|
The Digital Future Of The Telephone Network
A Study of Evolving Technology
By Lee Goeller
Originally published by Probe Research Inc. 1979.
Reprinted by permission
Chapter 5
An Idealized All-Digital Communications Network
Introduction
It is a basic fact that
one almost never has the opportunity to build a major engineering
structure from scratch, independent of the context into which it
will fit. Unlike novels or pictures, engineering creations are not
independent units; they must fit into a scheme of things that is
already widespread and well advanced. Nowhere is this more true than
in telephony. Each piece of the telephone system must work with all
the others, and even if something new and wonderful comes along, it
is almost impossible, by the sheer magnitude of the task, to replace
all the old equipment on short notice. As with a railroad, a highway
system, a power grid or a city, an appreciable time is required to
effect change in any direction.
All things considered,
this is not a disadvantage. When evolution is demanded and
revolution is, for all practical purposes, impossible, the infinite
facts of life that make even the best designers' plans go astray can
be discovered and corrected before errors are mass produced and the
entire future is committed to disaster.
In spite of this, it is
worthwhile to consider what an all-digital continent-wide telephone
network might be like, independent of the transition period during
which old and new equipment must co-exist. By ignoring the enormous
capital investment already in place, the facilities for
manufacturing, the training and retraining of personnel, etc., we
may still be able to gain some valuable insight by considering a
possible future system. Not the only one, or necessarily the best,
but a specific scheme that can, at the very least, be contemplated.
Basic Requirements
Of all the lessons we
have learned from the past, perhaps the most important relating to
telephony is that the value of the telephone system increases in
direct proportion to its ubiquity. If only a few rich people had
phones, for instance, society as a whole would get little benefit
from Bell's invention and, because of the small volume of traffic,
developments and improvements would find financing difficult.
In addition to ubiquity,
uniformity in certain specialized ways is desirable to permit each
terminal on the telephone network to communicate with each other
terminal. Signaling, transmission levels and other basic properties
must be planned and implemented as a whole.
Uniformity, however,
runs head-on into the next requirement: flexibility. Uniformity
achieved at the expense of limiting communications to voice only, as
at present, is totally unsatisfactory. Data and facsimile have
forced themselves into the telephone network, greatly increasing the
revenues of the telephone industry, but only at the price of
"modems" that transfer natural digital signals into pseudo-voice
patterns. Indeed, the unhandyness of the existing network for
digital communications may well be a major factor in holding back
data transmission. Flexibility in the future must include not only
new voice services, but also a convenient means for handling
non-voice communications.
The PCM Network
To achieve uniformity
and flexibility without sacrificing the present ubiquity, one big
switched network capable of handling all types of signals seems to
be indicated. Not two or three different networks for different
kinds of service, or separate networks for large customers competing
with the public network used by residential and small business
customers. But one big, universal network.
Needless to say, one
large network of this sort has never been achieved in the past;
countless dial tandem tie-line networks plus a smaller number of
CCSA nets have gone in, and many types of data network are common.
Indeed, it is reported in various places that as many as a third of
all intercity trunks are in private networks. To compound the
problem, packet switching and other advanced data techniques are
filling the void that the present voice network has generated.
To make a start at a
universal network, it would seem reasonable to consider PCM coding
to be basic, and to use digital switching and transmission as a
combined entity. Almost everyone seems to agree on this, but the
next step produces considerable disagreement: once a signal is coded
in digital form, IT SHOULD REMAIN INVARIANT. In particular, digital
pads should NOT be used to change levels. Digital manipulations
should NOT be used to go from one companding rule to another.
Digital translations should NOT be used to go from one type of
coding to another for specialized switching systems such as PBXs.
Only if the digital words can go end-to-end in their original form
can they be used most effectively for direct data transmission
without moderns.
Keeping the digital
coding invariant has a number of additional advantages. For
instance, VNL and TLP vanish in voice measurements, and a voice
signal, if expressed in PCM, is always at the same level. This
should simplify installation, line-up and maintenance, and take full
advantage of digital test tones and loop-around test techniques.
Naturally, ALL voice connections through the network will have the
same loss. A direct trunk will not come in with more level than
connection via a tandem/toll switch, and a local connection will not
be markedly louder than other connections, as is presently the case.
Since digital transmission makes circuit noise relatively
independent of circuit length, the signal-to-noise performance of
all calls, both local and long distance, will be about the same, and
better than at present.
The next aspect of the
PCM network is the need to move A/D conversion as close to the user
as possible. Whether the conversion should be in the telephone set,
on a per-line basis prior to concentration in the switching
equipment, or on a concentrated basis may well depend on the
particular installation; further, the use of -lumped vs. distributed
switching at the Class 5 switch will impact all such decisions.
In any event, it seems
clear that a digital Class 5 switch is needed and, if conversion
tricks are not to be played on the digital bit stream (which might
be desirable for analog loss but could destroy digital data), no
loss can be permitted in toll connecting trunks. The only place for
the loss required to control echoes in analog connections is on the
analog side of the A/D converters. With traveling class-marks,
digital pads could be inserted as required, but this would require
the station user, in many instances, to inform the system as to
whether a digital or analog call is to be made from the same
instrument. Since an analog connection might change to a digital
connection after the call is set up, the required operation might be
difficult.
As suggested above, the
A/D conversion might take place in the telephone set, at the switch
on a per-line basis, or on a concentrated basis. With conversion in
the telephone set, there would be no 2-wire segment at that end of
the connection and, as a result, no echo. Elimination of the hybrid
and the isolation of side-tone (from transmitter to receiver so that
the speaker can hear his own words), as shown in Figure 5,
would also reduce the cost of the per-telephone chip. It has even
been suggested that several telephones could be multiplexed on a
single telephone pair for greater economy. With suitable
multiplexing, both directions of transmission for each telephone
could be carried on one pair with little difficulty.
Figure 5. Possible
Electronic Telephone. Note that 4-wire integrity is preserved, and
data access can be provided.
Problems with this
approach are several. First, ringing must be altered to match the
new technology; power ringing would have to go. Second,
synchronization, even with a single telephone, might get a little
tricky. Third, bridged telephones (extensions) and multi-line
instruments would pose problems. Finally, supervision would have to
be considered carefully. Obviously the on-hook/off-hook information
should be digital, but should it be coded in with the speech (one of
the 8 bits every sixth sample) or should it be separate? Should it
be used for address information, as with existing dial pulsing only
faster, or should an analog DTMF signal be generated and converted
to a digital signal in the A/D converter? Common channel interoffice
signaling will ultimately allow interswitch trunks to work with full
8-bit coding, but if the signal were to reach the first switch with
something less than 8 bit coding, the 2 dB improvement expected in
signal-to-noise ratio by reclaiming the signaling bit could not
happen.
In any event, if a
practical means for performing the A/D conversion in the set and
getting the resulting signal to the switch is devised, it will be
quite simple to provide the set with an RS-232 or similar interface
as shown in Figure 5 and plug in data terminal equipment.
Alternatively, of course, the telephone set can be added quite
inexpensively to standard data terminals, making available the full
alpha-numeric capabilities of such equipment to the telephone
system, if desired. Communication with the station user as in some
of the more advanced telephone sets presently on the market would
then take on a new dimension.
However, in most
instances, digital conversion in the telephone set may not be
necessary. If conversion is done at the switch, concentration can be
carried out through a metallic matrix, as shown in Figure 6A,
so that conventional 2-wire telephone sets can be retained. Power
ringing could be applied from service circuits accessed directly via
the concentrator; similarly, DC dial pulses, coin control signals
and test access could be handled in the same way. Only when the
"talking" connection is set up would the line be brought to the A/D
converter and the path to the rest of the world.
Figure 6A. Metallic
Concentrator to minimize A/D and 2/4 wire conversion equipment while
allowing conventional telephone sets to be used. T-Carrier circuit
cards contain hybrids and supervision and are identical since
Service Circuits handle Ringing, Coin Control, etc.
To keep the 2-wire
analog loops short (to minimize loss and the variability in
impedance matching), such a system would almost certainly make use
of remote concentration. That is, T-lines from the main switch would
extend the T-format out to the concentrator where one or more
conventional T-carrier channel banks, as well as a selected group of
service circuits, would access a group of 2-wire loops via, perhaps,
a 5:1 metallic (reed switch, miniature relay, etc.) concentrator
switch. A data link would allow the central switch to control the
metallic matrix and to manipulate the service and test circuits.
Figure 6B. A
Conventional PBX providing concentration connects to a digital CO
via a standard T-Carrier terminal and span-line. With suitable
plug-in line cards (as in Vicom SM-T equipment, for instance)
customer data can access the digital network directly via T-channels
not needed as PBX-CO trunks.
Such a system could,
instead of interfacing a concentrator matrix, interface a
conventional PBX, as shown in Figure 6B. Since the PBX
concentrates traffic, high usage would be assured on the digital
channels. In addition, the PBX approach would bring digital channels
quite close to individual data stations. These might bypass the PBX
switch and enter the T-carrier channels directly, through data
interface line units rather than voice interface units.
Figure 6C. Electronic
Concentrator (could be a PBX) to minimize switch size.
Line cards must now
apply ringing, etc., and different types will be required. A/D and
2/4 wire conversion must now be carried out per line.
The third variation is
to have the concentrator follow the A/D conversion so that it, too,
can be digital as in Figure 6C. This allows a smaller and
probably less expensive stage of concentration, but may increase
overall costs in that per-line circuitry must now be provided. Such
circuitry can become quite expensive if conventional telephone sets
are to be used with their power ringing, and inventory gets to be a
problem if coin sets and other specialized equipment must also be
interfaced. However, the major class of such electronic
concentrators can be electronic PBXs. Indeed, many PCM electronic
PBXs can be considered concentrators of this sort today, and can be
connected via a T-line to the public network. Since they do not need
any channel bank equipment at the PBX end, savings are considerable.
With the PBX
Concentrator approach, it is once again economical to have four-wire
connections to the telephone sets and no 2-wire segment in the
connection at all. Most PBXs have appreciably shorter runs to their
stations than do central offices and, more important, most present
PBXs run 25 pairs or more to each key telephone set to provide the
required user features. Modern electronic telephone sets matched to
PBX capabilities are just beginning to come on the market. Such sets
usually require only 2, 3 or 4 pairs; typically, one pair is for a
voice connection to the switching matrix, one is a data connection
to a buffer for common control access, and the third is for power or
some other purpose. It would seem quite simple to have two pairs for
voice and have them also supply power. However, only two systems, so
far, do this: the very small Tele/Resources System 32, and the very
large Danray CBX (computer based exchange).
The Dimension Custom
Telephone Sets, with four-pair wiring, have extra pairs, but
maintain a 2-wire voice connection. This makes some sort of sense in
that the Dimension is one of the very few 2-wire time-division
switches available, and underscores the fact that being 2-wire is,
in the present world of analog tie lines, far more serious a design
flaw in the Dimension than the fact that it is not digital. When one
turns to the SL-1 or Rolm 4-wire digital PBXs, however, it is harder
to understand why 4-wire integrity to new electronic telephone sets
has not been preserved. Hopefully, the importance of 4-wire
integrity, particularly where it doesn't cost anything, will be
recognized by PBX designers in the future, and the traditional
choice of 2-wire voice paths for four-wire switches will give way to
more meaningful implementation.
It should be noted that
PBX and Centrex systems, although they serve only about 14% of the
telephones in the United States, originate or terminate something
over 60% of all calls. The statistical impact of 4-wire connections
all the way to the PBX telephone set should do a lot to improve
transmission in general.
With digital PBXs
accessing the public digital network on a fully digital basis, there
is no need to go digital all the way to the telephone set. Just as
the PBX itself can be thought of as a stage of concentration off the
local Class 5 switch, so individual sections of a PBX can be
concentrators and located remotely from the rest of the PBX.
Northern Telecom's SL-1 is already designed to provide just this
kind of service. One central control and distribution matrix
arrangement is fed by concentrators which may be adjacent, or may be
located miles away, interfacing via a T-line. One can easily
visualize, at some time in the future, digital line cards being
available in addition to analog line cards (as in T-carrier systems
used for transmission). These digital cards could then be plugged
into concentrators located near the people and equipment to be
served. One concentrator might be in the main computer room, another
in a terminal room, etc. Regular copper pairs should be sufficient
to bring even high speed data in DC form to a concentrator less than
300 feet away.
Another potential
advantage of remote concentration, as has been suggested earlier, is
the possibility of reducing the MDF (Main Distributing Frame)
function. Although a central office switch would, presumably, be
located to minimize the need for remote concentrators, the
existence, particularly in metropolitan areas, of large numbers of
concentrators already in place (PBXs) suggests that demodulation and
re-expansion just to provide the cross-connect MDF operation
approaches the absurd. Rather, the distribution stages of the CO
switch could well be designed to be nearly or truly non-blocking,
permitting the switch to patch through voice channels on a permanent
basis without coming back to analog.
With a T-carrier
terminal meeting a conventional PBX, it is relatively unlikely that
the required number of CO trunks would be a multiple of 24. Thus
other channels could handle data circuits or other point-to-point
lines directly. Such point-to-point circuits would, however, have to
be "pegged up" through the CO switching. Only if the switch's
traffic handling capacity would not be reduced by such a "permanent"
connection would such an operation be practical.
An MDF, like any other
switching system, becomes more complex as it increases in size. Many
MDFs are now backed by elaborate computer facilities to keep records
and assign jumper terminals. It would appear that many such records,
as well as the actual manipulations they define, could be handled
internally in the main switching system with savings all around. Any
additional cross-connections at remote concentrators would tend to
be quite simple, due to the relatively small size of most
concentrators compared with central offices. It seems fairly
evident, however, that any Class 5 digital switch should consider
incorporating the MDF function.
There is no reason why
all three of the above approaches could not be used simultaneously
on one switching machine. Further, for lines close to the central
office location, "remote" concentrators could be located adjacent to
the digital switching equipment, permitting relatively standard
operation. By combining these approaches, the 4-wire to 2-wire
interface would be eliminated altogether on many (of the busiest)
lines, and the impedance and loss variation on the remaining lines,
because of the short distance to either local or remote
concentrators, would be held to much tighter limits. This would
simplify the choice of a matching network and improve return loss
(to minimize echoes) where 2-wire lines have to be used.
With regard to the
matching network, it should be noted that some differences of
opinion exist already. The traditional 900 ohm resistor in series
with a 2.14-microfarad capacitor simply will not do in digital
systems (for reasons that we will discuss in the next section), and
various new networks have been suggested. AT&T recommends 1100 ohms
in parallel with .03 microfarads as a general network, with 800 ohms
in parallel with .05 microfarads for non-loaded lines if they can be
segregated from loaded lines. Loaded lines should be terminated in
1650 ohms in parallel with 5 nanofarads (which is a very small
capacitor indeed). Just to add to the confusion. North Electric (now
a subsidiary of ITT) likes 700 ohms in parallel with .05 microfarads
for non-loaded loops and 1400 ohms without a capacitor for loaded
loops. Doubtless other networks will surface shortly.
A four-wire circuit,
switched 2-wire through a metallic matrix to customer loops, has to
meet a variety of different loops and thus requires some sort of a
compromise network. Such a network, no matter how selected, will
never provide a very good match to the universe of existing outside
plant. However, if the switch is 4-wire, as almost all electronic
time-division switches except Dimension and No. 101 ESS must be, the
4-wire to 2-wire conversion has to be made on the line side. If the
conversion is made on a per-line basis, then it is possible to use
small switches or strapping points on the circuit card to tune the
matching network to its particular loop. Almost everyone seems to
feel that this is uneconomical in spite of the considerable amount
of hand labor already involved: running a jumper on the MDF, traffic
balancing, selecting class-marks, etc. But it would seem that echo
in the existing plant could be reduced considerably if an escape
from the ritualistic 900 ohms plus 2 microfarads could be effected.
Another advantage to the
possible use of remote concentrators and PBXs accessed via T-lines
is the potential for using fiber optics for transmission between the
concentrators and the central switch. It is unlikely in the near
term that fiber optics will be used to individual residential
telephones since light-pipes do not carry electricity and
reliability suggests that residential telephones should be powered
from a central telephone location where stand-by batteries and other
power systems are available in the event of the failure of
commercial power. Concentrators and PBXs, however, can be arranged
to have their own reliable power, perhaps supplied via separate
facilities from the central switch. Thus they could use optical
channels to the CO, enabling the telephone company to replace many
of their ancient cables with worn insulation and corroded conductors
without digging up the street to install new ducts. Although such an
approach is obviously desirable in major cities, another irony of
the present situation is the proposed use, so far, of digital local
switching only in rural regions.
In any event, the
technology is here now to provide most of the suggested approaches.
Non-Bell manufacturers are going ahead with their designs, and "No.
5 ESS," an all-digital local switching system, is rumored to be
under crash development at Bell Labs. But it seems very unlikely
that all the advantages of an all-digital system will actually be
available to anyone now living.
Getting There From Here
It took 25 years for
Crossbar systems to take over half the switched lines in the Bell
System. ESS switches are going in at the rate of one per business
day, or about 250 a year, but they have only begun to make a small
dent in the 11,000 Class 5 offices in Bell territory. Thus it is
easy to see why AT&T is starting to use digital switching in the
toll network. There are only a few hundred high order toll switches,
and replacing these will produce maximum impact. No. 4 ESS, combined
with CCIS and matching electronic controls on the remaining No. 4
Crossbar offices, will do a lot to upgrade telephony. However, as
has been mentioned, the long-haul trunks will, for very good
reasons, remain analog for many years to come, and the replacement
of Class 5 offices, including new and relatively "immortal" Nos. 1,
2 and 3 ESS, will take even longer. Thus digital Class 5 offices
will go into an environment that can only be described as hostile.
Nevertheless, it would
appear that digital Class 5 offices, serving digital PBXs and homing
on No. 4 ESS tandem and toll switches, could rapidly establish
little islands of all-digital circuitry that could be pressed into
service almost immediately on the basis of being digital as well as
being inexpensive. Toll connecting trunks between digital Class 5
and Class 4 switches would have to operate at no loss, as would PBX
trunks to digital PBXs. This would make available digital channels
that could then be used for digital as well as analog service.
The problems come in
interfacing the remaining parts of the network that are still
analog. Direct Class 5 to Class 5 trunks might well be digital on
one end and analog on the other, and the amount of loss to assign to
such a circuit is not obvious. Similarly, how much loss should a
switch insert into a connection between two long customer loops? The
approved answer at the moment is "No loss at all." This says that
intra-switch calls must be 6 dB louder than toll calls, and the
switch transmission properties must be completely altered for intra-
vs. inter-switch connections. This can lead to instability,
hollowness and even oscillations on local calls and add considerably
to the cost and operational complexity of the switch.
The more reasonable
approach would simply put whatever loss is required for stability
(presumably 3 dB per circuit end) on the 2-wire analog side of the
system and leave it there for all calls. If conventional hybrid
coils are used, they introduce the 3 dB loss automatically and no
additional equipment is needed. Note that 3 dB is only half a dB
larger than the present actual loss in a short toll connecting trunk
and the local switching system.
Trunks between digital
Class 5 switches and digital Class 4 switches would operate without
loss in such an arrangement, since the required loss would be
provided elsewhere. Analog offices, homing on digital Class 4
offices, would continue to operate within the VNL + 2 dB requirement
on toll connecting trunks. Since the installation of No. 4 ESS as a
digital Class 4 office will change much of the trunking already in
place, it might be desirable to use the opportunity to increase the
loss in such toll connecting trunks to Class 5 analog offices to
match the loss at a digital local switch. Then connections between
Class 4 and higher order switches would be simpler.
All digital trunks
between higher order digital switches should run at 0 loss. Further,
it would seem that all 4-wire trunks, whether digital or analog,
should also run at 0 dB loss. Long trunks with echo suppressors
already do this, and short trunks are known to have too much loss
anyway. Only in the intermediate range (1000 miles to 1850, say)
would there possibly be need for additional loss to bring the total
up to VNL +4. Complications would exist where 2-wire Class 4 offices
are still in place, and where operator positions are required, but
the oversimplified approach suggested above is not completely out of
the question.
To make possible
end-to-end digital connections for true digital signals, some means
would have to be provided fairly early in the game to interconnect
the various "digital islands." Digital interconnections would
parallel the existing long-haul radio and coaxial analog circuits,
and would require the use of traveling class-marks for access.
Digital terminals, on a per-line basis, could generate such class
marks, and CCIS could carry them away until some time in the misty
future when fiber optics replaces radio and other band-limited
facilities. The most likely candidate for such interim inter-island
digital paths might be satellite circuits, perhaps even used on a
DAMA* basis. A relatively small number of such circuits could handle
true digital signals until the effect of Parkinson's Law increases
the end-to-end digital traffic enough to justify more facilities.
[*Footnote: Demand Assignment Multiple Access. With DAMA, a radio
channel
is set up between two earth stations only while needed for a given
call.]
In any event, most of
the technology is available and the research has already been done
to provide, first, digital islands and then greater digital
opportunities. Working toward such an overall approach would seem to
be a good idea. But this is a game owned by AT&T, and most of us
will never be allowed to play.
[
Top ] [ Table of Contents ] [
Next Chapter ] |
