Voice
Communication in Business Volume 1
Essays on telecommunications,
1969-1980
Chapter 20
A Requiem For Step
With all
the wonders of the digital future spread out for study, I had to
return one more time to Step by Step. I kept running into articles
and books criticizing it for all the wrong reasons, and it seemed to
me that the record should be set straight. Sure, Step has to go, but
let's be certain we really know why. If more modern designers knew
what Step could do, we wouldn't have such limited modern systems
with their digital switching and their computer controls. "Requiem"
appeared in the May-June, 1980, Business Communications Review.
***
So it's 1980. 1980? To
those of us who remember the Trilon and Perisphere, and even A
Century of Progress, this is the future. Science fiction time. The
world of tomorrow. And somehow, it's just a little flat. True, the
bread man no longer has a horse pulling his delivery cart, as when I
was very small. And lo, the iceman hath long since goneth. But most
changes have been so gradual that it just doesn't seem like
tomorrow. We have super-highways and slums, pollution and gas lines,
slurbs and drugs; the family, when it exists
at all, huddles around the TV instead of the fireplace, hoping for
warmth that is no longer there. Maybe I'm getting old.
But we have had
progress. Particularly in telecommunications. We've gone through
five generations of switching machine (step-by-step, register-sender
control, crossbar and marker control, reed or electronic switch and
computer control, and all-digital), transmission is greatly
improved, and it costs less to phone than send a letter. Contrast
this with the automobile which is still in its first generation:
internal combustion engine with spring and dashpot suspension under
a creampuff of cosmetics at skyrocketing costs. Yes, some things are
moving ahead.
But why? What leads to
progress? There are those who insist that competition is
responsible, and they claim that competition in the telephone
industry has lowered costs, given us wider choices and improved
service. They cite the current rash of "modern" PBXs as proof,
assuming that competition resulting from interconnect has been
responsible.
Personally, I think this
absurd. It seems to me much more likely that the interconnect
business has been given a great shot in the arm by the telephone
industry, which is engaged in a major effort to eliminate Almon B.
Strowger's magnificent gay 90's invention, the Step-by-Step (SXS)
switch (see Fig. 1). By removing steppers, the telephone companies
are forcing customers to do all sorts of things to retain workable
telephone service at reasonable cost and interconnect is one of the
things they try. Unfortunately, as in most "competitive" situations,
most innovators copy the dominant supplier and the dominant
supplier, having abandoned SXS, does almost everything wrong.
Innovation and step-by-step
Are we really finding
all sorts of new and wonderful services in PBXs with 150 features
like direct outward dialing or station-to-station calling?
Unfortunately, no. Although some queuing and routing features now
available are highly desirable, in general new systems costing twice
as much haven't yet gotten back to where we were with good old
step-by-step PBXs such as the 701.
Note that a 701, backed
with a 607 or 608 manual switchboard, augmented with key telephone
equipment to provide user features, and assisted with a restrictor
to prevent toll abuse, was a formidable machine. A complete
installation was actually a family of dinosaurs: SXS switches,
manual switchboards, key telephone systems and a variety of add-on
boxes. And just as real dinosaurs ruled the earth for 50 million
years, so SXS in all its glory ruled the PBX field for 50. During
that time, it evolved to meet user needs as no direct development
could possible manage.
Now, don't get me wrong.
I'm not advocating everybody rush right out and sign up for a
stepper. SXS is, indeed, a concept whose time has come and gone. All
I am asking is that we kill it for the right reasons. In learned
books and articles, we see many statements about SXS that are
clearly preposterous: you can't use tie trunks two-way with SXS, for
instance, or you can't go off-net via a distant PBX. So what is the
truth? Let's take just a moment to see what SXS could really do, and
then let it rest in peace.
Ten outputs per level
One of the major
objections to SXS is the ten outputs on each switch level. This,
supposedly, limits trunk groups to 10 trunks, maximum. In reality,
there were two ways around the 10 output limitation: grading and
ROTS. Grading let each of several groups of switches have some
private trunks which it, alone, could access, plus other trunks
shared by all. The "private" groups would be first choice, and
overflow traffic from several private groups would share the common
group. For instance, two groups of switches could have five private
trunks each on their first five terminals, and, on the last five
terminals, have one additional group for a total of 15 trunks.
With grading, there is
always the possibility that a call from one group will find all its
private trunks busy and all the common trunks busy even though the
other group has some of its private trunks free. ROTS, or Rotary Out
Trunk Switches, solved this problem. With ROTS, all switches could
have full access to something over 200 trunks in one group if
desired. ROTS were single-motion rotary switches that could connect
one input to any one of 22 outputs. The input would be connected to
a terminal on an output level of a SXS switch group, and up to 21
trunks would be multipled past the outputs (one terminal, I am told,
was left as a rest position). The ROTS in a group serving 21 trunks
would all move in unison, seeking an idle trunk. Upon finding it,
they would stop and wait for a call origination. When a call on the
SXS switch hunted to the terminal connecting to an idle trunk via
ROTS, it would seize the trunk and the rest of the ROTS would go off
together to find the next idle trunk. After preselecting it, they
all would again lie in wait for a new call. With each output
terminal on a SXS group's output level expanded by 21, 210 trunks
could be given full access. The telephone company never liked to
talk about ROTS, but they were a most effective tool in large
systems.
Alternate routing
Another objection to SXS
is its inability to do alternate routing. It could, as a matter of
fact, do various kinds of alternate routing very well. One could
always put FBD WATS lines on the first several terminals on a level,
measured WATS later, and end up with circuits to DDD. This would
tend to pack the FBD circuits, skimming the cream from the offered
traffic where cost per minute decreased rapidly with use, provide
reasonable economies with MT circuits, and offer a good grade of
service with overflow to DDD. By backing off on the DDD and MT
circuits, retries could be used to force higher use on the FBD
circuits. Because loading was truly rotary, management of the
configuration was easy.
But tie trunk alternate
routing, fully automatic, could also be obtained using "digit
absorbing." Consider a small network with the hub at the main
executive offices in New York. The R&D lab in New Jersey often calls
the main hub, but it also has many occasions to call the sales
office in Philadelphia. When it does the latter, it dials "8" to get
to the hub, and then "5" to get Philadelphia, using two tie trunks
in tandem. The communication manager notes that tie trunks from the
New Jersey lab to the Philadelphia sales office cost about a third
as much as the two tie trunks in tandem cost. So he puts in, say,
two direct trunks between these locations.
The users at the R&D lab
are now given "75" as the code for Philadelphia. The direct trunks
are put on the first two terminals of the 7 level, and the rest of
the terminals connect to the same trunks which, on the 8 level, go
to the New York office. When a user dials "7," the PBX's rotary hunt
looks first at the two direct trunks. If one is free, it connects
through to the Philadelphia PBX. The "5," which follows, is ignored
by the incoming selector which is arranged to absorb the first digit
it sees. Actually, it goes up to the 5 level and falls down again to
await the next digit.
Now, suppose both direct
trunks are busy. The switch then hunts on to the trunks to the hub
and, upon finding one, connects through. But now the digit 5 makes
the connection through the hub and back to the Philadelphia office.
In either case, additional digits select a PBX extension; if a "9"
is dialed to reach CO trunks into Philadelphia, local off-net calls
can also be completed. A similar arrangement can be constructed for
calls from Philadelphia to New Jersey.
With digit absorbing and
rotary hunt, a SXS PBX could get good usage on direct trunks and
insure a good grade of service by letting overflow traffic join with
the remainder of the traffic to the hub. This hub traffic might
terminate on local extensions, go off-net into the city, or go on to
other tie trunks.
Satellite systems
A powerful SXS trick was
to use one switchboard for several reasonably close satellite PBXs.
One group of attendants at a 701's switchboard insured efficient
operation, particularly after hours. At the remote locations, 711
PBXs (just like the 701 but without switchboards) would be used.
Incoming calls would be handled at the switchboard; the attendant
would plug into the multiple to reach local extensions, or into a
trunk to a satellite location. Then two or three dialed digits would
select the desired extension. Note that the satellite trunks would
also appear on SXS switches at the hub which could concentrate
traffic for long-haul tie trunks as well as permit desk-to-desk
dialing in the local area. Attendants could also complete incoming
calls via long-haul tie trunks if desired. Many dial tandem networks
started as satellite systems.
From the remote PBXs, it
was desirable to have users dial "0" for the switchboard attendant
(operator), just as was done at the main PBX. But this posed a
problem because, to access the hub for direct dialing to other
extensions or access to tie or CO trunks, the digit "8" might be
used. Clearly, it would not be economical to have a second group of
trunks just for dial 0 traffic. The solution was simple: use a pulse
adder. Trunks on the 8 level would also be multipled via pulse
adders past the 0 level at the remote PBXs (see Fig. 2). Thus the
user, upon dialing 0, would arrive at the hub where the attendants
were located. However, to access the attendants, the pulse adder
would come into play. After the user had selected the 0 level, the
adder would insert a single pulse into the trunk as a second digit.
This pulse would cause the trunk's selector at the hub end to go to
the first level. and hunt for a path to the switchboard. Since 1
levels were almost always left free on selectors (the old
candle-stick phones often produced a false pulse when being taken
off hook), this level was usually available for switchboard access.
Switching speed
In certain special ways,
SXS is still the fastest switching system available. This speed is
associated primarily with dial tandem networks, where each trunk is
terminated in its own private selector at each end. If you can seize
a trunk in the first place, you can signal into it because the
selector at the far end, operated by the first digit into the trunk,
is part of the trunk itself. With a small tandem hub, you could make
a through connection with one digit and no dial tone delays.
If an ESS Centrex
replaced the PBX, all this changed. Upon accessing the ESS, you had
to wait for dial tone signifying that the ESS was ready to receive a
digit. Then, three digits would be required (1XX) to identify the
desired outgoing tie trunk group. Finally, you had to wait for dial
tone again, this time from the distant PBX, so that you would know
that the ESS had completed the path. When people tell me about the
speed of electronic switching, I sometimes laugh.
As long as you had SXS
all the way, you were connected to the called telephone as soon as
the dial returned to normal after the last digit had been dialed.
There was no "post dialing delay" for all practical purposes. After
a change from dial tandem to CCSA or EPSCS, everything immediately
slows down. At the user's completion of the last digit, the first
tandem switch starts to set up the connection. It may take 10 or 15
more seconds to complete via several hubs to the called extension on
a distant PBX. User training is required to prepare people for this
slow-down. Otherwise, they may hang up just as the call is about to
go through. Obviously, an electronic switch is actually faster than
a SXS switch. But if the hare doesn't even start the race until the
tortoise has already crossed the finish line, he can't win.
Matrix flexibility
Perhaps the main
advantage of SXS switches was their flexibility. Step is still the
only kind of switching system that can grow to unlimited size,
assuming you have the floor space available. But there was other
flexibility as well. This showed up in any kind of network, whether
it was just a local PBX, a satellite system of the type described
above, or a long-haul tie trunk network. Because hubs were usually
placed at major locations that originated and/or terminated the most
tie trunk traffic, and the hub switches were part of the PBX itself,
access lines of the type needed for stand-alone tandems could be
eliminated. This lack of access lines at each major location saved a
lot of money and helped prove in the network. In small networks,
outgoing tie trunk switching would be "free": tie trunks would
simply be accessed via unused levels on the first selectors: 2 and 3
might be the first digits of local extensions, and 4 through 8 might
pick up trunks to distant locations.
For larger networks, it
became necessary to "split" levels. Thus the hub, as well as the
remote satellites and more distant tributaries, would dial "8" to
reach the tandem selectors, and then dial one or two digits to pick
remote locations. Often the digit "2" on the "tandem firsts" would
reach back to the hub extensions for access by remote locations, 3,
4 and 5 would be reserved to reach other hubs or major locations,
while the higher numbered levels would go to a second rank of tandem
selectors. One might dial 62, 63, 64, 65 and 66 to reach five
smaller PBXs in the nearby area. These might use two or three digit
numbering plans internally, as opposed to 4 digits at major
locations, so that dialing two digits at the hub would tend to keep
the number of dialed digits uniform.
With a little care, one
could design tie trunk switching to work at very low cost and with
very simple dialing rules; for on-net traffic, one seldom needed as
many as seven digits. Further, addition of tie trunk switching would
not affect local switching at all; the switching matrix could be
built out selectively at exactly the point where more capacity was
needed, leaving the rest untouched.
Tie trunk traffic
capacity could be added as needed; for a single stage of tandem
switches, there was no matrix blocking. Even with two or three
stages of tandem switching, proper design could keep tie trunk
switching very nearly non-blocking. With regard to local traffic,
SXS could be arranged to carry very heavy loads if necessary.
Normally, 20 line finders are provided for 200 extensions; that
permits 10 percent of the extensions to be originating calls at the
same time. However, with very heavy traffic, additional line finders
can be added. In New York, I have seen as many as 33 line finders
for each group of 200 lines. At the other end of the matrix, one
might find only two groups of seven connectors for each 100 line
subgroup. Less than half as many terminating switches as those used
for originations. How come?
The answer is easy.
Incoming calls from the outside world are completed through the
switchboard directly to the called line, bypassing the automatic
switching. Thus, the switch can be much smaller. When a console is
used, whether in connection with some SXS systems or more modern
equipment, it can only direct the automatic switching which actually
does the work. Thus, the automatic switches must be able to handle
at least 25 percent more traffic than when an old fashioned cord
board is used. And, even so, the call is not set up any faster.
Cord boards
Cord boards were great,
too. They put the most sophisticated common control ever known in
charge of the system: a human being. The attendant could handle
incoming calls, assist with outgoing calls (including route
selection and call detail recording), and take messages. The
attendant would handle transfers (flashing recall made the cord lamp
flash continuously if the station user flashed just once) and
perform other services.
Because these things
could be done easily and quickly by experienced personnel, but could
only be done slowly if at all with the first generation of consoles
no matter how skilled the attendant, station transfer was invented
to reduce the console load. (Remember, consoles as of 1970 often
required dialing a connection with a rotary dial which took a lot of
time.) Station transfer implies that the time of company executives
is less valuable than that of the console attendant.
Cord boards were
particularly useful in controlling toll abuse. One common technique
was to have the "dial 9" trunks used one-way outgoing and toll
diverted at the central office. When the CO detected a toll call
(mainly from the need to keep a toll record), it would reject the
call by sending a reverse battery or other convenient signal down
the trunk to the PBX. At the PBX, the trunk circuit would call in
the attendant at the cord board. The attendant could then place the
call if necessary.
While doing all this, it
was not too hard to keep a toll record. just having a human ask for
the calling extension was usually enough to discourage abuse. Toll
records often used "mark sense" computer cards (marked with pencils
using magnetic lead) so that machine reduction of the call records
could be handled automatically. When the attendant dialed the call
herself, she had to have the called number and the calling
extension. The digital clock, a later switchboard feature, made
recording start and hang-up times relatively simple. Now that
digital clocks are the only kind of console clocks available, we
tend to forget how difficult it is to record toll tickets accurately
with "the big hand and the little hand."
When WATS lines became
available, it was just as easy to wire the switchboard to select the
proper facility as it was to put trunks in the right order on a
switch level. Further, the presence of the attendant and the ability
to keep call records made switchboard access the best way to go. The
approach was to modify the Idle Trunk Indicator (ITI) circuitry so
that it would select the least expensive trunk for the attendant.
Back in the days when it made sense to have Band 3 WATS lines, for
instance, one would put Band 3 Fulls first, followed by Band 5 Fulls,
followed by Band 3 Measured followed by DDD. The ITI lit a lamp
above the jack of a trunk; normally it was arranged to equalize the
load on each group of five trunks, but it could be arranged to
always light the lamp over the left-most idle trunk in a jack strip
containing 20 jacks. Equalizing the load on full and measured WATS
lines, then as now, defeated the whole purpose of "least cost
routing."
One could do even more.
At East Coast locations, time clocks could busy out the Band 5 fulls
in the Band 3 jack-strip when the West Coast people came to work (11
a.m. EST) and the Band 5 WATS lines could be used more productively
for longer distances (obviously, trunks could appear on jacks in
several different strips.) Provision could also be made to busy out
measured WATS in the evenings and on weekends. How many computer
controlled switches can do as well today? More to the point, when we
could do all this sort of thing 15 years ago, why do those
re-inventing the wheel spend so much time patting themselves on the
back?
Restriction
Toll diversion was a
handy tool, but it was often not suitable. When many toll calls went
very short distances, the attendants would be swamped. Further, when
multi-message-unit calls cost more than toll and require a 10-digit
number to be dialed (as in New York City, for instance), "local"
calls might need more control than tolls. Thus "restrictors" were
invented.
Restriction, unlike toll
diversion which is a CO feature, is a customer premises feature.
Originally, restrictors would monitor outgoing dial pulses and, if
the first three digits indicated a permitted call, they would drop
off. If, however, the user dialed a forbidden CO, the restrictor
would send the same kind of signal that toll diversion produced. As
DDD grew, three-digit restriction (on the office code or area code)
turned out to be insufficient and six-digit restriction was made
available. This gave the customer a means for tailoring the
permitted calling area fairly well, deloading the switchboard of
short-haul toll calls or, as the case might be, forcing
multi-message-unit calls through the attendants.
There was a problem,
however. Restrictors never knew who was placing a call. All they
could see was the dialed digits. To permit some unrestricted
stations, a trick very similar to that used with the pulse adder
could be applied. Any station dialing "9" to get out would have to
pass through the restrictor to the outgoing trunks. Restricted
stations could dial "9," but not "8," Unrestricted stations,
however, could dial 8 and access the outgoing trunks directly,
bypassing the restrictor. After many years of insisting, customers
have finally convinced some of the manufacturers of computerized
switches that this is a good feature. Certainly it is infinitely
easier to do with modern equipment, and extension class marks make
dialing "8" versus "9" unnecessary, but one should never assume that
it took computer control to provide the service. If the computer
types really want to show their stuff, they should stop copying the
6 digit restriction of the relay age and go after 7 or 10 digit
operation to block calls to "trash numbers," home phones, etc.
Stand-alone toll routers make a big thing of being able to do this.
To further simplify the
work of the attendant, toll recorders and, later, toll routers with
built-in recording capability came on the market. It is surprising
to see how many of these add-ons are used with the most modern PBXs.
It makes a certain amount of sense to have a stand-alone recorder
that accepts buffered information from a modern PBX, but it is hard
to understand the need for additional routing capability even on a
stepper, much less a Dimension 2000. Steppers could not provide
automatic queuing, of course, as the Dimension and other PBXs can,
but then, some stand alone call routers can't queue either.
Key systems
In the time of the
stepper, it was quite common to find about 90 percent of the PBX
stations terminated on key telephones. These key telephones provided
the required station features, almost totally independent of the
PBX. The user, in the middle of a call, could push the "hold"
button, select a second line by depressing another line pick-up
button, dial a new call (or answer an incoming call), consult with
the person on the second line, and then return to the first party,
either hanging up the second line or putting that line on hold with
the hold button. Further, it was always possible technically to
arrange the set to allow two line buttons to be pushed down at the
same time to make a conference call (at the expense of speech
volume).
With such features
available, the Series 300 features (consultation hold, three way
conference and station transfer), brought over from residential
service field trials, were not _viewed as much of a triumph by PBX
customers. Indeed, due to the complexity of operation, these
features were a giant step backward. Unfortunately, with crossbar
PBXs, popular in the late 60s and early 70s, consoles with rotary
dials, as mentioned above, were painfully slow. Thus, station
transfer was necessary to deload the attendant console, and the
other features, all being part of station transfer, came along for
the ride.
In crossbar PBXs, these
features were built into the incoming trunk circuits. Therefore,
they could be used only on calls coming in from the CO. Not being
available on intra-PBX, outgoing or tie trunk calls created
something of a problem. When computer control came along, designers
of at least one electronic PBX continued to associate Series 300
features with incoming calls only. Mysterious are the ways of the
great innovators.
With regard to
conferencing, the most common practice was simply to have someone
pick up the line at its appearance on another phone. By arranging
pickups properly, most conference needs could be handled easily.
When the high cost of Centrex and "modern" PBXs forced the wider use
of single line sets, loss of this ability to conference conveniently
and easily was a problem that designers never suspected.
The one way in which a
SXS PBX interacted directly with key telephone sets was through
hunting. In a stepper, all lines (compared to 30% in some
computer-controlled electronic switches) could be placed in hunt by
strapping certain terminals; extension numbers did, however, have to
be consecutive and different only in the last digit. The general
idea was to have the boss on one number and his secretary on the
next higher number. Boss and secretary could pick up both lines on
key telephones; if the boss's line was busy, the stepper would
deliver the call to the secretary's line. The secretary would answer
the call and, if it was urgent, she could put it on hold, use the
intercom to contact the boss, announce the call, and let him decide
whether to accept or decline.
Secretarial screening of
all calls was a major feature of this approach. Often, the boss
would not even have a bell on his own phone; he would just have a
buzzer on his intercom. All calls would be screened by the
secretary; the boss would never be trapped with an unwanted call, or
disturbed in the middle of a face to face meeting or a conference.
Obviously, this approach
cannot be used with the single line instruments recommended with
expensive "modern" PBXs. And camp-on, call waiting, and automatic
callback cannot be used effectively with multi-line instruments.
Which line do you camp on? Is the boss on the secretary's line and
is the secretary handling a new call on the boss's line? The
telephone industry has gone to a lot of trouble, through public
relations, advertising, training, etc., to try to eliminate
secretarial screening so that single line instruments and their
"modern" features can be used instead of the pick-up, hold and
announce features we have all grown up with. But it isn't easy to
warp the way America does business just to cover up a dumb design
decision. Designers lament the fact that users just aren't
sophisticated enough to take full advantage of their wonderful new
systems when, in actual fact, they themselves are not sophisticated
enough to understand the basic human-factors requirements of their
jobs. When I see all the time and money wasted in training and
retraining people to use these new systems, I wonder just how
expensive the old "intuitive" key telephones really are.
Northern Telecom
recognized the problem and designed the SL-1 around a key telephone
set that took full advantage of the PBX memory and control. Danray
offered a multi-line set with uniform station wiring (3pairs) as an
option. Bell added an electronic telephone set to Dimension (one of
the few bright spots in an otherwise discouraging 2-wire analog
system), and built Horizon around a set that is almost identical.
American Telecom's Focusphone is a standard two pair electronic key
telephone set from a stand-alone key system, with the key system's
common equipment replaced by the PBX. How long before the rest of
the industry catches up and provides at least SXS quality service
with their computer controlled dingbats remains to be seen.
One of the amusing
aspects of key systems versus single line sets behind PBXs is the
way key reduced the number of extensions required. Actually, three
lines on five key telephone sets will provide better service for
five people than five single line instruments. One might suppose
that five lines for five people would be non-blocking, but this is
true only for outgoing calls. Incoming calls are different. Three
lines in hunt, all available on all instruments, with intercom to
help in announcing someone else's calls, make service easy and fast.
Just try to provide comparable service with single line phones, with
or without hunting.
Hail and farewell
If SXS was so great, why
is it being phased out? Surely not just to make room for
interconnect. Well, it has its faults to go with its good points. It
is noisy, for instance, and has to be isolated in its own room.
Further, it gets to be quite large, physically. Thus, floor space is
sometimes a very real problem. And, in spite of all the tricks well
known to those of us who love it, it is not as flexible as a
properly designed modern PBX could be.
But the main reason why
SXS should pass into the great switch-room in the sky is that it has
outlived its era. It was designed for a different era, when people
were less expensive, and tender loving care from interested humans
was readily available. Today, young people do not want to train for
obviously dead-end jobs at manual switchboards, electromechanical
switch maintenance, or ancient key system manipulations. And they
are right.
Step has to go, but
let's not kid ourselves that call forwarding or automatic call back
did it in; its passing is making a market for modern PBXs, in spite
of their often silly features. Unfortunately, however, most "modern"
PBXs are no better than step at handling the needs of the future.
What, then, can SXS leave us as a heritage to prepare us to live in
the world of tomorrow?
First, it will continue
to stand as living proof that evolution works better than
revolution. Let's hope that modern electronics with stored program
control proves as adaptable as Strowger's switches, and chooses to
adapt itself to meet user needs rather than designer whimsy. Second,
SXS proves that almost anything can be made to work if you know what
you want do; this contrasts sharply with many modern PBXs that do a
variety of the wrong things very well indeed, but omit needed
features. Finally, the limitations of step show where we must take
advantage of new technologies: we need 4-wire switching with digital
end-to-end transmission as soon as possible; we need telephone sets
that match better with non-voice signals to say nothing of digital
transmission, and we need far better data storage and manipulation
to handle present and future information transmission as well as
telephone management and control. At the moment, however, I'd settle
for just a little less self-adulation from all the "innovators" so
busily reinventing Strowger's wheel.
[
Top ] [ Next ] [
Table of Contents
] |
