Background
for Telephone Switching
2nd Edition (Revised and Expanded)
Chapter 8
Operation, Administration
and Maintenance
OUTLINE
OBJECTIVE:
Once a system is in the field, operation,
administration and maintenance are the principal factors the
customer must consider over system life. This chapter discusses some
of the design requirements a system must meet to make the OA&M job
easier for the customer.
PREVIEW QUESTIONS
-
How can telephone systems report
troubles?
-
How can access for test personnel
be provided?
-
How do computers aid in the above?
OPERATION,
ADMINISTRATION,
AND MAINTENANCE
Traditionally, telephone switches last
a long time, growing and changing like living creatures. This means
that their design must give fundamental consideration to the needs
of the customer, whether an operating telephone company or a
business with a PBX, over the working life of the system. To obtain
the best possible service from such a major investment, the needs of
day-to-day operations, including craft access, must be built in;
both hardware and software design must be planned to handle all
administrative requirements including record keeping; and finally,
the ability to locate problems when they arise, continue operating
in the presence of such problems, and facilitate repair is
absolutely basic.
The designer must always assume the
implacability of Murphy's Law: if something can go wrong, it will,
and if it can find several ways to go wrong, it will choose the
least convenient. It is easy enough to design a system to "work," at
least under some circumstances. What is much more difficult, and for
which the newly graduated designer is often totally unprepared, is
to anticipate less obvious failure modes and then to design the
system to deal with them in a suitable way. The importance of field
trials cannot be overemphasized in this context. There is nothing
like actual operating experience in a live environment to test the
difference between theory and practice. Indeed, the more brilliant
the designers (or, for that matter, the more driving the
management), the more necessary the field trial.
Nobody can outsmart the universe. Even
the best designers cannot possibly guess all the ways their complex
systems can respond to the usage they will receive from both casual
callers and highly trained technicians. Sophisticated customers,
having learned the hard way that reality and time in the field can
force design improvements over the objections of ivory tower
designers, often prefer to stick with a proven system rather than
rush to something new. Those who wish to sell the latest state of
the art must be able to demonstrate that they have not ignored the
lessons of the past.
SXS systems, with their distributed
control located on each switch, tended to be quite reliable for a
number of reasons. First, the failure of any given switch did not
knock out the entire machine, or even very much of it. Second, the
path through the matrix, handed off from switch to switch, was
checked for continuity at every stage; there was no simple way an
open path could be turned over to a customer for a charged call. And
third, once the connection was up, every device used to establish it
was still in the circuit. If a bad connection could be trapped and
held, going "hand over hand" down the line ultimately located the
trouble.
The designers of crossbar systems
recognized all these factors. They knew that, once the originating
register, marker, sender and incoming register were used and
dismissed, there would be no way to identify them to find what went
wrong. So they designed elaborate routines to test all common
equipment between calls, and extensive marker-controlled tests on
each connection during call set-up. The false cross and ground (FCG)
test made sure a matrix path was not shorted, grounded or connected
to another path, and a continuity test not only checked the matrix
but also the cross-connect jumpers at the main distributing frame.
From the beginning of electronic
switching, when much of the system's structure was hidden in
abstract software rather than "real" hardware, it was obvious that
even more extensive self-testing would be required. In all working
systems, many forms of internal testing are carried out, both on a
per call basis and through special test routines carried out at
regular intervals. Once troubles are found, further testing is used
in to isolate the problem to the minimum number of PCBs.
As stored program control developed
ever more sophisticated software, more and more features and
functions were taken over by software alone, or software with
vestigial hardware; thus the testing of software, both during
development and in operation, took on increased importance. When
digital techniques dictated the merging of transmission and
switching, even more hardware could be replaced with zeros and ones,
and the importance of software administration and maintenance
increased even more.
What was not as generally appreciated
over the years was the role of the switch in testing connecting
circuits and systems. A local switching system with a metallic
matrix was a large access switch permitting measurements to be made
on outside plant conductors using dc meters, Wheatstone bridges, and
various other instruments, often from one local test-desk serving a
number of central offices. Need for this access was one of the most
important reasons why electronic switching matrices were slow to be
accepted for local CO switching. In PBXs, where station wiring was
mostly protected inside buildings, such testing could be ignored
with reasonable safety, and electronic switching flourished.
With a digital switching matrix, there
is no way to connect directly to customer lines to measure leakage
resistance to ground or to apply voltages to break down insulation
on lines with intermittent faults. Thus separate test access must be
provided if the advantages of digital switching are to be enjoyed.
Note that conventional 2500 type
telephones have little effect on line testing because everything
except the ringer is disconnected by the switch-hook when the line
is idle, and the ringer itself, in series with a capacitor, does not
draw dc or audio frequency currents and is essentially invisible to
most testing procedures.
One access technique, following the
approach of Northern Telecom, is shown in Fig. 1. Here a test relay
is provided on each line card. As long as the relay is unoperated,
tip and ring pass straight through to the line circuitry (battery
feed, supervision, ringing, codec, etc.). However, when the relay is
operated, the through path is broken, T and R are connected to the
"outside" test bus, and T1 and R1 are connected to the "inside" test
bus. By associating appropriate equipment with these buses, tests
can be conducted on the path to the customer and/or the customer's
line circuitry. Further, the outside T and R of one line card can be
switched, via the test buses, to the inside T1 and R1 on another
line card to provide emergency service (for a very limited number of
lines at any one time) until repairs can be effected.
With the coming of ISDN telephones,
where an electronic line circuit at the CO connects via tip and ring
to a customer owned and maintained NT1 or NT2 device, testing will
be further complicated by need to disconnect electronic circuitry at
both ends of the pair. In addition, the telephone company will no
longer be able to upgrade the customer's NT1 or NT2 device to
reflect new testing procedures or changes in the transmission
medium. Digital signals on customer lines will require even better
maintenance capabilities than those used for analog lines, but test
access will be more difficult.
A further complication can be expected
when optical fiber is inserted into the local loop in various ways.
The main purpose of "fiber in the loop" appears to be CATV delivery,
which will require CATV switching and voice/TV multiplexing to take
place between the CO switch and the outside world. Optical fiber,
however, will require tests quite different from those needed by
copper pairs.
On the trunk side, switching systems
have, for many years, interfaced carrier systems, first analog and
now almost all digital, rather than copper pairs. As a result,
transmission hardware stands between switching systems and the
outside world, and often provides its own test access, procedures
and personnel. On the other hand, today there is almost no trunk
hardware that is unique to a particular circuit, and as a result, a
switch can monitor one circuit or even a test channel in a
multiplexed group and be fairly sure that the information obtained
is pertinent to all channels in the group. Thus the switch still has
a major responsibility for monitoring its connecting circuits as
well as its internal components.
The net result of these various
requirements is that the switch, with its built-in intelligence,
must monitor test points in the outside world as well as within its
own equipment to locate troubles, and must have built into its own
programs enormous capabilities for locating and displaying problems
in both its own and connecting equipment when they arise. Indeed, it
is not unusual for the software that deals with maintenance and
administration to exceed that which is required to deal with
telephone calls.
Because system failures of various
sorts do occur, a variety of means have been developed over the
years to identify them and expedite their repair. Sometimes the old
reliable approaches are still valid even in the digital age.
Fuse alarms
Fuses are installed in power feeders
to various sub-assemblies to protect the central power supply and
with it the other sub-assemblies depending on it for power; a fuse
is not necessarily expected to protect the equipment it to which it
delivers power. Fuses in the telephone industry are generally rated
at the highest current they will carry indefinitely without blowing;
some other industries rate their fuses in terms of the lowest value
required to cause them to blow. In either instance, a fuse melts
under excess current to disconnect a short circuit or other
overload. This protection is one of the two things a fuse must do.
The other fuse function consists of letting somebody know that a
problem exists. Alarm fuses are quite generally used in telephony.
Alarm fuses are constructed in such a
way that the fuse element itself holds a spring compressed. When the
fuse element melts, the spring is released so that it can push a
contact connecting an alarm lead to the incoming power bus. In
actual fact, an alarm fuse is a transfer contact that disconnects
the overload and, at the same time, connects the alarm. Further, the
spring also provides some sort of visible signal at the fuse so
that, when the maintenance force arrives, the particular circuit in
trouble can be identified immediately.
Ideally, fuses should be located at
eye level to make them easy to inspect and replace. This is almost
never the case; fuse panels are usually at the top or bottom of the
frame or cabinet, depending on where the power enters. With seven
foot frames, location at the top isn't bad; with eleven foot frames,
spotting and replacing a blown fuse is more difficult. However,
fuses at the top of a frame are not subject to damage from
inadvertent kicks, cleaning machinery, etc.
Fuse alarms are arranged in a
hierarchy, depending on how many users will be inconvenienced by the
loss of a given circuit. In electromechanical systems, the loss of a
single trunk had far less impact than loss of an originating
register, and register failure was less critical than loss of a
marker. Different colored lights as well as gongs and buzzers were
often used to locate the floor and aisle on which the alarm
occurred, and to differentiate between major and minor troubles.
Electronic switching systems are much
smaller than comparable electromechanical systems, and are organized
somewhat differently. Typically, power is distributed to individual
cabinets at -50 volts dc, and power supplies in each shelf convert
it to the several smaller dc voltages needed for electronic
components. Often, each PCB (which may contain a number of lines,
trunks or other circuits) is fused at each voltage, each shelf power
supply is fused, and the -50 volts entering the cabinet has its own
fuse.
The system control, via the scanner,
can monitor alarm fuses for inputs to its automatic trouble
reporting and analysis system. These inputs are particularly
important when the office is unmanned and a telemetering system,
operated by the common control, reports regularly to a remote
maintenance center. Key fuses may have dedicated scan points for
their individual inputs to the system; less important fuses may be
monitored in groups by a common scan point. The use of fuses as a
source of alarm signals is as important as their protection
function.
It is not, of course, necessary to
scan all such scan points during peak busy hours when real time may
be scarce. However, such scans impose no time burden when carried
out during off-peak hours, or by a separate processor designed to
handle testing and traffic measurements. Output displays and reports
should be carefully formatted to give maximum information with
minimum possibility of error in interpretation.
Because common control failures are
possible, fuse alarms in electronic switches should also operate in
the traditional way so that a system failure will not cut off
necessary information that can easily be obtained directly. Lights,
gongs and other signals, independent of the common control, still
have their place.
Permanent signal alarms
Partial dials and permanent signals
are fairly common. Crosses to 60 Hz power lines can cause partial
dials in common control systems (few SXS selectors will follow 60
Hz, but mercury relays or electronic circuitry in dial pulse
detectors can), and leakage currents, produced by low insulation
resistance, lead to permanent signals. But the most usual source of
such troubles is the user. A phone knocked off-hook, someone wishing
privacy, or someone going back to look up the rest of the telephone
number are the common causes.
A permanent signal ties up a certain
amount of common equipment, and degrades the traffic handling
capacity of the system for the rest of the users. SXS had timers for
line-finders and first selectors; if the selector did not go off
normal after a certain amount of time, an audible signal was sounded
to alert maintenance. Crossbar systems often had timers built into
registers; the timers were reset after each digit. If the next digit
did not arrive before time-out, AND if someone else needed the
register, the non-dialer or slow-dialer lost the register to the new
origination. The timing interval could, as a refinement, be altered
as a function of traffic with 20 seconds being typical for a side
hour and 3.5 seconds for the busy hour.
Because electromechanical common
control systems were particularly subject to traffic overloads, the
clearing of permanent signals was of great importance. Some systems
were arranged to connect a "howler" tone for a given period of time,
and then route the line to a test position where it was stacked in a
queue. If the user found the phone off hook and hung up, the whole
connection was cleared automatically. But, until cleared, it tied up
a path through the switching matrix.
Line lock-out.
Rather than tie up a path through the matrix to a howler tone, some
early systems used line lock-out. Here, a line generating a
permanent signal was released from a matrix-connected resource and
returned to the line circuit for monitoring. As long as the off-hook
persisted, nothing further would happen because the signal lead to
cause the system to take action was inhibited. However, if the user
hung up and came off hook again, the inhibition was removed and dial
tone was returned.
In electromechanical systems, the line
circuit for a system with line lock-out was more complex than a
traditional line and cut-off relay. Usually an additional relay was
required to prevent the system control from recognizing the
continuing off-hook. Sometimes this relay was also used to return
howler tone without using a path through the matrix.
In most electronic switching systems,
the line circuit provides battery feed and supervision to analog
telephones during both idle and busy intervals; it also applies
ringing and coin control voltages. Thus, the line circuit is already
quite complex, and the incremental cost to add line lock-out and/or
howler, perhaps connected via the ringing access circuitry, is
minimal. With stored program control, it is a simple matter to have
the program recognize a permanent signal without any additional
hardware; the system can then ignore it until an on-hook is
detected. Howler tone can be returned via the matrix for a timed
interval and then dismissed to free the matrix path and tone port,
but in non-blocking systems, typically PBXs, there is little to be
saved by releasing the matrix path. In digital systems, where tones
are distributed on time slots to which ports needing tone can be
connected, there is no separate service circuit group supplying tone
which can be conserved. As a result, the line lock-out and howler
functions in digital systems need not follow blindly the approaches
used with metallic matrices.
Trunks can cause permanent signals as
well as lines; however, there is no advantage in returning howler to
a sender. It is urgent to clear the trouble on a trunk and return
it to service; as a result, trunks require their own alarms and
displays. Where switching and transmission are separate systems,
this is often the function of carrier system channel banks. In
digital systems, where the channel bank function becomes part of the
CO switch or PBX, the switch must alert others to the problem and
make use of the information itself.
Make busy.
In general, any circuit that is in trouble should be "made busy" to
prevent its seizure: an open line must not be seized to complete a
call, and, even more important, a defective trunk or service
circuit, intended for shared use by many lines, must be removed from
its hunt group. A system must be able to busy out circuits its
automatic test procedures find faulty; further, maintenance people
must have similar procedures for taking circuits out of service
manually.
Ideally, a "made-busy" signal, clearly
different from a regular busy signal, should be available for return
to calling users or repair staff, and made-busy circuits should be
printed out or otherwise displayed to provide an independent record
of circuits out of service. "Trap" programs should be constructed to
check the status of made-busy circuits to be sure information
provided to the maintenance force is up to date, and that records of
intermittent failures which may have cleared themselves are also
available for analysis.
As has been mentioned, it is not
unusual to find two, four or more circuits on one plug-in module.
When one circuit fails, the module must be replaced to restore
service. However, the module cannot be removed until all its
circuits have become idle, and all are made busy. Thus a make-busy
procedure must include a request to the system to make the
particular circuit busy with intent to remove; the system must then
busy out that circuit and all other idle circuits on the same
module. Circuits in use must be monitored for hangup, and then they,
too, must be made busy. Only when all circuits are made busy will
the signal for removal be given. This signal can easily be printed
out on a maintenance display, but a lamp or other visible signal on
the module itself is insurance against inadvertent termination of
active calls.
Because a trunk, unlike a line or a
service circuit, has a switch at both ends, it is harder to busy
out. Coordination between both ends is required, but common channel
signaling makes the job much easier today than in the recent past.
PBX trunks, whether digital or analog, pose special problems. When
the PBX customer makes a trunk busy, the matching make-busy function
at the CO is a service which requires compensation. If only one
trunk is found to be bad on subsequent testing, the local exchange
carrier may not charge to take it out of service during repair.
However, taking good trunks on the same PBX trunk card out of
service may add to the cost of maintenance.
Carrier group alarms
Although a back-hoe cutting a cable
serving customer telephones may not cause an immediate and massive
problem, other kinds of cable failure can produce large numbers of
permanent signals. It takes a tip-ring short (or, at least, a
ring-ground short) to cause an origination, and although it is hard
to short the several hundred pairs in a cable, it is possible. On
trunks using SF signaling, it was a different story. SF required the
presence of tone to show the trunk idle, while the absence of this
signal made it busy. If a cable containing such trunks was cut, or
if a microwave tower fell over, the idle signal, along with
everything else, went away, producing massive seizures at the
terminating end of all the trunks involved.
To deal with such problems, the
carrier group alarm (CGA), which originally alerted transmission
maintenance people to major system failures, was fed into switching
systems using stored program control. CGA could then be interpreted
so that massive seizures could be ignored, existing calls on the
same trunk group terminated and charging stopped, and all trunks on
the group made busy to further seizure.
CCIS solves this SF problem along with
others, but because CCIS usually uses a facility different from the
one carrying a trunk group, and thus may function properly even when
the trunks are knocked out, it is even more important to coordinate
CGA with specific circuits. Today, where digital carrier systems are
interfaced directly, switching systems must be able to extract
information from test signals on the transmission framing bits to
obtain the equivalent of a CGA signal. Even with this capability,
visual and audible alarms, independent of the switching system,
should be maintained for added reliability.
Many CGA systems must have their
carrier equipment in the same building with the switching system if
alarm information is to be passed along; this is not always the
case. When short-haul carrier systems are used to allow long-haul
carrier to terminate on a number of switches near each end, perhaps
with DACS "grooming" of which channels go where, the trunk that
terminates at a switch may very well not be in the carrier system
which failed. Thus failure of a large carrier system may affect
dozens of switching systems, all located many miles from the nearest
end of the particular link in trouble. Passing the CGA signal
forward to all of these remote offices is something of a challenge.
Because a CGA signal is not always
available, as with PBXs in particular, some kind of trouble
detection scheme should be built into switching system programs.
When large numbers of nearly simultaneous seizures are detected,
these seizures should be checked to see if they are associated with
a common trunk group. If they are, only one or two are given
incoming registers and the system monitors for time-out, temporarily
ignoring requests for service on the rest. This leaves the remaining
registers free to handle service requests from other trunk groups,
and provides immediate attention to a potential problem. Similarly,
on outgoing trunks, many sender time-outs in a short interval should
cause the system to check for facility failure common to the trunks
involved. By programming the processor to hunt for permanent signals
and time-outs on trunks in this way, making such trunks busy and
flagging maintenance, system reliability can be increased and
traffic overloads generated by carrier failures can be cleared, with
or without carrier group alarms.
Finding bad trunks with traffic
measurements. When trunks had individual plug-in units in channel
banks, external SF signaling units, and individual trunk circuits on
switching systems, all tied together with cable pairs terminated on
cross-connect frames, it was not unusual for an individual trunk to
fail. CGA does not work for individual trunks within a group, and
thus will do little to identify "killer trunks" and stuck trunks.
A killer trunk is one which looks all
right to the switch, and can be seized; however, the caller finds it
unsatisfactory due to noise, or some other problem. The caller hangs
up and tries again, usually getting a different trunk and completing
the call satisfactorily. A stuck trunk is one which appears busy for
hours on end, and is not released to be seized for another call.
Both of these faults can easily be located by means of traffic
measurements. A killer trunk will have much shorter holding times
than others in the same group, while a stuck trunk will have much
longer holding times. Stored program controlled switches can easily
detect and display such information. Although this sort of approach
would have been almost impossible with earlier generations of
switches, it is not uncommon to find modern switching systems
continuing to ignore an opportunity to take advantage of one of
their most obvious capabilities.
Of course, digital trunks between
digital switches have almost no hardware to fail on a per-circuit
basis, so killer and stuck trunks should slowly fade away. However,
hunt algorithms and other software features are sometimes arranged
inadvertently to avoid selection of certain trunks in a group; this
kind of problem can easily be detected with other variations in
recording and displaying traffic information.
Trunk make busy.
Failed trunks, either individuals or groups, must be made busy; that
is, they must be removed from the hunt procedure so that they will
not be selected for calls. At the same time, the distant end must
not see a seizure that will result in a permanent signal. In
electromechanical systems, an outgoing trunk was made busy by
putting a ground (or battery) on the sleeve lead so that hunting
would pass over it to the next trunk in the group; with no tip-ring
closure or M-lead operation, no seizure was sent forward to the
terminating end. With stored program systems, where hunting is done
on the map of available facilities in system memory, it is even
easier to mark the trunk busy. With several levels of busy possible,
and trap programs to inventory and display them for maintenance
personnel, stored program systems can greatly improve trunk
reliability.
Making a bad trunk busy at one end
implies a conjugate action at the distant end; for a two-way trunk,
seizure from the far end must also be inhibited, and either a
one-way or two-way trunk, at the incoming end, must be warned not to
accept a seizure which may be caused by the fault.
Today, common channel signaling, often
making use of centralized data bases for more advanced features,
changes the nature of trunk make-busy. With signaling and
supervision removed from the trunk itself, and the packet network
used for signaling able to update such data bases in real time,
taking a trunk out of service or restoring it should be much more
effective. It is likely, however, that many PBXs will continue to
use conventional circuits with in-band signaling for both PBX-CO and
tie-trunks, and older make-busy procedures will not fade away for
some time.
There are times when massive seizures
are not the result of a cable failures or other abnormal equipment
conditions. When it snows, when some kind of disaster take place,
when bets have to be placed with bookies, etc., there can be peak
loads on switching systems far in excess of those predicted by
averages where calls are originated "individually and collectively
at random." During such intervals, it is very important for doctors,
police, national guard and other emergency personnel to get through.
To this end, line-load control is provided.
Line-load control simply allows the
system to ignore originations from non-priority users when it is in
effect, while allowing emergency personnel to make calls as usual.
Actually, line-load control will give everybody a chance to
originate calls if they wait long enough for dial tone; the
procedure is to allow non-priority users on one line group at a time
to have an opportunity to place calls. Once the system has accepted
a call, line-load control does not interfere with the call's
completion.
Line-load control obviously requires
the use of class marks. When manual line-load control is activated,
there might be several different classes of priority users,
depending on the type of emergency condition in effect. If heavy
system traffic causes line-load control to be implemented
automatically, perhaps in response to a heavy snow-fall, priorities
might well be different from those imposed by a riot or invasion.
Line-load control is another function whose implementation is
greatly facilitated by stored program control.
It should be noted that having a
non-blocking matrix will not eliminate the need for line-load
control. With massive seizures, there may not be enough DTMF or dial
pulse receivers, and the call processing capability of the switch's
control itself may be overloaded. In addition, many of the emergency
numbers which a caller might wish to reach will probably be busy
already with higher priority calls.
One of the basic rules in the design
of large complex systems in general, and telephone switches in
particular, is always to close the feedback loop. That is, whenever
the control issues an order, make sure that carrying it out gives an
active indication which is returned to the control for positive
checking. The chain of events is usually fairly long:
-
The processor writes an order into
a buffer.
-
The buffer applies the order to
drivers.
-
The drivers activate the media in
the bus system which carries information to and from the
peripheral cabinets.
-
Buffer equipment in a peripheral
cabinet plucks the order off the bus and applies it to port or
matrix circuitry.
-
The addressed circuitry then
changes state.
It is not impossible for an order to
occasionally go astray under such circumstances.
An additional problem develops when
electronic controls are used to activate or monitor
electromechanical equipment. Because of the great difference in
speeds, there is no way for the response of an electromechanical
device to return to the control during the same, or even the next
several hundred, clock intervals. Obviously, the control cannot
stand around waiting, so one approach is to assume everything goes
all right and to watch for responses as they come in. Each response
is matched with an indication stored when the order was transmitted;
only when a match fails to come in after a reasonable time-out
interval does the system go looking for trouble. Even with
all-electronic systems, the transit times to and from distant
cabinets, to say nothing of remote switching modules, particularly
when measured in terms of the high speed clocks presently in use,
suggest the use of such a procedure.
Time-outs associated with permanent
signals, stuck senders and the like have already been discussed.
Another class of feedback for time-out checking includes actions by
the system control which cause a natural response to occur. For
instance, when a path is set up from a line through a metallic
matrix to a dial-pulse receiver or trunk, proper operation of the
system as a whole will cause the line sensor to be disconnected,
changing its output from active to passive, and the loop monitor in
the dial pulse detector or trunk to be operated, going from passive
to active.
This "transfer of supervision" acts as
natural feedback to check continuity of the matrix path and the
operation of the line circuit and the trunk or signaling detector.
Unfortunately, even in metallic matrices, there are instances where
transfer of supervision does not take place, and it does not occur
at all with paths through most electronic matrices.
Where natural responses are not
available, suitable responses can sometimes be constructed. For
instance, any kind of sender can be arranged to have a scan-point to
monitor up-checks and down-checks to show that each pulse or digit
has been sent. In E&M trunk circuits used for dial pulsing, one
could set the E-lead scan point to monitor the M-lead during
pulsing, at least when stop-go signals are not expected. The
outgoing digits can then be monitored with the incoming digit
program and the transmitted digit compared with the "received" digit
from the sender's output. In almost every type of operation, either
the sent signal can be monitored for checking, or the response to
the transmitted signal, when it returns, can used for checking as
well as for its intended purpose.
Common channel signaling, like most
data systems, has its own built-in checks. These may not, however,
be sufficient. Because call set-up information travels via a path
different from the one to be used by the customer, there is no
assurance that the talk path is actually available. As a result,
special tone senders and detectors are sometimes provided to
momentarily check the path through individual trunks. It would be
far more useful to set up the entire connection and have such a test
provided end to end prior to turning the multi-trunk path over to a
caller.
System time-outs should be included in
most programs to make sure that the goals required for successful
call processing are achieved on schedule--each digit received, all
digits received, wink start detected, answer obtained, hang-up
detected, etc. If any expected signal is not obtained before
time-out, checking should be instituted. Such an approach may also
catch program loops and other software difficulties.
In any kind of complex system,
component failures will take place sooner or later. Because this
cannot be prevented, overall design must take it into account when
system reliability is being planned. In telephone systems, extra
circuits and sub-systems are provided so that automatic switchover
can replace faulty circuits serving a number of customers. In
general, such duplication does not extend to circuits that serve one
or a small number of lines.
Several methods of providing extra
equipment are common. As has been mentioned, shelf power supplies
may be provided in pairs, sharing the load. If one fails, an alarm
is sounded and the other carries the whole load until repairs are
made. Another approach, often used with echo suppressors or
signaling sets in transmission systems, has one spare circuit for a
group of five to ten working circuits. The spare is kept in "hot
standby" and can be substituted for any of the other units in the
group that should happen to fail. Patch cords at large jack panels
have been used to make the substitution manually, but automatic
throwover has its advantages.
In switching rather than transmission
systems, it is often convenient to simply provide a few extra
circuits in the group, using the matrix for access. Senders, digit
receivers, tone circuits and other single-ended service circuits
work well this way; if any one fails, it is simply made busy and the
system continues unaffected except for a slight reduction in traffic
handling capability. This is called "graceful degradation."
In crossbar systems, where several
markers operating in parallel established connections through the
matrix, graceful degradation took place when a marker failed; the
remaining markers simply picked up the extra load. This approach
has, occasionally, been used with electronic common control systems,
but the speed and power of modern electronic processors makes the
complexities of parallel operation unnecessary. With only one
processor, the problem of keeping markers from fighting with each
other for control of specific resources is eliminated. The
popularity of "parallel processing" in the computer world, however,
suggests that the last word has not yet been said.
As was mentioned in Chapter 1, a
single common control makes a system vulnerable because its failure
will immediately take the whole switch out of operation. Thus common
controls are usually duplicated for reliability. Hot stand-by is
common; the two common controls run in parallel, each checking the
other but with only one providing outputs to the system. When the
active processor develops a problem, the other takes over and runs,
unduplicated, until repairs can be effected. The probability of the
second common control failing before the first is restored is,
hopefully, minute. The processors usually take turns being "main"
and "standby" so that there is no question that both are in working
condition. PBXs have sometimes used "cold" standby, with the second
processor doing other tasks (such as processing CDR information)
until it is needed. When computers were very expensive, this made
more economic sense than it does today.
In some systems, there are two
complete common controls consisting of processors with their
program, data base and current-information memories. When trouble
occurs, the entire active complex is replaced by the standby. This
makes it necessary for the standby common control to have its
current-information memory track the one in the working system so
that, at change-over, it will be ready to go. In other systems,
processors and memories are duplicated separately so that only the
faulty unit is replaced; the switch-over circuitry is more complex
in such an arrangement, but more flexibility is available to
configure a working control system.
Yet another approach to reliability
calls for each port group on a large switch to be autonomous with
its own redundant processors, and capable of completing internal
calls without outside help. If the event that a module fails, the
other port modules can continue operation unaffected. This approach
is particularly useful when remote switching units are supported;
neither a central processor failure nor the failure of the umbilical
from the RSU to the central location will cause customers to lose
service.
However, when there are N port groups,
the probability that a call originating in one will be completed
within that unit is roughly 1/N. Thus a means is required for the
originating module to find and obtain a connection, either direct or
via a group selector, to a terminating module. This can easily be
done with a central data base, perhaps associated with the group
selector, although other approaches have been used.
The switching matrix has always posed
special reliability problems. Because it was the biggest single item
in electromechanical space division systems, it could not be
duplicated; the throw-over circuitry alone would have been too
large. However, there were many paths through the matrix between any
two terminals, and because most trunk and service circuit groups
were well scattered over different switch frames, many kinds of
matrix failure affected only a small proportion of possible calls.
This has changed in digital switches.
The use of inexpensive RAM memory to build large TSIs, and the use
of inexpensive logic gates to make different space-division
connections in each time slot, encourage redundancy even for
switching matrices. Because most of today's digital switches trade
off time slots for crosspoints, the switching matrix has shrunk
almost out of sight. As a result, even very large CO switches, or
toll switches for 100,000 trunks, today duplicate their switching
matrices just as they duplicate control systems. Usually the group
selector is duplicated as a whole, while the line groups contain
duplicated concentrators.
Although line and individual trunk
circuits are usually not duplicated for reliability, multiplexed
trunks raise the ante. When one circuit board terminates a T-span,
all 24 trunks in that digroup depend on the single board; it is no
longer possible to "scatter" the individual trunks over a number
line groups for higher reliability as was common when trunks entered
the switch as individuals. However, both transmission and switching
are more reliable today than in the past, and trunk groups are often
large enough to require several T-spans in parallel. Thus T-spans
can scattered, at least until switches make a practice of
interfacing digital transmission systems at higher levels of
multiplexing.
Internal testers and testing
A large portion of system software is
devoted to checking the operation of the control and the
distribution of its signals to line, trunk, and matrix circuits. As
mentioned above, one approach uses both main and stand-by
processors, running in parallel; by matching the operation of one
control with the other, continuous checking is always in progress.
It is also desirable to run routine
tests on off-line sub-systems, including the processors, memories,
bus systems, etc., to locate faults that may not come up in the
general course of operation. Such tests are usually internal and are
clearly indigenous to the particular system. For externally driven
tests, specialized load boxes, test call generators, etc., are often
used during installation, and can be brought back at periodic
intervals for additional independent testing. Modern common
controls are particularly clever at discovering their own
weaknesses, but sometimes an external measure of their health is
desirable.
Trouble and maintenance routines can
be quite lengthy and complex. In larger systems, they are usually
part of the overall system program; in smaller systems, they are
sometimes kept off-line on tape or disk where memory is cheap, and
are only loaded into RAM when required. Such a procedure simplifies
the working memory and reduces costs; it also permits many more test
routines to be made available.
As the cost of microprocessors has
dropped, some systems have built in autonomous test processors to
continually generate a variety of test calls, monitoring the
response of the overall system to be sure it is working properly.
These processors remove the load of routine testing from the system
control, and can easily continue testing when heavy traffic would
cause cancellation of such functions when run by the main processor,
precisely the time when trouble analysis is most important.
For digit detectors serving customer
lines, special test transmitters can be connected periodically to
exercise them under extreme conditions such as maximum and minimum
percent break and pulsing rate, tone frequencies both just inside
and just outside the edge of each signaling frequency band,
(accepting the former and rejecting the latter), etc. When a form of
signaling such as MF has both digit transmitters and receivers
available, one can be tested against the other, preferably through a
worst-case artificial line, as long as different
transmitter-receiver pairs are used on each run. Upon detection of
a failure, further tests, using the suspicious transmitter and
receiver with other units, will be required to find which was in
trouble.
Where analog tones are generated
separately and distributed to tone ports on the switching matrix,
the tone distribution system can be protected with alarm fuses; loss
of dial tone, for instance, can make users think the whole system is
down. However, tone detectors, connected periodically via the
matrix, can also detect open circuits that will not blow fuses.
Special receivers can check both amplitude and frequency components
and, with a little care, their band-width can be adjusted to pick up
drifting senders, including DTMF telephone sets, before the drift
carries the signals out of the range of regular receivers.
Similarly, special senders can check receivers for bandwidths that
are too wide or too narrow, and amplitude thresholds that are too
sensitive or insensitive. It is just as important to check
signaling equipment for its ability to reject invalid signals as to
receive valid ones.
These test circuits, dealing with
audio signals, can be given ports on the matrix, connected as
needed, and cycled through their test procedures by the system
program for both automatic tests and tests selected by maintenance
personnel. The generation of digital signaling and call progress
tones, where amplitude is determined by PCM samples stored in ROM
and frequency is controlled by the system clock which is
synchronized via T Carrier throughout the country, should greatly
simplify testing in modern systems, as can digital signal detectors
which respond to signals in their digital rather than analog form.
Common channel signaling has already
gone a long way toward eliminating MF and dial pulse signaling on
trunks, along with traditional methods of supervision. Similarly,
digital signaling to electronic telephones, both ISDN and PBX
proprietary, is eroding the need for testing dial pulses, DTMF, and
power ringing, although new technologies have testing requirements
of their own.
Testing connections through metallic
switching matrices, for both continuity and crosses to existing
connections, was fairly simple. The false cross and ground ("FCG")
test simply connected battery and ground to tip and ring
respectively, through a detector, before cut-through to the trunk or
service circuit took place. Because tip normally went to ground and
ring to battery, a cross to another path (tip to tip or ring to
ring) or a short or ground operated the detector.
Metallic continuity was checked by
transfer of supervision, or by attaching loop closures and
supervisory detectors momentarily. Flow of ringing current was
monitored on terminating calls to check continuity through the
matrix, main distributing frame, and pair to the set, even when the
phone at the far end was on-hook. The no-test connection was also
used to gain test access from the line side of the matrix.
With most electronic and all digital
matrices, the problem becomes much more difficult. The FCG test
illustrates the point. One of the main faults picked up by FCG was a
stuck crosspoint, contacts welded together or otherwise shorted,
making a permanent path between a horizontal and a vertical in a
crosspoint array. A new call, arriving on the same vertical but
departing on a different horizontal, would actually be connected to
two horizontals at the same time; if the undesired horizontal was in
use on another call, high level cross-talk would result. With space
division electronic matrices, a shorted crosspoint is perfectly
possible, but connecting a detector with reversed battery and ground
would alter the bias voltages needed to keep the desired crosspoints
conducting. With a digital matrix using a Time-Space-Time
architecture, corresponding problems can exist and must be dealt
with if large numbers of calls are not to be affected.
Because all digital switches are
4-wire, and any number of "listen" paths can be connected to the
same time slot, the difference between a false cross to another
conversation and a desired connection to a common source of call
progress tones or recorded announcements must be distinguished.
Even continuity testing is more
difficult; with the exception of the analog line groups in AT&T's
5ESS, line circuits are AC coupled to the matrix path, and DC
supervision is not passed from the line circuit to another circuit
on the other side of the matrix. However, when the line circuit
remains in the talk path at all times, handling many functions on a
per-line basis, adding some means of applying a voice frequency tone
or a digital test pattern after the codec and detecting it at the
far side is only a small incremental addition to cost and
complexity.
In addition to per-call tests, a many
additional tests are required toward the customer loop and toward
the switch itself. A small relay per line, as was shown in Fig. 8-1,
is a convenient way to give a wide variety of test circuitry access
at any port, looking outward or inward as desired.
Test lines
Not so long ago, when all toll calls
were established by operators, the operators could check each
connection as they set it up. If the connection was bad, they could
immediately hold it for testing and connect a distinct tone to help
maintenance forces locate the bad circuit. "Tone and hold" was also
common in the tie-trunk networks of large companies when they used
their switchboard attendants to set up inter-location calls.
With the coming of DDD, the operator
vanished from the connection and with her the per-call monitoring
for suitable transmission. This accelerated the need for automatic
routineing of lines and trunks, and encouraged the provision of
"test lines," special service circuits to which lines and trunks
could be connected and which could apply tests and return test
results to the system.
Test lines are also needed by
telephone company installers to test the signaling and ringing
capability of newly installed or reconnected lines. Such circuitry
is quite similar to that needed for reverting calls, although
audible signals are helpful to indicate the nature of failure. Test
lines can also be used by the installer (or user) at the customer
location in cooperation with personnel at a test center; the latter
can walk the on-site person through various tests, and report back
results as displayed.
Several test lines have been
established over the years. Some of the simpler ones simply return
a tone--a 1000 Hz tone at 1 milliwatt, TLP, for instance. Dialing up
such a test tone allows the caller to measure the returned level
and, as a result, find the one-way loss in the facility. It is also
possible to dial up special quiet connections for making noise
measurements, and on trunks, to loop around to test both directions
of transmission or echo suppressor operation. Some test lines test
signaling and supervision while others are capable of testing data
transmission.
Test lines usually return answer
supervision so that 2600 Hz SF signaling, where still used, is
removed during the test. This poses some problems because SF
signaling, in the presence of another signal, tends to stay in the
state in which it is found. For this reason, test lines that return
tones usually have periodic intervals of silence long enough to
allow SF signaling to change from on-hook to off-hook or vice
versa. Without this refinement, release of the test line might be
very difficult. Such problems will vanish with the passing of SF.
Because customers in the United States
must own their own equipment, including PBXs and key systems in
addition to telephone sets, there appears to be a need for test
lines which can be used by customers to perform their own tests.
Indeed, this might produce a modest source of revenue for telephone
companies, and PBXs, in particular, could be programmed to make test
connections on each CO trunk at night to test transmission and
signaling.
ALIT.
Automatic line insulation testing goes
a step beyond line test circuits such as those used by installers to
check dial speed or the telephone's ringer. ALIT is a totally
automatic system that runs through all the lines in the office
checking for line leakage, either tip to ring or tip or ring to
ground. It is a very powerful tool for maintaining outside plant.
ALIT is often put into operation
during or right after rain storms so that cables affected by
moisture can be detected. It only takes the system a few minutes to
run through all the idle lines in an office; normally it skips over
busy lines, and gives an indication of the lines so missed.
When ALIT is first installed, its
threshold is set quite low; it just looks for lines with leakage
resistances less than 10,000 ohms, for instance. After these worst
offenders are found and improvements made, the threshold is set at
15,000 ohms. When these lines are cleared up, the threshold can
again be increased.
Obviously, ALIT is the kind of
function that can easily be built into a local switching system. All
that is needed is a service circuit containing a suitable tester,
and a program to cause it to be connected to each idle line in turn.
The busy/idle status of each line is already known by the system
control, and the maintenance display and print-out system is readily
available.
Although straightforward through
metallic matrices, ALIT access for digital CO switches requires the
kind of access made possible by the kind of circuitry in Fig. 8-1.
ALIT equipment can use the "outside test bus" to connect to each
line in turn under system control. Note that simply bridging the
customer line is not sufficient; the line circuit's shunt impedances
and battery and ground must be disconnected to protect them from
ALIT test voltages, and to prevent their interference with
measurements on the outside plant. Circuitry on the customer's
premises, no longer owned by the telephone company, must be designed
to permit ALIT measurements to be made, or an interface per line
provided to disconnect CPE during the test.
Other test boxes
As digital switching has been extended
to local central offices, ALIT is only one of the sets of tests
required at each line circuit; various tests must also look toward
the switch. Thus equipment associated with the "inside test bus" of
Fig. 1 can check line supervision, application and trip of ringing,
etc. on lines to analog phones, and the bit-stream and D-channel
signals used with digital phones. In addition to automatic tests,
such equipment must also be accessible to test personnel for manual
testing from a centralized test desk.
The development of such test
equipment, although still in its infancy, is one of the more
interesting applications of what is generally referred to as
"artificial intelligence." Careful studies have been made of
exactly what skilled craftspeople do to clear various kinds of
faults, including such subtle observations as how fast a meter moves
when connected to a circuit. Such information has been converted
into sophisticated computer programs designed to run a matching test
box which can be connected to the line to be tested and then
directed to do what the craftsperson of yesterday would have done
manually. Results are displayed for immediate use and storage in
archives. Test boxes, being relatively inexpensive, can be located
in each line group and with remote switching units.
As optical fiber in the local plant
increases, such boxes with a whole new array of tests suitable for
that medium will be needed. Test access will be more complex,
because CATV may well share the transmission medium, with its
multiplexer between the telephone switch and the outside world. Such
requirements are more of a challenge than a problem, and will
provide work for both programmers and hardware designers that may
turn out to be more useful than video games, interactive X-rated
entertainment, and other marvels of modern technology.
In the days of SXS and Crossbar,
Operations, Administration and Maintenance were carried out by a
number of different groups. Some would answer trouble reports from
customers while others might take customer orders for service
changes. A different group would write up work orders specifying
jumper changes at the MDF or wiring changes for class of service,
while still other groups might make tests on the switches, the
outside lines, and trunks to other central offices. In addition,
there were other groups that kept records related to all this
activity.
Stored Program Control, combined with
inexpensive memory in the form of magnetic tape and hard disks, has
made drastic changes. Although there are still craftspeople pulling
jumpers on the MDF, splicing cables, and changing circuit boards,
most of the OA&M effort can now be carried out by individuals
manipulating information that interacts directly with system
software. Most of the old interfaces such as jack fields, panels of
blinking lights, and rows of keys and switches are now gone,
replaced by glowing terminals or PCs sitting on desks in
conventional offices, operated by computer-trained people who cause
the system to carry out orders and keep appropriate records
automatically.
Traditionally, it has been the
practice in central offices to provide separate access positions for
handling switching systems and transmission facilities. There should
be enough access positions to accommodate several workers at the
same time to facilitate system debugging and occasional bursts of
trouble; because lines and trunks, although simpler, exist in vast
quantities, line and trunk test positions will usually outnumber
those intended for switching system access.
Even small switches have standard data
ports for plugging in a terminal to act as an interface. In PBXs,
the attendant console can sometimes take over the access role for
maintenance and administration, as can electronic telephone sets
with enough buttons and displays. However, the use of a VDT or PC is
preferred for interactive operations or to facilitate displays and
print-outs of related information. As the power of PCs has
increased, a whole industry has sprung up to provide software to
convert the austere information flows to and from the switch into a
more useful and friendly user interface for traffic, CDR and
management.
Master control
The master control for CO switches
(see Fig. 8-2) interfaces the switching system for internal
maintenance and testing. Elaborate status displays are usually
provided so that active and standby controls, memories, etc., can be
identified, along with made-busy circuits, areas with excessive
trouble counts, and the like. Maintenance personnel must have
available to them the results of routine tests run by the system,
both on test calls and regular calls. Naturally, it must be possible
for additional tests to be run under human direction.
Test result displays should be
designed for easy interpretation; systems of 1ESS vintage compared
cryptic test outputs with a "dictionary" listing the relationship
between test output and one or more system faults for any given
test. This sort of comparison can be done far better by a computer
than a human, with the computer providing a meaningful display.
For simpler interfaces,
teletypewriters are still sometimes used, largely because they are
well understood by the telephone industry. They have a keyboard for
input and a printer for output and a visual record of test sessions.
They can also run from pre-punched paper tape and can make machine
readable paper tape for later input to a computer or another TTY.
However, they tend to be slow, are relatively noisy, have high
maintenance costs, and can hardly compare to the convenience of a PC
supported by an inexpensive printer. Any terminal with a keyboard
has some disadvantages, particularly for those who do not
touch-type, but touch-screens and pointers such as the mouse or
trackball, coupled with menus for easy selection of desired
functions, leave older interfaces at the starting gate. Add the
intricate displays made possible by computer graphics, and the
direction for future user interfaces is obvious.
Line and trunk test panel
While the master control test center
may be optimized to handle the relatively few subsystems common to
large numbers of calls (digit receivers and senders, common
controls, memories, buses, switchover facilities, etc.), a line and
trunk test panel may be desirable to access the more numerous but
less complex lines and trunks. Again, access to and display of
system tests should be available, and any built-in routiners for
lines, trunks, and perhaps certain related service circuits should
also be available for human control. In particular, instruments for
checking telephone sets, coin phones, PBX trunks and the like should
have both automatic and manual control procedures.
In small telephone companies, a
separate local test desk may not be provided; thus the line and
trunk test panel may have to do the entire local test job. In larger
companies, a centralized test desk installation for many switches
may be used, as has been mentioned. Thus the line and trunk test
panel must be able to work under a variety of circumstances, and
share responsibilities with other test centers in a number of ways.
Remote test centers
One of the more important advantages
of electronic switching is the appreciably lower failure rates
encountered. There is a standard story that an electronic office
has a two-entity maintenance force, a human and a dog. The purpose
of the dog is to keep the human from fooling with the equipment.
There is a problem associated with too much reliability, however:
maintenance people do not get enough opportunity to practice their
skills and, as a result, tend to lose them.
Because it would be silly to increase
the number of troubles just to exercise the maintenance force, a
better alternative is being generally adopted for both CO and PBX
systems. The maintenance force is centralized so that it can serve a
number of switches, and these switches as a group generate enough
troubles to keep work force skills at peak efficiency. With suitable
terminals, this centralized work force can access the test and
administration functions associated with any switch as though it
were in the same building.
Major disasters such as the central
office fire in Hinsdale, IL, in May, 1988, which destroyed the CO,
access to the remote test center, and, most important,
communications back into the Hinsdale area including those from the
test center itself to fire, police and telephone maintenance forces
at the scene show the risks which remote test centers may incur.
However, it is far more important to deal with such risks
effectively than to permit hysterical politicians take over system
design.
Most PBXs are provided with an RS-232C
port for test access; used with a modem and a private line to the
CO, they can call and be called by a remote test center. A number of
manufacturers maintain such centers (which AT&T calls RMATS, for
Remote Maintenance, Administration and Test System), on-line 24
hours a day, to support the PBXs installed by their local
distributors. Occasionally, the modem is given an extension number
on the PBX; this saves the cost of a separate private line and, by
forcing calls to that extension to go through the attendant, offers
a degree of security. However, if the system is completely down, the
extension cannot be reached unless power failure transfer is used
from a particular trunk to the test port. When a maintenance group
has to go to the site to fix a problem, it is desirable for them to
have a separate access channel in addition to one used by the
central location.
Perhaps the greatest disadvantage of
remote test access, particularly in PBXs and voice mail systems, is
vulnerability to unauthorized use. Computer hackers, phone phreaques,
disgruntled employees and ex-employees and others can use the
maintenance port to play pranks with the system, steal phone calls,
and do other damage. Complex and frequently changed passwords are
often recommended as a solution to this problem, but limiting
dial-up access to specific times, personally verifying the caller,
and calling back to provide access only to authorized phone numbers
may be more effective.
Chapter 2 discussed some of the
advantages that stored program control offers telephone systems in
terms of record keeping for both customer billing and system
administration. Clearly, the switch itself must accumulate billing
information in order to set up a call, and can easily acquire and
store traffic information as it goes about its job. But this is just
the beginning. The telephone directory, whether on-line or off-line,
not only implies the nature of many system translations, but it also
has direct implications for billing in that it contains the
customer's name and address.
The deployment of 911 systems requires
customer records of this sort to be made available at emergency
switchboards as well as at system control centers; addresses are
particularly important here so that a specific apartment in a large
apartment complex can be identified immediately when necessary.
This problem is particularly difficult when customer service is
provided by the apartment management using a PBX with DID; in such
instances, it is not satisfactory to list the building manager's
name and location as the customer.
Internal to the switching system,
inventory records for circuit boards must be kept. Many systems give
each circuit board a serial number written in ROM and accessible to
the system control. Thus the control knows which PCBs are in which
slots, and can warn craftspeople when an attempt is made to
associate an analog telephone with an ISDN line card, etc. Such
serial numbers can also indicate to the system which release of PCB
is in place, important when software must conform to different
hardware releases.
Automatic and manual test results must
be stored for future reference, retrievable in ways that facilitate
the location of intermittent faults. The importance of historical
records for detecting seasonal variations in system traffic has
already been mentioned in Chapter 2.
It will take some time for the full
advantages of computer control, multiple processors and inexpensive
memory to even approach what it is capable of achieving in terms of
system operation and management.
Human factors requirements have been
mentioned from time to time throughout this book. Operator and
attendant positions, maintenance positions and other monitors and
display presentations all must be designed, checked by human factors
experts, and field tested if a system is to perform satisfactorily.
Short-cuts here, particularly when dictated by management or
engineering considerations, may save money, but often at the expense
of future reputation.
Other variables that need the
attention of the human factors expert are frame heights, fuse panel
locations, insertion and removal procedures for plug-in modules,
color coding for various reasons, labeling apparatus for type of
unit and specific identity, provision of proper illumination, etc.
In older systems, standard procedures developed over the years have
tended to make many requirements implicit. In new designs, if they
are not considered explicitly, decisions made without regard for
human factors may produce unsuspected problems later on.
Perhaps one of the most important
opportunities for system design to benefit from human factors
considerations lies in the training programs and instruction manuals
developed by manufacturers for users, installation and maintenance
people, sales personnel, and others who work with the system. Here
is where the payoff shows up directly.
If the system is designed to serve the
needs of the customers, and if reliability, maintenance and repair
are properly planned, the proper design of training aids will put
the frosting on the cake. The full advantage of good design will be
brought home to everyone involved, and labor costs and down-time
will be reduced and user satisfaction increased. Even the most
automatic equipment is designed to serve people, and must, sometime
or other, be served by people. Failure to consider this at all
stages in the design procedure can lead to disaster, even though the
system is a marvel of technology.
-
Permanent signal
-
Line lock-out
-
CGA
-
Test line
-
ALIT
Click Here for
Answers
1. Is built-in test equipment and
software worthwhile?
2. How are transmission and switching
related in terms of operations and maintenance?
3. How is line testing different from
trunk testing?
4. What does a fuse protect?
5. How are alarm fuses useful?
6. How are permanent signals and
partial dials detected?
7. What is "line lockout?"
8. If a trunk or service circuit is
found to be defective, what should be done?
9. Is CGA as important as it used to
be?
10. What would you think if one trunk
in a group has a very short holding time compared to the rest? A
very long holding time?
11. What is Line Load Control?
12. What is needed to make sure a
system control's order is actually carried out?
13. Because components will sooner or
later fail, how can a system be made reliable?
14. Give three ways circuitry can be
made redundant.
15. How can the path through a digital
switching matrix be tested for continuity and crosses to other
connections?
16. If you have the system connect a
trunk to a test line in a distant CO and you get back a tone at the
right level, is the trunk ok?
17. Will ALIT be important on copper
lines used for ISDN BRI and PRI?
18. How does a switch's master control
differ from the line and trunk test panel?
19. Which is more desirable: to design
a system with such complex displays and printouts that craftspeople
can have the satisfaction of passing a long and difficult training
program before they go to work, or to design a system with such
clear and logical outputs that almost no training is needed to take
responsible action?
[ Top ] [
Next Chapter ] [
Table of
Contents ] |
