|
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 7
Levels In The Switched
Digital Network
Since the majority of
carrier trunks in the public network are already T, and since the
economics of A/D conversion in No. 4 ESS will encourage an ever more
rapid expansion of T-carrier, even AT&T recognizes the digital
nature of the system that is developing. That the system appears,
from the outside, to be totally analog is beside the point. However,
new standards for transmission are required; since these standards
will be inflicted on Independent Telephone manufacturers in addition
to Western Electric, a quick look at them is very much in order.
In the ultimate
all-digital network as currently expounded by AT&T, connections
through the Toll Network will have a flat loss of 6 dB. Class 5 to
Class 5 office direct connections will have a loss of 3 or 6 dB,
presumably depending on trunk length. Conversion from Analog to
Digital will take place at 0 TLP, apparently to have the analog
signal/noise ratio as high as possible before encoding, and to
reduce any noise generated within the system, particularly noise in
the absence of speech. The 3 or 6 dB loss will be placed in the
terminating side of each connection, just before the four-wire to
two-wire conversion. Loss through Class 5 switches, even when they
are digital and, as a result, 4-wire internally, will be held to 0
dB. To maintain stability, particularly on local-to-local calls,
Digital Class 5 Switches are to be designed with new terminating
networks that, on paper at least, are supposed to provide a good
enough impedance match to prevent hollowness and singing around the
4-wire path through the switch. All intertoll trunks will, of
course, be 4-wire and lossless.
There will be a finite
period of time (would you believe 50 years?) before Digital Class 5
Switches have taken over the space occupied by Nos. 1, 2 and 3 ESS
in Bell operating companies. Thus somewhat different standards seem
to be required in the immediate future. Here analog Toll Switches
will continue to have their outgoing switches at -2 TLP, as at
present, and analog toll connecting trunks to analog Class 5 offices
will continue to have VNL + 2 dB loss built into them. The outgoing
switch (or whatever mythical point is used) in a No. 4 ESS will be
considered to be at -3 TLP. "Combination trunks"* (not to be
confused with 2-way CO trunks to a PBX) from analog to digital toll
switches will have 1 dB of loss. Analog trunks, even between No. 4
ESSs, will have VNL, and if either toll switch is analog, either
combination or analog trunks will be set up to incorporate VNL.
[*Footnote: In this discussion, a "digital" trunk is one that is not
demodulated back to analog for switching at either end while a
"combination" trunk is demodulated at one end only. An analog trunk
thinks it is analog all the way, even though it may be
T-carrier between two sets of channel banks.]
It is not immediately
evident how one goes from 3 dB loss in each end of each side of each
transmission path to 6 dB at the terminating end of each
transmission path when the millennium arrives, but we'll let that
pass. The much more serious requirement is for 0 loss through the
digital Class 5 offices that independent manufacturers are building
and marketing now.
The demand for 0 loss
seems to be based on the idea that local-to-local connections should
be 6 dB louder than toll connections or, as AT&T views it, toll
connections, to prevent echo or oscillation, must be at least 6 dB
below local connections. One can appreciate their reasoning, since
about 90 of all connections are classified as local. (That other 10
generate much of the revenue and most of the profits in the
telephone industry, however.)
To arrive at the proper
losses in a telephone connection, years of study and experiment have
been expended. The ultimate result is a model used for computer
simulation to relate amplitude of telephone signals to noise and
echoes. Rating of connections, with varying degrees of level, noise
and echo, is made on the basis of "excellent," "good," "fair,"
"poor," and "unsatisfactory." Combinations that, on a supposed large
sample of users, would produce a given percentage of goods and
excellents are then found.
This model is apparently
quite complex and is based on years of effort by many people, both
in the lab and in the field. To disprove it would take a somewhat
similar effort. In considering its use, however, one should keep
certain facts in mind. First, most of the people involved in
standardization of all the data were telephone employees, mostly at
Bell Labs, and hardly a random sample. Second, there are admitted
and unexplained differences between data obtained at the Murray Hill
and Holmdel Laboratories. Third, the Bell System has a curious track
record in human factors measurements: it predicted its customers
would love all number calling, for instance, and many years ago
demonstrated the absence of "perfect pitch" on the basis of a sample
that included not one musician. Further, so far as I know. Bell is
the only organization in the world that finds a good correlation
between grade-point average in college and performance in later
life. The possibility of self-fulfilling prophecies, the Hawthorn
effect and other well-known booby traps cannot be ignored.
In any event, based on
computer simulations, standards for the next several decades appear
to be based on the idea that, because 2-wire analog switches insert
no loss, 4-wire digital switches should follow suit. After all, Bell
is going to have a LOT of analog 2-wire switches for a long, long
time.
Bell articles seem to
imply that Independent Telephone Companies, when they buy digital
Class 5 offices, will continue to use 2-wire toll connecting trunks
and intra-local trunks to other nearby Class 5 offices; in addition,
long customer loops, more prevalent in Independent telephone
companies than in Bell areas, will not be impacted by the
introduction of digital switching. Thus the 3-dB loss (or VNL + 2
dB) for toll network stability must still be inserted in toll
connecting trunks, the outgoing (digital) switch must be at 0 TLP,
and customer to customer loss can only be minimized by keeping loss
out of the switch.
All of these assumptions
would appear to be false. All Independent digital Class 5 office
designs include remote concentrators, and some also include or plan
to add station carrier. Thus loss in long loops is greatly reduced
by bringing the switch to the customer. Once the A/D conversion is
made, there is no further loss on the T-span lines between the
concentrator and the main switch. (In rare cases where a very remote
customer cannot be served by station carrier or a remote
concentrator and, as a result, connects directly to the switching
system over a long pair, long-lines circuits or some such can be
used.)
On the trunk side, the
argument in favor of T-carrier moving the A/D conversion to the
distant office (if it is analog) is just as true for a Class 5
switch as it is for No. 4 ESS. When the digital switch homes on a
No. 4 ESS, it is hardly likely that an analog carrier system could
remain, even if already in place, when compared with simple T-span
lines. Even assuming an analog connecting office reached by analog
carrier, it is hardly likely that existing 2-wire terminals at the
Class 5 end would not be changed out or restrapped to at least
provide a 4-wire trunk-to-switch interface.
In view of the above,
standards should allow loss to be placed, not in the toll connecting
trunk, but in the Class 5 switching machine. The overall loss in a
toll connection would not be changed—the toll network could not even
tell that the loss is in the switch rather than the trunk. In a
local connection, the loss would remain to insure stability, but it
would be recovered by having shorter loops between customer
telephones and remote concentrators or station carrier. A meaningful
standard would limit customer-to-customer loss rather than dictate
how that loss should be distributed. There does not appear to be a
real need for AT&T to hang competitors’ switches on the edge of
instability, since proper system design will hold
customer-to-customer losses to present levels.
Of course if after years
of accepting 900 ohms + 2.14 microfarads, AT&T has a better
compromise balance network, one might as well use it. If remote
concentrators are used, both local and trunk transmission will be
improved. There seems little to gain by making local-local
connections 6 dB louder than toll connections if they are already
loud enough.
But the most important
requirement for the future all-digital network (for our
grandchildren) is end-to-end digital integrity. This suggests that
once an analog signal is encoded, it should not be altered or, if it
is, traveling class marks should be available to prevent a similar
fate for direct-access data. However, testing requirements would
suggest that a digital signal should be at the same level
everywhere, since there is no reason why this simplification should
not be used. Thus any stabilization loss should be "outside" the A/D
converters.
The system designer can
arrange the system to have fixed loss, to switch the loss on a
per-line/per-call basis (3 dB for inter-local, 6 dB for toll
connections, half in each end of the connection), or use digital
techniques with traveling class marks and require the user to inform
the system each time a change from voice to data is effected so that
CCIS can go hand over hand down the call to change pads at both
ends. Fixed switch loss seems simple. All other plans seem to add
vast and unnecessary complications.
Considerable effort has
been expended on a new balance network. However, little attention
has been given to maintaining full four-wire integrity to the
telephone set, particularly in PBXs where it is easy, cost effective
and has a maximum opportunity to improve transmission. Northern
Telecom is going a slightly different route. Apparently they are
working on the electronic residential telephone set that will
replace the 500 type subset. By eliminating the pre-electronic
components, they seem to be well on the way toward improving
return-loss balance from the subset end, improving compensation for
loop length, and improving frequency response. They are also adding
a flock of features including "hold" and repertory dialing.
But whatever approach is
used, an open ended future would suggest that every effort should be
made to preserve the possibility of effecting the A/D conversion in
the telephone set and providing at that point a parallel data input.
This would automatically eliminate the 2/4 wire conversion with its
attendant problems in return-loss balance, and would perhaps permit
telephony to enter the 21st century only 25 years late.
[
Top ] [ Table of Contents ] [
Next Chapter ] |
