Background
for Telephone Switching
2nd Edition (Revised and Expanded)
Chapter 4
Interfacing Trunks
OUTLINE
OBJECTIVES: The purpose of this chapter is
to discuss:
-
The difference between metallic and carrier trunks.
-
The different types of trunks often encountered.
-
The several types of supervision and signaling used on trunks.
-
The intent behind analog transmission theory and the problems it
sought to solve.
-
Some possibilities for future standards in transmission and
signaling which take into consideration modern technology.
PREVIEW QUESTIONS: As you read, watch for
the answers to the following important questions:
1. What are trunk signaling and supervision?
2. How does trunk transmission work?
3. How does digital technology change trunk design and
operation?
INTERFACING TRUNKS
Nowhere has the Glass Age had more impact than in trunking. Optical
fiber, with its huge bandwidth, has made digital long distance
trunks not only possible but economically inescapable; without
digital long distance trunks, there would be little thought of ISDN
or other approaches to unified information transfer.
Initially, T-Carrier systems had channel banks on each end to make
their individual circuits look to associated switches like analog
trunks with conventional supervision. When computer control was
applied to analog switches, it was possible to bring the per-trunk
out-of-band signaling information directly to the control in a
digital format, greatly simplifying the design of both trunk
circuits on the switch and channel plug-in units on the carrier
system. When digital switches using the same format as T-carrier
became available, starting with AT&T's 4ESS in 1976, it was possible
to omit all per-trunk hardware and bring the multiplexed signal,
including voice as well as out-of-band signaling, directly into the
switch. One or both ends could be handled this way; if one switch
was analog, a conventional channel bank could make the conversion at
that end. Digital switches, by omitting a great deal of hardware,
improved overall system reliability.
The
next step, starting with the replacement of 4XBAR with 4ESS toll
switches, was to omit out-of-band signaling entirely, eliminating
the "bit-robbing" of the 8th bit stolen from voice coding every
sixth frame for supervision, and go to CCIS, or common channel
interoffice signaling. Here, information was transferred from the
computer control of one switch to the computer control of another
using data packets on a separate signaling channel not necessarily
related to the individual trunks. The signaling network AT&T
developed for CCIS was updated in the early '90s to CCITT's
Signaling System No. 7 (SS7), and additional capabilities were moved
into the signaling network.
The
reader should know that new equipment being installed today is based
on digital technology and SS7. However, an enormous amount of older
equipment in both public and private networks will not be retired
for some time; thus it is still important to understand earlier
technologies. Further, if one understands how older systems worked,
it is easier to communicate with those who have been in the business
a long time and who still think in terms of earlier generations of
equipment. Thus this chapter will retain information about
loop/reverse-battery and E&M supervision, VNL transmission plans and
other dinosaurs, even as we all hope for a time when such concepts
can be left to historians.
The
last word, however, has not yet been said (let's hope it never is),
and many totally new approaches are already in advanced development
for eliminating individual circuit-switched trunk connections in
favor of various advances in packet switching. It remains to be seen
just how all these marvels will come to pass; in the meantime, we
have to live with what is presently available.
In
Chapter 3, we noted that "lines" connect customers to switching
systems and "trunks" make connections between systems. However, when
a customer owns a switch such as a PBX, the term "trunk" is applied
at the PBX end of a path to the CO while the CO still considers it a
line. Further, even at a PBX, an FX line is a CO trunk to a remote
CO switch, and private paths between PBXs have been called tie-lines
for years, although they are almost identical to trunks between
telephone company switches.
To
add to the confusion, paths between different stages of a SXS switch
were called trunks, while the military and now some telephone
companies refer to a trunk as a whole multiplexed channel between
two switches containing 24 or more individual circuits. Indeed,
digital multiplexing is such that several DS0 channels (each
equivalent to one voice circuit at 64 kBps) can be combined; a
contiguous group of 6, 384 kBps, is called an H0 channel, and a 24
channel di-group (DS1) can be made available as a single broadband
channel, with higher levels of multiplexing making even broader
channels possible. Obviously, channels at different bandwidths will
require different treatment in traffic studies, channel selection,
etc., as they come into more common use, and message (packet)
switching, when combined with circuit switching, will add to the
complexity.
In
what follows, we will use "trunk" to mean "one telephone
communication channel between two switching centers." Between a PBX
and a CO switch, we will use the term "CO trunk," and between PBXs,
"tie-trunk." With any luck at all, the context should also help make
clear what is meant. We will leave broadband switching for a future
book.
In
general, an analog trunk (or a digital trunk set up to emulate an
analog trunk) is terminated at each end in a "trunk circuit," part
of the switching system but considered part of the trunk when making
transmission measurements. The trunk circuit acts as an interface
for transmission, supervision and signaling between the transmission
facility and the switching system. The coming of digital trunks
interfacing digital switches directly has, of course changed all of
these concepts considerably.
Trunks tend to be expensive; thus they are provided only as needed
to serve the traffic* available, and usually have much higher
occupancy than station lines. For instance, a group of 30 trunks,
operating with the probability of three calls in a hundred being
blocked during the busy hour, will average nearly 70% occupancy. In
any new design for a local CO switch, plans should be made early to
permit trunk-to-trunk connections without appreciable concentration
and preferably on a non-blocking basis. CO trunks from PBXs, like
all other trunks, should not encounter concentration in the
switching matrix.
[*Footnote: See Tables for Traffic Management and Design, Book
1--Trunking, in the Lee's ABC Traffic Series.]
Metallic trunks
Until quite recently, many trunks, particularly in large cities
where central offices are in close proximity, were simply pairs of
wires between trunk circuits. For short runs, copper was less
expensive than electronics, and often more reliable. As with station
lines, two-wire facilities were used where possible, taking
advantage of their ability to transmit in both directions
simultaneously. On longer circuits, where amplification was
required, four-wire facilities were used even when it meant doubling
the amount of copper. Hybrids and unidirectional amplifiers were
generally more satisfactory than "negative impedance" amplifiers
that subtracted loss from two-wire circuits.
Metallic trunks, like lines, were characterized by series
resistance, shunt resistance, and shunt capacity. Because trunks are
usually much longer than station lines, inductance in the form of
loading coils was commonly added to improve transmission at the
higher frequencies. Because such trunks went from switch to switch,
using cable all the way in protecting ducts, shunt resistance was
much less important than in station lines; 30,000 ohms was typical
of the minimum value to be expected. Supervisory currents on trunks
between central offices and on tie-trunks between PBXs were often
limited to 50 mA rather than the 100 mA or more found on station
lines; this required the battery-feed resistance in "loop" trunks to
be on the order of 800 ohms rather than the 400 ohms used with lines
(prior to the coming of electronic battery feeds).
Analog trunk circuits generally provided idle circuit terminations
to transmission facilities to which they were attached. This
termination was nominally 900 ohms in series with a 2 µFd
capacitor for exchange trunks and 600 ohms plus capacitor for toll
trunks. Four-wire trunks do not need idle circuit terminations; if
there is no path from the receive to the transmit side, there can be
no echo. However, terminating the outgoing side of a four-wire trunk
tends to drain off any noise picked up in office wiring.
Metallic trunks seldom proved in at lengths longer than 20 miles
compared with short-haul analog O and N carrier (12 trunks on two
pairs); with the coming of digital T-carrier (24 trunks on two
pairs), the break-even point became as low as five miles. When
channel banks could be eliminated at digital switches, it dropped to
zero; that is, multiplexed channels on existing copper were less
expensive than adding more copper pairs, particularly when city
streets had to be dug up to install more cable ducts. Using
electronics to derive more channels on existing copper has been a
major factor, but replacing copper with optical fiber to increase
the available bandwidth in the same ducts is even more important.
Carrier trunks
Carrier trunks have, in the past, done their best to look like
metallic trunks to switching equipment, and switching systems have
supported this fiction from their side of the interface. The results
were often ludicrous because carrier systems cannot support loop
opens and closures for on-hook and off-hook, and have to substitute
signals such as audio tones instead. Dial pulse senders continued to
transmit opens and closures, all the while trying to maintain
suitable transmission terminations to prevent echo, and carrier
trunks then converted dial pulses into audio tones that could have
come from the sender in the first place. At the terminating end, the
tones were turned back into dial pulses to send through the metallic
switching matrix to a DP signaling receiver.
Because most analog switches had two-wire switching matrices, analog
carrier trunks, which had to be four-wire, converted the signal to
two-wire before meeting the switch. Naturally, digital trunks were
converted to analog before becoming two wire. Early digital PBXs,
when used for switching digital tie-trunks, would thus require a
digital-to-analog and four-wire to two-wire conversion in the
carrier terminal, followed by a two-wire to four-wire conversion in
the PBX trunk circuit to permit digital coding for switching. At the
connecting trunk, the process was repeated, and almost nobody
perceived the humor of the situation. The coming of AT&T's digital
toll switch, 4ESS, which interfaced digital trunks directly and in
the proper format, set an example which took others years to follow.
On
analog trunks, tone on and tone off corresponded to on-hook and
off-hook, respectively, and analog carrier systems developed two
approaches to supervision and dial pulsing: out-of-band and in-band.
Out-of-band used a signaling channel separate from but physically
associated with its voice channel; typically, a sharp filter allowed
the tone at 3700 Hz to sit just above the voice channel, isolated
from it so that it could not be heard and speech could not affect
its tone detector. Out-of-band signaling was part of the carrier
system.
Although built-in out-of-band signaling was inexpensive when a trunk
consisted of a single carrier system operating back-to-back with the
trunk circuits in the switches on each end, such conditions became
increasingly rare as the toll network grew. More commonly, channels
in two or more carrier systems had to be tied back to back to create
an individual channel to a particular switch. For instance,
short-haul analog carrier systems (O and N) might be used to bring
trunks from several different local switches to a central point
where they could be cross-connected to a long-haul high capacity
carrier system (L, TD2, TH, TJ) going to a distant toll switch. In
such instances, not only would speech paths have to be
cross-connected, but signaling leads as well. Further, the several
tone detectors in series introduced distortion in dial pulses,
switch-hook flashes, and other on/off-hook transitions. This led to
the development of in-band signaling.
With
in-band signaling, carrier systems only handled transmission and
omitted all signaling equipment, the leads connecting to it, and
records associated with those leads; this effected a considerable
saving in both hardware and operating costs. Instead, a single
frequency (SF) signaling unit was placed at each end to interface
the trunk circuits of the associated switching systems, improving
performance and reducing costs by eliminating signaling units in
tandem.
A
switch maintaining an on-hook condition toward the trunk caused
the SF unit to gate a 2600 Hz tone, well within the speech band,
onto the outgoing speech path; the far end's SF receiver detected
2600 Hz tone, and converted it to the kind of on-hook supervisory
signal the switch expected to see. When either switch wanted to send
an off-hook condition, it instructed its SF unit to remove the tone,
and the receiver at the far end converted tone-off into off-hook.
Digital T-carrier systems originally used out-of-band signaling,
with 7 bits coding speech and one additional bit added to each code
group for on-hook/off-hook. Later T-carrier systems wanted all 8
bits for speech coding, but had to rob one bit in every sixth frame
for supervision. With common channel signaling, a separate signaling
channel, not necessarily related physically to the medium carrying
the voice channels (and hence not classified as out-of-band), is
used. This allows all 8 bits in each sample to carry voice as long
as each group of eight bits contains at least one 1 (a pulse on the
line) to maintain synchronism in line repeaters. T-carrier line
drivers convert 0s to 1s and vice versa, so silence is all 1s on the
line, and all 0s can only be a 4 kHz tone which would be cut off by
low-pass filters associated with the codecs. Unfortunately, when
data or image are mapped directly into the bit stream, the all 0s
condition can exist causing the "ones density" requirement, needed
to keep line repeaters in sync, to fail. Various methods are
available to deal with the ones density problem, and common channel
signaling is rapidly becoming universal.
In
any event, a trunk circuit on an analog switching system (or a
digital switching system choosing to appear analog to the outside
world) connects to an SF set if in-band signaling is used, and
directly to the port circuit in the carrier system's channel bank if
out-of-band or common channel signaling is used.
Analog trunks are almost always de-multiplexed to individual voice
circuits to use a patch panel for cross-connection to one another in
built-up connections, or to an SF set at each end. Digital carrier
systems have the great advantage of being able to cross-connect
individual channels without converting back to analog, and to
interface with digital switches or higher level multiplexers without demultiplexing to individual channels. Further, it is possible to
make a channel of greater capacity, primarily for high-speed data or
compressed video, by simply combining two or more 64 kBps signals;
these options are not available with analog carrier systems,
although an entire 12 channel "group," omitting individual channel
units, can interface a V.35 modem to carry data at a 48 kBps rate.
Two-wire analog exchange trunks are nominally 900 ohms
characteristic impedance, while toll trunks are 600 ohms. Four-wire
facilities are 600 ohms in all cases. In carrier facilities,
characteristic impedance has nothing to do with the properties of a
transmission line; it is controlled by circuit elements in the SF
set or the analog port circuit in the channel bank.
Operator trunks
Not
so long ago, all toll calls were placed by operators working at cord
switchboards. Connections were established via manual trunks using
ringdown (spurt) supervision from operator to operator, and at the
terminating end, the final operator either connected to the called
party directly or through the local automatic switch. The original
ringdown signal used between operators was regular 20 Hz ringing;
because this signal wouldn't go through carrier systems, it was
replaced with a 1000 Hz tone modulated at 20 Hz.
No
matter how automatic telephone switching may become, access to a
skilled human being will always be necessary. Often, operators
(telephone company employees) or attendants (customer employees) are
associated directly with a switching system, as in the case of
modern toll systems or PBXs. However, in the past, operator
switchboards have been separate switching systems with their own
trunks.
Switchboard Trunks. A few manual toll
switchboards such as AT&T's 3CL (see Chapter 6) are probably still
in use somewhere to provide operator services. Trunks to jacks on
such switchboards are accessed by the local CO switch for dial 0,
coin calls beyond the basic charging area, etc. To simplify coin
collect and return, a third wire often accompanied the tip and ring,
although later versions used MF tone-pairs over the speech path
itself.
At
the 3CL, the operator would, upon seeing a lamp signal, plug into
the matching jack with one cord of a cord-pair, talk to the caller
to obtain the necessary information, and then, complete the call by
plugging the other cord of the cord-pair into an outgoing trunk. For
directory assistance and similar services, the second cord was not
used. However, on through connections, the operator actually
established the switched connection with the cords and had the call
available until hangup so that timing could be carried out for
billing. This arrangement also allowed the operator to re-enter the
call if further assistance was required.
The
interface from switchboard to human was primarily lamps, which could
be off, on, and flashing at various rates. The operator
interfaced the switchboard by inserting and removing plug-ended
cords and operating switches associated with each cord pair or the
operator position.
Other Operator Trunks. The purpose of a
cord board, even when used as the attendant position for a SXS PBX,
was to act as a switching system to connect one circuit to another.
With the coming of smarter automatic switches, not necessarily even
stored program, the operator position changed from a manual switch
to a control console to tell the automatic switch what connections
to make. As will be discussed in Chapter 6, operator positions
ultimately evolved into little more than electronic telephone sets,
associated with their switch via an ISDN basic rate interface (BRI);
the operator can still greet a caller and effect the proper
connections, but operator trunks as specific entities have vanished.
Automatic trunks
Although early long distance calls were set up by an originating
operator, a terminating operator, and various operators in between,
automatic switching equipment was introduced so that only one
operator, who made out the toll ticket for billing, could also dial
the connection without the help of other operators. Ultimately,
direct distance dialing (DDD) took over and the user was able to
dial toll calls without operator assistance. As a result, most
trunks today are automatic. They are accessed by automatic switching
equipment, signaling is totally automatic, and even maintenance is
automatic. Automatic toll switching, when it came, was not much more
complicated than tandem switching already widely used in major
metropolitan areas; automatic billing, called automatic message
accounting or AMA, was vastly more difficult and had to be developed
before the operator could be eliminated.
Automatic trunks may be classified as one-way and two-way. A one-way
trunk can be seized at one end only, while a two-way trunk can be
seized from either end. When one-way trunks are used between two
switches, two different groups are needed: one which can be seized
by switch A, and the other which can be seized by switch B.
Signaling and control are much simpler with one-way trunk groups; as
will be discussed, there is no need to worry about the same trunk
being seized simultaneously at both ends, producing a "glare"
situation.
PBX
tie-trunks, similar to telephone company interswitch trunks, are
usually operated two-way; when a circuit is purchased by the month,
it has to get the most possible use to justify its cost. From
traffic theory, it is well known that one group of 10 trunks will
carry a good deal more traffic than two groups of 5 at the same
grade of service. Where facilities are very expensive, or where
the number of circuits that can be justified is very small, even
telephone companies will use two-way circuits. Both tie-trunks and
telco trunks are sometimes arranged in three groups: a one-way group
from switch A, another one-way group from switch B, and a 2-way
group available to both and used when either one-way group is full.
This compromise gives higher occupancy than two one-way groups while
reducing glare (at least on the one-way trunks). It also guarantees
the possibility of originating calls from one end during a surge of
very heavy calling from the other.
Where possible, trunk hardware should be installed as two-way, even
if it is to be used one-way. In modern switching systems, the
direction of a trunk is determined by information stored in memory
and can be altered by simply changing its class of service unless
the trunk circuit, signaling circuit or port in the carrier
channel-bank prevent this. Naturally, a change in direction must be
made at both ends; if each switch thinks a trunk is incoming only,
the trunk will never be used.
Analog PBX-CO trunks
As
has been indicated, trunk terminology in connection with PBXs leaves
much to be desired. At the CO, trunks are just lines, usually
implemented ground start, but they carry highly concentrated traffic
from a much larger number of extensions. Prior to about 1980,
Centrex, defined as PBX features plus direct inward dialing (DID)
and identified outward dialing (IOD), was implemented, at telephone
company option, either by a CO switch or a PBX on the customer's
premises. When provided by a PBX, trunks to the CO were not billed
to the customer, and the telephone company took full responsibility
for providing the proper number. There were two separate groups: one
for outgoing and one for DID calls.
With
either form of Centrex (Centrex CO and Centrex CU, depending on the
location of the switching equipment), extensions were regular
loop-start lines. However, Centrex CO lines were, obviously, a great
deal longer than Centrex CU or PBX lines, and had to pass through
the hazards of the outside world. Eventually, the DID function was
unbundled and DID trunks were made available to customer-owned PBXs
as a separate feature of the CO. Today all Centrex service is
provided by CO switches; DID trunks to PBXs, like those for Centrex
CU, are a category separate from loop start and ground start, and
are implemented with different hardware and software.
The
CO ends of ground-start and loop-start lines have been discussed at
length in Chapter 3. In general, ground start lines are preferred
for use as PBX trunks because the PBX receives an easily detected
signal, ground on tip, when dial tone is returned on an outgoing
call or when the CO seizes the trunk for an incoming call; the
removal of that ground means the outside party has hung up. The PBX
originates a call toward the CO by applying ground to ring but, upon
seeing ground on tip from the CO, changes the ring-ground connection
to a connection from tip to ring. On incoming calls, the PBX again
goes to a tip-ring connection after detecting ground on tip.
Ground start PBX trunks used two-way have been called, in the past,
"combination trunks"; calls incoming from the CO are routed to the
console for answering, while PBX extensions can make outgoing calls
directly on a "dial 9" basis. The ground-on-tip signal from the CO
minimizes the possibility of the trunk being seized from both ends,
and removal of ground on tip when the outside party hangs up
facilitates release of the trunk when the call is over.
At
the PBX end, loop start trunk circuits are much simpler than those
for ground start. They simply make a tip-ring closure to originate a
call, and detect ringing when the CO calls them; they do not need to
detect or provide a ground signal toward the CO, or change to a
tip-ring connection after the ground is detected.
Because loop-start trunks receive no positive hang-up signal from
the outside party at the end of the call (at best, they may see a
momentary open in the loop), they usually depend on the PBX caller
to hang up to signal that the call is over. Loop start lines are
used to connect to conventional telephones and key telephone
systems, where the person originating an outgoing call has a high
probability of being the person for whom an incoming call is
intended in the event of a double seizure. Loop start PBX-CO trunks
can also be used one way, either outgoing or incoming, although the
lack of a positive hang-up signal is sometimes a problem.
DID
trunks are one-way from CO to PBX to eliminate the possibility of
seizure from both ends. Because few older COs were capable of "line
side outpulsing," or sending dial pulses toward station lines,
actual trunk circuits (loop start reverse battery, to be discussed),
are often the CO interface rather than a line circuit. The PBX DID
trunk circuit is different from loop- and ground-start PBX trunk
circuits in that it supplies battery and ground while the CO
supplies the tip-ring closure. Further, the PBX reverses battery on
tip and ring when the called party answers so that the caller can be
charged.
Signaling on DID trunks was, for some years, exclusively dial pulse;
because DTMF was not used on trunks and was only received from
customers, not sent to them, DTMF senders simply did not exist
(except for test and maintenance purposes) until quite recently.
Today, DTMF is the preferred way to tell the PBX which extension
a trunk is to be connected to for a particular conversation,
although there are many systems in place still using dial pulsing.
Because many older CO switches had no way of providing line-side
outpulsing, and also had no way of making a trunk-to-trunk
connection (incoming to DID), tandem switches, with more translation
and outpulsing capability, were sometimes arranged to bypass the
local CO and serve PBX DID trunks directly as though the PBX were a
CO itself. In general, loop supervision was used, although E&M
supervision was also common, taking advantage of standard PBX
tie-trunk circuits. The mechanics of loop, E&M and other supervision
techniques will be discussed below.
When
DID is provided, the CO should inhibit transmission toward the PBX
until after answer supervision is received; otherwise, a generous
PBX owner can make modifications to suppress the answer signal and
provide free incoming calls. (That few CO switching systems bother
with this refinement is a testimonial to the basic honesty of PBX
customers everywhere.) Transmission from the PBX should always be
possible so that call-progress tones (audible ringing, busy, etc.)
from the PBX can be heard by the caller.
Because of the high occupancy of PBX trunks and their need for more
complex signaling, new designs for CO switches should not try to
pretend that a PBX is a Princess telephone. Rather, all CO switches
should be designed to make trunk-to-trunk connections as readily
line-to-trunk, and PBX trunks at the CO should not be implemented as
though they were subscriber lines. However, trunks implemented on T
Carrier from a digital CO solve all these problems so easily that it
is unlikely that much effort will be applied toward improving analog
trunks.
In
modern electronic PBXs, there are usually at least four CO trunk
circuits on a circuit board, and possibly as many as 8 or 16. These
trunk circuits are usually arranged to function as either loop start
or ground start, depending on the way system software tables are set
up. Because an electronic matrix isolates trunks from lines, there
is no way to take advantage of the automatic level adjustment built
into 500 type telephone sets which worked fine with metallic
through-connections in the old days. Thus digital pads are often
inserted in the matrix path under program control to provide
different loss for intra-PBX and PBX-to-trunk connections. Further,
system software is sufficiently flexible to give incoming calls
access to the console attendant, internal hunt groups and a variety
of features, and there is no trouble providing transfer capability
from one extension to another on any call, something only available
to calls via incoming trunks just a few years (and several
generations of equipment) ago.
Digital trunks and digital signaling
When
digital trunks terminate directly in a multiplexed digital format on
digital switches, it is no harder for the switch to read the
incoming supervision on the bit robbed from voice coding in each
sixth frame, or to change the matching bit as required on the
outgoing side, than it had been for stand-alone channel banks to do
the same thing and interface the switch through 19th century control
leads.
Digital PBXs became available in 1975, but many of them used digital
formats alien to T-carrier and might as well have been analog. Those
that were digital in the proper format still had to depend on analog
CO trunks (often implemented on T-carrier) because the CO switches
on which they homed were, for many sound reasons, analog and
two-wire. However, during the past 15 years, digital (access and)
cross-connect systems (DACS or DCS) have made considerable progress
in augmenting and then replacing MDFs (as will be discussed in
Chapter 7), and digital switches are rapidly taking over local CO
switching. Further, direct digital connections to digital
long-distance carriers have become necessary in the same time-frame.
As a result, most PBXs today are digital and compatible with
T-carrier, and have circuit boards for trunks that allow them to
meet T-carrier on a multiplexed 24-channel basis.
Although digital PBX trunk circuits have been widely used, first as
tie-trunks and digital interfaces to computers, and then as CO
trunks (terminating on a channel bank at the analog CO, or via a
DACS to permit individual digital trunks to be diverted elsewhere),
the coming of digital COs has made digital trunks between COs and
PBXs the only reasonable way to proceed. Note that this is
independent of ISDN.
However, SS7 is rapidly taking over, and various features, marketed
under ISDN tariffs, are demanding common channel signaling to the
PBX itself. As a result, the ISDN primary rate interface or PRI has
been developed to replace bit-robbed supervision and conventional
signaling when digital facilities are used to a PBX. PRI is often
referred to as "23B+D" because it carves out one D channel at 64
kBps for signaling, and leaves the remaining 23 "bearer" channels
for voice, data, or whatever. The signaling messages on the D
channel are a subset of those used by SS7 (omitting many that relate
to telephone company administration, etc.), but allow a wide variety
of new services to be made available to PBX customers.
When
a PBX has more than 23 trunks to the CO, which is usually the case
for PBXs of even moderate size, the D channel in the first T-span
can handle a number of additional T-spans, allowing all 24
channels on those facilities to be used for customer traffic.
Indeed, D channel signaling allows one group of T-spans to support a
number of different trunk groups, and even reassign individual
trunks from one group to another in real time as needed ("call by
call service selection"). Note that SS7 for telephone company
signaling and the PRI version for PBX interfacing require a 64 kBps
channel most easily obtained on T-carrier. However, the trunks
controlled need not be on the same carrier system, and need not even
be digital.
Supervision
Obviously, the switch at one end of a trunk must know when that
trunk has been seized at the other end. Further, the originating and
terminating switches, and all the tandem and/or toll switches in
between, have to know when a call is over so that the trunks can be
released for use by others. It is also desirable to inform both the
originating and terminating line when a call is over so that various
customer owned equipment (such as answering machines) can be
released. The switch responsible for billing must know when
conversation begins so that charge timing can start; PBXs and
privately owned pay telephones also need answer supervision for
accurate call billing which they supply to their clients independent
of telco billing.
These are the principal functions of supervision. The
on-hook/off-hook signal from the calling telephone is transmitted
forward, and the on-hook/off-hook signal from the called phone is
sent back. On analog trunks, there are two common ways of doing
this: loop supervision and E&M supervision. However, common channel
signaling in the form of SS7, which can also handle address
signaling and a variety of other things which older approaches
cannot, will, with luck, soon be universal.
Loop/reverse-battery supervision. Loop
supervision, as used in switches with metallic matrices, is shown in
Fig. 1. It is one-way: switch X can seize the trunk but switch Y
cannot. X seizes the (outgoing) trunk by closing the loop, using
contacts K. At switch Y, the A relay operates over the loop. Various
unshown things happen to ring the called line. Upon answer, relay S2
at switch Y operates. This causes relay REV to operate to transpose
the connection to tip and ring, reversing the current flow through
relay CS at switch X. CS is a polar relay; it now operates because
current flows through it in the proper direction. The CS relay
causes charge timing to start.
At
the end of the call, either party can hang up first. If it is the
calling party, the S1 relay releases, causing relay K to open
the loop. This automatically releases the CS relay and the A relay.
REV will restore the proper polarity to the loop when the called
party hangs up and S2 releases. If the called party hangs up first,
restoration of REV releases CS to stop charging. The trunk will not
be released, however, until the calling party hangs up, releasing S1
and causing K to open.
Fig.
1 has been carefully drawn to omit the countless details that go
into a trunk circuit. For instance, slow release timing is needed to
follow S1 and S2 so that momentary hits on the line will not release
the trunk. CS is often followed by a timing device so that charging
doesn't start for two to five seconds after answer to prevent
shorter false signals from starting charging prematurely. In direct
controlled systems, the S1 relay must follow dial pulses and repeat
them into the loop; "ABC" timing is needed to differentiate between
dial pulses and release of the circuit. Ringing and ring-trip are
specialized problems that require considerable attention. Additional
system timing is required at hang-up to make certain both ends are
released before the circuit is re-seized.
Loop
supervision was developed with metallic switching matrices, which
could extend the customer loop through to the trunk, in mind. With
electronic matrices, the S sensors remain in the line circuits,
monitored by the system controls at switch X and switch Y. These
controls also operate K and REV respectively and monitor A and CS.
Timing functions are performed by software.
When
carrier trunks are used, the outgoing trunk circuit goes to an
outgoing SF unit or carrier-system input which contains the
equivalent of an A relay. A operated turns off the outgoing SF tone
or changes the state of the supervisory bit in every sixth frame of
a T Carrier system. At the terminating end, a K relay in the SF or
carrier trunk unit closes a path toward the switch to repeat the
off-hook to another A relay in the incoming trunk circuit. When the
called party answers, a reversal is sent to a polar device in the
terminating signaling equipment which relays the off-hook signal
back over the trunk to signaling equipment which repeats the
reversal to operate CS in the outgoing trunk circuit. Signaling
sets, whether separate external units or built into the transmission
channel, are quite complex, and contain various timing and guard
functions. When MF pulsing is used, SF sets, in particular, can be
simplified by omitting circuitry required for dial pulse correction.
One
major use of loop supervision today is in DID trunks to PBXs, where
the central office is Switch X in Fig. 1, and the PBX is Switch Y.
E&M lead supervision. E&M lead supervision
is two-way, but it can be operated one-way by the switching system
if desired. A very simple version is shown in Fig. 2. The M lead is
switched between battery and ground; battery sent toward the trunk
means the local user is off-hook. The E lead receives an open or a
ground from the trunk; open says the distant party is on-hook, and
ground says off-hook. It should be noted that the E lead from one
trunk cannot operate the M lead of another when two trunks are
connected back to back, bypassing a switching system; a "pulse-link
repeater" must be inserted.
Obviously, an M lead from a trunk circuit can turn an SF tone on and
off in a signaling set, or change the state of the T-carrier
supervisory bit every sixth frame. Similarly, the E lead from the
trunk toward the trunk circuit can be controlled by tone or the
supervisory bit on the other side of the trunk. Both ends of an E&M
trunk are completely symmetrical; in either direction, the status
applied to the M lead comes out on the distant E lead. In carrier
systems, supervision only sends on-hook/off-hook; thus the presence
or absence of tone or the supervisory bit does not really care if
the terminating equipment is loop, E&M, or something else. As a
result, it is not uncommon to have E&M plug-in units on one end of a
trunk, and loop signaling units on the other, with the carrier
system acting as a loop-to-E&M converter.
E&M
lead signaling is a hold-over from the very early days of telephony.
Battery, through a resistance lamp, was applied to a copper trunk
conductor so that it could charge up as quickly as possible,
minimizing dial pulse distortion. The resistance lamp was used so
that, if a short occurred, the lamp would heat up, increase its
resistance, and limit current flow to a safe value. Normally, it
runs cool and has a low resistance. Because almost all trunks (other
than PBX CO trunks) run on carrier and there is no copper path to
charge up, the resistance lamp today is usually replaced by a much
less expensive resistor.
Traditionally, the M lead may have only 25 ohms of resistance
between the trunk circuit and the signaling set. This says the
carrier terminal or SF set must be in the same building with the
switching system, and as close as possible. When the switching
system is located in a distant building, as is often the case with
tie-trunk switching at a PBX, means must be found to get the E&M
lead signals from the PBX to the signaling set.
CX
and DX sets have been used for this purpose. These signaling sets
were developed for long metallic trunks, and use polar relays and
other devices in ingenious ways. CX (which stands for "composite")
and DX (which stands for "duplex") sets differ largely in that CX is
used when three voice channels are placed on two pairs, while DX
works over a single pair, serving one circuit.
In
composite systems, one voice channel was put directly on each pair,
and the third used the "phantom" circuit. Three of the four wires
were used for supervision, one for each voice channel, and the
fourth wire was either used for ground-potential-difference
compensation or else for a telegraph channel. DX sets, using the
same conductors as their associated voice channel, were sometimes
built directly into trunk circuits, making them particularly useful
in the PBX-to-CO connection for tie-trunks. We may hope, however,
that such ancient hardware is now phased out of the telephone plant.
When
SF sets became generally available, it was logical to omit CX and DX
sets by putting SF sets near the switching system rather than
adjacent to the carrier system. But just as this became possible,
digital carrier terminating directly in the switching system instead
of a channel bank came along, allowing the switch to access
out-of-band supervision bits directly (note that these bits could
have been used as a 1.67 Kbps data link rather than on-hook/off-hook
if anybody had really wanted to). Then, when designers were in a
position to exploit the possibilities of this mode of signaling, SS7
and its customer interface, PRI, came into vogue. With luck, SS7 and
PRI will stick around long enough to let designers to take full
advantage of modern technology.
Other supervision methods. Other
supervision methods have been used in the past, and are still
doubtless in the plant. "High-low" is very much like loop
supervision except that the CS relay has two windings, one with a
very high resistance, and one with a low resistance. The relay K,
instead of opening and closing the loop, shorts out the high
resistance winding to operate the distant A relay. The REV relay at
the other end, upon operating, provides the right polarity to
operate the CS relay on both windings or the low-resistance winding
alone. Thus "High-low" supervision is two-way rather than one-way.
"Wet-dry" is also mentioned in industry literature. A circuit is
"wet" when battery is applied at the called end, and "dry" when
battery is removed. The circuit must be wet to detect a seizure
(tip-ring dc connection), but it then goes dry during the call.
Opening the loop should cause release, but it cannot do so until the
called end makes the loop wet, indicating that the called party has
hung up.
Hang-up timing. At the end of a call,
either party may hang up first, or both parties may hang up nearly
simultaneously. There are often several trunks in tandem in
interswitch connections which must be released at both ends before
they can be re-used for another call. It is not enough to know the
distant end has gone on-hook. There must be a high probability that
the path through the switching matrix has been released and the
trunk is free before it is again made available for seizure.
On a
one-way trunk, the problem is relatively simple. When the callING
party hangs up, the connection is released and the originating
switch will not seize the trunk again until a timing interval
elapses, giving the far end time to disconnect. If the callED party
stays off hook, the trunk-to-callED-line connection is dropped. If
the callED party hangs up first and the callING party stays off
hook, there is usually an interval to allow the callED party to pick
up on a different extension. When manual switchboards were used as
PBXs, this also gave time to transfer a call by moving its cord from
one jack to another. If the call is not picked up in 10 seconds or
so, a real hang-up is assumed, the matrix path from incoming trunk
to callED party is released and the path from the callING party to
the outgoing trunk is released.
In
electromechanical systems, a party remaining off hook after the
other hung up would be given dial tone after the appropriate
line-to-trunk connection was released; this practice was copied in
many stored program systems which were smart enough to know better.
Today, the preferred practice is to withhold dial tone until the
caller hangs up and comes off-hook again, but implementation is by
no means universal.
On a
two-way trunk, reseizure at both ends must be prevented for a timed
interval long enough to assure each end that the other end has
released before reseizure is permitted. Because one end will very
likely time out before the other, a system must be able to respond
to a far-end seizure before it has made its end of the circuit
available for near-end seizure.
In
the era of electromechanical switching, where relays detected
supervision and used their own contacts to pass it along, hang-up
propagated quickly through a built-up connection of many trunks in
tandem, while slow release relays in each trunk effectively released
in parallel to make individual circuits available. In electronic
switches where sensors are scanned for hang-up, there may be up to
half a second time-out before a switch decides the hangup is valid
and causes the on-hook signal to be propagated on the connecting
trunk. Such delay is, of course, cumulative, and shows the
importance of common channel signaling in bringing modern switching
systems back to the speed that came naturally with early
electromechanical systems.
PBXs
have their own release problems which are reduced with ground start
trunks. The ISDN BRI and PRI interfaces between telephone company
and customer equipment will, with luck, resolve the remaining
difficulties, although having PBXs access CO switches via trunks
rather than lines, technically possible years ago, could have been
almost as effective.
Make-busy. Trunks are shared facilities,
and a bad trunk can cause no end of difficulty until it is detected
and taken out of service. Automatic maintenance procedures as well
as manual testing, discussed in Chapter 8, are absolutely necessary
to help find such problems; once found, the bad trunk must not be
used until it has been repaired. This means that some provision must
be available to make a bad trunk appear busy to the switch so that
it will not be seized (two-way trunks have to be made busy at both
ends), and the difference between make-busy and regular busy must be
clearly visible to maintenance and administrative personnel.
Start-pulsing signals
None. In SXS systems, each tie-trunk and DID trunk had its own
incoming selector to which it was permanently connected. Thus, if
you could seize a trunk to a SXS system you could start sending dial
pulses immediately (actually, a 70 milli-second delay was built into
senders to permit selector relays to soak, or become fully
operated). There was no need for dial tone to hold off humans, or
other techniques to delay outpulsing from senders.
In
some electromechanical common control systems such as 5XBAR,
designed to work with nearby SXS COs, a "by-link" was used, making a
high-speed connection from an incoming SXS trunk (with user dialing)
to a dial pulse register until the regular path could be
established. If the first dial pulse came in before the by-link was
up, reorder tone was returned to the dialer. In switches like 1ESS,
controlled from a central computer, a different approach was
possible: just scan the supervisory signal on incoming dial-pulse
trunks fast enough to pick up the initial seizure and the following
dial pulses. Software could then assemble the digits and store them
in common memory, making the incoming dial pulse register of 5XBAR,
which stored the called number in its own relay memory, unnecessary.
Unfortunately, in large systems, scanning idle trunks fast enough to
catch seizure followed by dial pulses took up a great deal of the
central computer's "real-time." Only with the coming of distributed
processing, where each module had its own front end to handle
scanning and other routine chores, was the real-time problem
resolved. By then, however, SXS central offices were almost
completely gone. In private tie-trunk networking, PBXs, having
perhaps two orders of magnitude less scanning to do than CO
switches (500 lines vs. 50,000), were, at least in principle, less
affected by this real-time problem.
Only
dial pulses could be detected in the trunk; tone signaling required
special analog-tone detectors to convert their information into
something a common control could use. Ultimately, chip sets for DTMF
detectors became inexpensive enough for some PBXs to build them into
incoming trunks directly, permitting both DID and tie-trunk
signaling to use DTMF rather than the much slower dial pulsing,
while retaining the ability to detect incoming address digits
without the assistance of some form of start-pulsing signal.
Stop-go. Until recently, there were a few
locations in the country where non-senderized SXS tandem switches
selected trunks to local common control COs. A senderized system
originating such a connection had no problems seizing a trunk to
tandem and outpulsing enough digits to select a path through the
tandem switch. However, when the tandem seized a trunk to the
terminating common control switch, the senderized originating switch
had to wait, right in the middle of the train of digits, for the
terminating switch to attach a register. To produce this wait, a
"stop" signal (off-hook) was returned by the terminating office via
the tandem as soon as the tandem seized its outgoing trunk, followed
by a "go" signal (on-hook) when the terminating switch was ready to
accept the remaining digits. Senderized switches were only required
to handle one stop-go signal in a train of digits.
We
may hope that SXS tandems without registers and senders are long
gone, but one never knows. There are advantages to designing
networks where the originating switch controls call set-up rather
than using the "hand-off" procedure typical in North America prior
to the coming of common channel signaling. In particular, the
originating switch, upon encountering an ATB condition, can back off
and retry via a different route or even a different long distance
carrier. However, something similar to stop-go might be needed at
each common control system.
Wink-start. Because of their innate
slowness, senderized systems have to control the flow of signaling
information to enable a register to be associated with an incoming
trunk after seizure of that trunk has been detected. Although dial
pulses can, if necessary, be scanned at the trunk, DTMF and MF
signals require digit detectors which are generally shared by a
large number of trunks and which, during the busy hour, may all be
in use when the next call arrives. In switches like AT&T's 1ESS with
2-wire metallic matrices, start-pulsing signals tend to be a
function of registers and senders rather than trunks to reduce
costs, but when electronic matrices are used, start dial signals,
which are usually variations of DC supervision, are activated and
detected in trunk circuits by the common control (or one of its
distributed front-end processors).
Wink-start is used on one-way trunks, and is most understandable
when a metallic matrix is used to associate the sender with the
outgoing end of a loop trunk and a register with the incoming end.
When a circuit is idle, the terminating end sends back an on-hook
signal (loop polarity normal). At the originating end, the system
makes the trunk busy to other calls and attaches a sender. When the
connection to the sender is complete, the tip and ring conductors
within the trunk circuit are extended directly, without series or
shunt impedances, to the sender which then closes the loop.
The
trunk circuit at the distant end responds by requesting a connection
to a register. There may be a delay in the busy hour, but,
ultimately, the register, like the sender, is connected directly to
the trunk conductors. It, however, returns a battery polarity
opposite to that returned from the trunk. The sender, seeing this
reversal, knows that the register is attached but not yet ready to
receive digits. After a nominal 200 milli-second "wink" interval
(140 mS minimum), the register reverses the line polarity from
off-hook to on-hook, and the sender knows, from this transition,
that it can out-pulse.
As
soon as transmission of the address signal is finished, the sender
and the register return control to the outgoing and incoming trunk
circuits, respectively. Note that the register, on-hook, is at the
same polarity as the trunk circuit. This polarity on the trunk will
not change until the called party answers to provide another
off-hook. At the originating end, the outgoing trunk circuit, having
been bypassed for the entire signaling operation, now monitors the
trunk to detect answer. Only then will it see off-hook polarity.
Delay-dial. As has been mentioned, two-way
trunks in senderized systems pose problems. First, the trunk has to
be made busy at both ends as nearly simultaneously as possible to
minimize double seizures and second, because there may be a
considerable delay in attaching registers and senders during the
busy hour, with no way of knowing which will be connected first,
great care must be taken with the start-pulsing signal. Third, it is
obviously desirable to be able to use the same senders for both wink
and delay-dial trunks, and, finally, because it is impossible to
eliminate double seizures, the "glare" problem (following) must be
considered.
The
delay-dial procedure is as follows. The originating office seizes
the trunk to be used, making it busy, and sends a "connect" signal
(off-hook) to the other end from the trunk circuit used outgoing (on
this particular call). The trunk circuit used incoming, at the
distant end, immediately acknowledges the connect signal by
returning off-hook and making itself busy to seizures by its own
switching system.
Next, both switching systems look for service circuits. The
originating switch finds an idle sender and attaches it, and the
terminating switch finds and attaches an idle register. There may be
several seconds difference in time, either way, in attachment during
the busy hour. However, the off-hook from the outgoing trunk circuit
is passed to the sender, and the off-hook polarity at the incoming
trunk circuit matches the register's polarity. Thus no change takes
place on the trunk itself at the time of attachment at either end.
If
the sender is attached first, the off-hook interval returned to it
will be lengthened by the nominal 200 mS wink interval from the
register. Then on-hook from the register will indicate that pulsing
can begin. If the register is attached first, the wink interval may
be completed before the sender is available. Thus the sender will
miss the transition from off-hook to on-hook. Electromechanical
systems were set up for the trunk circuit to remember the transition
and pass it on to the sender, or the sender simply started pulsing
upon observing an on-hook condition after an initial timed interval.
In computer controlled systems, where call memory keeps track of
these changes whether they take place in the trunk or service
circuits, the problem is greatly simplified. Note that operators
require delay dial; clearly, a wink is too short to be seen reliably
on a console lamp.
Double-seizure and Glare. That a two-way
circuit will be seized from both ends simultaneously is minimized by
careful design, but never reduced to zero. Because the connect
signal and the delay-dial signal are both "off-hook," there is no
way for one switch, after sending a connect signal forward, to know
for certain whether the off-hook it receives in response is another
connect signal (double seizure) or the start of a delay dial (proper
response to the original connect signal). To differentiate,
time-outs must be designed into the call processing. If the returned
off-hook comes back in less than the round-trip signaling time or
does not change back to on-hook before time-out takes place, the
sender will assume it has a double seizure and release. Because such
time-outs are not exact, one sender may time out first, and the
other will be left in possession of the trunk.
The
sender that gives up will either cause reorder tone to be returned
to its caller, or will have the common control come in and try to
set up the connection on another trunk. The abandoned trunk will
then go on-hook, see the connect signal still present from the other
end, and again return an off-hook delay dial signal while it goes to
find a register.
When
a sender detects the off-hook to on-hook transition that is supposed
to be a start pulsing signal, it must not outpulse immediately. For
dial-pulsing, a 70 mS interval is usually provided, and with MF
pulsing, an interval as long as 300 mS may be needed. If a glare
situation has occurred, premature outpulsing must not be started by
the winning sender when it detects the momentary on-hook as the
losing sender is dropped; if a new delay-dial signal related to
attaching a digit receiver appears quickly, it should continue to
fend off the sender until a register is attached and ready.
Dial tone and other approaches. It should
be noted that most of these problems could have been avoided if the
start-dialing signal were different from the connect signal. Use of
dial-tone from the receiver, for instance, would have greatly
simplified matters; it could easily have been detected by senders
and could have been used by humans as well. Although some PBXs
followed this procedure, most central office switching systems chose
not to. The coming of common channel signaling makes the point moot.
For
private tie-trunk networks among PBXs, the dial-tone approach can be
particularly useful. A number of smaller PBXs, many still in use,
only operate "cut-through" into the local CO (see Chapter 2). Thus a
connection reaching the PBX via a tie-trunk can not use the PBX's
digit receiver for other than the 9 to get "outside." That requires
a sender, after it accesses the PBX and gives it a 9, to be able to
recognize CO dial tone before sending the rest of the digits. To
save money, what is usually done instead is to insert a pause after
the 9 to give the PBX time to access a CO trunk and the CO to attach
a register. The problem with a time-out is that it may be too short
in a busy hour when the CO has calls in queue waiting for an
originating register; further, when light traffic minimizes the
delay for a register, post-dialing delay will always be increased by
almost the full duration of the time-out.
Most
of the wink start or delay dial manipulations grew up in crossbar
systems where appropriate trunk leads could be switched metallically
to registers or senders as needed. With only a single pair available
through its switching matrix, 1ESS could handle loop trunks but not
E&M in this way. Thus the system control had to scan the E lead and
manipulate the M lead directly, both for supervision and dial
pulsing. Further, MF signaling, discussed below, needing tone
senders and receivers connected via the matrix, had to leave
supervision with the E&M trunk circuit. The common control had no
difficulty dividing its attention between trunk and service circuits
as it processed the call, and this approach is, of course, used
today with most electronic and digital switching systems for loop as
well as E&M trunks._ _
Digital trunks with bit-robbed supervision, terminating directly on
digital switches, can interface analog circuits at the far end of
the T-span via a conventional channel bank with a variety of plug-in
units (loop out, loop in, E&M, FX line, etc.) as long as the digital
end associates the right software with each trunk. T-carrier,
however, offered some unused possibilities that are interesting to
contemplate.
In
the original version of T-carrier, where each frame included 7 bits
for voice and one bit for supervision each way in each channel, the
signaling bit could have been switched through a digital matrix from
incoming trunk to outgoing, providing an 8 kBps data channel
end-to-end paralleling the voice channel once the connection was set
up. Such a channel would have been available throughout the call to
let both intermediate and terminating switches monitor supervision
and signaling packets for feature applications and hang-up, and
could also have been used for customer data transmission.
Unfortunately, digital switches were not yet available and long-haul
microwave, in addition to blocking end-to-end digital connections,
required A/D and D/A conversions when cross-connected to T-carrier.
These conversions increased quantizing noise, a problem greatly
reduced by using the 8th bit for voice coding. This led to a
superframe of 12 frames, developed as part of the D2 channel bank
program (circa 1969) to allow five frames out of six to use all
eight bits for voice, with only one frame out of six to carry
supervision.
When
digital switching became available, a new problem appeared: too much
buffering delay would have been required to align the incoming
trunk's superframe (and, thus, its supervisory frames) with those of
the outgoing trunk, even if such buffering had been available. Thus
five times out of six, a connection through a digital switch
replaces a voice bit with a signaling bit, a process referred to as
"digit robbing." Clearly, digit robbing reduced the number of 8-bit
voice-only frames which D2 had gone to such efforts to provide.
By
the time optical fiber for digital long-haul trunks eliminated the
need for intermediate A/D conversions in most connections, and 7 bit
voice coding, inadvertently reappearing as a result of digit
robbing, would have been perfectly satisfactory, the improved D2
coding, now unnecessary, was firmly in place. Then common channel
signaling took over and today all 8 bits in each frame became
available for voice or data (ones density permitting). Although the
final result is highly desirable, it is interesting to speculate on
the uses which could have been made of an 8 kBps channel, for
signaling and data, associated with every telephone connection at no
additional cost. We have lost that particular opportunity, but ISDN
may someday turn out to be better.
Address signaling
On
trunk connections, it is necessary to pass forward the called number
to enable call routing; the calling number, often transmitted to a
central point for toll charging, can now be carried via CCIS to the
called party for calling number ID. Other information, related to
features and "traveling class marks," is also sent forward in many
instances. Thus "address" signaling will have a continually
broadening definition, particularly useful in an ISDN context.
Dial
pulsing. Dial pulsing on trunks is not very different from dial
pulsing on lines, except for the use of wink start rather than dial
tone; most of it is 10 pulses per second with 600 milli-second
interdigital timing as a memorial to vanished SXS offices. Central
offices have used dial pulsing to and from toll networks, and
tie-trunk networks have also been heavy users of dial pulsing.
Outside the U.S., dial pulsing to match SXS was widespread prior to
the coming of ISDN-compatible digital systems.
When
dial pulsing was used between senderized switching systems, the
interdigital time was sometimes cut to 300 mS and pulsing speeded up
to 20 pps. "Battery-ground" pulsing was also available on metallic
loop trunks (when they existed): instead of opening and closing the
loop, loop closures were made through a battery added during pulsing
so that 100 volts was available around the loop. The extra battery
increased the pulsing range of loop trunks, but not the range of
supervision; when the called party answered, the reversed battery in
the terminating office would buck the battery in the originating
office, causing the trunk to release.
As
mentioned in Chapter 3, dial pulsing on T-carrier trunks has very
little distortion internally. By foregoing the opportunity to look
like a copper trunk with loop or E&M supervision between digital
COs, DP could have run far faster than 20 pps. When SF supervision
was used on trunks, ac (tone) dial pulsing was another possibility
left unutilized. SF sets without pulse-correcting circuitry (common
on trunks using MF) were appreciably less expensive; it would have
made sense to let tones representing dial pulses pass through the SF
sets from sender to register to improve signaling and reduce costs;
but as with high speed dial pulsing, CCIS has made the issue
irrelevant.
Multi-frequency (MF) pulsing. MF is used
for trunk signaling, and should not be confused with DTMF which is
used primarily for signaling from customer to switch. It is faster
than conventional dial pulsing, the rate being seven digits per
second and limited primarily by certain echo considerations on long
trunks. A somewhat less strict limitation is imposed by the need for
the signal to persist long enough to facilitate differentiating a
valid signal from a voice simulation. Although MF represented an
early (1943) improvement in high speed data transmission, its rate,
just under 28 bits per second, shows how far we have come.
Senders for MF pulsing, particularly at operator positions, are very
inexpensive; ac tones are simply gated into the trunk in question.
Receivers are much more costly, requiring rather elaborate
electronic circuitry to detect valid digits while rejecting
"talk-offs." Holding times are on the order of 3 seconds, so few
receivers are needed.
MF
uses six frequencies, spaced 200 Hz apart between 700 and 1700 Hz.
For a valid digit or other signal, exactly two of the six
frequencies are used; 15 combinations are possible. Two-out-of-six
symmetric checks on relay contacts were easily implemented to prove
validity of received digits. Checking in stored program systems is
even easier, although solid state hardware does not deal elegantly
with symmetric functions.
The
principal problem with MF is the need for the receiver to deal with
both signaling tones simultaneously; in DTMF, the high-band and
low-band tones are separated before detection is attempted. An MF
signal, passing through a compandor (automatic level adjusting
circuit) in an analog carrier system or in the MF receiver itself,
tends to produce harmonics while the level is changing. These
harmonics, plus sums and differences, can make trouble because some
of them are signaling frequencies. The "2A-B" modulation products
are particularly troublesome because 2 x 900 Hz - 700 Hz = 1100 Hz,
and the receiver may see 700, 900 and 1100 Hz when only the first
two were sent, causing the two-out-of-six check to fail. DTMF took
advantage of this knowledge and used carefully arranged
geometrically spaced frequencies. Unfortunately, DTMF was not
suitable for use on trunks because it would have provided a large
number of users with a "Blue Box," a device to obtain free toll
calls fraudulently.
In
MF, an additive code is used where the frequencies, from low to
high, are assigned the weights 0, 1, 2, 4, and 7. The digit 1 is 0 +
1, 700 and 900 Hz; 6 is 2 + 4, or 1100 and 1300 Hz, etc. The only
exception is 0, which is represented by 4 + 7. Two additional
signals are used: KP and ST. KP (for key pulse) is an unlocking
code, consisting of 1100 and 1700 Hz. Receipt of KP activates the
other frequency detectors in the receiver. ST (for start) is an
end-of-dialing digit and consists of 1500 and 1700 Hz; this tells
the receiver it has all the digits it is going to get and the
switching system can start to set up the connection. When ST was
sent from early operator positions, it also disconnected the
operator's key set from the call.
Certain frequency pairs are used in connection with operator
positions, and others with CCITT Signaling System No. 5 (SS5). When
coin control is exercised from a distant switchboard position, 700 +
1100 Hz indicate "collect" while 1100 + 1700 indicate "return."
Tones are normally transmitted at -6 dBm per tone, measured at the
zero transmission level point (TLP, defined later in this chapter).
Frequency accuracy of ±1.5% of nominal is expected. Receivers are
designed to work at signals as small as -22 dBm per frequency.
In
the United States, the sender, upon receipt of wink start, outpulsed
the required digits at the rate of 7 per second and then turned the
trunk over to the caller. In Europe, where "Compelled MF" was used,
the sender transmitted each digit until it received an
acknowledgment from the digit receiver. This eliminated the need for
timing digit duration at the sender which sometimes speeded up the
signaling process. However, on a noisy channel, compelled MF took
longer and could set up connections on channels unsuitable for
conversation.
CCIS. Common Channel Interoffice
Signaling* or CCIS, similar to CCITT Signaling System No. 6 (SS6),
was installed by AT&T in connection with its 4ESS toll switches
starting in 1976. Its purpose was to provide a packet signaling
network, independent of the trunks used by customers, over which the
common controls of switching systems could communicate directly at
speeds much higher than the 30 to 40 bits per second of MF and DTMF.
The CCITT standardized the speed for SS6 at 2400 Bps, suitable for
use on analog voice circuits dominant in the age of microwave, but
AT&T's digital D2 channel banks were planned to provide a 4000 Bps
signaling channel to accommodate CCIS when T Carrier became
practical for long-haul trunks. With CCIS, out-of-band bit-robbed
supervision was no longer needed; thus the "alternate framing bits"
which D2 used to identify those frames in a superframe where
supervision would have been, became available for use as a "high"
speed data channel that cost nothing and didn't even need a modem.
[* Footnote. It should be noted that CCIS was originally called
CCS until confusion with a well known unit of telephone traffic
was pointed out. Today, some writers refer to Signaling System 7
as CCS7, while others use SS7, the practice followed here. It is
not evident which usage will win out, but there are already too
many acronyms with two or more meanings.]
The
advantages of CCIS, in addition to its high speed and ability to
eliminate per-trunk supervision equipment, signaling transmitters
and receivers, and the control complexity to manipulate all these
devices, are several. In the first place, talk-offs, the bane of SF,
vanished, and hackers with blue boxes (which depended on SF and MF
tones) found themselves locked out. Then, CCIS could send in both
directions simultaneously, and communication could take place at any
time, not just when an MF sender and register were connected to a
trunk. Finally, CCIS could send all sorts of network management
messages in addition to the messages needed in conventional and
advanced call set-up.
Where there was a very large trunk group between two switches, it
was possible to have a direct CCIS channel called a "Fully
Associated Link" to serve that trunk group alone; however, the more
general practice was to establish duplicate packet switches called
Signal Transfer Points (STPs), physically separated for reliability,
in each area, and provide a data network among these pairs of STPs.
Every switch was to home on both STPs in its area, and each STP had
connections to all other STPs in remote areas, as well as a special
connection to its local mate. As a result, either STP could carry on
alone if the other failed, and if an STP lost one or more of its
links to the rest of the network, it could access those of its mate
and keep on working.
The
idea was to have every local switch, as well as those in the toll
hierarchy, use the CCIS network for inter-switch call set-up.
However, the coming of competitive long-distance networks and the
divestiture of the Bell Operating Companies from AT&T made this plan
impossible. To the newly liberated "Baby Bells," the CCIS network
was perceived to give AT&T a competitive advantage over other
networks still clinging to SF, MF and DTMF. However, AT&T completed
the implementation of CCIS within its toll network, reducing call
set-up times between what had formerly been Class 4 toll offices.
With
CCIS, intelligence remained incorporated in the toll network
switches. Upon receiving call set-up information from a local CO,
the toll switch would select an appropriate outgoing trunk and then
use CCIS to pass signaling information to the switching system at
the far end where the process was repeated, just as had been done
with DP or MF signaling.
When
the customer dialed an 800 number, this procedure was insufficient.
Before routing could commence, the switch had to first obtain a
"real" telephone number with an area and office code, as described
in Chapter 2. Putting the required translations in all toll switches
and keeping them up to date would have been an administrative
nightmare. CCIS allowed a few centralized data bases to be
interrogated as easily as a switch's own memory.
When
the standard for Signaling System 7 (SS7) was released by CCITT, it
became the approved signaling method for local and long distance
companies all over the world. SS7, operating at 64 kBps, needs an
entire T-Carrier DS0 channel but can support analog trunks as well
as digital. Like any form of common channel signaling, it need not
follow the route used by its trunks. Thus a direct trunk group
between switches A and B needs no signaling of its own; SS7, running
from Switch A to an STP and then back to Switch B can handle the
whole job, and signaling to Switch C for alternate routing as well.
The
next step, now being implemented under names such as AT&T's
"Advanced Intelligent Network," is to expand the data bases required
for such things as 800 numbers to include routing of all interswitch
calls. Trunk busy-idle status is kept updated at these centralized
points, along with routing algorithms. SS7 is used to forward the
destination code to the data base which then selects the required
trunks; using SS7, the centralized data base then issues orders to
each switch to make the appropriate connection between an incoming
and an outgoing trunk. Because of the Balkanization of the telephone
industry, the number of inter-exchange carriers with various
vintages of equipment, and the independence of local exchange
carriers, it will take some time before the full advantages of
Intelligent Networks can be realized. However, SS7 will play a major
role in advancing the possibilities, assuming local and long
distance carriers all use it the same way.
SS7
and Intelligent Networks probably have more to offer local telephone
companies than long distance carriers. Customer features such as
those marketed under the heading "Citywide Centrex" use several
switching systems to serve scattered customers as though they were
at a single location. This requires close coordination among system
controls and data bases, both for using and administering the many
features, and may well turn out to be more complex than handling
call routing.
With
only a few data bases for a large number of local CO switches, new
features can be tested and then implemented and administered easily
and economically. Centralized Service Circuit Nodes, mentioned in
Chapter 3, can provide the complex functions these new features need
(voice processing, text to speech, etc.) without modifying large
numbers of local switches. Equipment manufacturers are designing
such hardware, along with programming tools to allow telephone
companies and business customers to create their own features.
Limitations on advanced routing and features are not a result of SS7
capabilities, but of legal restrictions placed on the geography in
which such approaches can be offered.
Transmission
Analog transmission has been extensively studied in terms of
transmission lines and carrier systems, and also in terms of human
factors requirements. The transmission properties of analog
switching systems have received somewhat less attention. Even the
largest switches seldom have more than 1000 feet of cable, MDF to
MDF; thus the problem may appear to be trivial. Unfortunately, it is
not. There are many types of facilities to be interconnected,
supervision and signaling must be accessible, and test access must
be provided. Thus transmission and switching have to be designed to
work together.
The
telephone industry grew up with these requirements and met them well
in terms of analog transmission. However, the coming of digital
transmission in 1962, followed by digital switching in the same
format in 1975, changed the whole situation beyond recognition.
Unfortunately, many of the solutions to analog transmission problems
had become so much a way of life for the people charged with
successful operation of the telephone system that it is very
difficult for them to accept the full advantages and simplifications
made possible by digital technology. As a result, it will first be
necessary to review analog transmission before we consider the
impact of digital transmission and switching.
The
transmission parameters that have been important in analog switching
systems are insertion loss, return loss, longitudinal balance, and
crosstalk and noise. Harmonic distortion and overloading are also
factors to which, with the coming of modems for data transmission,
must be added envelope and absolute delay, and the effects of
quantizing noise due to conversions between digital transmission and
analog switching. All of these factors have been specified in terms
of traditional metallic matrices with supervisory bridges (circuits
from tip to ring leading to supervisory sensors in such a way that
their impact on transmission is minimized).
When
both series and shunt losses are small, the switching system tends
to become invisible. However, when electronic space division
crosspoints are introduced, particularly in transmission paths that
are unbalanced to ground, the picture changes markedly. Electronic
crosspoints require biasing circuits, have relatively high series
impedance compared to relay contacts in the operated state, and
present wiring difficulties when a matrix must be constructed in two
or more cabinets.
Digital switching, on the other hand, simply manipulates code groups
that lock in both amplitude and phase. PCM tends to be relatively
insensitive to noise and crosstalk and the exact shape of the signal
need not be preserved; all the receive end has to do is detect the
difference between zeros and ones. Finally, digital switching must
be 4-wire; this is a great advantage when dealing with trunks, but
is sometimes a problem when dealing with 2-wire analog lines to
customers.
The
down-side impact of digital switching lies mainly in the flat delay
introduced by time-slot interchangers and conversions between serial
and parallel operation. This delay can be a major factor in human
perception of echo performance.
Insertion loss, TLP and VNL. The ratio of
voice frequency power levels is measured in decibels (abbreviated
dB), or tenths of a bel, named after the inventor of the telephone.
The bel, which says the power of a signal is ten times larger that
the power of the signal to which it is being compared, turned out to
be too large for practical work. The decibel, however, is just about
at the threshold of difference in loudness that the human ear can
detect, and is thus a much more useful unit.
A
signal 3 dB louder than another has twice as much power, while one 6
dB louder has 4 times as much. Obviously, if it is 10 dB (or 1 Bel)
louder, it will have 10 times as much power and if 20 dB louder, it
will have 100 times the power. The ear does not respond to sound
power in a linear way; rather, each dB of increase is (more or less)
perceived as equal. Thus use of decibels, which go up much more
slowly than power ratios, makes more subjective sense.
Because a measurement in dB always represents a ratio, one cannot
talk about a 10 or 100 dB signal all by itself. In much telephone
work, 1 milliwatt (.95 volts across a 900 ohm resistor) is used as a
reference, and one can speak of measurements relative to a milliwatt
in terms of dBm. Thus 10 dBm is a signal 10 dB louder than a
milliwatt, and a signal at -8 dBm is 8 dB lower in level than a
milliwatt. With this background, we can approach the concept of the
Transmission Level Point or TLP.
The
outgoing switch of a local analog switching system is generally
considered to be at 0 TLP. The incoming side of an analog carrier
circuit arrives at +7 TLP, and the outgoing side is at -16 TLP. What
this means is a 1000 Hz tone applied at the 0 TLP at 0 dBm must
arrive at the outgoing side of the carrier at -16 dBm. Further, at
the distant end of the carrier trunk, the tone will be +7 dBm. The
point is, a standard test tone can take on a variety of levels,
depending on where it is measured. If all measurements are referred
to 0 TLP, appropriate comparisons can be made.
The
23 dB amplification or gain between the signal level entering a
carrier channel and leaving it at the far end comes from the early
days, before transistors, when carrier systems using vacuum tubes
were a good and convenient source of gain to compensate for
cross-office losses. A carrier system must have amplifiers anyhow,
so it was decided to use those amplifiers effectively. Frequently
the carrier system would end in a toll building, but the trunk would
terminate on a switch in a local telephone building some blocks or
even miles away. Thus the gain was needed. Further, going from a
four-wire circuit to a two-wire circuit required a hybrid (see
Chapter 1). The 3 dB loss for each direction of transmission through
the hybrid also had to be made up.
The
outgoing switch of a toll office is considered to be at the -2 TLP.
This comes from Via Net Loss theory which says that any toll
connection from local office to local office should contain VNL+4dB.
VNL or Via Net Loss is the minimum loss in any given circuit
required to minimize the subjective effects of echo; the 4 dB is
split evenly between the two toll-connecting trunks from the local
offices to the toll network. Actually, the toll-connecting trunks
are supposed to contain 2 dB+VNL, but because they are usually very
short, VNL is practically zero.
VNL
is a function of distance and the type of facility used. For all
carrier systems, the VNL factor is 0.0015 dB per mile. Thus the VNL
for a 600 mile intertoll trunk (or an intertandem tie-trunk) would
be 600 x 0.0015 or 0.9 dB. This loss is inserted at the output of
the carrier system; a measurement at the -2 TLP in a toll office for
a 600 mile trunk would be 0.9 dB below -2 dBm for a milliwatt tone
(0 dBm) inserted at the 0 TLP in the distant office, or, as would be
more likely, a -2 dBm tone inserted at the -2 TLP. Note that VNL is
inserted at the "listen" end of each of the two sides of the 4-wire
carrier trunk.
Trunk transmission is measured outgoing-switch to outgoing-switch.
Thus, for each direction, one switching matrix is included: the one
at the "listen" end for the given measurement. Attenuation is
inserted at the output of the carrier system to make the total loss,
including matrix, office wiring, etc., come out at the right value.
The
outgoing-switch to outgoing-switch arrangement comes from many years
ago when SXS switches, manual toll switchboards, and test desks all
competed for the same trunks; that is, the input of the trunk
circuit was multipled past any switch or jack that might require
access to it. Tests were performed from the test desk by seizing the
trunk at the same point the outgoing switch would seize it, dialing
a code or whatever was required, and reaching another test desk at
the distant end. This was, quite properly, outgoing switch to
outgoing switch; note, in particular, that the trunk circuit at the
originating end was included.
In
more modern analog systems, operators, test desks, etc., all gained
access to trunks via the switching matrix, and the trunk circuit,
physically part of the switch rather than the transmission facility,
was wired directly to its matrix appearance. Because the interface
between the switch and the trunk circuit was not readily accessible,
the outgoing-switch to outgoing-switch standard had little
relevance. With digital switching, where the switch and the
transmission system become a single unit, individual trunk circuits
do not exist, the signal reference level is replaced by digital
code, and the whole concept changes from irrelevant to meaningless.
With
regard to insertion loss: for a line-to-trunk analog connection,
insertion loss from the MDF through the matrix and the trunk circuit
is usually specified as 0.5 dB nominal at 1000 Hz, with a maximum of
0.8 dB. It can be twice as great, line to line. Obviously, the less
loss there is in the switch, the more loss can be allocated to a
line for a given performance. With a metallic matrix, the 0.5 dB
figure really does little more than permit maintenance people to
pick up shorted turns in coils or open capacitors. When an
electronic analog matrix consisting of two isolation transformers
plus electronic cross points is used, appreciably more loss will be
encountered. Its effects can be minimized by specifying shorter
lines, an approach considered too costly to contemplate, or adding
amplification, which may produce instability.
When
a digital local CO switch is used to interconnect conventional
2-wire lines terminating in 2500 type telephone sets, the hybrids at
the interfaces with 4-wire digital paths through the matrix require
gain if analog standards are to be maintained. This is easily
provided by op-amps performing the hybrid function (discussed
below), with digital pads (which can introduce either positive or
negative loss), or by setting the analog level at the codec output.
In all instances, the gain is easily produced, but at the expense of
stability. Instability is most troublesome on short loops; station
carrier, whether analog or digital, suffers from this same problem
for the same reason.
The
bandwidth of a metallic matrix is quite flat from dc to about 1 mHz
if care is taken with the wiring. The limiting factors are usually
in the trunk circuit where battery is applied to the customer's line
and the line and trunk are supervised. The calling line and called
line, or the line and trunk, must be separated at dc so that each
can be supervised separately, and a short, low-resistance line won't
drain current away from a long, high-resistance facility. This
implies repeat coil (transformer) coupling between the two halves of
the circuit, or capacitor coupling, as suggested in Fig. 3. Thus the
low frequencies will tend to roll off, due either to the increase in
loss with decreasing frequency at the capacitors, or the increasing
shunting effect of the relay coils or transformer. In any event, at
200 Hz, the line-to-trunk-loss can be 0.5 dB greater than at 1000
Hz, and it can roll off faster at lower frequencies. At 3200 Hz, it
should be within 0.3 dB of the 1000 Hz value. Again, these losses
can be doubled for line-to-line connections.
In
the early 1970s, the bandwidth of space division matrices in systems
such as AT&T's 1ESS and several PBXs in the 800 series, independent
of the format of the signal to be switched, led to one of the many
attempts at Picturephone as well as the hope of switching broadband
data. Because of their smaller physical dimensions, space division
systems using electronic crosspoints can handle an even wider
bandwidth, and have been used for switching both analog and digital
signals consisting of many individual channels multiplexed together.
Digital systems, by contrast, require information to be mapped into
their particular bit-stream. This can be an advantage in some
instances and a disadvantage in others. A digital signal bypassing
the PCM codec can travel at 64 kBps for the price of a regular phone
call; this is faster than is practical on a current dial-up voice
channel using modems, and is the hope of ISDN. However, a
Picturephone signal at 1 mHz, circa 1970, required over 6 mBps (the
equivalent of 96 voice channels in T-carrier) for coding, and
commercially acceptable TV required a lot more. Today, of course,
video compression techniques are well advanced and video
teleconferencing systems using only two DS0 channels are common.
Even so, space division switching for broadband digital signals is
an option waiting to be exploited.
In
an analog switch, whether PBX, local, tandem or toll, insertion loss
for voice should be made as small as possible, consistent with the
economics of loss in the overall system. Uniformity of loss,
independent of the several possible paths which can be chosen
through the matrix, is perhaps even more important. An analog switch
should not limit frequency response to less than the bandwidth
normally used by carrier systems and, consistent with the technology
employed, should handle the widest possible bandwidth to allow for
future services. Although an analog signal must be bandwidth-limited
to be encoded into a digital format, once encoded, neither loss nor
variations therein are a factor, either within the switch or in
transmission paths between switches. There is no need to consider
TLP, VNL or other cherished concepts which no longer apply.
Return loss. Return loss is a measure of
how well impedances are matched. If a signal source's impedance is
exactly equal to the load being driven, the load will absorb the
maximum power and will "reflect" none. Thus the return loss is
infinite, because no energy at the destination is lost by being
returned to the source. If the load impedance does not match the
source impedance, less than maximum transfer will take place and the
energy not absorbed will be reflected back. This, as has been
emphasized, causes echoes, a major transmission impairment.
When
a four-wire signal entered the switching area at +7 TLP and left at
-16, the actual loss in the echo path from incoming to outgoing
included incoming and VNL level adjusting pads, approximately 3 dB
loss through the hybrid, the return loss in matching the connection
to a 2-wire line, another 3 dB loss through the hybrid, and further
level adjusting pads to the -16 TLP. Toll offices that switched on a
two-wire basis were required to provide at least 27 dB of return
loss, but this was not difficult because such trunk-to-trunk
connections were much less variable than trunk-to-line connections.
At a local CO, the average return loss of the local lines was
supposed to be at least 11 dB, with a standard deviation of 3 dB,
between 500 and 2500 Hz. From 250 to 3200 Hz, it could be at least 6
dB on the average with a 2 dB standard deviation. These measurements
were made against a 900 ohm resistor in series with a 2.14 µFd
capacitor as a standard. Even with two-wire toll switching, it is
evident that most echo came at the line-to-trunk interfaces.
The
return loss problem became a little more difficult when PBX
connections were considered. PBXs often have tie-trunks to other
PBXs at great distances; these tie-trunks usually allow callers to
make off-net calls into the public network via the PBXs at each end
in addition to calls to PBX extensions. Tie-trunk networks, like the
public network, used VNL concepts to optimize one connection,
end-to-end. When a connection via a private network was connected
back to back with a similar connection through a public network,
there was a possibility of the total connection having VNL + 8 dB
loss. This was more than twice as much as a single direct
connection, and added stability problems related to impedance
matching at the 2-wire interface.
In
the early days of long distance competition, where the then
"specialized" common carriers accessed the public network as though
they were local lines rather than trunks, it was not unusual for
calls to experience VNL + 12 dB of loss (three multi-trunk
connections back to back). Regulatory authorities, disregarding the
decades of human factors work that had gone into VNL design, felt
that "freedom of the market place" should determine telephone
transmission quality. Later, local telephone companies were
required, at enormous cost, to provide trunk-side "equal access" to
all long distance carriers to reduce this loss and provide a "level
playing field."
Analog service circuits used in 2-wire switching systems in
connection with trunks (senders, digit receivers, tone circuits,
recorded announcement circuits, etc.) were designed to have as high
a return-loss as possible; this was not difficult because the only
variable involved was the length of the path through the switching
matrix.
Longitudinal balance. Transmission
engineers often use the word "balance" to refer to return loss, as
in "a well-balanced office" where there is little echo. Longitudinal
balance is something quite different. Residential telephone lines
often share poles with power lines and in business districts, some
modern building codes allow steel building frames to serve as a
partial ground return for unbalanced three-phase electrical power
systems. Both situations subject telephone lines to inductive pickup
at 60 Hz and its harmonics. Open circuit measurements from one
telephone wire to ground will sometimes show signals as large as 100
volts peak, and it is not unusual to find several milliamperes of
current flowing through a 1000 ohm terminating resistor to ground.
The
only way to make the effects of such signals negligible is to
balance both sides of a circuit to ground so that both wires "go up
and down together," producing no net voltage between tip and ring;
if the longitudinal signals on each side of the pair are equal,
there will be no "metallic" signal, audible to the user.
Cable pairs tend to be quite well balanced but circuitry in the
switching system can cause unbalances that produce noise. To see if
a circuit is properly balanced, a measurement can be made as
suggested in Fig. 4. The longitudinal signal, introduced equally
into both sides of the circuit through the repeat coil, must be at
least 55 dB greater than the resulting "metallic" signal measured on
the volt-meter at 1000 Hz, and 53 dB greater at 3000 Hz.
Fig.
5 shows how various forms of switching system impedance to ground
can affect longitudinal signals. If impedance to ground is very high
or very low compared with the impedance of the line, exact balance
may not be necessary. A very high resistance to ground will not
change the voltage on either conductor very much, and a very low
resistance will short out the longitudinal voltage to such an extent
that the difference on each wire is negligible. However, very high
resistances to ground, producing a "floating" circuit, can attract a
static charge and be noisy. A very low impedance to ground can, of
course, short out the desired signal as well as the undesired
longitudinal; thus a "longitudinal drain" is usually an inductor
with a high impedance to voice frequency currents but negligible
impedance to dc and the first few harmonics of 60 Hz. Further,
longitudinals go through a drain coil in opposite directions, each
canceling the inductive effect of the other, while the desired
signals are "metallic," go through both halves of the coil in the
same direction, and get the full inductive effect of both windings.
Thus the drain coil is a high impedance from tip to ring for
metallic currents, but very nearly a short circuit to ground for
longitudinals.
As
we have seen, most switching system impedances between tip and ring
are supervisory sensors, sometimes inductors themselves, and
sometimes in series with inductors or located at the center-tap of
repeat coils. These sensors tend to have intermediate resistances,
and require careful attention to balance, particularly at low
frequencies.
A
repeat coil will block the transmission of longitudinal signals by
transformer action; if there is no current through the primary
winding because the voltage on each side is equal, there will be no
voltage induced in the secondary. However, there is capacitance
between the primary and secondary windings, and that capacitance can
couple longitudinals fairly easily. When a repeat coil is used to
convert a balanced line to an unbalanced electronic circuit, this
capacitive coupling can be particularly troublesome. In such cases,
a grounded electrostatic shield or some other means of reducing the
effect of the capacitance must be used.
The end of the iron age. As can be seen in
Figures 3, 4 and 5, the inductor, a coil with one or more windings,
often used as a transformer (repeat coil), is a thoroughly useful
electrical component. It is suitable for coupling ac signals while
blocking dc, going from balanced to unbalanced transmission,
changing ac voltage or current levels, providing the hybrid
function, subtracting the effects of capacitance, etc. On top of all
this, it is a completely passive device, requiring no power supply.
Another factor of considerable interest is that inductors can be
arranged to convert electrical energy to mechanical, becoming the
basis for relays, motors, etc. Thus relays were particularly useful
in supervision bridges where the inductance of their windings kept
them from shorting out audio signals, the resistance of their
windings could limit dc loop current, and their contacts could
provide a supervisory signal in a circuit completely isolated from
tip and ring.
Unfortunately, inductors used at audio frequencies tend to be fairly
large, consisting of many turns of copper wire wound around an iron
core. Even though the iron core allows a high value of inductance to
be obtained with a modest number of turns, inductors are expensive
to make and bulky to use; as a result, they are generally considered
prime candidates for replacement by modern electronic devices on
printed circuit boards. In particular, analog line and trunk
circuits terminating in digital switching systems make extensive use
of fairly complex approaches based on operational amplifiers or "op
amps." Op amps can be arranged to provide the balanced to unbalanced
and two-wire to four-wire (hybrid) conversions for ac signals and to
detect dc supervisory signals, using circuitry more suitable to
circuit board mounting. Because the quantity of line circuits
produced is so large, chip costs are low and their complexity trades
off easily against higher circuit densities and other advantages.
Most
such analog chips designed for customer line interfaces depend on
resistors to terminate lines and trunks and to act as one of several
means for limiting supervisory current. These resistors now become a
major factor in controlling both longitudinal and return loss
balance and, unlike inductors, cannot provide a low longitudinal
drain or a high bridging impedance. As a result, they are typically
specified at 1% accuracy, and balanced to 0.1%. It will be
interesting to see if these resistors and their supporting
electronics can maintain both kinds of balance over a twenty year
period as inductors did in the past.
Although great strides have been made in ridding telephone switching
systems of devices with iron cores, there is at least one place
where inductors still seem to be the best solution. Analog PBX trunk
circuits, both loop- and ground-start, "float" in the center of the
loop (that is, tip and ring are isolated from ground at the PBX end,
while battery and ground connections are made only at the CO which
may be several miles away). This requires the tip-ring connection to
be made through a transformer winding or an inductor, suggesting
something like the CS relay of Fig. 1 which also provides an
isolated supervisory contact. Similarly, a relay contact is a good
way to provide loop closure, although opto-isolators are also used.
Doubtless electronic devices could ultimately be devised to replace
the CS relay, but the PRI will probably eliminate the problem before
we get an iron-free solution.
Other factors. When many pairs of wires
are in close proximity, as in cable racks and at cross-connect
frames, any given pair is likely to pick up speech from others
(crosstalk) and noise. Many components have non-linearities that
cause distortion of the transmitted signals and add frequency
components not in the original signal. Circuitry can cause flat
transmission delays as well as delays that are a function of
frequency, particularly harmful to data transmission between modems.
Modems also impose more stringent requirements than voice on
quantizing noise at analog/digital conversions.
Many
specifications have been used to guide designers, but they do not
always provide the desired results. For instance, crosstalk loss
based on electromechanical systems was specified as 75 dB minimum
(that is, the crosstalk power picked up in a pair being tested had
to be less than one 31 millionth that of the disturbing signal in an
adjacent pair). Although this specification worked well in its
intended area where masking noise was generated by such events as
relays operating and releasing, it left the crosstalk clearly
audible in early electronic systems which, having no relays, did not
produce masking noise.
The impact of digital technology. The
abrupt change from analog microwave to digital transmission on
optical fiber brought about a number of important changes in long
distance telephony in the years between 1980 and 1990. In the first
place, optical fiber is not affected by electrical noise and, as a
result, can ignore many of the problems that have always beset
traditional transmission systems. Second, optical fiber has a huge
bandwidth, particularly when compared with microwave, and by
installing more glass fibers at small incremental cost, available
bandwidth can be multiplied. This encourages the use of digital
modulation which, itself, even when not used on optical fiber,
eliminates noise, distortion, phase shift and other problems of
analog transmission. Indeed, a major advantage of optical fiber is
the way it allows PCM to be used on long-haul trunks.
Third, optical fiber does not conduct electricity; this is an
advantage in that it does not attract lightning and magnetic
induction, but a disadvantage for intermediate or terminal equipment
traditionally powered via the transmission facility. Finally,
digital signals can be multiplexed far less expensively than analog
signals, and the high speed bit streams so obtained match well the
capabilities of optical fiber; on the other hand, the high bit rate
of optical fiber can be shared in a number of ways other than
straight digital time division multiplexing to provide a wide
variety of different services.
To
take full advantage of fiber's capabilities, a very close
examination of past design is required. It is not enough to simply
be able to ignore noise, distortion, phase shift, echo, etc. What is
needed is to redesign the worldwide telephone network so that
obsolete specifications will not prevent new networks from reaching
their full potential.
The
entirety of VNL theory, for instance, is based on obtaining enough
amplification for two parties to hear each other over a long
distance connection while, at the same time, minimizing the effects
of echo where 4-wire trunks connect to 2-wire station lines. It
should be evident that if the connection were 4-wire end to end, the
echo problem (except for acoustic coupling through the air between
the handset's receiver and transmitter) would be eliminated rather
than solved, and transmission design would have different
priorities.
All
carrier systems, whether on microwave or cable, whether analog or
digital, have to be 4-wire. Digital switches, which appear to be the
only option available when 100,000 or more trunks must be served by
one machine, must also be 4-wire. Today, almost all tandem and toll
switches are 4-wire, and local central offices using digital
switching have moved the 4-wire to 2-wire interface to the customer
line itself. When 4-wire channels are extended to 4-wire telephones,
a promise of ISDN (and a reality in most PBXs since about 1980), VNL
will have to be relegated to the museum along with the SXS switch
and the vacuum tube.
In
any rational world, digital telephone transmission should, in the
opinion of the author, be based on the following rules:
1. Once a signal is encoded into a digital format, that format
must not be changed at any point, but must be delivered intact
to the instrument at the far end of the connection.
2. The TLP should be replaced by a comparison signal based on a
digitally generated tone where frequency and amplitude are
locked in by digital techniques.
3. All connections between 2-wire analog lines should contain 6
dB of loss in each direction, 3 dB at each end (the equivalent
of loss in transformer-based hybrids). Measurements should be
made from MDF to MDF (the 2-wire side of the hybrids).
The
advantages of these relatively simple rules are several. First, rule
1 allows any channel to be used for voice, data, digital video,
etc., and permits switching from one to another at any point in a
call. Rule 2 makes level definition easy, simplifies measurement,
and takes advantage of digital capabilities. Rule 3, however, is
more subtle in that its intent is to allow existing analog switches,
until they are replaced, to access without modification digital
inter-switch networks.
It
should be noted that the 3 dB loss at each end of a connection to
analog lines is slightly larger than the 2 dB loss in toll
connecting trunks in the VNL network plus the loss through a 2-wire
analog matrix; it is almost what we would expect if we moved the
point of measurement from an inaccessible or non-existent location
(outgoing switch) to one readily available. Thus analog switches
could connect lines from all kinds of customer equipment to local
and toll networks by accessing T-carrier channel banks just as they
do today. No loss changes would be required and the system would
work just like VNL from an analog switch's point of view.
When
an analog CO such as a 1ESS is changed out to a digital switch such
as a 5ESS or DMS-100, the T-carrier channel bank would be removed
and the digital trunks connected directly and without loss. With
ISDN's BRI and PRI to interface small and large business customers,
and the BRI for residences, the 4-wire digital path can be extended
to the customer's premises, maintaining 4-wire integrity all the
way. Where analog phones are still used, the 3 dB loss in both the
talk and listen sides would produce about the same levels as the
analog toll network, and help insure stability on intra-switch
connections between two analog lines, something much harder with a 0
loss requirement.
Note
that under rule 1, all inter-switch trunks would run at 0 loss, so
that both local and long distance connections would have the same
level. A major objection to this approach is that the telephone
industry expects inexpensive local calls to continue to run at a
level 6 dB higher than premium long distance calls, as they have
since about 1950. Considering how the ratio of long distance to
local calls has increased since the coming of DDD, this is curious,
indeed. Another objection is that reducing the analog signal by 3 dB
before encoding could reduce the signal to noise ratio.
If
and when analog telephones built to 1950s specifications can be
phased out and digital phones, either ISDN-compatible or proprietary
behind PBXs, replace them, the problem will fade away. With the
codec in the telephone set rather than on the line card, there is
little exposure to noise. Further, most digital phones allow the
incoming sound to be adjusted to meet the listener's needs, making a
"standard" level as meaningful as TLP. With voice level adjusted by
the listener after digital-to-analog conversion, signals such as
data or video mapped directly into the bit stream would not be
endangered by digital pads or other network adjustments.
The
principal network adjustment remaining would be µ-Law to A-Law
conversion on international calls. This is presently carried out at
international gateway switches between µ-Law countries and the rest
of the world, and is one of the main impediments to world-wide ISDN.
Note that µ-Law and A-Law apply ONLY to speech, and non-voice needs
are presumably the main growth area in telephone communication. The
solution would seem to be a more complex codec that can decode both
µ-Law and A-Law. Normally, it would use the local standard but on
international calls, a traveling class-mark would switch it to the
other mode for that call only. Actually, codecs which can handle
both µ-law and A-Law are available, but selection on a per-call
basis is not implemented.
To
keep ISDN telephones simple, traveling class-marks will apparently
not be used for level adjustment, digital coding will be done at 0
TLP, and a compromise value of 3 dB loss will be built into the
"listen" side of ISDN telephones. Traveling class marks will be
available to identify a call as voice or data. There seem to be no
plans for alternate voice/data on a given connection, and
µ-Law/A-Law conversion at international gateways will take place
only on voice calls. Further, during the transition era, while
analog CO switches are being phased out, some interesting problems
in level adjustment will be encountered.
Obviously, CCIS (SS7) solves most of the signaling and
administration problems we have discussed at length: glare, called
party answer and hang-up, calling number identification, locally
returned busy, system management, verification of 800 numbers, etc.
It could also help solve transmission problems with traveling class
marks, assuming such solutions were desired.
The
transition from 2-wire analog (metallic) to 4-wire digital
(electronic) switching in local COs, the change from 2-wire analog
phones to suitable digital instruments, and overall standardization
among hundreds of manufacturers world wide will be a challenge our
children and grandchildren can wrestle with. If they understand the
objectives as well as they seem to grasp the potential of hardware
and software, the future can be bright indeed.
-
Trunk
-
Signaling/supervision
-
Glare
-
Group/digroup
-
DP/MF/CCIS
-
TLP/VNL
-
Return Loss
-
Longitudinal balance
Click Here for
Answers
1.
What is a trunk?
2.
When is a trunk a line?
3.
How do metallic trunks differ from lines?
4
List some factors encouraging the use of trunks on carrier systems.
5.
Discuss the differences in terminating digital trunks on 2-wire
space division switches and digital switches.
6.
Compare in-band and out-of-band signaling on analog trunks.
7.
What's the difference in one-way and two-way trunks?
8.
When is a program change from incoming to outgoing, or from either
to two-way, impossible?
9.
What is the major difference between regular CO trunks to PBXs and
DID trunks?
10.
Describe briefly the operation of E&M supervision. What must be in a
pulse link repeater between two E&M trunks directly connected back
to back?
11.
How might a digital switch connected directly to a digital facility
handle supervision?
12.
What is "wink start?"
13.
How can glare be dealt with?
14.
How is MF different from DTMF?
15.
What is the difference in bit robbing and digit robbing?
16.
Give some advantages of common channel signaling.
17.
What is TLP?
18.
What is VNL?
19.
Transmission people talk about two kinds of "balance." Identify
them, and explain how they are different.
20.
If digital pads were used to insert VNL into every digital long
distance trunk, what impact would this have on transmission?
21.
If signals, whether voice, data, video or something else, are
encoded to and decoded from a T-carrier bit stream in the terminals
which generate and use them, what is the function of transmission
and switching systems?
[ Top ] [
Next Chapter ] [
Table of
Contents ] |
