplexed links between remote line
groups and the CO group selector is beginning to be used to good
effect. However, "fiber to the curb" and "fiber to the home" are
also being seriously discussed. Fiber to or near the home would
appear hard to justify for conventional telephony which can only use
a microscopic portion of the bandwidth available. The twin problems
of getting the customer to buy expensive customer premises equipment
(CPE) to terminate the fiber and then provide reliable operation
(including power for the terminating equipment and the telephone
instruments as well) would appear to be formidable. Using radio,
similar to that in cordless telephones, to complete the path to
subscribers from an SLC, concentrator or RSU "at the curb" has been
suggested as a possible approach.
The
only possible rationale for running optical fiber into each home
would be to allow telephone companies to take over the delivery of
information which, unlike voice or data, might use some reasonable
portion of the bandwidth available; television is the only candidate
in sight, and the CATV industry may not look with favor on such
competition. On the other hand, there are efforts afoot whereby the
telephone companies will offer to use their advanced technology to
deliver the signal for CATV companies, relieving the latter of the
cost and effort of managing complex outside plant. If this does not
work, the telephone companies may rent CATV channels to competitive
"information providers." Thus fiber to the curb or even to the home
offers many new possibilities.
With
or without optical fiber, extending digital technology to the
telephone itself is technically simple but economically difficult.
Following modern PBX practice, the codec can be moved from the
switch's line card to the telephone set so that digitized voice can
be multiplexed with digital signals from computers and other devices
and transmitted between the set and the line card; once internal to
the concentrator, whether local or remote, the digital signals can
be connected to other digital lines or digital trunks in precisely
the same format whether voice, data or image is transmitted.
ISDN
is laboriously hammering out a way to do in general what a dozen or
more PBX manufacturers have been doing with proprietary equipment
for years. The path between the switch and the user's location is
defined by the Basic Rate Interface, abbreviated as BRI, and
includes two "B" channels, each at the T-carrier DS0 rate of 64
Kbps, for voice or data, and a separate "D" channel at 16 Kbps for
signaling. The combination is referred to as "2B+D." During the
course of this chapter, we will compare BRI with the de-facto
standard of the past 40 years required by 500 or 2500 type telephone
sets. Clearly, a new standard is desirable. The possibility of
sending information at 64 (or 128) Kbps for the price of a phone
call is something of more than passing interest, although many data
professionals consider it an idea whose time has gone.
It
should be clear that the communication channels between subscribers
and switches are going to become a lot more complex and a lot more
interesting in the next few years. However, the billions of dollars
worth of copper pairs presently connecting to analog telephones (and
capable of supporting digital telephones as well) will take some
time to replace. We need to understand them as well as the more
intricate possibilities of the future.
Subscriber lines, as they have been used for a hundred years, can be
classified in many different ways, and described in terms of many
different parameters. But first and foremost, their electrical
nature must be considered. It is these electrical properties which
set limits on the area a switching system can serve and,
consequently, control the overall economics of rendering telephone
service.
Description of subscriber lines
A
subscriber line consists of two copper conductors, insulated from
each other and twisted together to minimize the pickup of noise from
electrical and magnetic fields through which the wire must pass.
Today, pairs of wires are seldom run singly; most wires run in
cables composed of hundreds or thousands of pairs. These cables
begin at the main distributing frame (MDF) in the CO (see Chapter
7), and extend out into the environs. From time to time, they are
spliced to smaller cables and, ultimately, to cables composed of a
few tens of pairs provided with periodic distribution terminals to
which "drop wires" to individual homes or offices can be connected.
Drop wires terminate in telco-provided jacks such as the RJ-11 or
some other approved CPE interface.
The
MDF is where this outside plant is cross-connected by means of
jumper wires to switching equipment and other inside plant. The MDF
also contains protection devices to limit the effects of crosses to
power lines, hits by lightning, etc., and it affords convenient
access for service observing and making tests and measurements.
The
two wires that compose a subscriber pair are named "tip" and "ring"
after the matching portions of the plug used to make connections in
manual switchboards. The tip side usually goes to ground through the
CO equipment, while the ring goes to battery (-50 volts). Within
electromechanical switches such as SXS and crossbar, a third wire is
run for control purposes. It is called the "sleeve," again after the
part of the plug to which it connected in manual days. The sleeve is
used to hold relay-like switches operated once a path from one line
to another is established; at such times it is either grounded or
connected to battery, depending on system design, and a "busy test"
is performed by checking for the presence of the appropriate
voltage. The tip, ring and sleeve notation is not universal; some
systems used +, - and p for private in older circuits or c for
control in newer ones to describe the same three wires.
Lines from a PBX to its station equipment are quite different from
central office lines. In the first place, they are usually inside a
building rather than outside, and thus are not exposed to weather
and lightning. Second, their physical length is shorter, perhaps
2000 feet and often a great deal less, typically serving only a
floor or two in an office building. And finally, PBX wiring often
uses more than the single pair typical of residential wiring,
because a business telephone system is required to perform services
unnecessary in residential use and multiple pairs of wires
simplified such services in the days before inexpensive and reliable
electronics.
Only
a few years ago, the principal business telephone was part of a 1A2
Key System, to be discussed in Chapter 5. A regular 6 button
telephone set (hold, line 1, line 2, line 3, intercom and buzzer,
for example) required 25 pairs, and sets with more buttons required
even larger cables. Key systems could be used behind PBXs, or
directly on CO lines. In small PBX installations prior to about
1980, it was often desirable to locate the key telephone control
equipment, called a Key Service Unit or KSU, in the same room with
the PBX switch itself, to simplify maintenance and assure
flexibility. A single pair per extension number, obviously quite
short, ran from the PBX to the KSU, while the wiring from the KSU to
the multi-button telephone sets consisted almost exclusively of
25-pair (or larger) cables, one cable per instrument.
In
larger PBX installations, or when key equipment was served directly
by a central office, multi-pair cable length was minimized by
locating KSUs in closets near their instruments, and the single pair
per line was extended to the switch. Whether the KSU was near the
PBX or remote, PBX wiring tended to be shorter than CO wiring, and
wiring for businesses tended to require more pairs than residential
service. As will be elaborated in Chapter 5, modern PBX and CO
switch design attempts to provide the features of 1A2 while
minimizing the number of wires and the amount of hardware actually
used as well as the cost of administration in the face of constant
changes.
ISDN
BRI wiring between switch and telephone is initially intended to use
existing copper wire as much as possible when serving residential
and small business customers. For these, preliminary planning calls
for a single pair per line from a port on the switching matrix to
and through the MDF to a "BRI-U" interface at the customer's
premises. ISDN telephone sets will require two pairs for
transmission, one pair in each direction. These pairs are inside
wiring only; they may not be exposed to the outside environment.
They terminate on an S or T interface.
Between the S/T and U interfaces, there is a Network Termination or
NT. An NT1 is a single circuit board per line which makes the direct
T-U conversion. NT2 describes a PBX or RSU which includes switching
capability with the S interface as part of its line cards. The S/T
interface can be arranged to serve a single telephone, or the
so-called "passive bus" that permits a BRI to serve up to 8
telecommunications devices. Set power can also be supplied from the
NT via a third pair. In general, PBXs and RSUs will connect to a CO
switch via T-carrier, ultimately standardizing in a Primary Rate
Interface (PRI), to be discussed in Chapter 4; only small PBXs or
those with special requirements are likely to have matrix ports for
a BRI-U interface to CO lines.
Unfortunately, this relatively simple picture is clouded by a number
of possible variations. In the first place, telephone sets can be
built to meet the BRI-U interface directly, eliminating a seperate
NT device. Second, power can and probably will be made available to
the telephone set via the single pair U interface, or simplexed onto
the two pairs of the S/T interface. Finally, there are many digital
line coding schemes to contend with the one presumably selected
(2B1Q). Thus one must approach ISDN standards with great care.
Electrical properties of subscriber lines
Copper conductors resist the flow of electrical current. The amount
of this resistance is measured in "ohms," where an ohm is the amount
of resistance that will limit the flow of current from a one volt
battery to one ampere. Most switching systems today use a 50 volt
battery with its positive terminal grounded, and current is more
conveniently measured in milliamperes (abbreviated mA), one
one-thousandth of an ampere.
When
a conventional analog telephone's handset is picked up, the
hook-switch completes the path between tip an ring to power the set
and request service from the CO. The carbon microphone requires a
minimum of 23 mA to function properly. From ohm's law, the maximum
resistance of a subscriber line, including the telephone set and the
central office switching equipment, is nominally 50 volts divided by
23 mA, or about 2200 ohms. For many years the central office
"battery feed" resistance was made about 400 ohms and the telephone
set included another 200 ohms, so only about 1600 ohms, maximum, was
left for the copper in the "loop" or pair of wires from switch to
phone. The actual limit was usually given as 1300 ohms to allow for
battery voltage variations, temperature, etc. There were many
efforts to extend this range, but they were generally uneconomical,
particularly in competition with modern subscriber loop carrier and
concentrators.
Wire
size used in telephone cables ranges from 19 to 26 gauge, where the
higher number indicates higher resistance (smaller diameter wire).
For 19 gauge wire, the loop resistance per mile (that is, the
resistance of both conductors, added), is about 85 ohms at 68
degrees F.; for 26 gauge, it is 431 ohms. With 19 gauge wire, a
range of 18 to 19 miles would be possible if 23 mA is to be
available at the set; with 26 gauge, range drops to less than four
miles. Note that there are many other factors that control loop
length.
Perhaps the next most important such factor is "leakage resistance."
The insulation between two conductors is not perfect and, under
trouble conditions, can become quite low. Even dust on the terminals
of the MDF can have an effect. Leakage between tip and ring can
steal current away from the telephone set but, more important, it
can trick the switching system into thinking a line has originated
or answered a call when it has not.
Because high leakage resistance is expensive to maintain, leakage
resistances on the order of 15,000 to 20,000 ohms are permitted.
When open wire lines existed (bare conductors supported by glass
insulators on telephone poles as opposed to insulated twisted pairs
in cables), leakages as low as 10,000 ohms were encountered. A line
with 10,000 ohms leakage will draw 5 mA from the 50 volt battery,
even when the phone is on hook. Add to this the current induced in
the telephone line by nearby power lines (longitudinal induction)
and the alternating current that is used to ring the telephone bell,
and the problem of differentiating between on-hook and off-hook can
be appreciated.
A
third electrical factor that enters into subscriber line limitations
is "capacitance." When two conductors are separated by an insulator,
an electric charge is placed on this insulation by any voltage
between the two conductors. The amount of charge that a battery of
given size can put on the insulator depends on capacitance, the
capacity of the insulator to accept charge. Air, a very good
insulator, has a very low capacitance. Other insulating materials
such as cloth, rubber, plastic, etc., have much higher capacitance.
The unit of capacitance is the microfarad (millionth part of a
farad), and abbreviated µFd
("u" can be used when the Greek mu is not available).
Telephone cables for customer lines designed to have about 0.08µFd per
mile.
The
significance of capacitance is that it tends to act like a leakage
resistance for voice frequency currents, and the higher the
frequency, the larger the leakage current. If high frequencies are
selectively shorted out, it is evident that distortion will occur.
Further, the same capacitance can also distort dial pulses. Note
that the longer the line, the more capacitance there is to attenuate
higher frequencies. Obviously, capacitance sets a limit to the range
of frequencies that a pair of given length can carry, or how long a
pair can be if it is to carry a given bandwidth. Capacitance will be
particularly important when high speed digital signals are needed
for ISDN.
There are other sources of capacitance that affect design. The
cables that run through the exchange area are wired in various ways
to insure maximum use and minimum cost. For instance, one pair can
be accessed at several different points along its length by a
customer drop or a pair in another cable. When a pair is tapped
before its end, the unused length constitutes a "bridged tap" and
acts as lumped capacitance at the point where the working pair is
connected. This lumped capacitance, like the distributed capacitance
of the working pair itself, is harmful to transmission, particularly
at higher frequencies.
Another source of lumped capacitance is the telephone set. When
extensions are present or party line service is provided, a number
of instruments will be connected ("bridged") across a given pair.
Each has a ringer across the line, with a capacitor in series to
prevent the ringer drawing current from the CO battery. The ringer
and its capacitor don't affect voice currents, but can distort dial
pulses.
Students of electricity know that there is a third electrical
element, inductance, which acts, in many ways, as the opposite of
capacitance. In particular, inductance acts like a resistor that
gets larger as frequency increases; the ringer, for instance, is
mostly inductance and acts like an open circuit to voice
frequencies above about 300 Hz. In telephone lines themselves,
inductance is negligible. In 1900, however, Pupin showed that, by
adding small elements of inductance (loading coils) at regular
intervals along a line, the effects of distributed capacitance could
be greatly reduced. Loading coils today are still used on long
lines, particularly in rural areas, but advances in subscriber set
design coupled with various CO advances tend to reduce the need for
loading. Unfortunately, both loading coils and bridged taps
sometimes escape telephone company records and remain in place,
causing enormous difficulties for data transmission and ISDN.
Audio compensation in the 500/2500 telephone set is based on current
in the line. Long loops (high resistance) draw small currents when
the phone is off-hook, while short loops draw high current. On short
loops, the high current causes both the carbon transmitter and the
DTMF signaling pad to generate a relatively low level signal. On
long loops, they operate at a higher level which is attenuated back
to standard by loop loses. Similarly, level in the receiver is
reduced on short loops, and not reduced on long loops.
With
PBX switching, this compensation leads to some interesting problems.
PBX telephones are generally quite close to their switch; therefor
local battery for intra-PBX calls, in older systems, caused very
nearly the maximum value of line current to flow and reduced the
audio level appropriately. On external calls, however, the distance
to the CO was often large, and a small current was needed to
increase the speech level to compensate. In PBXs with metallic
matrices (see Chapter 1), this was accomplished by simply making a
direct connection between tip and ring to the telephone and tip and
ring to the CO so that the telephone obtained current from the CO
rather than the PBX (a relay was inserted in series with the line to
detect hang-up). Thus the telephone automatically produced the right
audio level depending on whether it was connected to PBX or CO
battery.
PBXs
with electronic matrices usually provide power to the set from their
line cards; when line cards are designed to serve 8 or 16 lines,
they can easily overheat when several phones are off hook at the
same time. As a result, such line cards limit current to 30 or 40 mA.
This, however, makes the audio level too high on intra-PBX calls,
and the matrix must introduce about 6 dB of loss. However, when a CO
connection is made, matrix loss is reduced to allow for the longer
circuit. One of the advantages of digital telephones with codecs in
the set is that audio level, once encoded, is not affected by loop
length.
Unfortunately, certain trouble conditions can ground even a short
telephone line; when this happens, excessive current will flow. In
older switches, half the battery feed resistance can be shorted out
by such a ground, leaving only 200 ohms to limit trouble current.
Under these conditions, 12.5 watts must be dissipated in the line,
trunk, or connector circuit, or whatever connects battery to the
line; this requirement obviously places a minimum limit on the
physical size of a particular component if it is to get rid of the
heat and be ready to work when the short is removed. In modern
switching systems with electronic matrices and line circuits, more
sophisticated means are used to limit currents and make higher
circuit densities possible.
There is one additional parameter of a telephone line to consider:
characteristic impedance. Impedance is like resistance, but it also
takes into account both capacitance and inductance. Characteristic
impedance of a line is the impedance seen when a measurement is made
between tip and ring at the central office. It is important because
hybrid circuits associated with 4-wire trunks (see Chapter 1)
contain a terminating network that must be equal to the
characteristic impedance of a two-wire line to produce a high
"return loss balance." If a good balance is not obtained, echoes
will be returned to the far end of the trunk.
The
characteristic impedance of a telephone line is a complicated
function of its length, its lumped and distributed capacitance, its
loading coils (if any), and the one or more telephone sets that may
be on-hook or off-hook. Further, it is different at each frequency
at which the measurement is made. Thus, as discussed in Chapter 1, a
compromise network used when a trunk is switched through a two-wire
metallic matrix to any line on the switch leaves something to be
desired. Local COs use 900 ohms in series with 2.14
mFd, while two-wire toll
switches, when they existed, used 600 ohms plus the capacitor.
When
a digital switching matrix is used, as with modern COs and PBXs, the
hybrid problem is somewhat different. A digital matrix cannot extend
the two-wire line through to the trunk. The matrix itself must be
four-wire and, as a result, the hybrid has to be provided on a
per-line basis. Thus, within limits, its balance network can be
tailored to the particular line it serves. Several different balance
networks have been suggested. If the line is four-wire all the way
to the telephone set, as in certain military networks, many PBXs,
and at the ISDN S/T interface, there is no hybrid and no electrical
echo path, eliminating much of the problem. There is, however, and
acoustic echo path from the receiver to the transmitter which must
be considered in set design, and hands-free telephones and sets
designed for teleconferencing require echo suppressors or, better,
echo cancelers, to deal with acoustic echo.
Use
of a separate pair for each direction of transmission can be thought
of as space division. Time division can also create a 4-wire
facility, but needs only a single pair. In some PBXs, "Time
Compression Multiplexing" sends information in alternate directions
in different time slots (ping pong; digitized voice is sent twice as
fast each way for half as long). Loop length is limited by the time
it takes each sample to travel between line card and phone, but this
is usually no problem with PBXs.
The
apparent winner in the race to produce four-wire transmission on a
single pair, 2B1Q, is a "higher level" hybrid approach. Pioneered on
some PBXs, it will be the ISDN standard between the BRI-U interface
and the transmission channel to the matrix port. The digital bit
stream is sent in both directions simultaneously, and an electronic
hybrid at each end, coupled with echo cancellation, provides
separation. The channels so derived are four-wire end to end,
assuming the digital pulse streams behave properly.
Optical fiber to the curb or home can be installed in a number of
ways. In most systems suggested for telephone company use in the
early 1990s, CATV justifies the bandwidth available, and the channel
selector, controlled from the customer's premises, is located in the
CO or some other centralized distribution point. Thus the DS0
channels from the telephone switch and the TV program chosen by the
channel selector must be merged into one bit stream (in each
direction) on optical fiber, and separated at or near the user's
location.
Some
plans have called for a pedestal at the curb serving perhaps 8
residences, delivering digitized TV via very short but traditional
coaxial cable, and analog voice with conventional battery feed and
ringing over short copper pairs. The path from the pedestal to the
distribution point is via two optical fibers, one for each
direction. These distribution points, which serve as concentrators
for pedestals, home in turn on the CO. Their hardware interfaces
customer DS0 channels to the CO switch like the remote unit of a
digital subscriber loop carrier system on copper or glass, and
brings multiple TV channels from a central information source to the
channel selectors via optical fiber.
This
sort of approach, when carefully planned, allows optical fiber to
compete financially with copper; when the cost of optical fiber
drops, we can expect a single fiber from each home to the CO, using
either frequency division multiplexing (called wavelength division
multiplexing in optical work) so that different colored light can be
used in each direction, or with directions separated using the
optical equivalent of a hybrid on each end. Ultimately, it is
possible that a single CO switch will be able to handle both
traditional telephone traffic and TV program distribution.
Factors limiting range
The
maximum length of copper subscriber lines used with analog phones is
controlled by four factors: supervision, signaling to the CO,
signaling from the CO, and transmission. Each range is measured
separately so that if improvements are made in the factor known to
be limiting, the point where the other factors take over can be
compared. Measurements are made under worst case conditions, and
both the ability to respond to a valid signal and to reject an
invalid signal must always be checked. It must be remembered,
however, that station carrier, concentrators and RSUs, based upon
inexpensive and reliable electronics available just as labor costs
are soaring, are changing the traditional economic assumptions about
local plant that have been standard for a hundred years. It is no
longer obvious that squeezing the last few hundred feet out of
copper wires is worth the design effort, or that installing new
pairs can compete with new channels produced by adding electronics
to existing pairs. As a result, increasing the range of subscriber
lines will almost certainly take second place to "distributed"
switching in new switch designs.
ISDN
BRI lines (to the U interface) were limited to about 12 kilofeet
(with no bridged taps or loading coils) by their need to handle high
speed pulse trains in both directions simultaneously using an early
line coding called AMI. The current standard for line coding, 2B1Q,
can reach to about 19 kilofeet. In 2B1Q, two binary digits are coded
into one quaternary signal: the digits 00 are coded as a -3 volt
pulse, 01 as -1 volt, 11 as +1 volt, and 10 as +3 volts, measured at
the coder. To handle the BRI's 2B channels plus a D channel (2x64+16
Kbps), 144 Kbps are required. To these must be added another 16 Kbps
for frame synchronization and an M (for maintenance) channel,
bringing the total to 160 Kbps. By using 2B1Q coding, where four
possible signals can be sent but each signal contains twice as much
information, only half as many signals need be sent per unit of
time. Thus signal speed on the line is reduced to 80 kilobaud in
each direction.
Range is limited only by the ability to maintain the 2B1Q bit
stream; transmission for voice and data is incorporated into the two
B (bearer) channels, and the D channel conveys the packets used for
signaling, supervision, alerting and, in some instances, customer
data. Call progress tones and recorded announcements will still be
provided from the switch or the connected network for the
convenience of humans, but D channel messages will be required for
the benefit of computers, terminals, and ISDN telephones with their
buttons and displays. The point, however, is that once the 2B1Q bit
stream (or whatever successors follow it in the future) is
established, all of the limiting factors so important to analog
telephones cease to apply.
From
a functional standpoint, subscriber lines can be classified in a
number of different ways. There are business and residential lines,
individual and party lines, private and public lines. These
categories overlap and have to be sorted out.
Business vs. residential lines
The
great majority of lines are residential, but business lines,
although numerically small, generate far more daytime telephone
traffic, both local and long distance, establishing the proportions
of the public telephone networks. Indeed, long distance rates are
lowered at night to encourage residential customers to place their
calls after business hours, reducing the size of the networks
required for both business and residential calling, and making
better use of equipment which otherwise might lie idle in off-peak
hours.
Business lines are tariffed differently from residential lines. It
is often stated that "value of service" makes business lines more
expensive than residential lines and thus that business service
subsidizes residential. Certainly a modern business would find it
difficult to get along without telephones, but because businesses
generate more traffic per line, on the average, than residences, the
concentration ratio in the CO switch must be appreciably smaller,
making the cost per port of the switching matrix greater. Further,
businesses need many features that are not meaningful in residential
service. Thus higher costs for business lines seem inevitable, based
on what it costs to provide the required services.
Flat-rate local service for business lines is now almost extinct;
business customers have little choice but message-rate where they
pay so much per line plus so much per local call. Residential
customers seem to prefer flat-rate service (at a higher monthly cost
but allowing unlimited local calling) to message-rate, but
geographically larger local communities of interest produced by the
automobile, the ease of per-call billing in modern systems (see
Chapter 2), the need to provide "life-line" service at the lowest
possible monthly cost for the poor, and the telco-perceived fairness
of usage-sensitive pricing are slowly eroding what flat rate service
is left.
There are occasions where both business and residential customers
need to have service provided from a central office other than the
one nearest them. Lines to such a remote CO are called FX lines, for
foreign exchange. Because they usually have to be extended to the
distant office through one or more carrier systems, they pose
special signaling problems. The user pays a monthly fee based on
mileage in addition to the distant office's flat-rate or
message-rate charge.
Lines connecting a CO to a customer's PBX are called CO trunks at
the PBX end and will be discussed further in Chapter 4. Because the
PBX concentrates traffic from its many extensions, CO trunks tend to
have a very high occupancy. Centrex, which is PBX service (plus DID
and IOD) provided by a CO switch, does not have CO trunks; rather,
each extension has its own pair from the user's premises to the CO.
These centrex lines have about the same traffic as PBX extensions,
usually more than most residential lines but far less than PBX-CO
trunks.
Just
as a CO switch may have FX lines to serve distant customers, a PBX
may have off-premises extensions or stations, referred to as OPXs or
OPSs, to serve users at other locations, nearby or remote.
Connections from the PBX to remote telephones (which often require
special line cards to interface outside wiring) are usually provided
by the local telephone company, and are routed back to the serving
CO where they are extended to or toward the OPX. In a certain sense,
all centrex lines are OPXs and tend, on the average, to be much
longer than PBX lines. However, for multi-location customers, with
few
lines concentrated at any one place, Centrex can have shorter loops
than a PBX with a large number of OPXs. One example is local
governments, with offices, courts, police and fire departments,
schools, libraries, etc., scattered over the municipality; Centrex,
without OPX backhauls via the central office from one PBX location,
can often realize economies.
Individual and party lines
An
individual line serves only one customer; a party line serves two
customers or more. Party lines have been used since the earliest
days of telephony; indeed, there was once a time when it was not
unusual to have a party line without a central office. A party line,
by sharing one pair of wires among several customers, makes more
efficient use of the copper; however, there is an obvious loss of
privacy and loss of accessibility inherent in sharing which prompts
an excess of customer complaints and trouble calls. Further, higher
occupancy means higher probability of incoming calls not getting
through, reducing service to the customers and revenue to local and
long distance telephone companies.
From
the designer's perspective, party lines present difficulties in
identifying the calling customer for billing, ringing the called
customer, and facilitating connections between two parties on the
same line (reverting calls). When deregulation forced customers to
own their own telephones, even Washington lawyers recognized that
customers could not be expected to set their phones up to receive
party line ringing or to provide billing identification; thus party
lines are not being encouraged.
There is an interesting similarity between party lines and "bridged
extensions" in both residential and business service. In residences
today, now that users do not have to keep paying a monthly fee for
extra instruments, it is not uncommon to have two or more phones
bridged across the line to the CO. All the phones ring with an
incoming call, and calls originated on any phone are billed to the
same number. Now, however, any conveniently located phone can be
used to make or answer a call, and several phones can be used at
once to make a "conference call" with the distant party.
Although analog 2-wire phones work quite satisfactorily in bridged
connections, the proprietary digital phones designed for specific
PBXs do not. Each digital PBX phone must have its own port on the
switching matrix, and the conference function inherent in analog
bridging is carried out via a PBX conference circuit.
ISDN
BRI standards allow an S/T interface to support up to eight devices
(digital phones, computers, terminals, printers, etc.) in an
arrangement similar in some ways to a party line. Signaling on the
separate D channel greatly simplifies most party-line problems, but
one phone cannot use a "bridging" approach to "pick up" on a call
already in progress. When this function is required, each phone uses
one of the two B channels, and both local parties and the distant
party are tied together through a CO conference circuit.
Multi-line groups
Just
as a party line used to serve several subscribers, multi-line groups
allow one subscriber to be served by several individual lines. When
several lines are used as a group, whether they are CO trunks to a
PBX or key system or a group of extensions behind a PBX or Centrex
switch, it is often convenient to have one directory number identify
the group. This leads to line hunting and the similar call
forwarding features which will be discussed in detail in Chapter 5.
Private vs. public lines
All
the lines described so far may be thought of as "private" in the
sense that they are provided for the service of a particular
residence or business. (The term "private line" is also applied to
tie trunks between two PBXs and to point-to-point connections which
do not terminate on a switch, either PBX or CO. We will not concern
ourselves with this usage here.) There are, however, "public" lines,
provided for the use of the general public. The familiar pay
telephone is the most common of these, but there are other types as
well. House phones in a hotel may be considered public phones; other
public phones are found in airports and bus and train stations as
"hot lines" (see Chapter 5) for calling taxis, hotel reservation
desks, and the like.
Starting in June, 1984, it became possible for anybody to install
and operate pay phones, just like any coin-operated vending machine.
Pay phones operated by the telephone company can take advantage of
the switching system's capability and the availability of operators.
Privately owned phones had to have elaborate computer control built
into them just to handle coin calls, adding to their cost. Then, the
various competing long distance carriers demanded that they each
have a share of the long distance revenue generated from pay phones.
This
was not too complicated with coin calls, but when customers insisted
on making credit card, collect, and person to person calls, as well
as calls billed to a third party, often involving a different
carrier than the one assigned to the coin phone, considerable
difficulty arose in providing appropriate operator assistance. At
the present time, the local telephone companies, the long distance
carriers and the owners of private coin phones all provide
operators, and programs for CO switches have become unbelievably
complex (at rate-payer expense) to provide the various vendors with
a "level playing field."
Functions such as supervision, dialing, ringing, ring trip and
charging are basic to all switching systems. Because they involve
direct interfacing with lines, they are discussed in this chapter;
transmission is discussed in the next chapter along with trunk
functions, while the more complex and exotic functions are taken up
later. ISDN phones will work quite differently from analog phones,
taking full advantage of the possibilities implicit in 2B+D via the
2B1Q 80 Kbaud bit stream. However, 500 and 2500 type phones, and
phones made to the same specifications with more modern hardware,
now owned by subscribers, will be with us for at least another 50
years. Thus it is important to understand the complexities their
supposedly simpler functions impose on the designer.
Supervision
Line
supervision determines the busy or idle status of each line served
by a switch. As has been discussed, the transition from idle to busy
or busy to idle is what causes the switching system to initiate some
action. The terms on-hook and off-hook (referring to the action of
picking up or replacing the receiver of a candle-stick phone, or the
combined handset holding both the transmitter and receiver of a
"French" phone) are generally used interchangeably with idle and
busy, even though a ringing phone is on-hook and busy. Such terms
seem inappropriate when hands-free phones do not go "off hook,"
features such as on-hook dialing do not require the handset to be
picked up until the called party answers, and computers, fax
machines, and a variety of other devices using the telephone system
go from busy to idle and vice versa without operating a conventional
switch hook at all; never-the-less, standard usage will be followed
here.
For
analog phones, sensors are associated with each line at the CO to
discriminate accurately between the two states, whatever they are
called, independent of line length, noise pick-up, leakage
resistance and a wide range of other parameters. Because the
connection between physical lines and ports on the switching matrix
is constantly being modified at the MDF, sensors must be designed so
that the same one will work properly on any line to which it may be
assigned.
Origination. In all systems with metallic
switching matrices, only call originations are detected on a
per-line basis. Once the origination has been acted upon, the
line-circuit sensor is disconnected by "cut-off" contacts and
supervision is passed through the switching matrix to other circuits
(selectors, connectors, originating registers, trunk circuits,
etc.). This checks the continuity of the matrix path, and also
allows line-circuit sensors to be relatively inexpensive because
they have only one function and are never in the circuit during
times when transmission takes place.
Traditionally, line relays have been used to detect originations. A
relay has four parameters of importance; for reasons of economy, it
is highly desirable that as few of these parameters be specified as
is consistent with system operation.
1.
Operate current: the minimum required to operate any properly
functioning device of specified type.
2.
Non-operate current: the maximum current that will not operate any
properly functioning device of the specified type.
3.
Hold current: the minimum current that will keep any previously
operated device of the specified type operated.
4.
Release current: the maximum current that will allow any previously
operated device of the specified type to release.
Because the magnetic structure of a relay changes when it changes
state, "hysteresis" effects make the operate current greater than
the hold current, and release current less than non-operate. Other
types of sensors, including some types of electronic circuits, may
not be as adversely affected. All sensors of a given type, however,
differ somewhat from one another, and all have some region of
uncertainty where it is not clear which output a given input will
produce. Because it is equally important that a sensor not operate
on leakage or noise at maximum battery and operate properly on the
longest line when the system battery is at its minimum value,
reduction of the region of uncertainty for a family of sensors is
required for greater range. More sensitivity will not work because
it will only make false operation more likely.
A
line relay that detects originations needs only the operate and
non-operate currents specified as long as it releases on zero
current when it is disconnected by the cutoff contacts. Other
sensors, supervising a call for switch-hook flashes, dial pulses or
hang-up, must operate and release over the loop in the presence of
leakage current, but there are many fewer of them; thus they can be
given a more complex design and still save money.
When
supervision is switched through the matrix, a minor but interesting
problem called "showering" or "cascading," can occur. Showering
takes place when the sensor in the line circuit is more sensitive
than the sensor in the originating register or other circuit to
which the line may be connected. In trouble situations (moisture in
a cable, wet leaves on open wire, etc.) the line sensor sees an
origination and flags the system. The system establishes a
connection to a selector switch or register, but the sensor there
does not respond. The connected circuit releases and the line
circuit is reattached. It sees off-hook again, and the cycle repeats
indefinitely.
The
term "showering" originated in SXS systems where linefinders would
hunt to supposed originations and cut through to selectors; when the
selectors failed to accept the originations, the linefinders would
fall down in a readily observable shower.
To
deal with showering, it was customary to make matrix-connected
sensors operate on less current than line sensors. Then a "permanent
signal" would result which could be detected by a time-out. In a
stored program system such as 1ESS, where control intelligence
observes both the line and the connected circuit, showering is
handled by deliberately making the line sensors more sensitive than
those connected via the matrix. Then, leakage currents force
showering which the system control can easily detect and deal with.
Because all digital matrices and most electronic analog matrices are
unable to extend a dc connection to a digit receiver or trunk
circuit, showering is no longer a problem. However, their line
circuits must be designed to detect originations, flashes, dial
pulses and hang-up, provide for test access and ringing application,
offer toll-grade transmission when the phone is in use, etc., all of
which lead to a much higher cost per-line than the old line and
cut-off relays. This cost is more than compensated by the advantages
of electronic over metallic matrices as discussed in Chapter 1.
Northern Telecom's DMS-100 CO switch has a selection of 3"x3" line
cards, each of which terminates a single line. There are different
cards for conventional telephones, coin phones, proprietary
multi-button telephones (with an analog voice channel and a control
over voice (COV) signaling channel), and ISDN telephones. AT&T's
5ESS follows this procedure when ISDN phones are to be provided, but
designed a space-division two-wire concentrator for conventional
phones using a special electronic crosspoint that behaves almost
exactly like a metallic crosspoint. The new crosspoint can pass dc
supervision and battery feed through the matrix, along with ringing
and test signals. This preserves the traditional economies of simple
line circuits followed by concentration to a smaller number of smart
circuits, but different line groups are needed for different kinds
of customer equipment.
Loop start lines. The great majority of
analog CO lines and all analog PBX extension lines are loop start.
That is, battery and ground are supplied at the switch and, when a
connection is made between tip and ring at the station, current
flows. This current operates a line sensor to inform the system that
a call is being originated. In many systems, the line sensor has two
equal windings, one in each side of the line; this balances out
"longitudinal" noises by having them affect both windings equally
but in opposite directions. Only current that flows from ground
through one winding to tip, around the loop returning on ring, and
to the -50 volt battery via the other winding will be recognized.
Because balanced sensors tend to be expensive, single winding
sensors in the ring-side of the line alone, specially designed to be
unaffected by 60 Hz longitudinal induction, have been used with
metallic matrices. A single-winding sensor has an additional
advantage: a tip-to-ground short will not reduce its effectiveness
and thus deny service to a line that has such a trouble. After an
origination is detected, the switch goes to a balanced path by
transferring supervision through the matrix to a trunk or service
circuit.
Ground start lines. Coin telephones and CO
trunks from PBXs often use ground start lines. Here a one-winding
sensor is placed in the ring side (to battery) at the CO, and the
tip-side path to ground is left open. An origination is produced by
user equipment connecting ring to ground rather than to tip. After
origination is detected, the CO goes to a balanced connection,
adding a ground on tip. A PBX trunk circuit will then remove its
ground on ring and change to a tip-ring connection. Coin-first pay
telephones used the coin to make a ground connection via the coin
control magnet in the set. This was a large inductance in series
with several thousand ohms resistance; connected to a center-tap in
the voice circuit, it was not removed after recognizing ground on
tip as the CO's acknowledgment of seizure.
Ground start lines are valuable in that they leave the tip side of
the line free for independent signaling to the station from the CO.
In coin phones, the system can apply the coin collect or coin return
signal to the tip side of the line to operate the coin magnet,
regardless of whether the coin phone is on-hook or off-hook. On
outgoing calls from a PBX, ground on tip is the equivalent of dial
tone and can be detected by a relay rather than a tone detector. On
incoming calls to the PBX, a tip-side sensor sees this ground at
once and makes the line busy to its own outgoing traffic; it does
not have to wait for ringing to start. This is important so that
users dialing 9 will not meet incoming calls that should go to the
console attendant. At the end of a call to or from a PBX, the CO
removes ground from tip to send a positive signal that the connected
party has hung up. If the CO reseizes the trunk very quickly, the
PBX may not see the open tip; thus detection of ringing is a valid
check of a new incoming call.
It
the days of metallic matrices, it was typical practice to make about
20% of the lines on a CO switch ground start. This means the tip was
left open in the line circuit, and usually that a different sensor
was connected from ring to battery. The DMS-100, with its electronic
matrix, can plug in appropriate line cards as needed. The 5ESS, with
its space division concentrator, has separately controlled contacts
to cut off the ring-side sensor and the tip-side ground. Both are
operated and released together on loop-start lines, but on ground
start lines, the tip-side cut-off contact is left open, under
software control. Thus system software can configure the same line
circuit as loop-start or ground-start as required.
CO
trunk circuits in PBXs have to be able to work properly with lines
that are loop-start or ground-start at the CO. Although most CO
trunks are ground start, there are occasions when the PBX needs a
loop start line. It is current practice to make PBX CO trunk
circuits software configurable to be loop-start or ground-start as
needed.
With
the coming of ISDN and its D-channel signaling, loop-start and
ground-start, at both the PBX and CO, can be relegated to the
museum. This may be accomplished by 2025.
Holding a connection. Once a connection is
established, it must be monitored for hang-up. Because there is no
way to predict which party will hang up first, or what momentary
hang-ups from either end may be encountered even though the call is
expected to continue, the problem is by no means trivial. In
general, the "calling party hold" rule applies; in the days of
manual PBX switchboards, when the attendant had to transfer incoming
calls by moving a cord from one jack to another, it was important
not to lose the caller. Similarly, in residential service, it is
important for the called party to be able to hang up one phone and
pick up the call on a different extension.
Calling party hold used to be absolute until it was noticed that one
person could tie up another's line (usually a competitor's) by
dialing the number from a pay phone, waiting for answer, and then
leaving the calling phone off-hook (this is sometimes called "the
delicatessen effect"). Present practice calls for the connection to
be held for about 10 seconds minimum after the called party hangs up
when the calling party stays off-hook; after the time-out, the
connection is released. If the caller hangs up prior to the time
out, the connection is released immediately. At one time, the called
party, staying off hook when the caller hung up, would get dial tone
immediately; similarly, dial tone would be returned to the off-hook
caller after called party hung up and the system timed out. Many
central offices still follow this procedure, although the
recommended approach, easy with modern switching systems and highly
desirable in the face of increasing toll abuse, is to not return
dial tone to either party until they hang up and come off hook
again.
When
transfer (and other features) are activated by a switch-hook flash,
either party must be able to flash without losing the connection.
With calling party hold, this is easy enough for the called party,
but when a secretary places a call and transfers it to a principal,
the calling party is flashing the switch-hook and can lose the
connection. Indeed, many electronic PBXs in the 1970-80 period
simply could not transfer outgoing or intra-switch calls, limiting
their utility and confusing their users.
Certain emergency numbers, typically those of fire and police
departments, and 911 answering points when available, are provided
with "joint holding," so that the called party can hold the
connection when the caller hangs up in a panic without providing
sufficient information. Provisions are also made to ring back on the
calling line, whether on-hook or off-hook. The need for joint
holding is doubtless being reduced by the feature "Calling Number
ID" which, through the use of modern signaling, allows the calling
number to be shown on a suitable display in or associated with the
called phone. When the system's data base can also provide the name
and address associated with the calling number, 911 and other
emergency service will be greatly improved; however, the loss of
information from callers who want to be anonymous must also be
considered.
When
a call originates or terminates behind an electronic PBX via a loop
start CO trunk, the PBX may not get a signal from the CO when the
distant party hangs up. At best, it may get a momentary open when
the CO changes from monitoring for hang-up at the trunk circuit or
connector to monitoring for a new origination when the cut-off
contacts reconnect the line sensor. Thus the PBX must depend on its
connected extension to know when the call is over, a problem when
the PBX happens to be making a connection between two CO trunks. If
the PBX extension is on hold or connected to an answering machine or
a fax, it may stay off hook indefinitely, holding the PBX matrix
connection to the CO trunk. In general, ground start trunks with
their open tip after far-end hang-up are vastly superior to loop
start trunks, but ultimately, improved signaling via the D channel
in an ISDN BRI or PRI should eliminate the whole problem.
Ring-down supervision. When each telephone
set had its own battery and, as a result, did not draw power which
could be monitored for supervision from a central point, ringing was
used to establish connections. Each telephone set and switchboard
had its own "magneto," or small ac generator which produced about 80
volts at 20 Hz; the magneto was hand-cranked to produce coded
ringing (long short short, or short long, for instance) at all the
phones on a party line. Subscribers, upon recognizing their ring,
would then answer. A single long ring was used to call the operator
at "central."
At
the switchboard, each line was terminated in a jack and bridged by a
"drop," or ac operated relay-like device not to be confused with a
"drop wire" from a telephone junction box to the customer's
premises. The drop, in the presence of a sustained burst of ac
current from the magneto, would pull up on a small catch that let a
target pivot out where it could be seen. Once released, the target
stayed in position until restored by the operator plugging a cord
into the jack to respond to the caller. Note that the drop was "slow
operate" so that it would not be affected by party-line ringing.
"Ringdown
supervision" was also used on trunks; it is interesting in that only
a "spurt" rather than a continuous condition was sent to the far end
to signify the need to take action. A spurt of ringing would get an
operator or another subscriber to answer, and when the call was
over, another spurt of ringing, generated by the user, indicated
hang-up. Operator cord circuits had "clearing out" drops built into
them to respond to this signal so that the operator would know when
to pull down the cords; other party line members (those not
listening in) heard the ding on their ringers, and knew the line was
free. The expression "ring off" comes from this procedure.
In
the modern telephone plant, the only remnant of ringdown supervision
is between central offices and PBXs and other customer premises
equipment such as answering machines and modems. Drops and
cord-boards have long since passed away, but ac operated relays or
other sensors that latch upon detecting ringing tell customer
equipment that a call is coming in.
Flashing. In the early days of telephony,
the switch-hook flash (a half-second depression of the switch-hook)
was used to recall the operator for assistance or some special
service. Although automatic switching initially caused the flash to
vanish in the public telephone networks, it is now used extensively
in PBXs and Centrex.
In
manual PBXs and SXS systems with cord boards for attendants, it was
not unusual for an incoming call to reach the wrong party. To effect
a transfer, the PBX user flashed the switch-hook to signal the
attendant to re-enter the call so that instructions could be given
to move the cord to the jack of the correct telephone. Because an
attendant might not be looking at the cord lamp when the wrong party
flashed, a feature called "flashing recall" was provided on the more
intricate manual switchboards; here a single flash by the user
caused the cord lamp to flash continuously until the attendant
responded. Note the similarity to "spurt" signaling above.
PBX
systems with consoles and, of course, centrex systems, must also
transfer calls that reach the wrong party. (Transfer in not
generally provided in CO switches, and early Centrex, based on
5XBAR, had to be hurriedly corrected.) The original procedure
required the user to flash the switch-hook to request the system to
bridge on an attendant; the attendant would then obtain the needed
information, release the incorrect party, and key in the correct
extension number so that the switch could make a new attempt. To
reduce the work load on the console attendant, various special
features, described in Chapter 5, have been developed to allow the
user to perform such tasks directly. The switch-hook flash calls in
a signaling detector (which returns a special dial tone); by dialing
or keying codes or extension numbers into the system, the user can
instruct it to do a number of things in addition to transfer.
Although residential and small business customers with only one or
few lines usually have no need for transfer, a number of other
modern features, invoked with a switch-hook flash and a feature
code, have much to offer.
With
simple analog phones, the basic problem is the need for the user to
provide additional information to the switching system. The flash is
detected by standard supervisory sensors during the course of the
call. When the system decides the momentary on-hook it has received
is real--neither a "hit" on the line nor a hang-up--it can call in
the more complex detector to accept DTMF signaling and convert it to
digital information that can be used by the control. Allowing the
"stupid" monitor to call in a "smart" monitor is one of the basic
functions of supervision in most modern switching systems, and will
remain so until the last 2500 set is retired.
Systems with only rotary dial telephones can avoid the flash and
accept dialed digits at any time during a call if the supervisory
sensor in the line or trunk circuit can respond directly to dial
pulses without expensive pulse correction circuitry and if the
system can poll all calls in progress often enough to detect and
store dial pulses as they come in (one scan every 10 milliseconds is
about right). Many new features were introduced in AT&T's Morris,
Illinois, field trial in 1960 using this approach, and several
interconnected PBXs followed suit 15 years later. Unfortunately for
the latter, when DTMF phones were mixed with rotary dial, each kind
of phone required a different procedure with matching training.
A
scanning approach has occasionally been tried with DTMF; all feature
codes are arranged to start with # or *, and a non-blocking
electronic matrix repeatedly connects a DTMF receiver across active
lines just long enough to see if a # or * is being sent. If so, the
detector pauses long enough get the subsequent digits of the feature
code, plus whatever else (such as an extension number) is needed to
do the job. DTMF detectors are sufficiently expensive to make one
per line impractical, but even if the price were lower, continuous
monitoring would almost certainly produce a considerable number of
"talk-offs."
ISDN
telephones, like the proprietary electronic phones designed for
PBXs, use the "D" or signaling channel to send a digital message
from phone to line card which is interpreted by the system as a
flash. This signal is usually activated by pushing a button labeled
"flash"; depressing the switch hook sends a different digital
message on the D channel, and may cause the connection to be
dropped.
Although electronic phones run directly by the switch seem logical,
it often happens that centrex CO switches and some older PBXs do not
support electronic phones. To meet this need, small PBXs, called
"hybrids" or electronic key systems (see chapter 5), provide
multibutton electronic telephones and interface the larger switch
via loop start CO trunks. In such circumstances, it is necessary to
have one flash to signal the small PBX to activate its features, and
a different flash which can be identified and repeated on the
appropriate CO trunk to activate features from the Centrex or large
PBX.
Returning answer supervision to calling lines.
Almost all SXS switches, whether used as COs or PBXs, returned
answer supervision to the callING party when the local callED party
answered. The supervisory relay toward the callED party in the
connector for local calls or the trunk circuit to other switches
operated on answer to cause a second relay to reverse the polarity
on tip and ring toward the callING party via the metallic path
through the switch. This reversal incremented a message register
bridged across the line for local billing. Panel and the various
Bell crossbar CO switches used more complex means not involving the
tip and ring to score users' message registers.
Reading message registers and calculating the number of calls was a
major expense item to telephone companies, so flat rate and extended
area service (EAS) rather than message unit billing was, for a time,
fashionable. The coming of AMA automated long distance billing and
then, twenty years later, stored program control made usage
sensitive local billing practical and profitable. As a result, most
modern switching systems choose to pick off answer supervision and
generate billing information internally; they do not use it to
reverse battery and pass the answer signal back via tip and ring to
message registers and, inadvertently, the calling party.
Not
sending answer supervision toward a subscriber line via a metallic
matrix path meant not needing a relay in every trunk circuit to
reverse battery to tip and ring. This, of course, implied a
considerable saving in hardware. When digital switches began to be
used in central offices, metallic matrices were replaced with
electronics and supervision had to be done on a per-line basis. By
not including a per-line relay to return answer supervision, the
saving was increased by an order of magnitude.
The
major problem caused by lack of answer supervision is the need for
accurate billing carried out at PBXs or in privately owned pay
telephones. Starting with PBXs in hotels and motels, where guests
make both local and long distance calls and must be billed for them
when they check out, a means of accurate billing (including a
positive indication that the call was answered) became necessary.
Although ground start lines, when used as CO trunks to a PBX, give
positive hang-up supervision, as discussed above, it is not usually
economical for a modern central office to return answer supervision
to a PBX or a coin phone.
In
addition to hotels and motels, businesses have found it desirable to
provide internal billing for cost allocation to various departments
for both local and toll calls. Originally, separate machines were
developed to provide Station Message Detail Recording (SMDR) from
calls originated by PBX extensions, or Call Detail Recording (CDR)
which also captured billing information from calls arriving via
tie-trunks or incoming CO trunks devoted to 800 service or direct
inward system access (DISA). These systems bridged extension lines
and trunks to obtain the information they needed to prepare a call
record. Eventually, it became obvious to some PBX designers that a
separate system paralleling a PBX to obtain information the PBX
already had was ridiculous, and they began to include the generation
of billing records as a premium feature at extra cost.
Unfortunately, other designers refused to acknowledge this mandate
from customers and went out of business. However, without answer
supervision from the CO, accurate billing was still impossible.
One
obvious solution will be the use ISDN interfaces with their D
channels to handle answer supervision along with a variety of other
things. This is beginning to happen, but it could have happened long
ago if PBXs of, say, 50 lines or more interfaced CO switches as
interoffice trunks rather than as ground-start lines and took
advantage of trunk signaling. This has been resisted by the
telephone industry because (a) many early CO switches could not make
trunk to trunk connections easily and a PBX must, of necessity, make
calls to both local lines and trunks to distant switches, and (b),
billing has always assigned charges to CO lines, not trunks. These
reasons are not valid any more; they just represent policy left over
from the iron age.
Most
computer controlled CO switches and all digital switches can make
line-line, trunk-trunk, and line-trunk connections with equal
facility and, at least in principle, can handle billing for calls
originating on trunks as well as lines. The divestiture of the Bell
operating companies from AT&T seems to be producing some movement in
this direction, although the D channel associated with PRI and BRI,
now often available, is a better solution. In the meantime,
entrepreneurs are developing schemes for simulating answer
supervision. PBX and coin phone designers are using various timing
procedures, usually assuming answer occurs so many seconds after the
last digit was dialed, while external circuitry, inserted into the
PBX's CO trunks, looks for the absence of ringback tone for more
than 4 seconds, the presence of speech on the line, etc.
Northern Telecom's DMS-100, with its variety of plug-in line
circuits, actually has one that offers the traditional battery
reversal toward the customer to return answer supervision, available
as a premium service to customers who need it. AT&T's 5ESS will not
easily be able to duplicate this approach in line groups set up for
traditional lines. However, its ISDN line groups are similar to
those of the DMS-100, support line cards for analog lines, and could
easily support a card to return answer supervision.
Note
that a PBX, on incoming calls, whether via the console or DID
(Direct Inward Dialing), is required to return answer supervision to
the telephone company so that parties originating these calls can be
properly billed. PBXs do not, however, return answer supervision to
their own extensions.
Toll diversion. One reason why answer
supervision was not generally returned to PBXs by Bell System COs
through 1, 2 and 3ESS is that another service, "toll diversion" (see
Chapter 2), had usurped the battery reversal. To prevent toll abuse
by station users, the PBX customer paid a small amount extra per CO
trunk to have that trunk class-marked as local only. When the
signaling receiver in the CO, while checking digits as dialed,
discovered the presence of a toll call, it rejected the call by
returning a battery reversal to the PBX via tip and ring through the
metallic matrix. At the PBX, the reversal caused the caller to be
connected to the console attendant or to a distinctive tone. The
attendant, if so instructed, could advance the call by means of a
trunk not toll diverted; a user receiving tone had to call the
attendant for assistance. In metropolitan areas, it was common
practice at one time for a PBX to have two CO trunk groups, one for
"dial 9" outgoing calls, toll diverted, and one for incoming calls,
which could also be used outgoing without toll diversion, but only
for calls set up by attendants.
Toll
diversion was a useful customer service from the central office only
when PBXs were relatively dumb. Today, PBXs have a number of classes
of service or classes of restriction which control calling range as
a function of the extension placing the call. Although it would seem
unthinkable for the owner of a modern PBX to have to pay extra for
CO toll diversion, many PBX designers have chosen to implement
restriction functions in ways that require the use of CO toll
diversion as well. When a CO trunk uses battery reversal for toll
diversion, it can't use the same signal for answer supervision.
Release of user equipment. Various pieces
of user equipment such as modems, answering machines and holding
bridges in 1A2 key telephone systems can only be released when the
distant party hangs up by interrupting the flow of line current
provided to them by the CO or PBX switch on which they home.
Momentary opens at hang-up were naturally available in
electromechanical switching systems between the time when the
switching matrix released the path to a connector or trunk and then
reconnected the line sensor to watch for the next origination.
Designers of key systems, in particular, believed this open to be a
natural function of switching systems and took advantage of it.
Designers of CO switches were totally unaware of this use and,
during the development of AT&T's 1ESS, found out that they had to
add timing to the call release programs to make sure the open
existed.
Switches with electronic matrices detect originations and supervise
lines from a line circuit which is always present whether the line
is busy or idle. Thus there is no way that changing the matrix
connection can provide an open to release station equipment.
Consider the instance where someone has reached an answering
machine, left a message and hung up. If the PBX or CO switch does
not send it an open, the answering machine may keep listening,
probably to dial tone, until its tape runs out. Designers of
answering machines have become quite clever in identifying such
situations and releasing on time out, dial tone detection, absence
of speech, etc., but there are certain kinds of modems which do not
work this way, and the simple holding bridge of a 1A2 key system
will send an off-hook toward its line circuit forever.
To
add a relay per line circuit to release station equipment would be
as expensive as adding a relay to return answer supervision, so
other means are sometimes found if a release signal is to be
provided. Because most analog telephones require ringing, and
ringing has to be applied via the line circuit in such a way that
its high voltage does not harm the line circuit's electronics, it is
sometimes possible to use the ringing access circuitry to provide a
station release signal. However, it is easy enough to do with
electronics what is no longer economical with relays, and many
modern line circuits have "power down" circuitry that can be
arranged to turn off the battery feed under program control. Such a
feature is normally used when a PBX is being started up or restarted
after a failure; because many phones may be off hook, the system
activates them in small increments to minimize the power surge that
otherwise could occur. However, software can be used equally well to
momentarily turn off power at the end of each call.
Solutions using the ISDN D channel will, of course, require customer
equipment capable of responding to digital signaling. Thus today's
most modern CPE, including voice and fax mail systems as well as
modems and answering machines, all designed to look like 2500
telephone sets, will have to be upgraded to be used effectively in
the modern world.
Address signaling
Transmission of the called number from an analog telephone to the CO
or PBX can be done in a number of ways, but only two are of
importance at the present time: dial pulsing and DTMF. Dial pulses
are out-of-band, while DTMF is in-band. That is, dial pulses are
electrical signals on the physical channel that will be used for
voice, but well below the limit of human hearing (although their
harmonics are clearly audible), while DTMF signals are in the band
of audio frequencies which can readily be heard and, more to the
point, transmitted via voice frequency telephone channels.
Since 1975, various PBXs have offered electronic telephones, both
analog and digital, with a separate out-of-band digital signaling
channel similar to the ISDN D channel. Digital signaling,
independent of the communication channel it serves, has a lot to
recommend it, but it must be compatible with the CPE it controls.
Unfortunately, each manufacturer of proprietary equipment has a
different protocol for signaling and the exchange of other
information between set and line card, and ISDN, which is supposed
to provide standardization in this area, is lagging sadly in its
job. Until a truly standard BRI D channel protocol is agreed upon,
phones based on the older rules designed for 2500 sets will
dominate.
Dial pulsing. Dial pulsing was developed a
century ago to work with SXS switches. Thus it is based on two
parameters of particular importance: the rate at which the pulses
representing digits can be sent, and the time required to
differentiate pulses within a digit and successive pulse trains
representing different digits. The mechanical inertia of SXS
switches made 10 pulses per second (PPS) the maximum practical pulse
rate, while interdigital timing required the system to determine
that no more pulses are coming in the current digit, switch to the
rotary mode, and then, in selectors, hunt over as many as ten
terminals to find an idle path to the next stage of switching. Over
the years, the nominal interdigital interval has become standardized
at 600 milliseconds.
When
a telephone is taken off hook, it completes the path from tip to
ring through the dial mechanism. To send dial pulses, the user turns
the dial to the point that corresponds to the desired digit while
winding up a spring. When the dial is released, the spring causes it
to return under control of a governor, and the dial mechanism opens
the loop a number of times equal to the digit. Ten opens represent
the digit "0" in the United States, while in some foreign countries,
0 is represented by one open, 1 by two opens, etc. There are other
variations.
Dial
tolerances are normally taken as 9.5 to 10.5 PPS, and the percent
break is permitted to range from 58% to 64%. Break is longer than
make so that ringer capacitors, at the far end of the line and
shorted out quickly by the immediately adjacent dial, will have
enough time to charge up during opens to make the line appear open
at the CO; charging on a long loop may take time because these
capacitors have to receive energy all the way from the CO battery.
What
the dial (or sender) transmits is not necessarily what the A relay
(Fig. 16, Chapter 1) at the CO sees, and the A relay or sensor can
introduce distortion of its own. Lumped and distributed line
capacitance converts the square dial pulses to a signal that looks
almost like a sine wave by the time it gets to the CO, and ringer
capacitance can alter the percent break. The ringer inductance or
other coils can oscillate, particularly on short loops, to produce
split pulsing (twice as many pulses as the dial transmits), and
holding bridges in early PBX trunk circuits, shorted out during a
train of dial pulses, can produce an extra (false) pulse when put
back into the circuit. Elaborate pulse correction circuitry is used
in dial pulse receivers in common control systems where only a small
number of detectors is required. In SXS systems, where every
selector and connector must respond to dial pulses, and in switching
systems with electronic matrices where dial pulses must be detected
at each line circuit, such correction becomes costly and encourages
the use of DTMF signaling (to be discussed below).
The
CO must be able to tell the difference between abandoned calls and
dial pulses (loop open), and interpulse and interdigital intervals
(loop closed). As has been described, these are the traditional
functions of the B and C relays. These relays remain operated for
about 200 milliseconds after their operate path is opened. This
gives them the ability to hold over the longest dial pulse make
(42.4 milliseconds at the dial) or the longest break (67.4).
Actually, the A relay will modify both these values, and in some
systems, a 10 millisecond minimum make or break at the A relay's
contacts is used as the design parameter. However, 10 PPS implies
100 milliseconds per pulse and a slow release (SR) relay with a
nominal release time of 200 milliseconds will hold very well, even
in worst case situations, and then release when required.
Variations among slow-release relays make timing relatively
inaccurate. In electronic systems where timing is based on the
system's accurate clock, the timing interval can be safely
shortened. Further, in common control systems where matrix path-hunt
is carried out later as a completely different function, there is no
need to allow for the time required by a SXS selector to hunt to its
last rotary terminal. From this, it can be seen that an interdigital
interval for 10 PPS dialing could be shortened to as little as 250
milliseconds with complete safety in non-SXS systems. Of course,
there is no reason why any form of common control switching should
be limited to 10 PPS. Operator toll dialing many years ago used 20
PPS dials to load senders. PBXs, too, homing on Panel and XBAR COs,
were equipped with 20 PPS dials to reduce attendant and originating
register holding time. Further, faster dial pulsing permits even
shorter interdigital timing. AT&T's electronic switching field trial
at Morris, IL., 1960-61, used 20 PPS dials. For users, they were
just about as fast as DTMF, and worked over long loops.
Until quite recently, it was much less expensive to generate dial
pulses than DTMF signals, and in the early days of deregulation and
interconnect, many telephones came on the market with push-buttons
replacing the dial, inserting digits into a buffer memory which then
activated a relay to produce trains of dial pulses. Many modems
followed the same procedure, accepting the called digits from the
keyboard or memory of their associated PCs and sending out dial
pulses.
Today, the cost of generating and detecting DTMF has fallen as the
demand for the chips to perform these functions has grown. Thus most
phones, modems, fax machines, etc., dial with DTMF, but many still
offer a dial pulse option. In the early 1990s, it was widely
reported that between 30% and 40% of the CO lines in the US did not
accept DTMF; as telephone companies increase the cost of DTMF lines,
we may even expect a modest increase in dial pulsing. Unfortunately,
the telephone operating companies have abandoned the 20 PPS dialing
option in their CO switches, even omitting it in connection with
digital subscriber loop carrier (SLC) and trunks which could easily
support 40 PPS dialing with negligible pulse distortion. Why improve
dial pulsing (which is, after all, digital) when analog DTMF can
make more money now and D-channel signaling is almost here? It is
safe to predict that analog DTMF will coexist with the digital
D-channel for even longer than dial pulsing has coexisted with DTMF.
DTMF. DTMF stands for dual-tone
multi-frequency, and is the generic term that covers such
proprietary service marks as Touch-Tone (AT&T), Touch Calling (GTE),
Digitone (Northern Telecom), Tel-Touch (originally ITT) and
Tone-Dial (originally Stromberg Carlson). The acronym is unfortunate
because there is already a tone-signaling system, used throughout
the world on trunks, called MF (for multi-frequency); it, too, uses
two tones. Both systems were developed at Bell Labs. DTMF is
discussed here, and MF will be covered in the next chapter. They are
quite different; MF came first, and DTMF profited from experience.
DTMF
uses two groups of four signaling frequencies, each spaced, within a
group, 11% above the next lower tone. Spacing between the two groups
was selected to minimize harmonic relationships and thus make a
valid signal, consisting of one tone from each group, as different
as possible from sounds produced in human speech. Nominal
frequencies in the low group are 697, 770, 852 and 941 Hz. In the
high group, they are 1209, 1336, 1477 and 1633. A total of 16
combinations is possible; normally 12, as shown in Fig. 2, are used
at telephone sets, although the military and various PBX
manufacturers have used the full 4x4 array.
Digit detectors are tuned to respond to signals that are within 2%
of the nominal value, while oscillators in the telephone set are
supposed to maintain their accuracy to ±1.5%. Tones generated at the
set are approximately -6 dBm each; receivers respond to tones that
range from 0.095 to 1.26 volts (-20 dBm to +2.45 dBm across 900
ohms), and tones must be within 18 dB of each other at the receiver.
As
is usual in signaling systems, recognizing a valid digit is no
problem; what is difficult is rejecting invalid signals that look
good. Because DTMF uses frequencies in the voice band, and is
connected when the user may be talking or when a TV may be going in
the background, "talk-off" protection is vital.
In
addition to geometric frequency spacing, techniques have been
developed to insure rejection of invalid signals. First, a
valid-appearing signal must be present for a minimum of 40 mS before
it will be recognized. Second, only one tone can be present in each
band, and third, signals outside the signaling bands, if present,
will block recognition of in-band signals (guard action). Fourth,
the receiver is usually desensitized at the nominal signaling
frequencies. Once a signal is present, outputs are held over
momentary breaks to prevent double registration of the same digit.
And at the telephone set, the microphone is blanked during the time
a valid digit is being sent. Even so, voice simulations of digits
sometimes get through.
Dial-tone poses an interesting problem. It must be applied to let
the user know the system is ready to receive digits and, because it
comes on at the CO, the dial-tone to signal ratio is quite high. To
prevent guard action from interfering with DTMF, filters are used in
digit receivers for 2-wire CO switches to block dial-tone from the
receiver while passing DTMF from the user. Had the traditional
dial-tone, 600 Hz chopped 120 times a second, been used, the complex
spectrum produced could not have been filtered economically; thus
precise dial-tone consisting of exactly two frequencies, 350 and 440
Hz, was developed to permit simplification of the filter. Precise
dial-tone also makes the design of dial-tone detectors easier.
In
digital switching systems serving analog lines, filtering dial tone
is less difficult because of trans-hybrid loss introduced in the
line circuit between the dial tone being sent to the customer and
the DTMF digits coming back. Today, call progress tones are
generated digitally and DTMF detection is carried out using digital
signal processing; D/A and A/D conversions take place in the codec
in the line circuit.
Although DTMF is a potent merchandising tool, its principal
technical advantages are end-to-end signaling capability and high
speed (compared to dial pulsing; compared to even rudimentary data
transmission, 40 bits per second can't keep up with a
teletypewriter). DTMF reduces holding times of signaling detectors
for telephone companies and unproductive use of customer facilities.
However, because users, punching buttons on a phone, cannot approach
the 10 digits per second of machine sending, DTMF is at its best
when it comes from a repertory dialer or a sender. Some such systems
seize a trunk, time out for three seconds to allow the far end to
get ready, and then outpulse. They could reduce their holding time
by almost 75% if they could outpulse a pre-stored number upon
detection of dial tone. Clearly, a three second time-out for a
one-second transmission is silly but cost effective.
End-to-end signaling is becoming increasingly useful as the cost of
voice-response units drops. Countless systems, from automated
attendants in PBXs to voice mail to electronic banking, ask the
customer verbally to use DTMF digits to select what is wanted from a
voice menu. Most of these systems terminate in 2-wire analog form
even on digital switches, as do most customer lines calling them.
Because they require callers to send digits during such voice menus,
and speech, unlike dial tone, cannot be filtered out, talk-off
protection poses some interesting challenges.
Telephone dictation and early voice-mail systems, without the
benefit of voice response, have DTMF commands a customer can use to
leave a message, replay it, edit it, and send it to someone else.
One may hope that such systems will be upgraded to use ISDN
signaling in the future, but even with universal D channels and
protocols, analog DTMF signaling via the voice path will be
attractive for many years to come. By going directly rather than
wandering through a packet network such as SS7, it may even reach
its destination faster.
Alerting
Telecommunication terminology is confusing in the way it uses the
same words to mean different things. For instance, there are token
ring data networks, tip and ring wiring to conventional telephones,
and the ability to ring the bell which was invented by Mr. Watson,
not Alexander Graham. Technically, announcing the presence of an
incoming call is "alerting," and may involve a variety of different
techniques.
Power ringing. The great majority of
telephone lines use power ringing: 20 Hz at 86 volts RMS. Central
office ringing is usually on for two seconds and off for four in the
United States, permitting one generator to ring three times as many
lines as would be possible if ringing were continuous. On long
lines, 105-volt ringing is sometimes used. The 86-volt signal is
superimposed of the regular -50 volt dc battery but the 105-volt
signal is not; because dc current is needed to signal off-hook, it
follows that some long lines cannot trip in the two-second ringing
interval but only during the four-second silent interval. This
situation is permitted only where absolutely necessary; in general,
it must be possible to trip ringing in both the ringing and silent
intervals.
On
individual lines, the ringer at the telephone set is connected from
tip to ring; on two-party lines, one party has a ringer from ring to
ground, and the other has one from tip to ground. Ringing is applied
selectively at the CO by choosing one side or the other to which to
connect the signal. Individual lines and the first party on
two-party lines are rung on the ring side; the second party is rung
on tip.
On
lines with more than two parties, ringing becomes more difficult.
Independent telephone companies have used harmonic ringing at
frequencies such 16-2/3, 33-1/3, 50, and 66-2/3 Hz, with tuned
ringers at the subsets responding only to one frequency. The former
Bell System used a scheme with gas tubes in series with the ringers,
and bias batteries at the CO; four-party full-selective ringing was
provided with two bias polarities and two sides of the line, making
a total of four combinations.
For
more than four or five parties, coded ringing has been used; with
five different codes (long and a short, two shorts and a long, etc.)
and two sides of the line, up to ten-party ringing was common until
fairly recently. Almost all multi-party lines will undoubtedly be
phased out in the near future, if they are not already casualties of
deregulation. Although it is not hard to complete an incoming call
by ringing the right phone, identifying the calling party for
automatic billing of outgoing calls is much more difficult.
Early ringers consisted of a three-pronged permanent magnet, with
the north pole in the middle and two south poles right and left,
combined with a two-winding electromagnet on the south poles. When
the 20 Hz ac current passed through the electromagnet, one winding
strengthened the permanent magnet's field at one south pole, while
the second winding bucked the field from the other. When the current
reversed, the electromagnets reversed their roles. An iron bar,
pivoted over the north pole, would be attracted first to one south
pole and then to the other; this rocking motion moved an attached
hammer to strike two gongs alternately.
Ringers in 500/2500 type phones differ only in details. Their coils
have a little less than 4000 ohms resistance, and a very large
inductance. For 20 Hz ringing, a 0.02
mFd capacitor is placed in
series to achieve resonance and also to block the flow of dc. When
harmonic ringing is used, ringers are, of course, tuned to other
frequencies. These ringers draw about 10 mA of ac current; if more
than five ringers are operated at one time on the same line, there
is a high probability of pre-tripping, as will be discussed.
Tone ringing. Although the principle of
the ringer has changed very little since Watson invented it,
advances in magnets, springs, acoustics and electricity have been
enormous and the ringer has been refined to the point where it does
the best possible job at the minimum possible cost. Thus it took
modern electronic sounders, which chirp with a pleasant tone, quite
a while to become competitive in price. Tone ringers using
electronic devices that respond to the power ringing signal are now
widely used, but they do nothing to eliminate the principal problem
of power ringing, the 86 volt 20 Hz ac ringing signal itself.
Power ringing was designed to work on copper wire lines and has, for
years, reached these lines via a path through a metallic switching
matrix. With the coming of electronic switching in general and
digital switching in particular, the switching matrix is almost
always incompatible with the power ringing signal; as a result, the
ringing signal must be applied through a small relay or high-voltage
electronic switch on a per-line basis.
OPX
and FX lines, like digital switching, cannot pass power ringing when
they must run through carrier systems. Usually, the power ringing
signal operates a relay which controls a supervisory signal which,
at the far end of the carrier system, operates another relay to
apply power ringing locally. Subscriber loop carrier (SLC), now
widely used in metropolitan as well as rural areas, must also have
power ringing available to apply to the called phone. One possible
improvement over power ringing would be to use a voice frequency
tone to activate a tuned ringer in the called telephone; this
signal, like DTMF, could then be distributed through electronic
systems as easily as a voice signal itself.
Tone-ringing has been the subject of considerable study for years;
it was even field-tested in the AT&T Morris field trial mentioned
earlier. The actual sounder was an acoustic horn driven by a
transistorized amplifier tuned to a specific audio frequency
(actually, eight different frequencies were available to provide
party line ringing). When bursts of tone at the ringer frequency
were sent down the line to an on-hook phone, the sounder reproduced
them acoustically. Customers loved tone ringing, but cost and
compatibility made it hard to justify in 1960.
The
compatibility problem can be understood in the context of
electromechanical switching systems. A connector or a trunk circuit
had to apply ringing to any one of a large number of lines; in
5XBAR, for instance, seven different ringing signals could be
applied to either side of the line from any terminating trunk
circuit. Unfortunately, adding additional ringing signals was almost
impossible.
In
1ESS, ringing is applied from service circuits independent of the
trunks but accessing called lines via the switching matrix. This
permits, in principle, any kind of ringing to be returned to any
line. Although the 1ESS had to work quickly after answer was
detected to drop the ringing connection and establish a completely
different connection for talking, the flexibility obtained was
justified by the enormous simplification in trunk circuits and in
the ringing circuits themselves. To add tone ringing to 1ESS, all
that would have been necessary was a small group of service circuits
to apply tone ringing and detect answer on lines class-marked
appropriately. However, 1ESS, unlike the electronic Morris switch,
had a metallic matrix which made conventional ringers too
inexpensive to replace.
Using a voice frequency tone for ringing has several advantages over
the use of power ringing. As has been indicated, it can pass through
any voice-frequency channel, including SLC, carrier and electronic
switching matrices. Second, because the ringing signal is well
separated from dc, the problem of ring-trip is greatly simplified.
Third, when the "call waiting" feature (see Chapter 5) is offered,
the same ringing signal can be used whether the called line is busy
or idle. And finally, tone ringing offers a possible way of
obtaining calling party identification on multi-party lines: the
tuned circuit that selects the ringing signal can be arranged to
provide a party-identification tone on outgoing calls.
The
form of tone ringing most appropriate to modern technology is found
in some electronic PBXs and uses a completely different approach
based on the separate signaling channel between set and line card.
Via this channel, a digital message is sent to the set to tell it to
activate its sounder; when the phone is answered, the off-hook
signal is sent to the line card as another digital message. This
approach requires no connection through the switching matrix for the
ringing signal, allows the user to tune the ringer so that it can be
differentiated from others in the office, and obviously eliminates
the major problem, the incompatibility of the power ringing signal
with electronics. It also eliminates all problems involved with
ring-trip and pretripping prior to answer (discussed below). ISDN
telephones are expected to follow this approach.
Immediate ringing. Most modern switching
systems are arranged to provide "immediate ringing." That is, they
apply ringing to the called line immediately upon connecting to it,
rather than connecting it to a ringing signal that may be at any
point in its two-seconds-on/four-seconds-off cycle. Immediate
ringing reduces call set-up time (post-dialing delay), because the
odds are two to one that ringing will be in the silent part of its
cycle if the connect time is chosen at random.
Immediate ringing is easy to implement, even when power ringing is
being supplied, because a modern system can be arranged to control
the application directly on a per-call basis, or monitor the ringing
plant and choose a ringing source that is either ringing already or
is just about to start. These approaches should not be confused with
the "splash ring" of earlier systems where a burst of ringing was
applied upon connection to the line prior to picking up regular
ringing wherever it might be in its cycle.
In
addition to reducing post-dialing delay, immediate ringing has
another useful function. It can operate a ring-up relay in a PBX
trunk circuit immediately upon seizure, reducing the probability of
a collision between an incoming and outgoing call, particularly if a
loop-start trunk is used. Ground-start trunks remain more desirable
for use as PBX "combination trunks" because they also provide a
clear indication of dial-tone and hang-up.
Ringing-trip. Detecting answer in the
presence of power ringing is possibly the most difficult problem
confronting the switch designer trying to increase the range over
which an analog line will operate. Five ringers, operating at one
time, can draw 50 mA RMS of ac current; upon answer, however, dc may
be as low as 23 mA. A half-period of 20 Hz ringing is 25
milliseconds; during a half-cycle, the switch can't tell the
difference between ac and dc. Thus answer detection must be averaged
over at least one full cycle (50 mS), and the detector must be able
to "see" dc which is less than half as large as the ac.
Slow-operate relays have been used for years to respond to dc
without responding to ac. With electronic circuitry, it is easy
enough to build circuits that ignore 20 Hz and respond to dc, but
only recently have such circuits been competitive in price. There is
one further aspect to the problem. Just as the answer detector must
not respond in less than one ac cycle, it must be certain to respond
in something less than four cycles, or 200 milliseconds. A PBX or
ACD attendant or other person wearing a head-set will have the
ringing signal delivered right into his or her ears in the interval
between answer and ring trip. Such a "bat" in the ear, at a
relatively high level, can impair hearing unless it is removed very
quickly. Once ringing is tripped, the system must remember that
answer has taken place so that ringing won't be reapplied if the
called party hangs up first.
Tone
ringing distributed via the switching matrix is easier to trip than
power ringing because the audio tones used are quite different from
dc. With care, the regular line-supervisory sensor should suffice.
There is one problem here, however. The amplifier in the telephone
set that receives the ringing signal and drives the sounder draws
power from the line only during ringing. Because it draws power from
the dc battery, this current, added to leakage and noise, may cause
pre-tripping. Thus a non-operate current of perhaps 5 or 6 mA may be
a necessary specification. Tone ringing activated via a signaling
channel, as has been pointed out, solves all these problems but
cannot be used with the millions of existing analog telephone.
Distinctive ringing and line displays.
Distinctive ringing has two purposes. One is to provide some useful
information about the incoming call, while the other is to provide
information about the called phone. As we have seen, certain older
types of party line used a distinctive (coded) ringing signal to let
all parties know who an incoming call was for. A similar feature is
being marketed today for individual residences, saving parents from
contact with friends of their teen-age children. In PBX and centrex
systems, ringing often has a different cadence for internal and
external calls (single vs. double ring); in key systems, the
intercom buzzer is quite different from the ring of the phone. In
such instances, something is known about the call prior to answer.
When
there are many phones close together, as in an office, a distinctive
sound for each allows users identify their own phone's ring, even
when they are several desks away or down the hall. It was quite easy
to modify the sound of the electromechanical bells (put tape on the
gongs, remove one gong, etc.), but tone ringers are harder to
change. Some electronic systems allow the user to select one of
several different ringing sounds, either in the set itself, on via
PBX control.
When
loud-ringing is required for outdoor use or in a noisy location like
a loading dock, special bells and tone ringers can be substituted.
However, there are limits to the ringing power available from the
line, and to go to higher sound levels, or to operate a flashing
light (as for the hearing impaired), an ac relay may have to be
substituted for the ringer in the telephone set so that its contacts
can operate a separately powered device. Tone ringers amplifying an
audio tone via the voice path or activated by a digital signaling
channel require a similar approach based on a relay that works on
the ringing signal provided.
When
several lines appear on a telephone set, as in 1A2 key telephone
systems, to be discussed in Chapter 5, lamps are used to show which
line is ringing, and a variety of bells and buzzers can be used in
each telephone set. Over the years, these key systems have also
developed many ways to handle external devices such as loud-ringing
bells, paging systems and the like.
The
separate signaling channel to PBX and ISDN phones is designed to
help them to replace 1A2 key systems and do more besides. Not only
can a signaling channel control lamping, ringing and external
devices, it can also transmit text to a small display on the phone.
Thus Caller ID goes well beyond distinctive ringing.
Party identification and reverting calls
On
calls originating from party lines, the calling party must be
identified for billing. It is prohibitively expensive to do this
manually, even on toll calls, and for usage sensitive pricing of
local calls it is out of the question. Thus automatic identification
is mandatory.
The
problem is fairly easily solved for two-party lines. One party has a
resistance to ground when the phone is off-hook, and the other does
not; the ground is detected by special circuitry in the signaling
receiver. The resistance to ground is in the 1000 to 4000 ohm range,
made up of the ringer winding. Difference in ground potential
between the CO and station ground sometimes becomes a factor. The
very high inductance of the winding is, for all practical purposes,
an open circuit to voice frequency currents, so transmission is not
appreciably affected. "Ring" party, without the ground, is always
the first to be assigned on a party line; minimizing the number of
"tip" parties, requiring a subset modification, is desirable on an
overall basis.
On
multi-party lines, the problem is much more difficult. "Spotter
dials" have been used; these provide one or more pulses to ground as
the dial returns to normal, identifying the party by means of ground
pulses while simultaneously opening and closing the loop to generate
regular dial pulses.
Less
costly in equipment is the "circle digit" approach. Each party is
assigned a party number, shown on the dial by a digit in a circle.
The user is instructed to dial the circle digit in addition to the
called number. The system is arranged to use the circle digit in
conjunction with the line identity for billing, and the other digits
for establishing the call.
There was a period in the late 1970s and early 1980s when customers
were expected to dial their entire telephone number into some sort
of machine used for toll billing. This was done with many of the
early competitive long distance carriers, and with toll routers and
recorders used in connection with PBXs. The coming of "equal access"
has made automatic number identification (ANI) available to all long
distance carriers from the originating local telephone company, and
reasonably smart programming in PBXs has made external add-on
devices unnecessary. About the only vestige of circle digit and its
cousins today is in connection with credit card calling, and even
this is automated in some public phones which read a magnetic stripe
on the back of the card.
As
mentioned earlier, tuned tone-ringer amplifiers at subsets have been
made to provide party identity in the laboratory. For example, the
amplifier can be changed to an oscillator at the ringer frequency
while the dial is off-normal, or the DTMF buttons can activate a
ringer-frequency tone that dies out quickly in the interdigital
interval. Many possible technical solutions have been rejected
because of cost.
Whenever party lines are provided, the problem of reverting calls
must be faced. Sooner or later, one party on a line will call one of
the others. This used to be common in the early days of telephony;
it is so much less common now that special precautions are
necessary. Even though complete instructions may be provided to the
user in the front of the telephone book, the odds are greatly
against them having been read. Thus the system must be able to deal
with the problem in spite of complete ignorance on the part of the
user.
Obviously, the system cannot ring the called party as long as the
calling party is off-hook. Thus the calling party must hang up, the
system must then ring the called party, and the calling party must
be informed of answer so that the phone can be picked up again.
Probably the best approach is to have the system, once it knows it
is dealing with a reverting call, connect the calling party to an
operator or a recorded announcement. A message to the following
effect is returned: "You have dialed another telephone on your own
party line. Please hang up and listen for ringing. When the ringing
stops, the other party has answered. You may then pick up your
telephone to complete the connection. If the other party does not
answer, please lift your own phone momentarily to terminate
ringing."
With
luck, the user will then hang up. The system proceeds to ring the
two parties alternately, which is not too hard on two-party lines.
On multi-party lines where coded ringing is used, it is not much
harder. If both parties are on the same side of the line, the
calling party will hear the coded ring of the called party
automatically; if calling and called are on opposite sides of the
line, a special very short bat of ringing is returned to the calling
party's side during each ringing cycle.
When
multi-party lines with full selective ringing (ringing heard only by
the called telephone) are used, the problem becomes most difficult.
Here, the system must be able to identify the calling party so that
it can return the correct ringing signal; it knows the called party,
of course, from the dialed number. When the calling party is
identified so that the proper ringing signal can be selected, the
line is rung alternately with the calling and called rings. If
either party answers, ringing is tripped and the talking connection
is established. With an electronic matrix, no path needs to be set
up; talking battery is returned to both parties from the line
circuit. With a metallic matrix, only one path is set up to talking
battery and supervision, because only one line is involved.
As
we have seen, an ISDN S/T interface can be thought of as a
mini-party line in that it can support up to eight devices, although
no more than two of them can make a circuit-switched connection at
any one time. Use of the D-channel for signaling in both directions
has great potential for alerting or identifying specific entities
(such as phone vs. PC vs. fax), and the two B channels allow one
phone to be connected to another via the switch, just like any other
connection. Further, the D-channel can support packet connections in
addition to signaling. Unfortunately, the variations in
implementation previously mentioned may make these desirable
possibilities difficult to standardize.
Coin control
A
central office switch must be able to collect or return coins on
deposit at a pay telephone provided by the telephone company.
Privately owned pay telephones usually contain circuitry built
around a microprocessor to do this, along with many other things
that the telephone company does in the CO to provide payphone
service. Here, we will consider only payphones provided by the
telephone company.
In
addition to collecting and returning coins, a CO must also be able
to test for the presence of coins on deposit without collecting or
returning them. Collect, return and test functions must be possible
whether the telephone is on-hook or off-hook. Collect and return
voltages in the 120 to 130-volt dc range are applied at the CO;
different polarities are used for collect and return at the option
of the local telephone company. The collect voltage causes the coin
magnet at the pay phone to tip deposited coins into the coin box;
the return voltage causes the coins to be released to the coin
return chute. The regular 50-volt battery is used for testing; when
coins are on deposit, the coin magnet is connected to ground. If no
coins are present, either due to failure of the user to deposit them
or because the deposited coins have been collected or returned, the
path to ground is open. As can be seen, there is a similarity
between party testing on two-party lines and coin testing.
Ground-potential differences may be a factor in both.
Until fairly recently, most pay telephones required the phone to be
off-hook and a coin to be on deposit before a call could be
originated. Ground start lines were used, and the off-hook plus the
coin-in-slot completed the ring-to-ground path at the station. The
advantage to "coin first" operation is the elimination of a
"permanent signal" when the pay phone is knocked off-hook
accidentally or by vandals. Its disadvantage is the inability to
place emergency calls without a coin, and the increasing
inconvenience to users and the telephone company alike in the
unnecessary depositing and return of an initial coin on toll,
credit-card and 800 number calls.
Some
small telephone companies have always provided "dial tone first"
operation. The user simply dialed a local call, listened for answer
and, if the correct party was reached, deposited a coin. The coin
activated the coin phone's microphone, and two-way communication was
possible. Long distance calls were routed through the operator who
got the called party on the line and then requested the user to
deposit the correct amount. The advantage to this system is the
complete absence of coin collect, return and test functions. Once
the coin is deposited, it is placed directly in the coin box and
cannot be returned. Although inexpensive in equipment, this approach
is non-standard; it also adds to personnel costs on long distance
calls.
Modern dial-tone-first operation returns dial tone when the user
goes off-hook. For a local call, the user then deposits the minimum
interval payment, dials the number, and converses as always.
However, if the user wishes to dial 911, the (more or less)
universal emergency code, a coin is not needed. There are other
codes that can be dialed without a coin, as well. The most important
are 0 or 00 for local and long distance operator assistance. In
addition, many areas are already adapted to the procedure of dialing
0 plus the ten digit called number if the call is to go beyond the
local area, or is to be billed to a credit card even if local. On a
"0+" call, the operator, human or automated, enters the connection
only to obtain the credit card number. There is also considerable
effort being made to automate collect calls and those charged to a
third party.
The
operator originally checked the deposited amount by listening to the
coin gongs at the payphone as the user put in the money; more modern
phones use a "coin totalizer" to activate a display at the operator
position. With a combination of voice response units and digital
recording, technology is now available to automate collect, third
party, and person to person calls. The ISDN BRI should make both
coin and non-coin operations appreciably easier from pay phones.
The
public telephone picture in the 1990s is relatively complex in that
the courts have decided that all long distance carriers, and not
just AT&T, have a right to share the revenues of long distance calls
paid for by coin, and customers have the right to use their own
carrier for credit-card, collect, and third party calls. This makes
necessary access to operator services provided by local and long
distance telephone companies, and similar services provided under
contract to privately owned pay phones and alternative long distance
services. Taking full advantage of all these choices is not an
unmixed blessing from the customer point of view.
Out-of-Band and Common Channel Signaling.
Although ringing, coin control and dial pulsing signals are all out
of the audio frequency band, out-of-band signaling is usually
thought of as signaling on a separate channel that uses the same
physical facility as the communication channel it serves. Common
channel signaling, on the other hand, is a separate signaling
channel serving a group of communication channels in common; it may
or may not share the same physical facility as its communication
channels, depending on overall system design.
As
will be discussed in the next chapter, AT&T began using common
channel signaling in its toll network in 1976. Signaling which might
equally well be called out-of-band or common channel, between the
telephone set and its line card, began to be used in connection with
electronic PBXs about the same time. Some of these PBXs used one
pair for voice and one for signaling; others used a digital carrier
frequency well above the voice band for control over voice (COV) on
a single pair. In such instances, the voice signal used the
transmission standards of the 2-wire 500 type telephone. However,
when per-line codecs became economical, the codec was moved to the
telepone and digital speech was multiplexed with the digital
signaling channel. In most cases, a second channel, usually intended
for data transmission, was also included, served in common with the
voice channel by the signaling channel.
As
has been illuslrated in this chapter, all the traditional problems
related to line-side signaling, alerting and supervision are
eliminated with a separate signaling channel, and transmission losses
and echo are also nearly eliminated with the codec at the telephone,
always assuming that the two-way bit stream between set and line
card works as expected when used on real-world loops. Because PBX
loops are usually very short, proper operation is not too
difficult. The major problem is that no two PBX manufacturers
use
the same approach, and some manufacturers manage to find several ways to do the same thing
within their own product line.
When
CCITT standards for ISDN began to be considered, it was hoped that
the Basic Rate Interface, or BRI, with its 2B+D arrangement coded
into a 2B1Q bit stream would bring some order out of chaos; indeed,
it may yet do just this when the signals using the D channel (the
signaling protocol) become standard. However, the telephone industry
is far more interested in CO switches than PBXs, and residential
rather than business customers. As a result, most ISDN efforts are
directed toward future services for residential customers who,
content with 2500 type telephones, don't need them, while business
customers, who could take advantage of ISDN, are investing in PBXs,
LANs, and private digital networks to meet their existing needs.
There is little that BRI can do that hasn't been done for more than
a decade, with limited commercial success, by proprietary PBX
telephones. What ISDN can actually offer is standardization so that
those who wish to do so can buy digital telephones from the
manufacturer of their choice and use them with the same assurance
they experience today with phones made to 1950s specifications. If
all digital telephones used the same signaling messages (described
in Standard Q.931) and also handled data packets (standard X.25) on
their signaling (D) channels in the same way, we might actually
begin to experience some of the benefits of the digital revolution.
Line
and trunk circuits are part of a switching system, interfacing the
switching matrix and system control to related transmission
facilities. Service circuits, on the other hand, may be thought of
as matrix terminations that do NOT serve as an interface for
transmission facilities. Rather, service circuits are connected to
lines and trunks via the matrix or auxiliary switches as needed to
assist with establishing or disposing of calls. Service circuits
include registers, senders, tone sources and conference bridges,
circuits for application of ringing and coin control signals, etc.
SXS,
which dominated automatic switching for decades, had no service
circuits. Each switch accepted and used dialed digits directly,
connectors applied ringing to called lines, and call progress tones
were returned from the switches involved in the connection.
Register-senders, part of the Rotary and Panel Systems and added to
SXS in Director Systems, were inserted by an auxiliary switch into
the paths between linefinders and first selectors. Dial tone was
usually returned from the register-sender rather than the first
selector, but other tones and ringing came from the matrix switches
as before.
Crossbar systems, because their simplified switches required a
marker to establish a connection, separated registers from senders
and used the latter only for calls to other switches. Markers
connected calling lines to originating registers via the switching
matrix, but connected trunks to incoming registers or outgoing
senders via separate auxiliary switches, allowing a large number of
trunks to share a much smaller group of more complex circuits.
Markers, in addition to setting up these connections, moved
information stored in registers to senders when necessary.
Because there were only a few originating registers, as compared
with selectors and connectors in a SXS system, considerable
attention was directed toward dial pulse detection and correction,
eliminating the effects of split pulsing, chattering relays, pulse
distortion, etc. Crossbar systems also used separate tone circuits
as alternate destinations for connections blocked by ATB or busy
lines, but left ringing built into each trunk circuit.
1ESS
went to separate service circuits for ringing, coin control, etc.,
simplifying trunk circuits and increasing system flexibility.
However, the elimination of the metallic matrix with the coming of
digital switching forced many of the functions of service circuits
back into the line and trunk circuits themselves. Fortunately,
digital lines to customers and digital trunks to other switches,
with their separate signaling channels, offer opportunities to
regain the hardware simplification and savings of earlier systems.
For
the foreseeable future, service circuits will include digit
receivers and senders for DTMF and possibly MF, access circuits for
call progress tones, recorded announcements and computer synthesized
messages, and conference bridges. Switches with matrices capable of
handling high voltage signals (including the concentrator in AT&T's
5ESS) may also include service circuits for dial pulsing, ringing,
coin control and test access. When the main switching matrix cannot
be used, smaller adjunct matrices designed to support the particular
signal on per-port connections will be required.
In a
new version of AT&T's SS7 signaling system, called AIN for Advanced
Intelligent Network, Service Circuit Nodes expand the general idea
of service circuits: they are separate switching systems which give
callers access to intelligent service circuits providing a variety
of functions such as text-to-speech, interactive speech, and
automatic speech recognition, all under the control of a powerful
computer. Using centralized equipment for such advanced services is
similar to centralizing the routing intelligence of the network. In
what follows, we will deal with the older and more conventional
service circuits.
Registers and senders
Concentrating signaling in registers and senders led to common
control which, in turn led to modern computerized switching systems.
A register extracted incoming digits from a line or trunk and stored
(or registered) them in sets of relays where the whole number would
be available to the system's control. Senders reversed the process,
accepting a telephone number from control, storing it in relays, and
sending it, one digit at a time, into the connected transmission
facility. Clearly, registers and senders had two separate functions
to perform: line interface and storage.
Computer-controlled systems, with large quantities of inexpensive
read-write memory, have taken over storage. Interfaces, however,
still require specialized hardware. Thus registers and senders, with
"registering" done elsewhere, have become signaling receivers (or
sometimes "decoders") and transmitters, handling one digit or one
change in line or trunk status at a time, under processor control.
Even so, the terms "register" and "sender" are still frequently
used.
DTMF
and MF tone signaling are essentially analog, and originally used
relatively simple oscillators in senders (or telephone sets) to
convert a digit to an analog tone-pair, and elaborate receivers to
convert the tone-pair back to something digital the register could
store. Today, tones are generated digitally and detected directly in
digital form using digital signal processing (that is, the digitized
form of the analog signal is converted to the kind of digital signal
the switch control can understand). From transmitter to reciever via
digital trunks, DTMF and MF digits may never actually be "tones" at
all. Although DTMF and MF receivers and transmitters today are quite
different from the analog versions in wide use only a few years ago,
they are still service circuits and are connected to the line or
trunk via the switching matrix.
Traditionally, dial pulsing made extensive use of registers and
senders and then digit receivers and transmitters when a metallic
matrix was available to connect them directly to tip and ring. There
were, however, certain kinds of trunks (such as E&M, described in
Chapter 4) which required more than tip and ring to be switched
through; because the 1ESS matrix switched only a single pair, digit
transmitters and receivers could not be employed in this situation.
As a result the common control bypassed service circuits, scanning
incoming supervision directly while using a distributor output
associated with the trunk to apply outgoing supervision and modulate
it into dial pulses.
Digital matrices, of course, cannot convey dial pulses directly, but
some systems have chosen to code dial pulses into a signal that can
be passed through their matrix to a service circuit which counts
them and assembles digits, relieving the common control of this
routine work. However, many such systems simply scan for dial pulses
in each line or trunk, and distribute dial pulses to trunks on an
individual basis. Such scanning or signal distribution requires
considerable effort on the part of the system control.
When
computers were expensive, and one computer was supposed to handle
all telephone interfaces and operations in real time, such scanning
and out-pulsing could waste a large proportion of the real-time
available. In small systems like PBXs, this was seldom a factor, but
in large CO switches, it created a problem. As computer prices
dropped, distributed "front end" processors were often used to good
effect, each handing its own line group's signaling, converting a
wide variety of information to and from the outside world into a
standardized format for efficient interaction with the central
processor. Fortunately, CCIS makes even better use of computer
control, and will ultimately eliminate the need for dial pulses and
MF and DTMF digits, along with most of the service circuits they
presently require.
Tone circuits for call progress tones
Call
progress tones include, among other things, line busy tone, reorder
or overflow tone when ATB is encountered, ringback (for a time
called audible ringing), and dial tone. The characteristics of these
tones are shown in Table 1. They are returned to the user to ask for
information, as in the case of dial tone, or to provide information,
as with busy, reorder, and ringback. Many systems return dial tone
from a signaling receiver. Ringback tone, in electromechanical
systems, was mixed with ringing for distribution within the switch.
At the trunk circuit or connector, the 20 Hz ringing voltage was
sent to the called line but only the voice-frequency ringback tone
was able to pass through a capacitor to the caller. This gave the
caller almost positive proof that the called line was being rung.
| TABLE 1: Characteristics
of Common Call Progress Tones |
| |
Constituent Frequencies |
Interruption Rate |
| Tone |
350 |
440 |
480 |
620 |
|
| Dial |
x |
x |
|
|
Off at first digit |
| Busy |
|
|
x |
x |
60 IPM |
| Reorder |
|
|
x |
x |
120 IPM |
| Ringback |
|
x |
x |
|
Follows ringing |
|
Note: IPM=Interruptions
Per Minute. |
Busy
and reorder tone, although returned from switches in SXS, had
separate network terminations in crossbar and reed switch systems.
In 1ESS, ringback was also returned from such circuits, completely
separate from ringing. Because tone circuits supervised the line to
which they were connected, they could only serve one caller at a
time, and many were required. However, these circuits matched the
characteristic impedance of trunks as well as lines, assuring high
return loss and minimum echo.
Originally, call progress tones and the ringing signal were
generated by rotating machines which were part of the power plant.
With the coming of transistors, analog tones were generated separate
from ringing, but were distributed to tone circuits through special
wiring. Digital technology offers a new way to generate call
progress tones and tone-ringing signals: PCM-coded samples
representing a brief segment of tone are stored in read-only memory
(ROM); when the tone is required, the ROM is read out over and over,
producing a signal that a digital-to-analog converter will
reconstruct into the desired tone. Both the amplitude and frequency
of tones generated in this way are quite stable; amplitude is fixed
by the digital coding, and frequency is locked to the system's
clock.
Even
with digital matrices, most telephones are still analog and are
reached via two-wire lines. However, a digital matrix makes separate
one-way connections in each direction; a caller cannot talk to call
progress tones, and any number of callers can listen to the same
tone without any problems concerning echo, line supervision or loss
of volume. This allows the matrix itself to turn on and off call
progress tones to individual users as needed, and use the other
direction of transmission independently. In particular, a caller,
upon hearing dial tone from a tone circuit can start keying DTMF
digits into a digit receiver. Upon detecting the first digit, the
system can release the dial-tone connection through the matrix while
the caller continues to send signals to the receiver. If a rotary
dial is used and dial pulses are picked off in the line circuit, no
"talk" path is established through the matrix during dialing, and
the "listen" path from dial tone can be opened after the first dial
pulse is received. Actually, steps are taken to make sure the caller
hears "quite tone," the PCM encoding for a signal of 0 amplitude.
Just
as dial tone can be connected and removed by the switching matrix,
interruption rates can also be generated by matrix operations.
Although most matrix connections will be continuous, they can also
be turned on and off 60 or 120 times a minute. This allows the same
source to be used for busy and reorder, while the matrix connection
itself provides the appropriate interruption rate.
Recorded announcement circuits
A
recorded announcement source is just like a centralized tone source,
and the recorded announcement can be distributed to the matrix ports
by protected wiring. However, recorded announcements differ from
tones in one very important respect: they generally have to be heard
from the beginning. This requires the system to return ringback
until the announcement is ready to start, and then shift the
caller(s) to the recorded announcement, often by changing matrix
paths.
Memory prices had dropped so low by the early 1990s that it became
possible to use standard computer memory for storing digitized
voice. Thus recorded announcements, particularly useful as part of
an automated attendant for PBX and ACD service, can be stored on a
tone circuit as easily as a digital tone source. Digital recorded
announcements do not have to be rewound or cycled up to the starting
point; however, if the message is already in progress when new calls
come in, the switch must know when the new calls can be connected.
Conference circuits
Conference circuits are used when more than two lines or trunks are
to be connected together simultaneously. In analog systems, their
purpose is to provide the amplification necessary so that all
listening parties can hear the speaking party, and to prevent echoes
when many inputs imply many echo paths associated with
amplification. Any party hanging up must be disconnected from the
conference circuit promptly and that port must be terminated, again
to prevent echoes.
Digital systems provide conferencing quite differently. Each port of
the conference circuit has an input for its talk signal and an
output for its listen signal. The conference circuit delivers the
talk signal from any one individual to the listen side of all, but
subtracts the talk signal from the listen side of the talking
person. When two or more people speak at the same time, the digital
signals, coded as numbers representing the amplitude of each
person's speech, are added together to produce a resultant; however,
each speaker's own signal is subtracted from his listen path to
eliminate echo. This procedure is made difficult by the non-linear
coding of PCM, where companding is used to let 8 bits provide the
same quality as 13 bits would provide in a linear system. All
signals are expanded to linear as they enter the conference circuit;
in this way, standard adders and subtracters can be used in the
combining process, and the resultant is then compressed back to 8
bit samples at the "listen" connection to each participant.
It
is not necessary to have a separate conference circuit to do this.
Each port circuit in AT&T's Definity PBX, a.k.a. System 75, can
accept up to 5 digital signals from the switching bus (5 time-slots
per frame) and combine them into one signal to which it listens. It
can also combine the single-frequency constituents of call progress
tones, available on individual time slots, and connect them,
continuous or interrupted, to the caller.
It
must be kept in mind that conference bridges in 2-wire systems,
while elegant, are not always necessary. Indeed, most conference
connections are made by simply picking up another phone on the same
line and joining in. But there are other ways. In SXS areas,
conference connections used to be a favorite teen-age prank. Members
of a group arranged to call the same line at the same time; once the
line was busy, everyone else was connected to busy tone returned
from that group of switches. So many connections reduced the tone
level, and a multi-party conversation was easy. Separate 2-wire tone
circuits eliminated the fun in XBAR and ESS, while listen-only tone
sources, as in digital systems, keep it from springing up again.
Circuits for access to external equipment
Telephone dictation, voice mail systems, paging and code calling are
typical of external systems that require access circuits to
interface with a telephone switch. In general, conventional analog
line circuits are used, and the inputs to these external systems are
designed to look like 2500 type telephone sets.
Telephone dictation, available for decades, was originally
controlled by digits from a rotary dial which SXS systems passed
easily through a trunk circuit interface to the dictating equipment.
Because modern matrices cannot pass dial pulses, most telephone
dictation systems answer a power ringing signal when called, and
accept control signals via a built-in DTMF signaling receiver.
Paging and code calling are relatively simple. For paging, the
attendant dials the extension to which the page amplifier is
connected; the page amplifier either answers when it detects
ringing, or is associated with a permanently off-hook line to which
the switch connects without making a busy test. Once connected, the
attendant speaks through the page amplifier via the telephone
connection.
With
code calling, the called party must be identified by some signal
passed through the switching matrix so that the proper chime or gong
signal can be sent out. The problem is not unlike that encountered
with control of telephone dictation, and DTMF is a good solution at
the present time.
Voice mail is a little more complex. Not only do voice mail systems
perform the same functions as telephone dictation, but they also act
as a group of telephone answering machines. Thus they have to accept
information from the telephone system to identify which client's
message they are about to store, and they have to be able to tell
the switch to light a specific client's message waiting lamp or
otherwise indicate the presence of one or more messages. Although DTMF signals can be sent both ways on a telephone line when digit
transmitters and receivers are attached at both ends, a data link is
sometimes used instead.
Voice store-and-forward is also available when a large customer has
several voice mail systems. Unfortunately, as of the early 1990s, no
two manufacturers of voice mail systems use the same compression
technique for digital voice storage, so even when end-to-end digital
channels are available, messages have to be returned to analog prior
to transmission from one voice mail system to another. The Audio
Message Interchange Specification (AMIS) has made a start in
standardization here.
When
end to end digital connections via the public networks become
available, connecting external systems will be much easier. If an
ISDN BRI or PRI is used as the interface, two-way control signals
for manipulating telephone dictation or identifying the caller to
the voice mail system can take full advantage of the D channel, and
compressed voice messages can be sent from one voice mail system to
another via a B channel in perhaps a quarter of the time an analog
message would take. Even without ISDN, some voice mail systems have
options that let their ports look like proprietary digital PBX
phones. When call forwarding (see Chapter 5) transfers a call to
such a phone, the signaling channel causes the number from which the
call has been forwarded to be displayed; this is exactly the
information a voice mail system needs to store the call properly.
In
the meantime, digital PBX phones either have to be able to generate
DTMF as well as their own digital signaling, or else the PBX has to
be able to inject DTMF into the connection under digital signaling
control. Otherwise, users with the newest and most modern phones
cannot access equally new features and services which choose to look
like 1950.
Ringing, coin control and test access
With
matrices capable of handling high voltages, ringing, coin control
and other signals that differ from speech in amplitude or band-width
can be connected to lines as needed from specialized service
circuits. Digital switching and most analog electronic switching
block this approach, requiring individual line circuits to include
the ability to apply these signals, although sometimes a substitute
approach such as tone ringing or DTMF tones for control purposes can
be found.
One
reason why electronic matrices were slow to replace metallic
matrices in local central offices was the need to use the matrix
itself for test access and automatic line insulation testing (ALIT).
In such instances, ALIT interfaced the matrix as a service circuit.
The problem is more complex with digital matrices, and will be
discussed further in Chapter 8.
-
Line
-
Trunk
-
CPE
-
MDF
-
ISDN
-
Ohm/volt/milliampere
-
Loop
-
Return loss
-
2B1Q
-
Loop start/ground start
-
Flash
-
Alerting
Click Here for
Answers
1.
Why is a pair of copper wires used to a telephones rather than a
single wire with ground return, as in telegraph, or two pairs, one
for talking and one for listening?
2.
Name the wires used in a connection to an analog telephone. Where do
they begin and end?
3.
How is ISDN wiring at the BRI-U interface different from that at the
S/T interface?
4.
Name the electrical properties of a telephone line.
5.
What is a bridged tap? A loading coil? Who cares?
6.
Where is a 4-wire to 2-wire hybrid located when the switch has a
metallic matrix? A digital matrix?
7.
In 6, which location provides the best echo performance?
8.
How is the range of analog phones limited?
9.
Are there technical reasons for charging more for a business line
than a residential line?
10.
What is the difference between an OPX and an FX line?
11.
Suggest some reasons why party lines should die.
12.
What are the basic functions an analog line must perform?
13.
How will these functions be different with an ISDN BRI?
14.
What is a "bridged extension?" Can digital phones be bridged like
analog phones?
15.
A switch vendor says his switch can support longer customer lines
because it uses more sensitive detectors for supervision. How do you
know he is wrong?
16.
Why are ground start lines used?
17.
Does a 911 system need joint holding?
18.
Will ISDN phones provide a switch-hook flash?
19.
Do customer lines receive answer supervision?
20.
Why would a customer line want answer and hang-up supervision?
21.
Can dialing at 20 pps be used on your phone line?
22.
What are the advantages of DTMF?
23.
Why is power ringing still so widely used?
24.
Why is ring-trip a problem?
25.
Identify three types of tone ringing.
26.
If party lines are being phased out, why would a designer be
interested in reverting calls?
27.
How does an ISDN S/T interface correspond to a party line?
28.
How can many lines on a digital switch listen to the same call
progress tone?
29.
What kind of service circuits are not normally found on a digital
PBX or CO switch?
[ Top ] [
Next Chapter ] [
Table of
Contents ] |
