Voice
Communication in Business Volume 1
Essays on telecommunications,
1969-1980
Chapter 16
Some Odds & Ends
The
following article, from the November-December, 1979, issue of
Business Communications Review, contains a variety of minor topics,
some of which are of major importance. For the benefit of the
historical record, Harry Newton, mentioned in the opening paragraph,
is one of the more interesting consultants working in the field
today. He writes a column in BCR, does a smash-bang seminar, and
runs the Telecom Library.
***
For some time, I have
had a number of thoughts (and second thoughts), witticisms, gripes
and the like floating around in my head. Unfortunately, none of them
is substantial enough to warrant an entire article in a prestigious
journal. (Ouch! You're hurting my arm!) So I've decided to get them
all off my mind (or is it my chest?) at the same time. In what
follows, then, I ramble on about a variety of unrelated topics,
Harry Newton style.
MISCELLANEOUS UPDATES
You win some and you
lose some. During the past year, I have written a couple of things
that need correction and, at the same time, I have thought of a few
things that I wish I had said but didn't. So now is the time to set
the record straight.
Inexpensive electronic telephone sets
In "Tutorial on Two
Toys" (Chapter 15) I lamented the lack of
certain features in automatic dialers: the ability to send a
switchhook flash and the ability to detect dial tone. My point here
was that automatic dialers could easily send feature codes and could
convert regular single line telephones into electronic telephone
sets that would be easier for the customers to use.
Turns out we don't have
to wait for the toy makers to see the light. Any repertory dialer
that can send dial pulses is ready to go already, as long as it is
used with the Womack PBX. The Womack PBX, sometimes known as the ITT
CS 1024, appears to be unique in that it was designed to read dial
pulses in its line circuits. Further, it is always monitoring its
lines for dial pulses while a call is in progress, even when DTMF
phones are used. It doesn't need a switchhook flash to alert it to
future dial pulses.
What I did, with the
help of the local ITT distributor's Bill Buckley, was simply to put
the Womack's single digit feature codes into a Radio Shack automatic
dialer. With a call in progress, I could then push a button on the
dialer labeled "hold" and put the other party on hold. The system
would then give me dial tone and I could call up a second party. By
pushing a button labeled "conference," all three of us would be tied
together. In the Womack, "hold" is the digit 3, and "conference" is
the digit 2.
But there are lots of
other feature codes in the Womack. The digit 6, for instance,
invokes call forwarding. I programmed in a 6, followed by the four
digit extension number of the secretary serving the principal who
was letting me play with his phone. Now, at the push of a button,
incoming calls would be diverted to the secretary. The digit 1
cancels call forward. Needless to say, I programmed it onto another
button.
Actually, I programmed
in several call forwards; one to the secretary, one to another
secretary, and one to the console. But to allow for a perfectly
general call forward, I put a 6 on one key all by itself. It turns
out that the Womack programming automatically attaches a DTMF
"decoder" to the line when call forwarding is requested, so that the
DTMF key pad can be used to key in the extension to which a call is
to be forwarded, even though the 6 was sent in dial pulses from the
repertory dialer.
Call pick-up works the
same way. I put in a 5, plus a four digit extension number. Using
several keys, I arranged for single button answering of each of
several different lines. With the particular programming in the
system, a dial pulse 5 all by itself can be followed by an extension
number from the DTMF key pad.
There was just one
problem. The Womack can only read dial pulses at the rate of 10 per
second. The fast dial-pulsing option available on the Radio Shack
dialer wouldn't work. For a ten digit number, this can be a pain,
but for four or five digits at most, it isn't bad.
There may be other PBXs
on the market that can read dial pulses without a preliminary
switch-hook flash. If so, they will work with any repertory dialer
that simply sends feature codes as trains of dial pulses. To work
with a code plus an extension number, as with forwarding and
pick-up, they may require either rotary dial phones or programming
as in the Womack that connects a DTMF detector (or decoder) whenever
needed, even when dial pulsing comes first.
Hopefully, manufacturers
of repertory dialers and PBXs will take note of this experiment.
Repertory dialers in quantity should cost maybe $50 or less; this is
quite inexpensive when compared to some of the "electronic"
telephone sets already on the market that don't do a heck of a lot
more.
Before we get too
carried away, however, it is important to note that all I have
demonstrated with my little field trial is that system features can
be implemented much more easily with a repertory dialer than with a
flash and feature code routine. I have not demonstrated that the
important features, those associated with real key telephone
systems, can be done the same way. And, of course, they cannot. With
key systems, either traditional or electronic, a visual display is
provided, and the station user can select one of several lines or an
intercom on which to conduct a conversation. Even with my little
dialer, I can't do that. The real electronic key telephone with line
pick up buttons and good visual displays is still required at many
if not most stations in a business environment.* .
[* FOOTNOTE: In 1980, several
repertory dialers designed for feature code use hit the market.]
The dark side of telephone toys
Alas, not all my field
trials are so successful. My other automatic dialer died on my desk.
It died quietly, and I didn't even know it had passed away. I had to
find out the hard way. On a Saturday morning, I lugged myself out of
bed and went off to an early meeting, only to find myself alone.
When the guy I was supposed to meet finally showed, he apologized,
but pointed out that my phone was busy. I called it immediately, and
sure enough, it was.
Upon returning to my
office, I found the trouble and disconnected the dialer. The
"permanent signal" went away, and my phone was ready to receive
calls again. But my hundred dollar toy was dead: no stored numbers,
no nothing. And it was putting a short on my telephone line, faking
a call origination, until I pulled it off.
I took it back to the
dealer. He listened to my sad story and said, "It was probably
struck by lightning." But he wouldn't give me a refund. The warranty
calls for the gadget to be sent back to the factory for repair, and
he said I could do it as quickly as he could.
Reading the owner's
manual again, I discovered an 800 number to call in case of trouble.
The bright young lady who answered, listened to my- story and said,
"It was probably struck by lightning." This unanimity of opinion
seemed to me to indicate that my problem was not uncommon. When I
pointed out that telephone equipment was supposed to be reliable in
the face of known conditions, the bright young lady told me that
since some dialers had not been damaged, the design was perfectly
reliable.
Let this be a lesson to
us all. For all its faults, the telephone industry, from the very
beginning, made a tradition of designing really reliable equipment.
Since design is not taught in engineering schools and few people
have a chance to learn it on the job, the toymakers invading the
telephone business simply have no way of knowing how to do their job
properly. Just because telephone equipment manufactured by most of
the traditional telephone supply groups is reliable is no reason to
assume equipment manufactured by others will follow suit. And if a
failure causes us to call the phone company with a trouble report,
the phone company is perfectly justified in charging us when they
find our toy is at fault.
I have sent the dialer
back to the factory to see if they can fix it under warranty. But I
am not sure I will trust it back on the one and only telephone line
that I depend on for business and personal communication.
Using single line phones
I have frequently
pointed out, based on the experience of those who don't believe me,
that single line telephones cannot replace multi-line instruments in
the great majority of business situations. Visual cues, the need to
handle three or more lines, and the several intricate operations
involved in secretarial screening all militate against success in
such a pie in the sky dream.
However, if you are
determined to try it, let me offer a couple of suggestions in
addition to the repertory dialer trick. First, either eliminate
hunting or arrange for it to be avoided for internal calls and calls
from the console. This will allow camp-on, call waiting and
automatic call-back to know which phone to deal with. It is always a
strain for the system to decide whether to hunt to the secretarial
phone when the boss is busy, or camp-on, call back or whatever. And
what if both phones are busy? With multi-line traditional key
phones, you don't know whether the boss is on the secretary's line
or what. With single line phones, you know, but you may not always
be able to do something about it.
There are two
arrangements that help out here. One is the approach that allows
hunting if you dial a pilot number but does not allow hunting if you
dial an extension number. Another allows hunting on direct in-dialed
calls or console calls, but not on internal calls (note that
call-back only works internally). With call waiting, the secretary
has a chance to reach the boss when an outside call hunts past his
busy line, but the procedure is a lot harder than using a hold
button and intercom.
Another big help with
single line phones is the proper design of call forward all calls.
CFAC has to allow the receiver of forwarded calls to call the
forwarding line. Only a few PBXs do this, but they give a big
advantage in secretarial screening. When the boss doesn't want to
answer his phone, he simply applies CFAC toward the phone of his
secretary. She gets his calls but, if one needs his attention, she
can flash to put the call on hold and consult with him, as always,
and transfer the call to him if he wants it. Further, when the boss
wants to answer his calls, he just cancels CFAC, a procedure he
could not follow if, as is often done, his secretary's number is
listed for him and his number is unlisted.
The many PBXs designed
on the theory that CFAC should simply allow you to have your calls
follow you as you visit the offices of your associates (even though
security requires you to return to your office after each visit to
change the forwarding), do not permit this. And, when you don't get
to the office intended, the person you started to visit cannot call
you to complain. But "standard" CFAC can be even worse when it works
in the intended way. You used to visit in person so that calls could
be held and the meeting could be uninterrupted. Now, the wonders of
modern telephony assure interruptions by calls normally intended for
the visited office plus the calls chasing you. And even if the
person you visit does you the courtesy of forwarding his calls to
his secretary, in most systems your calls forwarded to his phone
will continue to ring it.
Make no mistake about
it: the main purpose of CFAC is to facilitate secretarial screening
with single line phones. We still don't have any visual cues or the
ability to handle three calls at a time, but at least we have a
fighting chance to screen calls in some circumstances.
Companding, the mu-law and the SL-1
In one of the
interminable parts of "Those Awful PBX Proposals (May-June, 1978,
Chapter 9 in the present volume), I made
the statement that Northern Telecom's SL-1 uses the European (CCITT)
version of T-carrier. Northern Telecom's H.E. Theloosen wrote
immediately to point out my error. SL-1 does, indeed, use part of
the CCITT format, but not all. That is, it uses 30 voice channels
plus two signaling channels as a basic building block, like CCITT,
but it uses the so-called mu-Law in its companding as in American
T-carrier. Thus, individual voice channels are compatible with
American T-carrier as they would not be if the CCITT A-Law
companding plan were used. Further, by selecting 24 of the 30
channels to interface an American T-carrier line, SL-1 can be just
as compatible with the ultimate all digital network as any of the 24
channel switches. SL-1 can also be compatible with European
T-carrier by simply changing its companding approach, while PBXs
using 24 channel building blocks cannot. This is an example of the
basic scientific principle that allows you to cut a string shorter
but not longer. But the important thing here is that SL-1 is
compatible with American T-carrier.
Did I hear somebody ask,
"What in the 0*!$ is companding?" Of course not, but I will explain
anyhow. Companding is a technique, used for years in analog as well
as digital carrier systems, to make the signals we want to transmit
override the noise that is normally present in transmission
channels. The approach in analog systems is to make weak signals
louder prior to transmission so that they can swamp any noise they
may tend to pick up, and, then, at the receiver end bring them back
to their proper level. Mid-range signals are usually left alone in
such systems, and very loud signals are compressed for transmission
so that the amplifiers will not be overloaded. In any event,
companding compresses loud signals for transmission and expands them
at the far end. Conversely, it expands weak signals for transmission
and compresses them at the far end.
PCM digital systems,
however, are supposed to avoid noise by converting a signal to a
digital word, passing this word end-to-end intact, and then
reconstructing the signal. By regenerating pulses (or the lack
thereof) rather than amplifying the signal (and whatever noise
happens to be riding on it), there is little opportunity for noise
to build up. Sounds like utopia, right? But we never get to choose
the most best — only the least worst. PCM encoding makes a different
kind of noise all by itself. This noise is called "quantizing
noise."
To convert an analog
signal to a series of digital words, we have to make an
approximation. We measure each sample, and then code it into a
particular binary word that defines a range of signal amplitudes
between two thresholds. At the far end, we reconstruct the sample to
the exact center between the two thresholds. Thus, any given sample
at the receiving end can be off by as much as half the distance
between the thresholds. This is the "quantizing" error.
To minimize the effects
of the quantizing error, we can make the difference between
thresholds very small. With 8-bit coding, we can define 256
different values. This many discrete levels should let us have very
small quantizing errors and, indeed, it is almost impossible to tell
the difference between the "grainy" digital signal and the "smooth"
analog signal it represents. But there is a problem. Not all sounds
we want to transmit are of equal amplitude. A shout is as likely as
a whisper in a telephone conversation, and a couple of hundred
levels handling the full range of a shout will leave only a few
levels (near zero amplitude) to express the full variation of a
whisper. Something has to give.
In the earliest version
of T-carrier, analog corn-panders were used to change the signal
level to make it more suitable for encoding. However, the difficulty
of making analog companders track at the two ends of the connection
was a problem and an all-digital approach was sought. The technique
ultimately adopted was to make the coding levels unequal — small
(with more levels) near 0 and large (with fewer levels) near the
positive and negative maximum values to be encoded. This is shown in
Fig. 1. Thus, a small signal would have a quantizing error that was
about the same percentage of its amplitude as the larger quantizing
error would be to a large signal.
There is only one
remaining complication. The spacing of levels is different in
America and Europe. In America, we use what is called the mu-Law,
while Europe uses the A-Law. This says that when we have digital
trunks across the Atlantic, gateway toll offices will have to have a
translator to make the conversion. At present, of course, analog
transmission is used both via cable and satellite; the signal is
decoded from one kind of T-carrier at one end and recoded into the
other kind at the far end.
All this is lots of fun,
but the ultimate payoff can be lots of trouble. Someday, we are
going to have an all digital network which permits us to send 8-bit
bytes from anywhere to anywhere at the rate of 8,000 per second.
This is just how T-carrier operates, and fortunately it fits
perfectly with the ASCII code which uses seven bits, usually with an
added check bit, to represent each character. When that day comes,
data will enter the T-carrier bit stream without the modems which
are needed in analog systems. However, when it goes through a mu-Law
to A-Law converter, it will be converted to garbage. The technical
types shrug this off — all we need, they say, is traveling class
marks. Maybe so. But it is more likely that digital signals may
simply be barred from the public network forever — unless they
continue to be converted to whistles through modems. After all, a 50
Kbps channel in analog terms uses a 12 voice channel "group," and we
are conditioned to think it should cost a lot more than a dial up
path through the public network.
SL-1 and 4-wire switching
In the same article, I
also pointed out that the SL-1, like all digital switches, is 4-wire
internally (separate paths for each direction of transmission), but
had no 4-wire tie trunk circuits. Almost as soon as the magazine
came out of the mail chute, I got a call from an SL-1 distributor
demanding a retraction. I told him that nothing would make me
happier than to know that the SL-1 could switch 4-wire, and if he
would give it to me in writing I would be happy to retract. About an
hour later, I got a call from one of the distributor's assistants,
somewhat sheepishly admitting that I was right. Two-wire tie trunks
only.
Well, I am now very
happy to say that the SL-1 has had 4-wire tie trunk circuits for
about a year, and they are selling like hotcakes. Northern Telecom
is happy, the customers are happy, presumably the distributor who
called me is happy, and I, as I say, am happy. It turns out that it
was letter I wrote in January, 1977, to Northern Telecom that gave
the marketing types a little extra ammo to get engineering to
recognize the obvious: 4-wire tie trunk switching is an absolute
necessity, particularly in a 4-wire machine.
SOME
TELEPHONE HUMOR
One seldom thinks of the
telephone business as a source of humor. The chaste white Bell on
the somber blue background just doesn't give rise to laughs. But
there is humor to be found if one digs. There is the famous Dimond
ring translator in the No. 5 Crossbar CO switch, named after Tom
Dimond who used something much larger but quite similar to the
magnetic cores of a later decade to translate equipment numbers to
directory numbers for Automatic Message Accounting. Then, there is a
circuit in No. 1 ESS that generates two unequal pulses — it has two
different time constants. I am very pleased with myself for naming
it the "Two-timer." But I only dared do it after enlisting the
support of my then department head.
But possibly the worst
was reported in Business Communications Review in the
January-February 1979, issue. We were treated to a detailed
description of an on-site call processor that sits, figuratively, at
the rear of a PBX and processes call data. It is called, naturally,
the Ascend processor, by Bitek. If you can dig that, you can dig
anything.
Things are seldom what they seem
We have already noted
that CFAC has, as its most useful function, the handling of
secretarial screening on single line phones. But there are also lots
of other things that are not what they seem. Speed calling, for
instance, does not save enough time to be worth the effort but is a
dandy form of restriction to limit toll abuse. And restriction,
surprisingly enough, is a pretty good form of least cost routing.
And least cost routing is, or course, a misnomer.
Keep looking at all the
"modern" features with a critical eye. Many of them may turn out to
be useful, but almost certainly not in the way their designers
planned. If you can't find a few new approaches on your own, the
joke may be on you.
Network harms
The National Academy of
Sciences, the FCC, the telephone industry and just about everybody
else under the sun has been concerned with interconnected telephone
equipment inflicting "harms" on the public network. I have already
given one example of a very real harm inflicted on the network by
interconnected equipment, but many of the "harms" that have caused
so much worry seem to be somewhat less than realistic. Let me cite a
particularly absurd instance.
There is an organization
I know that operates a private microwave link. The equipment was
made by Lenkurt, a subsidiary of General Telephone. The organization
has a large PBX, made by Automatic Electric and supplied by the
local operating company, both GTE subsidiaries. But to protect the
telephone company (GTE) from the evil interconnect company (GTE), a
special interface must be provided on each microwave trunk. These
interfaces are made by Wescom, a company heavily into the
interconnect business.
High speed signaling on CCIS
One of the funniest of
the items in current telephony is the high speed connections that
Common Channel Interoffice Signaling (CCIS) is supposed to produce.
CCIS is supposed to be able to set up calls all the way across the
country in two seconds or some such. Just why this is so important
is hard to understand since there is a great deal of evidence to
prove that something less than one call in four during the business
day actually reaches the person for whom it is intended. But even if
we could get calls to the right person every time, CCIS wouldn't
help much. During the day, the majority of calls are going to PBXs
and PBXs are going DID, Direct Inward Dialing. DID calls go from the
CO (or tandem office) into the PBX with the called extension
identified by dial pulsing, the slowest method of signaling
currently available. Thus CCIS will have to "run and run just to
keep us where we are." As CCIS goes in, DID will also be going in at
an even faster pace. And the speed picked up by CCIS will be lost as
three, four or five digits are dial pulsed into the terminating PBX
at the leisurely rate of one digit per second. The Bell System won't
even bend to permit sending at 20 pulses per second, twice as fast —
the standard rate that has been used for 30 years from PBXs into
common control central offices.
Equalizing the load
As a final chuckle, let
us contemplate the process of "equalizing the load." In most
electromechanical switching systems, matrix switches and trunk
relays at both ends of a trunk wear out in proportion to how much
they are used. Further, in systems like step-by-step (SXS) and
Panel, it takes time for a switch to find a free trunk if it has to
look at one terminal at a time. Manual jacks on a switchboard wear
out in much the same way. Thus, to minimize wear on these old
fashioned devices, pains have been taken, both at the design and
operating level, to "equalize the load" and make sure all trunks in
a group carry about the same amount of traffic. It has become an
ingrained habit approaching religious conviction.
Slipped and reversed
multiples were used in SXS systems, and Idle Trunk Indicators (ITIs)
lit a lamp over the manual jack to be used for the next call. When
common control systems became available, they used wonderfully
intricate techniques to equalize the load since they did not have to
depend on stepping switches. And computer controlled switches give
even more possibilities.
But consider WATS.
Customers use full business day WATS as first choice, measured time
WATS as second choice, and DDD as a third choice. The last thing in
the world the customer wants is to equalize the load because he
wants to put a lot more traffic on the FBD circuits than on the MT
circuits, etc. Further, since the breakpoint is easily found between
full and measured, he has to know how the traffic is actually
distributed on a per trunk basis to administer his system. The
telephone company's efforts to equalize the load are not
appreciated, and when these efforts put more traffic on the MT
circuits than on the FBD circuits, the customers often become
excited, demand refunds, and ,call the public utility commission.
Knowledgeable
consultants and communication managers deliberately arranged for SXS
switches and ITIs on manual switchboards to hunt in the right order
for loading WATS. This worked so well that new systems like the
Dimension were designed to prevent hunting "from left to right." The
Dimension can hunt over several different trunk groups in order, but
it insists on equalizing the load within each group. Some
communicators, as a result, put each WATS circuit in a separate
trunk group. Because Dimension is electronic, the extra wear causes
little trouble. At the far end, however, where obsolete systems
based on reed switches are used, some extra wear over time may show
up.
But now consider
Centrex. In a PBX, you have to have specific circuits to get to the
telco, and a WATS line is a physical circuit to get to a telephone
switch that is smart enough to do the routing, restricting and
billing required for WATS service. However, if you are served by an
ESS Centrex system, you are already in just such an office. You
don't have to go somewhere else. Thus, you don't have any real
facilities that can be identified as WATS lines. The system just
sends your call out over the public network, blended in with
everybody else's traffic on large trunk groups, and handles the
billing in a way that is appropriate to your WATS configuration. The
same program that would have hunted over fulls and measureds now
simply selects the appropriate billing approach for each call. You
are served by "virtual WATS."
With virtual WATS, your
bill shows the load equalized, and there isn't much you can do about
it. The telephone company has won... but it is an empty victory.
There aren't any actual circuits on which the load is equalized! No
switch wear has been reduced. All that the telephone company has
done is deprive the customer of information he needs to manage his
communications. But perhaps that is victory enough.
Oh, well. Centrex is on
its way out, and someday digital switching will be as common in
central offices as it is in PBXs today. Then T-carrier will run
directly from PBX to CO switch, and all trunks will be intervals
that exist in time rather than hardware that exists physically in
space. There will be nothing anywhere to wear out, but we can be
sure of one thing: the telephone companies will still break their
collective backs to equalize the load on these non-physical
facilities. Old gods die hard.
***
The
Womack PBX was a joint venture between Womack and Solid State
Systems. When SSS bought out Womack, the PBX was renamed "The Smart
Telephone System."
Full and
Measured WATS vanished in the summer of 1981, replaced by "Tapered
Rate" WATS.
[
Top ] [ Next ] [
Table of Contents
] |
