The PBX: What It Is, How It Works

Lee Goeller's "The PBX: What It Is, How It Works" was Chapter One of The BCR Manual of PBXs, a ring-bound information service published by Business Communications Review from 1980 to 1990.

To speed downloading, we have posted the chapter in five parts, and configured diagrams to open as separate windows.

• Part 1 The PBX and Its Competitors
• Part 2 How a PBX Works
• Part 3 The User Interface
• Part 4 System Control
• Part 5 The Switching Matrix; Conclusion

The PBX: What It Is, How It Works

Part Two: How a PBX Works

A PBX system, as shown in Figure 1, is composed of four basic parts; switching matrix, control, user terminals and trunks. The PBX proper includes the control and matrix, along with matrix interface units, usually called line and trunk circuits, which terminate the transmission facilities that extend to user telephones or other switches. In addition, one usually finds "service circuits" in a PBX to apply ringing and call progress tones (busy, reorder, etc.), and to assimilate caller signaling information from telephone dials and tone pads, converting this information to something the control can use.

[ Figure 1. The basic elements of a PBX ]

No matter how complex a modern PBX may become internally, it still works very much the way the old manual switchboards did in years gone by. In a manual switchboard, each line and trunk terminated in a suitable interface circuit which controlled one or more signal lamps to alert the attendant and passed the voice path to a "jack" that permitted easy inter-connection.

When a user picked up his phone, the PBX supplied power to the instrument and, by monitoring this flow of power (usually with a line relay), knew to light a lamp to tell the attendant that assistance was needed. The attendant had, in front of her, a number of pairs of "cords" which could be used to make connections to lines and/or trunks. These cords performed the precise functions that a modern PBX performs with its switching matrix. The attendant plugged one cord of a particular pair into the jack associated with the calling line and operated a switch that connected her headset to the cord circuits. She then said something like "Number, please" so that the caller would know she was ready to respond to his communication needs. A modern system, of course, returns "dial tone" to perform the same function, even though most modern systems have replaced rotary dials with tone or digital pads.

The caller then gave the attendant the number he wished to reach. If it was another extension on the PBX, the attendant took the other cord of the pair and made a "busy test" on the called line to make sure it was not busy on another call. If the called line was idle, the attendant then completed the connection to the appropriate jack; in very early systems, she would then ring the called party by operating a ringing key that applied the ringing signal to the called telephone via the cord circuit. Later systems started ringing immediately upon completion of the connection.

Either the attendant or automatic circuitry detected answer, terminated ringing, and left the calling and called station users to converse. If either of them needed further assistance, they flashed their switch-hook. That is, they depressed and released their switch-hook. Each of the two cords in the pair making the connection had a lamp associated with it. Momentary depression of the switch-hook caused the lamp to come on momentarily. This "flash" signaled the attendant to reconnect her headset to the cord circuit to see what assistance was required. When the call was completed and the parties hung up, the cord lamps lit to let the attendant know that the call was over and the cords can be "pulled down" and made available for another call.

There are still a few manual PBXs in use. They serve their customers well and, because they consist primarily of a very small switching matrix (the cords) and simple line and trunk circuits, they occupy little space and are relatively inexpensive. But perhaps their main advantage, in addition to low cost, is their control. They have the most advanced control available today, one that is quite flexible and very smart: a human being. A human attendant at a manual switchboard can provide almost all the "modern" features associated with the new, computer controlled PBXs. Attendants, available in common to all lines and trunks, meet all the requirements of what is called "common control" in modern switching systems.

However, people are getting more and more expensive, they like to go to lunch in the middle of the day, they want to go home at night, and they require training. Further, they sometimes forget to perform some of the more complex functions, and occasionally arrive at work late or, when sick or indisposed, don't get to work at all. Thus, the trend today is to duplicate as many of the functions of the attendant as possible with automatic equipment, often in the form of a computer acting as a common control.

Extension-to-extension calling

Many modern PBXs still detect the flow of power when the telephone user picks up the handset. The calling line is then identified and its Class of Service (COS) for originating calls is checked. The PBX then establishes a connection automatically to a digit receiver (dial pulse, DTMF, digital, as identified by COS), which returns dial tone. The caller then dials or keys the called number into the digit receiver (sometimes called a register, a decoder or something similar). The system makes a busy test on the called line and, if it is free and the call permitted, rings the called line. Upon detection of answer, the calling and called parties are interconnected via the switching matrix for conversation; the system monitors for switchhook flashes that indicate the need for some additional service, or hangs up to free the portions of the switching matrix used on this call for future calls. The several parts of Figure 2a and Figure 2b suggest the detailed operation of an extension-to-extension call.

[ Figure 2a: PBX Operation for an Intra-Switch Call ]
[ Figure 2b: PBX Operation for an Intra-Switch Call (continued) ]

A major point of complexity, far more important in PBXs than in most central office switches, lies in terminating the call to some line other than the one requested. The called line's COS for terminating calls, in addition to identifying the kind of ringing to be used, alerts the system to ways in which the call may be completed. Hunting is usually available and widely used in PBXs so that, if the boss's line is busy, the call will be routed to his secretary. Modern PBXs have added a variety of call forwarding features that can be invoked and canceled by the station user. Sometimes privacy features require the return of a special tone to indicate that the called party does not wish to be disturbed, or that data or facsimile is being transmitted. Pick-up is a relatively new feature that allows any station in a previously defined pick-up group to come off hook and snatch away a call that is ringing unanswered at another station in the group. Camp-on and call-waiting may stack up calls waiting for an existing conversation to end; automatic call back may have one or more callers waiting for an existing call to be completed so that the system can call them back and then complete to the called party. Obviously, a call encountering camp-on, call waiting or call back at the terminating line will be dealing with a situation that has several levels of "busy" to contend with. A simple test to see if the line is off-hook is no longer sufficient. From all this, it is easy to see that the terminating half of the call set-up procedure can be quite complex.

Outgoing CO trunk calls

Trunk calls are a little more intricate. As Figure 3a and Figure 3b show, an outgoing call starts off exactly the same way an extension-to-extension call does, but now the control can complete its part of the job by connecting to any one of several trunks to the local central office (or, perhaps, some other switch). Further, it can, if it wants to, do this long before the caller has finished dialing. Traditionally, callers on a PBX dial 9, get dial tone from the CO, and then dial the telephone number of the called party. This approach was very important in early automatic PBXs that were constructed from relays and other electromechanical components. It let the PBX complete its part of the job quickly and easily and turn all the hard work over to the CO. The CO would then store the called number, set up the connection, handle the billing, etc.

[ Figure 3a: PBX Operation for an Outgoing Call ]
[ Figure 3b: PBX Operation for an Outgoing Call (continued) ]

There was just one little problem with all this. Even in the dim misty past, PBX station users tended to call Aunt Minnie in St. Paul, Uncle Harry in Los Angeles, and the bookie joint in the nearby center of iniquity. They ran up, as a fringe benefit, a considerable bill in message units, short-haul tolls, and even long-haul tolls. This led to the development of toll diversion and restriction. Toll diversion was and is a CO feature. When the CO's class of service for the PBX trunk indicates toll diversion and the caller tries to make a toll call (which the CO can readily detect), the CO returns a signal to the PBX that terminates the call. The signal (often a battery reversal) can cause the PBX to either connect the caller to the attendant, or to a tone.

Toll diversion, in many instances, was not a very good feature. Often, station users had to make many legitimate calls to customers, vendors, etc., served by COs near enough to be close but just far enough away to require a toll connection. Toll diversion meant all such calls had to go through the switchboard attendant, and the whole purpose of having an automatic PBX was to eliminate as many attendants as possible. Thus the restrictor was born.

A restrictor was a box full of switching components located at the customer's premises. Each trunk to the central office passed through the restrictor so that it could monitor the digits sent to the CO and check them against a "list" of approved numbers. If someone dialed a call to the wrong central office, the restrictor would bounce the call very much the way the CO would have with toll diversion. The difference, however, was in the way in which the permitted calling area could be specified. It wasn't a matter of toll or local. Selected COs could be called, even though toll charges were involved, and other COs, even though local (often residential) areas, could be blocked to staunch the outflow of message units.

The first restrictors did "three-digit" screening. That is, after the PBX caller had accessed the CO, the restrictor would monitor the first three dialed digits as they passed by. These would normally be the "office code" of the called number. Later, "six-digit" screening became necessary: the restrictor looked at an area code as well as an office code to see if the call should be allowed. Restrictors got more and more complicated, but they provided some measure of relief from employees placing personal calls.

Once one started adding the complexity of a restrictor to a PBX installation, it became fairly obvious that a new PBX design might well incorporate the restrictor function into each line's originating class of service. Further, as foreign exchange (FX) lines and WATS lines became more and more popular, it became evident that a PBX might examine the dialed number even before seizing a trunk to the CO. In that way, it could select a trunk itself, taking the choice out of the hands of the caller.

There is an additional advantage to performing these functions internal to the PBX: the PBX knows which extension is originating the call. Thus, in addition to having the potential for billing calls to specific extensions, the PBX can automatically perform restricting and routing functions in different ways for different people, based on the class of service assigned to their extension. External restrictors treated all calls alike since they had no way of identifying the caller, and external call routers, quite popular during the last ten years, required the user to identify himself by dialing in an additional group of digits.

Two ways of handling outward calls became possible; cut-through and register-sender. With memory and control capability being relatively inexpensive in modern PBXs, the register-sender approach has much to recommend it. However, many manufacturers still cling to cut-through. Note carefully the difference: with cut-through, the system connects as soon as possible to the CO and, in certain circumstances, monitors the digits as they go past. With register-sender operation, the system takes in the entire number; if the caller's class of service does not require restriction, the PBX then chooses the proper trunk and sets up the call.

In setting up an outgoing call, a register-sender PBX must know when the CO is ready to receive digits, when the called party answers and when the distant party has hung up. In cut-through operation, the user hears CO dial tone and, ultimately, the called party answer (or busy tone or what-ever); this relieves the PBX of the need to do anything. Further, the calling party knows when the distant person says goodbye. With register-sender operation, the PBX cannot depend on the user to dial into the CO, but must know when the CO is ready to receive digits. Further, the PBX must be able to detect hang-up at the far end automatically, particularly if the internal extension has been put on hold; otherwise, a path through the switching matrix and a CO trunk may be tied up indefinitely.

When call detail recording is added to the PBX, the inability of the system to determine when a call is answered is a severe problem in making accurate toll bills, whether cut-through or register-sender is used. On the other hand, machine detection of hang-up is fairly simple if "ground start" rather than "loop start" trunks are obtained from the CO; ground start trunks also work well to provide an automatic start-dialing signal in register-sender operation although, for a variety of reasons, detection of dial tone directly has much to recommend it.

Incoming CO calls

Traditional incoming calls from the CO are quite different from the intra-PBX and outgoing calls we have examined so far (see Figure 4a and Figure 4b). It must be recalled that a PBX trunk is, as far as the CO is concerned, a station line just like the one that goes to a telephone. Thus the CO, when it wants to complete a call to somebody served by a PBX, will send ringing down the "line" (viewed from its end) or into the "trunk" (when viewed from the PBX end). Ringing (which is a large, high power signal at 20 Hz and 86 volts) operates a "ring-up relay" or similar device in the PBX's trunk circuit to make some kind of indication to the system that a CO call is coming in. This indication is forwarded to the console position where the attendant can see and respond to it. When the attendant signals the system control that the call is to be accepted, ringing is tripped and the call is connected to the console. Tripping ringing causes the CO to put its equipment into the talking state and to start charging the caller. Thus, the calling party pays for the call if it reaches the PBX attendant.

The attendant obtains the called extension number or, more often, the name of the called party, and instructs the control to manipulate the switching matrix to complete the call. The control causes ringing to be applied toward the called extension, (generally) causes audible ringing (ringback) to be returned to the calling party, and watches for answer. It is highly desirable that a PBX monitor this ringing situation and return the call to the console after a timed interval if no answer is obtained. It is also desirable that the PBX be able to monitor the trunk circuit for abandon in case the calling party hangs up. "Ground start" trunks are required for this to be possible. As on any call, the system must monitor the internal party for switch-hook flashes and hang-up.

[ Figure 4a: PBX Operation for an Incoming Call to the Attendant ]
[ Figure 4b: PBX Operation for an Incoming Call to the Attendant (continued) ]

In many systems, the same CO trunks used for incoming calls are used for outgoing "dial 9" calls. This poses a problem that must be understood if serious trouble is to be avoided. The trouble concerns seizure of the same trunk simultaneously by both the CO and the PBX. This situation puts the PBX caller, who has just dialed 9, in direct contact with an incoming call which is almost certainly intended for someone else.

Simultaneous seizure can be minimized at the PBX if the trunk circuit is made busy to dial 9 calls as soon as the CO has seized it from the far end. This cannot, in general, be done if all the CO does is apply ringing, as on a "loop-start" trunk. Ringing, in most central offices, is on for two seconds and off for four. Thus, if the CO seized the trunk at the start of the silent interval in the ringing cycle, up to four seconds could elapse before the PBX trunk circuit would know. In a busy hour, four seconds is eternity and many dial 9 calls would have a crack at the "idle" trunk. Thus, once again, "ground start" trunks are necessary. They tell the PBX immediately when the trunk is seized from the CO, and thus allow the PBX to direct outgoing calls to other circuits. Trunks used in the outgoing direction for dial 9 and incoming for completion to the attendant are usually called "combination trunks." With present PBX and CO designs, it is imperative that the CO use ground-start circuits. With digital trunks under ISDN standards in the near future, where signaling is based on CCITT No. 7, a standard data link between the PBX and CO control computers, all this may become irrelevant.

"Immediate ringing" is a feature frequently used in modern PBXs and COs. If a CO has immediate ring, ringing starts as soon as the trunk to the PBX is seized. This might make it appear that ground start on incoming CO trunks is not needed. Unfortunately, this is not correct. The alternative, use of loop-start trunks, does not send a hang-up signal. Thus the PBX has no way of knowing when the CO-connected party abandons before answer or, after answer, goes on hook. If the PBX party remains off-hook after the outside caller has hung up, the CO trunk can stay hung up all night.

Immediate ringing is a good PBX feature, particularly when attendants complete incoming calls and where station users frequently transfer calls. It is usually desirable for the attendant to stay with the call until the first burst of ringing is heard, and such delays can accumulate in the busy hour. If ringing at the called line starts immediately, the attendant can release from the call with some assurance that the system can handle things from there on out. When a station user transfers, immediate ringing not only permits quicker release but also provides assurance that the procedure has been carried out properly. In general, it is no harder to provide immediate ringing in a modern PBX than to have traditional random connection to a 2-on, 4-off ringing cycle. In certain instances, the ringing equipment is used to light message lamps; then immediate ring is traded off for this additional feature. In such cases, immediate application of ringback tone is sometimes used to aid the attendant or transferring party.

DID calls from the CO

Direct Inward Dialing is rapidly making Centrex unnecessary, and most modern PBXs offer it. As mentioned above, DID is not particularly difficult at the PBX. Any PBX that can handle dial repeating tie trunks is ready to go immediately. All that is needed is for a CO to be able to send dial pulses; this is harder to come by. In some metropolitan areas, New York City in particular, tandem offices in the public network bypass the local (Panel) central offices and connect directly to DID PBXs. This lets each PBX be treated as a small central office for incoming calls and everything works fine.

There are, of course, some practical details. All the old-fashioned step-by-step (SXS) PBXs were quite fast in that they could accept dial pulsing on a tie trunk as soon as the trunk was seized at the distant end. Unfortunately, most modern COs and PBXs are much slower. After detecting seizure, they must find a digit detector to attach to the trunk, or arrange their internal operation to examine the trunk circuit fast and often enough to catch all the dial pulses as they come in. This may take a while under heavy traffic conditions. Thus, the distant end must be held off until the PBX is ready. The technique used is called "wink start." The CO (or tandem office) is psyched up to watch for a momentary "off-hook" signal from the PBX. At the end of this 1/5 of a second interval of off-hook, the telco end knows it can start sending dial pulses.

Some modern PBXs, with microprocessors built into each trunk card, are always watching their DID trunks (and tie trunks) for incoming dial pulses, and thus are as fast as SXS systems. However they sometimes have to return a wink start signal anyway to make the CO happy. Thus, we have two requirements here: we should know if a PBX is able to send wink start to satisfy the CO, and if it must send wink start to fend the CO off until the PBX is ready to receive digits.

When digits are sent to a PBX in a DID situation, the telco sends them at the slowest possible rate: 10 pulses per second, with something more than half a second between trains of pulses that constitute a digit. Thus, to send a four-digit extension number to a PBX, the CO will require at least four seconds. If DTMF were used, about half a second would be needed. There is a standardized form of dial pulsing, long used between common control switches in the public network and between PBX attendants and common control COs, in which the time could be cut to two seconds. But for reasons that are unknown, the slower form of dial pulsing is used instead. This is particularly amusing in that AT&T is installing, as rapidly as possible, CCIS, or Common Channel Interoffice Signaling, a form of data link that is similar to and ultimately may become identical with CCITT No. 7 used in ISDN. At present, CCIS makes call set-up time in the public network, CO-to-CO, less than four seconds, but without ISDN digital trunks from CO to PBX, DID dial pulsing increases set-up time by just that amount. We run and run just to stay where we are. Maybe ISDN can save us.

A DID call seizes the trunk, gets wink start (if required), and sends the PBX the extension-identifying digits it needs. The PBX then completes the call as indicated in Figure 5a and Figure 5b. That is, the PBX rings the called extension, sends audible ringing (or, if appropriate, busy tone) to the calling party, and monitors for answer. When the call is answered, the answer signal is returned to the CO and the talking connection is established. Note that charging does not start on a DID call until the called party (or somebody else) answers. The telephone company, to be sure it gets paid for the call, will not make the trunk work in both directions until the answer signal is received. The trunk must work in one direction so that the caller can hear audible ringing or busy tone - but it won't work in both directions until after answer has taken place and charging begins.

Note that DID calls use a separate DID trunk group that is one way only—from the CO toward the PBX. However, these trunks can complete to any PBX extension if the control knows what is wanted. In particular, a DID number can hunt to a non-DID extension, and a regular extension can use the call pick-up feature to respond to a call to a DID extension in the same pick-up group. This allows a considerable saving in DID numbers with a negligible degradation in service. With Centrex, one usually had to supply all phones with extension numbers that were part of the nationwide numbering plan. That this was a waste is even recognized to some extent by the telephone industry, particularly as they run out of telephone numbers.

[ Figure 5a: PBX Operation for DID Call ]
[ Figure 5b: PBX Operation for DID Call (continued) ]

Tie trunk calls

Tie trunks have been used between PBXs for many years. In their simplest form, they ran from switchboard to switchboard, and required an operator on each end to set up the connection. In systems using consoles, this is a particularly difficult method of operation, and dial-up tie trunks are now used almost exclusively, when they are used at all.

In electromechanical PBXs, one usually dialed an access code for the tie trunk group desired, and the system operated cut-through to complete the call. In SXS systems, this was a good idea, as a complicated tie trunk network would use a variable number of digits depending on source and destination; when the caller dialed the last digit, he was connected to the called telephone. Any system that operates in any other way must have a numbering plan containing a (more or less) fixed number of digits so that the system knows when it has all the digits it can expect. Only then can it start to set up the call. The need for a uniform numbering plan, imposed by this limitation, is touted as a highly desirable feature. It is somewhat helpful to users of complex tie trunk systems in that they can dial the same seven (usually) digit number to reach a given telephone from any point in the network; it leaves something to be desired, however, in that it usually requires several of these digits to be dummies to fill up the register to let the control know dialing is complete.

Tie trunk calls work very much the way DID calls do. One PBX, analogous to the CO in DID, seizes the trunk. The connected PBX fends off the signaling until digit reception is ready, and then sends wink start. The called number is sent from one PBX to the other, and the connection is set up. As long as a PBX extension is called, that PBX knows when answer takes place. It can then send answer supervision (an off-hook signal) down the tie trunk toward the calling PBX. This answer signal lets call detail recording equipment, if used, know that charging can be started. It also does something else.

On long-haul tie trunks and other circuits such as FX lines, implemented with analog transmission facilities, on-hook and off-hook signals are exchanged between switching machines by means of audio tones in a single frequency (SF) signaling system. Tone on means on-hook, and tone off means off-hook. Note that the tone is not present when people are talking on the circuit. Off-hook is sent forward when a trunk is seized, a momentary off-hook transmits wink start in the reverse direction, and a steady off-hook is returned when the called party answers. It is important that the answer signal be returned so that the SF tone can be removed during conversation. The tone is blocked by filters and cannot be heard, even when present, but the filters block out a considerable part of the frequency band in the trunk (you can hear a squeak of tone at hang-up before the filters cut in, if you listen for it).

The frequencies blocked out aren't all that important in voice communication, but they are quite important in most forms of modemized data communication. As long as one dials up a computer that has a regular appearance on the PBX, all is well. But when one reaches the distant PBX, dials 9 or the equivalent to go into the public network, and then the number of a time-sharing service, answer supervision will not be available because it is not, as has been discussed, sent to the PBX in the first place. PBXs must be able to simulate answer supervision on off-net calls (they often time out and, 30 seconds after the last dialed digit is noted, pretend the call is answered) and return that signal by knocking off the SF tone and thus removing the filters.

CCSA, an early form of tie trunk network using stand-alone tandem switches, solved this problem very simply: it did not permit the caller to go off-net through a PBX. One could only go off net through an "ONAL" or Off Network Access Line from one of the tandem hubs. This added considerably to the expense of a tie trunk network and, obviously, to the revenues received by the vendor supplying the circuits.

It should be evident that the design of tie-trunk networks is a relatively complex business involving transmission, signaling, switching, traffic, cost and a variety of other factors. We will not try to cover all these factors here. But something must be said about transmission if one is to discuss tie trunks at all.

The major transmission problem in any voice communication system is echo. All long-haul trunks are "four-wire”; that is, they are made up of two one-way transmission facilities. Customer lines, including CO trunks from a central office to a PBX, are "two-wire," or a single channel carrying information in both directions. Echo takes place at the "hybrid circuit" (no relation to a hybrid key-system/PBX) that interfaces four-wire with two-wire facilities. See Figure 6. The signal incoming from the distant end of the four-wire trunk is reflected from the two-wire trunk with a certain amount of "return loss." The reflected signal then goes back to the distant end of the four-wire trunk on its outgoing channel. This reflected signal is perceived at that point as echo. How loud the echo is depends on the loss in the four-wire trunk plus the return loss at the interface to the two-wire trunk.

[Figure 6. A hybrid circuit to convert between 2-wire and 4-wire transmission.]

If the return loss were very high, there wouldn't be much echo left. Unfortunately, return loss depends on the ability of the network associated with the hybrid to match (or correspond, electrically) the properties of the two-wire line. Since a long-haul trunk may connect to any line on the terminating switch, and variability of the electrical properties of these lines is considerable, return loss is often quite low. Thus the echo it produces is controlled by loss purposely introduced into the four-wire trunk if that trunk is less than about 1,800 miles long, and by echo suppressors or echo cancellers if longer.

Echo bothers a caller more and more the longer it is delayed. Since delay is a function of how long it takes a signal to go down the four-wire trunk to the far end, bounce, and come back, the effects of echo are worse for long trunks than short ones. This is a subjective factor, and has been studied for decades. The loss required (which is added to the statistically distributed return loss of the connecting two-wire circuits, is called Via Net Loss (VNL), and can be found for any given distance and type of facility. Unfortunately, to reduce the echo sufficiently, VNL is not quite enough, and an extra amount of loss, approximately 4 decibels (dB), must be added. This is divided up to provide 2 dB in each end of the connection, and loss in an end-to-end tandem connection through several trunks is described as being VNL + 4.

(A decibel is, very roughly, the minimum change in loudness that a human ear can detect. Further, each 3 dB of loss means the power of the signal has been cut in half. One uses decibels to describe very large power ratios with relatively small numbers. For instance, 60 dB of gain or loss means a million to one ratio of power level between input and output.)

The problem comes when a PBX must switch one tie trunk to another on some calls, while switching the tie trunk to an extension on other calls. In the former case, the extra 2 dB loss must be omitted, while in the latter, it must be inserted. Loss used to be provided by "pads" or attenuators in the old days. Thus, the act of changing loss for a terminating verses a through connection is called "pad switching." In any PBX that is going to be used to connect one analog tie trunk to another, some form of pad switching is required. The nature of the problem, however (to say nothing of the nature of the solution), is not always understood by PBX designers.

It should be noted that, in the all-digital world of the future, none of the above applies. The ISDN will maintain bit and byte integrity end to end, permitting data as well as voice to move through the system transparently, with no change in coding which night represent an amplitude or a character in a code like ASCII. When analog two-wire voice lines or telephones terminate a digital connection, echo-controlling loss will be introduced as needed. But then or now, perhaps the most satisfactory solution to the problem is simply to eliminate echo by maintaining the four-wire connection all the way to and through the telephone set itself. If transmission from mouth to ear going west went by a channel completely separate from a similar channel from mouth to ear coming east, there would be no echo except some minor acoustical feedback that might come out of the receiver and get back into the transmitter of the four-wire telephone set. (Such four-wire integrity all the way to the telephone set is another part of the ISDN spec.)

Meanwhile, back at the ranch, there are still a great many PBXs using 500- and 2500-type standard telephones, 1A2 key systems that contain the same two-wire transmission package, and electronic sets, either proprietary or general purpose, that are still based on two-wire principles ("that's the way we've always done it!"). Even when a fifteenth generation PBX is used, these sets will cause echo problems, and will require echo-controlling loss to be switched in, even with digital trunks.

In our present old-fashioned analog word, there would have been an enormous cost penalty for using four-wire transmission to residential telephone sets. The reason is, of course, the additional copper wire needed to go from the central office the several miles, on the average, to each telephone. The capital investment for this wire is something like 60% of the total value of the local telephone company. Thus each phone uses a single pair, and local central offices (SXS, Crossbar, and Nos. 1, 2 and 3 ESS) use two-wire switching matrices to make connections. Two-wire space-division matrices of these types cost much less than four-wire matrices.

For a PBX, however, going four-wire to the set would not impose such a cost burden. In the first place, most PBX extensions are within several hundred feet of the PBX, so the amount of wire is greatly reduced. Second, the cost of PBX wiring lies mostly in the labor to install it and administer it in perpetuity, not in the copper. Third, most business telephones were, until recently, 1A2 key systems, even when behind PBXs, and they used cables with a minimum of 25 pairs. And finally, all digital PBXs must do their switching on a four-wire basis; unlike space division switches, they cannot be made cheaper by being two-wire. However, only two PBXs of the new wave in 1975 came on the market with four-wire telephone sets, and they were both analog: Danray and Tele/Resources.

As we will see, most of these PBXs, and all of the newer ones coming on the market in the 1981+ time-period, now offer at least some electronic telephone sets that maintain the four-wire integrity end-to-end. Because of the advances made in digital technology, they often do it on single pair wiring. ISDN standard channels to telephone sets also imply single pair wiring for full duplex voice and data to the telephone set. The most interesting show to watch during the next few years will be the rush to ISDN standards from the hodge-podge we currently enjoy.

[ Top ] [ Next ] [ Table of Contents ]

Copyright 2006 Lee Goeller. All Rights Reserved.