Voice
Communication in Business Volume 1
Essays on telecommunications,
1969-1980
Chapter 17
Sometimes It's Better The Second Time Around
For
history buffs, 1976 was a dandy year. It was the Bicentennial for
the United States and the Centennial of the telephone. Furthermore,
the magazine Telephony was 75 years old. While at Bell Labs, I had
spent many lunch hours in the library, looking through Telephony's
early volumes. I found the story of how Kempster B. Miller had taken
North Electric back from the evil telephone monopoly and many other
adventures worthy of epic motion pictures. I tracked the "Automanual"
system from its triumphant entry to its sad fading away as a one
inch ad in the back of the journal several years later. And there
were other items. It is not generally realized that bound volumes of
old magazines are really time machines, just waiting for those who
want to live in another era.
With this background, I was very pleased when
Telephony asked me to do a
piece for their July 5, 1976, special issue that was commemorating
so many things I admired. Because one must know where one is coming
from if one wants to know where one is going, it seems fitting to
start a look at the digital future with a brief survey of the analog
past. Here is the article, but with my original title.
***
In a hundred years,
almost everything has been tried at least once in telephony. But for
a concept to catch on, it must come at a time when technology,
economics and user attitudes converge to produce acceptance. Some
good ideas have never quite made it, while others have triumphed two
or three times, only to fade out once again as something else a
little better came along. Innovation is what makes the telephone
business the most interesting way to make a living that I can
imagine; historical perspective, however, adds insight that can give
deeper meaning to each new development as it appears.
Pushbutton calling
As an example, consider
pushbutton calling. Almon B. Strowger used it in his first
installation at La Porte, Ind., in 1892. His basic patent shows
three pushbuttons at the subset for controlling a thousand point
"connector" switch. Depressing the first button caused the switch to
move vertically, depressing the second button caused the switch to
step past ten-line groups with a rotary motion and by depressing the
third button one line was selected, again with a rotary motion, in
the chosen group of ten. To call line 345, the user would push the
first button three times, the second button four times and the third
button five times. A fourth button was provided which, when pressed,
released the switch. Each of these four buttons had a wire going
from the subset to the central office; a fifth wire was used for
talking. La Porte was a 100-line system, and actually used switches
without the vertical movement. Presumably that meant only four wires
were needed in the La Porte system.
The Strowger system
required a large number of wires and a heavy battery at each station
to operate the switch at the central office. User problems
(including errors produced by losing count, not holding down the
button long enough on each pulse and failure to hold the release
button down long enough to completely restore the switch at the end
of the call) also complicated its operation.
This form of pushbutton
calling was phased out with the invention of the dial by Keith and
the Erickson brothers in 1896.
Pushbutton calling seems
to have surfaced next in such "semi-automatic" systems as Clement's
Automanual (1906) and an early version of Western Electric's rotary
system (circa 1910). Here the operator rather than the user had the
pushbuttons; they were mounted on a console, another invention that
has come and gone several times. The operator would obtain the
number from the user and key it into the system which would complete
the connection automatically. In Automanual, all four digits could
be entered simultaneously. An operator was said to be able to handle
over 400 calls per hour without difficulty. The WE system, using
teamwork between two or more operators, claimed 500 calls per hour
per operator.
In both of these
systems, four rows of ten keys each were provided to set up a
number; in the WE system, additional keys for selecting different
exchanges were provided. Once a key was depressed, it locked until
the systems was finished with it. Automanual had a very simple
out-pulsing system, and the keys stayed down until all the switches
were set; for this reason, each operator was given three sets, so
that she could have three calls in the works at one time. The WE
system had two registers associated with each key set; as soon as
one was loaded, the other would be connected in its place to accept
the next number while the first register set up its call. This way,
the operator needed only one set of pushbuttons.
The ten-button key set
was first used in toll dialing in a Bell installation in Detroit in
1930. Actually, 12 buttons were used since KP (key pulsing) and ST
(start) were added to the ten digits. The operator would push KP to
unlock the system, transmit the digits one at a time to a sender by
means of DC key pulsing, and then depress ST to signal the sender to
start. ST also released the operator from the connection.
Pushbutton calling with
voice-frequency tones was tested in Baltimore and then was used
extensively with the first No. 4 crossbar toll switch in
Philadelphia in 1943. The advantage of multi-frequency (MF) tone
pulsing was that the local sender was not needed; an operator at
some distant point could key the called number over the trunk
directly to a receiver in the Philadelphia No. 4 crossbar.
Pushbutton tone calling
by users was tested in Media, Pa., in 1948. Tones were generated at
the subset by plucked metal reeds resonant to the desired
frequencies. These vibrating reeds induced tones in coils for
transmission to the central office where a receiver detected them.
In these pre-transistor days, "The economic aspects of the system...
were not attractive and further work was deferred."
Later that year,
invention of the transistor was announced. It made a whole new look
possible, and ten years later, Meacham, Power and West, in a Bell
System Technical Journal article, discussed laboratory work on a
transistorized system. The vibrating reeds at the subset were
replaced by a transistor oscillator, and a transistorized central
office receiver replaced the vacuum tube circuitry used earlier. The
system used two tones, one for party identification and the other
for digit identity. Both were damped sine waves, following the
plucked reed example, and were transmitted simultaneously. The
receiver separated the eight tones used for party identity in the
"low group" from the ten tones used for digits in the "high group,"
and then detected each signal independently.
It was only a short step
to the Touch-Tone system described by Schenker in the BSTJ two years
later. A transistorized oscillator in the subset still created two
tones, one in a high and one in a low group. The CO receiver split
the two groups apart and detected the tones independently. But there
were differences, too. The subset tones were continuous, not damped,
and only eight tones in all were used. Party identification was
abandoned in favor of a more robust digit signal that could be
simulated only with great difficulty by the human voice. By
representing each digit by exactly one tone in the low band and one
in the high band, the ten digits plus six other signals are
available. This is the system in use today.
As can be seen,
pushbutton calling went through a number of trials before it reached
its present state of perfection, and it needed a radical change in
technology to become practical. Is it the last word? Probably not.
Digital signaling is already being used in some key telephone
systems, not only to transmit digits to the control equipment but to
return information as to which line is calling, which is on hold,
etc. Nothing stands still in this business.
Pre-set signaling
There are, of course,
other approaches to user signaling. Why shouldn't the user, for
instance, set up the called number before coming off hook and let
the system accept the number in some high-speed way as soon as the
handset is lifted? With a visual indication provided, the called
number could be checked to reduce errors, and register holding time
at the CO could be greatly reduced.
An early example of this
method was invented by the Lorimer brothers of Brantford, Ontario,
prior to 1900. Brantford, of course, was the home of Alexander
Graham Bell when he wasn't busy teaching speech or inventing the
telephone in Boston. The Lorimer system used a calling mechanism at
the subset based on four pivoted levers that could be set to any one
of ten positions. Moving the lever grounded a terminal corresponding
to the digit desired; at the same time, it placed the digit on view
in a small window above the lever.
To place a call, the
user would set the four levers to the appropriate positions, check
the number and then wind up a clockwork spring with a crank. When
the crank returned to normal and the user picked up the receiver,
the CO would send a train of pulses to the station to step an
escapement that allowed the wound-up spring to move a wiper across
the terminals associated with the first digit-lever. As soon as the
wiper found the ground established by the lever, that ground would
be sent to the central office to tell it to stop. Each pulse sent to
the subset had also been stepping a CO switch. The stop pulse told
the switch it had reached the right terminal. The switch would be
locked at the desired point, and the next switch would return
another pulse train to interrogate the following digit at the
subset. There is a remarkable similarity to the "revertive pulsing"
later used by the Bell System in the rotary and panel systems, and
at one point in its development, rotary used a call mechanism that,
from photographs, looks very much like the Lorimer equipment. When
the Lorimer patent was issued, after some 16 years in the Patent
Office, it was assigned to the Western Electric Co.
The Bell System obtained
the rights to SXS (step-by-step) at about the same time, and the
first Bell stepper was installed at Norfolk in 1919. With the dial
now standard, more exotic calling devices apparently were not
pursued. The next pre-set telephone that I can find is part of an
experimental electronic system described by Malthaner and Vaughan in
the May, 1952, BSTJ.
The calling mechanism
had finger wheels corresponding to the Lorimer levers; eight were
provided, however, to allow for two letters, five digits and a party
line letter. The idea was to explore means of storing the called
number in the telephone set so that memory requirements in the CO
could be reduced. The position of the finger wheels stored the
called number; the switch would interrogate each wheel in turn by
activating a small stepping switch in the subset. The subset would
return the digit in the form of two pulses, the spacing between them
corresponding to the stored number. The pulses were generated by
means of small pulse transformers which were switched from one
direction of saturation to the other and back again, just as
magnetic cores are read in computer memories.
Repertory dialers using
solid-state circuitry are now are readily available to store
frequently called numbers. Most of these also store the last number
dialed so that it can be left, pre-set, ready for the next attempt.
The revolution in hand-held calculators shows what can be done with
modern electronics. It seems very likely that the pre-stored
telephone number, either by itself or as part of a repertory dialer,
has hardly begun to be exploited. A light-emitting-diode display at
the subset, a dial tone detector and Touch-Tone out-pulsing (ten
digits in one second) would reduce errors and cut holding time of
registers by more than 80%. This sort of thing is just a step away.
Order wire or CCIS
Signaling between
central offices is similar to user signaling in some instances;
indeed, in step-by-step systems, there often is no difference at
all. Dial pulsing, under such circumstances, is the fastest
signaling method yet produced. By the time the user has finished
telling the system the number he is calling, he is connected to it.
However, as systems grow in size, the nonproductive use of trunk
facilities while the user is dialing can become quite expensive.
Further, alternate routing, limited to the tricks that are possible
with digit absorbing, leaves much to be desired. Thus, common
control systems often get the entire number from the user before
they even start to set up a connection. This delays the user
slightly but, due to the much faster MF pulsing on trunks, reduces
their non-productive holding time.
To go further, it has
been known for many years that, if the called line on a distant
switch and tandem trunk groups on the way could be tested for busy
before the call is set up, elimination of non-productive attempts
would make trunks even more efficient. This has led to common
channel interoffice signaling, or CCIS. CCIS is particularly
suitable for allowing the common controls of electronic switching
systems to communicate; not only can they send supervisory and
signaling information, but they can handle traveling class marks
(there is already an echo suppressor in this built up connection)
and traffic management information so that calls can be routed
around congestion.
CCIS has been here
before, however, at least twice. The "Law System," in New York, when
it replaced telegraph instruments with telephones in the late 1870s,
used a common channel for user signaling. Each user had his own
private wire to the switch and also had a common wire that he shared
with many other users. To place a call, he would connect his
telephone to the common channel, listen to see if it was in use and,
if it was free, he'd tell the operator at the CO what connection he
wanted. The 1879 patent covering this system listed all the problems
of calling by name and claimed great advantages for using numbers.
With this invention of the telephone number, the user would instruct
the operator to connect line 47 to line 53. The operator would call
these instructions to the switchmen who would make the connection
manually and ring the called party. At the end of the call, the
common channel would be used to signal disconnect.
In 1879 there wasn't
much trunk calling, but the patent also mentioned the use of an
order wire between exchanges. This is where the common channel
turned up again, long after the difficulties of educating users and
the competition of new methods had made it obsolete within an
exchange. In large centers such as New York, where separate A and B
boards were needed since only a negligible proportion of traffic was
completed within an exchange, the order wire worked very well. The A
operator would obtain the number from the caller and would then
operate the order wire key to the called office. She would repeat
the four digits of the directory number and the B operator would
choose the trunk to be used since they were all plug-ended before
her. The A operator would then connect the calling line to the
designated trunk, and the B operator would connect the trunk to the
called line. If the called line was busy, the B operator would plug
into a busy jack; some systems had a phonograph record playing into
a telephone connected to the jack repeating over and over, "The line
is busy."
When many A operators
shared an order wire, when many different order wires were needed to
other exchanges, and tandem switching complicated matters, the order
wire again became too cumbersome and "straightforward signaling"
took over. Here the originating operator chose the outgoing trunk
and only passed the called number forward to the tandem or B
operator upon the receipt of a zip tone.
Today's CCIS goes a bit
beyond these early ideas, but it is clearly a giant step in the same
direction.
Some switching concepts
In central office
switching equipment, special problems occur. Historical sources tend
to be of two types: non-technical, where the writer does not know
how the system works, and technical, where the writer is so wrapped
up in the mechanics and circuitry that the general principles are
obscured. Principles, when discovered, tend to be quite simple.
Implementation is almost always complex.
Detecting originations
is a case in point. Most electromechanical systems sit quietly until
a user comes off-hook. Then the line relay operates to inform the
system that action is required. The system responds in any one of a
number of highly complex ways to return dial tone. In No. 5
crossbar, some 2000 relays operate in the process.
There is an alternate
approach, however. Almost all electronic switching systems busily
poll each line and trunk every few hundred milliseconds to see if
they require attention. This polling or scanning goes on night and
day, in periods of both light and heavy traffic. When a whole system
is being operated by one control, this appears to be the best way to
let the control know what is going on as it shares its high-speed
operations among many needs on a time division basis.
Back in the last half of
the 1950s, the scanners built for the Bell System's field trial at
Morris, Ill., and the earlier lab models, seemed like novelties—a
complete departure from earlier practice. They weren't, of course.
They had been preceded by the Lorimer brothers by a good
half-century.
The Lorimer system used
a mechanical scanner for each hundred lines. Each line appeared on
one segment of a fixed commutator, and a rotating brush operated by
means of a friction drive revolved steadily, night and day, wiping
over the commutator segments. When a phone (with the called number
loaded and its spring wound up) came off-hook, it grounded one line.
The brush on the commutator searched for just such grounds. The
ground signal, connected by the brush from the commutator segment to
a relay-like ratchet, caused that device to operate, moving in a
catch to stop the brush's movement immediately. The rest of the
machine would then fling itself into action to connect the calling
line through to the next stage of switching so that pre-stored
digits could be obtained. Then the cutoff relay would be operated
and the scanner released so that it could continue its search for
originations.
The Lorimer system, like
Babbage's early computer, was full of ideas ahead of their time.
Unlike Babbage's computer, however, it worked successfully. A number
of systems were sold in Canada and one in England.
Large motion switches,
such as those used in the Strowger, rotary and panel systems, tended
to be complex mechanically by their very nature. Thus, simpler
switching mechanisms were constantly being sought. North Electric's
all relay system was quite successful, but the crossbar switch
ultimately worked out better. Invented in 1913 by Reynolds at
Western Electric, apparently for use as a line switch, it was
shelved while the panel system was developed. Betulander, in Sweden,
seems to have taken the next step about 1917, and later, the Swedish
Telephone Administration developed a number of step-by-step crossbar
systems for small offices. In 1930, Mattheis, of Bell Labs, visited
Sweden and was so impressed that he managed to revive Bell System
interest in the device. No. 1 crossbar resulted, going into service
in 1938; the peak of crossbar art was reached with No. 5 crossbar a
decade later. But the crossbar switch itself corresponds directly to
the peg switch used by Western Union for telegraph connections in
the late nineteenth century and similar manual telephone
switchboards. These usually had a number of horizontal bars for
interconnection, with the lines to be connected coming in on
vertical bars. To make a connection from an incoming to an outgoing
line, a metal peg would be pushed through the vertical and
horizontal bars to make half of the connection from the incoming
line to the horizontal. Another peg would connect the horizontal to
the vertical associated with the outgoing line.
We could also follow
"translation," one of the basic ideas in modern common control
switching, from Molina's 1906 patent to the panel decoder to
crossbar number groups and the Dimond Ring translator in No. 5
crossbar. Ultimately, however, translation is a process so natural
to computers that, in its current uses, it doesn't seem to be
anything special at all. But prior to computers, it was quite a task
to be able to relate the user's directory number to his line's
position on the switching matrix and vice versa.
Stored program control
is another feature that is natural to modern systems built around
computers. To anyone who has ever watched a player piano, there is
nothing new about the concept, but the nature of electromechanical
switching does not offer many examples of stored program precursors.
The only possibility that stands out is the sequence switch used in
both the rotary and panel systems. A sequence switch consisted of a
shaft turning up to 25 "contact cams" or disks with conducting
material on both sides. By cutting away the conducting material in
various patterns, four springs contacting each cam could be
interconnected in a variety of patterns in each of the 18 positions
a switch could assume during one revolution. Sequence switches saved
a number of relays, but they also recorded the normal and alternate
operating sequences of various switching processes.
One final switching
concept worth reviewing is "call back." In the days of manual
switching, where separate recording and completing operators handled
toll calls, the recording operator would obtain the calling and
called number, make out the ticket and pass it to the completing
operator. The completing operator would then call back the person
making the toll call and set up the connection. The recording
operator, who also handled local calls, had a 24-volt talking
battery; the completing operator, to increase the level from the
telephone set, used a 48-volt battery.
In automatic switching,
establishing a connection was a very complex operation, particularly
with any form of common control. Thus many systems would set up the
connection to a register-sender and then let the register-sender,
once it had obtained the called number, continue the connection from
that point. Dropping a connection through one or more stages of
switching was unthinkable after all the intricate operation required
to set it up.
Then came No. 5
crossbar. Copying an earlier PBX that had not been put into
production, it used the callback principle to the hilt. Two thousand
or more relays would operate to establish a connection through line
link and trunk link frames to an originating register, store the
calling switch position and class mark, test the connection and
return dial tone. Then, when the called number was registered, the
whole thing would be dropped and a brand new connection,
line-to-line or line-to-trunk, would be established. Wasted effort?
Not really; that's what gave No. 5 crossbar its great flexibility.
No. 1 ESS does the same thing, and any switching system using an
electronic matrix follows suit. But again, with electronic
components, the method of operation is so natural that one hardly
notices it. Because connections can be established and changed so
quickly in an all-electronic system, no other approach need be
considered.
Remote concentrators
Perhaps the most
intriguing idea to recur throughout the history of telephony is
remote concentration. It is evident that most lines are idle most of
the time; thus concentration is built into most local CO switching
matrices. A modern line-finder, for instance, will concentrate 200
lines to approximately 20 first selectors, and only about ten
connectors are needed to serve 100 lines. Since only 40 trunks are
needed to connect the line-finders and connectors for 200 lines to
the rest of the switching matrix, it should be possible to locate
the line-finders and connectors close to a group of customers and
save 80% of the pairs to the CO.
Before World War I, such
an approach was apparently used fairly extensively with both
automatic CO switches and manual switchboards. Typically, plunger
line switches and connectors for 100 lines would be located
remotely; all calls would be carried back to the main switch for
completion. In addition to saving copper, this arrangement
simplified trunking between switches. Only the main COs had to have
connecting trunks; the remote concentrators, called "automatic
district stations," were trunked only to their mains.
The disadvantage comes
when the local users have a high community of interest. When each
intraconcentrator call takes up two CO trunks, it doesn't take many
calls to lock out the majority. And it never has been easy to beat
the cost or reliability of plain copper pairs.
Common control equipment
didn't lend itself too well to this approach; the need to ingest and
store a certain number of digits before any kind of path was
selected and the need to associate a register with a trunk for
incoming calls led to problems. Community dial offices (CDOs) were
designed, of course, but not as concentrators. In Europe, mit
Laufer ("with-runner")operation on outgoing calls set switches
in both the CDO and the main at the same time, ultimately dropping
the switch train that turned out not to be needed, but concentrator
development lagged.
The coming of the
transistor suggested that remote concentrators should be given
another chance. The germanium transistors available in the early
days were highly temperature sensitive and there were many other
difficulties. But the potential seemed to be very great. An
electronic concentrator was developed by Bell Labs, but it never
made it to its field trial. While being lowered from the lab on a
floor well above street level, it slipped its moorings, breaking
away from the hoist and landing in the street below. Solid-state
circuits are rugged, but not that rugged.
For a while, foreign
concentrators were imported, and efforts were made to realize the
potential of modern components in large COs. Initially, it took a
large number of lines to make the per-line cost of electronic
control equipment low enough to use.
In the meantime,
business traffic was growing. A PBX is a form of concentrator, but
half its calls are internal. This poses no problem since an access
digit (dial 9 for outside) can separate the concentration function
from the internal calling. Where a step-by-step PBX is served by a
SXS central office, the automatic district stations idea springs up
once again. In 1958, DuPont, in Wilmington, Del., obtained direct
inward dialing (DID) by the simple expedient of running trunks from
a CO selector level to PBX selectors graded in with the tie-trunk
incoming selectors, effectively making a portion of the PBX the
final stages of CO switching.
Bell's No. 101 ESS is an
interesting combination of concentrator and PBX. It features one
large common control in the CO, and a number of small time-division
switching matrices serving individual PBX customers. The small
switches are controlled over data links; it is tempting to think of
the same control also operating the CO switch so that inter- and
intra-PBX calls could all be handled as part of one big system.
Clearly, such an approach would also work in a concentrator/main
situation to provide CDO service.
The remote concentrator
idea has been around for a long time, but it never has quite
overcome the low cost of copper or the high cost of maintenance. The
tide may be changing, however. Electronic switching is now becoming
so reliable that maintenance forces have to be centralized to serve
a number of offices if enough troubles are to be obtained to
maintain troubleshooting skills. A centralized maintenance force
could as easily serve, concentrators as main switches.
And as for the copper,
that too may be licked. Digital Telephone Systems has combined
T-carrier with remote concentration to serve 100 customers on 24
T-carrier derived channels on two pairs. And if this isn't enough,
there is one more step that is sure to be taken in the immediate
future. The remote concentrator requires a large amount of CO
equipment that brings the two pairs back to 100 lines for connection
to a standard telephone switch. With the right kind of switch, this
would not be necessary; the PCM signals could be switched directly
in digital form.
Such switches already
exist. They generally use T-carrier channel banks to provide analog
to digital conversion, some nearby for lines near the CO, and some
remote. Where the channel banks are remotely located, space in the
CO is conserved. Digital PBXs and remote concentrators can be served
in the same way. And, of course, trunking can go direct to higher
level switches; No. 4 ESS is just the first of these to come along.
My guess is that the time is almost right for remote concentrators
to come into their own. I'm on their side.
Telephony is a vital,
exciting field. We've tried almost everything, but each time around,
new possibilities open out before us. The business may be 100 years
old, but "you ain't seen nothin' yet!" Watch out for the next 100
years!
A note on sources
Arthur Bessey Smith's
"History of the Automatic Telephone," serialized in Sound Waves
between October, 1907 and January, 1908, then in the American
Telephone Journal through most of 1908, and finally, in
Telephony through 1909 is one of the better sources. Smith and
Campbell's Automatic Telephony (McGraw Hill, 1914) is also
excellent, as is Kingsbury's The Telephone and Telephone
Exchanges (Longmans, Green, 1915—London). A good biography of A.
B. Stowger, with an interesting description of the La Porte cutover,
will be found in J. Hartwell Jones' Telephony article,
"Industry Honors First Automatic Inventor," October 15, 1949.
Certain articles in the
Bell System Technical Journal were also used. Oscar Myers'
"Common Control Telephone Switching Systems," Nov., 1952, contains a
wealth of information. Other articles include "Tone Ringing and
Pushbutton Calling," by Meacham, Power and West, March, 1958;
"Pushbutton Calling with a Two-Group Voice-Frequency Code," by
Schenker, January, 1960; and "An Electronically Controlled Switching
System," by Malthaner and Vaughan, May, 1952.
The Lorimer System is
described by Smith, Smith and Campbell, and patent No. 1,187,634.
The Law telephony system is covered in Kingsbury; the patent cited
there is No. 220,874.
***
The
section on remote concentrators showed just how history flows on,
with human ideas evolving from one step to the next. Within a month,
several major manufacturers announced their new digital Central
Office product lines (see Telephony, July 19, 1976) including
stations carrier and remote concentrators. It's not all that hard to
predict the inevitable, but it is gratifying to have things work out
so promptly.
[
Top ] [ Next ] [
Table of Contents
] |
