Voice Communication in Business Volume 1
Essays on telecommunications, 1969-1980

Chapter 9
Those Awful PBX Proposals:
System Descriptions

It is just possible that there are some people out there in business communication land who are not as fascinated by switching systems as I am. Hard as it may be to understand, there are things these poor unfortunates don't want to know about the systems they are considering for their very own. Although sometimes they are right (there are things about the inner structure and workings of a PBX that boggle even the minds of system designers), there is a need to know certain basic things about what is being purchased. One must never allow the need to avoid the infinite unimportant details to team up with a vendor's inability or unwillingness to provide vitally important facts to obscure the difference between quantity of paper and amount of information in a proposal.

In chapter 8, we saw what a proposal probably ought to contain. In this thrill-packed episode, we'll take a closer look at details: what should a system description include? We'll leave features for later, and concentrate now on the system itself. Of course if you'd rather not know ... but that is defeatist talk.

Table 2 listed some of the more important aspects of PBX systems: size, matrix, control, console, trunks, instruments and station wiring. We'll take a look at each of these items to see how it affects your costs, your operation, and your possibilities in the future.

Size

In a PBX, size is important, but there are many different ways that it can be measured. How many lines can the system handle at maximum? How small does the manufacturer feel he can go and still be economical? How many central office trunks are possible? How many groups of trunks can be provided?

We have already considered the importance of knowing the difference between "wired" and "equipped" in a specific proposal. Related to this, we need to know how many line and trunk circuits are on each plug-in unit, and how much such a unit costs. This gives us an appreciation of short-term growth costs. For longer intervals, it is important to know the system's maximum size. But you must have a feel for the way in which this size is reached. In particular, how many lines and trunks, along with the related switching matrix and service circuits (tone sources, DTMF receivers, etc.) will fit in the first cabinet and each additional cabinet? Is more control needed? Power? How many cabinets, maximum, does it take for a full sized system, and how big are the cabinets?

In addition to the PBX cabinets, the switch room will require a main distributing frame on which to terminate wires to the station equipment. How big is the MDF? Is it wall mounted, or does it stand alone? This all leads to the space requirements for the PBX switch. How much space, and what shape? Will the present switch room do? Will the present switch have to be removed first, and if so, is service interrupted?

There are other properties of the switch room itself. How much weight will it support? Modern packaging techniques permit a very high density of circuit components, and this leads to weight in terms of pounds per square foot. If a switch cabinet, fully equipped weighs more than 160 pounds per square foot, you may be in trouble. In any event, you need the total weight per cabinet and the weight per square foot to check with your building manager. If your PBX ends up on the floor below, it will be a hell of a surprise to the people working there.

Similarly, heat may be a problem. Most electronic PBXs generate a good deal of heat, and are particularly susceptible to excess heat and humidity. Thus, a small closet, suitable for an electromechanical system, may not do at all for an electronic machine. Note also that, come winter, air conditioning is turned off and heat is turned on. Thus, even if the closet is air conditioned in summer, the building heat plus the system heat may be too much for the PBX in the middle of winter.

Traffic loading is another "size" that should be known. There are several ways in which traffic can limit the utility of the machine. You need to know the maximum number of calls that the processor can handle in the busy hour, the average traffic per extension line that the system is designed for (less than 6 CCS, 10 minutes, or 0.167 erlangs per busy hour is not acceptable in most business systems), the number of digit detectors provided for dialing or DTMF keying, etc. You also need to know if traffic balancing is required. We'll return to that one later.

Matrix

Turning next to the type of matrix, there are several kinds of description that can be applied. Is it space-division, frequency-division or time-division? Older systems tend to be space-division, while time-division appears to be the choice of most of the newer systems. There is only one frequency division system, so far: the Collins ATX-101, and it is no longer available.

Space-division systems tend to have somewhat larger switching matrices than time division systems, but a more interesting distinction can be based on the components of which the system is constructed. A "metallic" matrix, using crossbar switches, reed switches, miniature relays or some such, will be much larger than an "electronic" matrix using transistors, LSI, SCRs, etc. However, what is given with one hand is taken back with the other. Electronic matrices must have much larger line circuits, and the net size saving is often a standoff.

For those who are interested, line circuits are much larger in electronic systems because they must perform all the per-line functions required by telephone service: detecting originations, monitoring for flash and hang-up, applying and tripping ringing (particularly if conventional power ringing is used) and, sometimes, providing test access to the pair of wires to the telephone set. Further, an electronic matrix must be isolated from the outside world by a transformer on each end of the connection. This transformer, of course, will not usually pass DC dial pulses or DC line supervisory current or operating power, and it will not pass power ringing. In a metallic matrix, by contrast, all a line circuit must do is detect originations and then be disconnected. The line can be connected directly through the matrix via a balanced, metallic path to ringing circuits, dial pulse detectors, trunk circuits that can feed battery to the station and monitor for hang-up, etc. There are no transformers in the path, and the continuity of the path can be checked as a function of setting up the call. Thus, the metallic matrix has some advantages, even in the modern world.

If we have a time-division system, there is even more that must be known. Is the system analog or digital? When pulse height represents the amplitude of the voice signal sampled at regular intervals, and that height can vary continuously between the maximum and minimum limits, the system is analog, even though it is sampled and pulsed. Dimension is a case in point, using pulse amplitude modulation or PAM. The No. 101 ESS also used PAM.

Another analog time division system is the Chestel PBX, which uses pulse width modulation (PWM). Here the width of each pulse represents the amplitude of the voice signal at the sampling instant. PWM can be used with space division as well as time division systems, as in the case of Danray. PWM coding makes use of electronic switches more satisfactory.

Delta modulation is one form of true digital operation. In delta mod, each pulse is either present or absent; if present, it says that the voice signal is bigger this time than the last time, while if absent, the voice signal is smaller. If the system looks at the voice signal 32,000 times a second, for instance, the increase and decrease can be tracked quite accurately and economically. The PBXs developed by Digital Telephone Systems and marketed by Executone use delta mod.

The big runner today, however, is pulse code modulation, or PCM. Here the voice signal is sampled somewhat less frequently than in delta mod (usually 8000 times a second), but the height of each sample is coded into one of several hundred levels. These levels are then defined by a binary number consisting of pulses present and pulses absent (1s and Os). Note that by having discrete levels rather than continuous variation as in PAM, we don't reproduce exactly the input signal. However, by going from analog to digital we can greatly simplify noise elimination, and there is no further deterioration of the signal due to attenuation, amplitude distortion or phase shift.

Unfortunately, there are several kinds of PCM. Wescom and Automatic Electric, for instance, use PCM similar to that found in T-carrier, a widely used American transmission system. Northern Telecom, in the SL-1, uses PCM of the form found in the European version of T-carrier. Rolm, on the other hand, uses a completely different system which requires 12 bits rather than 8 to code each sample, and uses 12,000 samples per second rather than 8000. All of these systems should be able to interface T-carrier "span lines" directly at some time in the future, eliminating the need for carrier terminal equipment at the PBX and permitting a great reduction in wire between the PBX and the telco central office.

The point to remember, however, is that all time-division systems are not alike, all digital systems are not alike, and all PCM systems are not alike. Fig. 1 may help to clarify all this.

There are other ways of describing switching matrices that may have profound effects on how you implement your system. For instance, most time-division systems are, by their nature, four-wire internally. That is, they have one talking path from the calling to the called party and another from the called to the calling party. Thus, if you have tie-lines, you should be able to make through connections on a full four-wire basis. This is not, however, always possible.

Consider the SL-1. It is a four-wire system internally, but has no four-wire tie trunk circuits available as yet.* There may be some offered late in 1978 if you can wait, but in the meantime, tie trunks enter the switch-room four-wire (all transmission facilities going any distance at all today have separate paths for each direction of the transmission). The signal then passes through a "hybrid" coil to become two-wire to enter the SL-1 where another hybrid splits the signal apart for switching. At the other side of the matrix, two more hybrids back to back are provided, defeating the whole purpose of four-wire transmission.

The Dimension, a two-wire, analog time division system, accepts tie trunks on a four-wire basis, but switches them two-wire. The Danray switch and the Tele/Resources System 32 have four-wire station sets, and make four-wire connections between all possible connecting circuits and devices. This is the ideal situation, since the microphone and ear-piece in the telephone set are separate; when echoes are eliminated by four-wire end-to-end transmission, both voice and data transmission are greatly improved.

The last thing to mention here about the switching matrix is traffic balancing. In most small systems, traffic balancing is not required since all inputs have equal access to paths through the matrix whether these paths are time slots, frequency channels, or actual physical connections. As systems grow larger, however, lines and trunks are arranged in subgroups, and these subgroups act partially autonomously. In particular, if you put a lot of heavy telephone users in one subgroup, they may not always be able to get a connection through the matrix; at the same time, light users, on another subgroup, may never fully exercise their part of the equipment. Traffic balancing is mixing light and heavy users in such a way that each subgroup handles about the same amount of traffic—with that amount kept within the capability of the system.

Small systems such as the Dimension 400, the Tele/Resources System 32, and the smaller Automatic Electric GTD systems do not need traffic balancing. The Womack, however, does. The SL-1 is very traffic sensitive and requires even greater care in traffic balancing. The Wescom 580, a large PCM time division switch similar to the SL-1 in that it has many subdivisions, has been designed to be non-blocking and does not require traffic balancing for that reason. With ever increasing traffic loads produced by data circuits as well as voice and facsimile, the ability to assign any line or trunk to any matrix port without worrying about traffic balancing can be a real advantage. Thus, you have two questions to ask: Does the matrix require traffic balancing? Is it non-blocking?

Note that a non-blocking matrix may still have trouble handling traffic. If all trunks are busy, for instance, the call won't get out even if the matrix doesn't block. Similarly, if there are no DTMF digit receivers available, a call will have to wait for dial tone. And, finally, if the processor can't handle the number of calls requesting connections, the non-blocking nature of the matrix won't help.

Control

Turning now to the system's control, it is a good idea to know the type of processor being used; if it is duplicated for reliability (usually only in larger systems); the amount of memory used; and what the strategy is when power is lost. Unless back-up batteries are provided, loss of power will kill the system, momentarily losing all calls in progress. If the system's working memory is "volatile," all the call-in-progress data will have to start from scratch.

Some systems use volatile memory for operation, and wired-logic or read only memory (ROM) for the basic program structure. The Digital/Executone systems work almost exclusively with ROM and class-mark switches on line cards. This is good in that loss of power only erases current call memory; the entire operating program is intact and ready to go when power comes back. However, program modifications are relatively hard to make. New ROMs have to be prepared at the factory and plugged in on site.*

[*FOOTNOTE: Newer versions are more flexible]

Dimension, SL-1 and others have a back-up tape and volatile memory. After a power failure, the program is reloaded from tape and the system can take off. This takes time, however (perhaps as long as five minutes in a large system), and may be necessary after even a momentary "glitch" in the commercial power.

The Womack has a tape memory, but also monitors its power supply so that power failure, which takes several hundred milliseconds to become complete, is detected early. In such an event, the system, stops and puts its entire memory contents into nonvolatile memory to await restoration of power. When power returns, it takes off from the exact spot where operations ceased.

The Rolm CBX and the Siemens SD 192 use yet another approach: a small battery supply is available for the system memory. As long as the battery is available, the memory is, for all practical purposes, non-volatile. The battery is normally kept charged by the commercial power, and only goes into action when commercial power fails.

From the above, it may appear that it is more important to know how a control system handles power failure than it is to know how it works. This is largely true. Component reliability and system operation are pretty well worked out and aren't half bad these days. And besides, there isn't much the user can do about them, anyhow. But power failure, like the poor, we have with us always.

Control architecture has many variations. Sometimes one processor handles the whole system; as has been indicated, for larger systems, it is duplicated for reliability. Other systems have a small processor for each cabinet, under control of one larger processor. It is also possible to have several computers sharing the load. Wescom has taken a novel approach: distributed processing is used where each processor handles a particular group of functions—monitor lines, monitor trunks, monitor matrix, monitor signaling, monitor consoles, and monitor data base, TTY and other access terminals. Having separate processors for each area presumably simplifies the over-all programming.

Consoles

Consoles are important in all PBX systems, so one question that should be asked is if the system, particularly when equipped with universal night answer or direct inward dialing, can operate without a console. Sometimes the system alarms appear only on the console and, without it, are unavailable. It should be noted that in modern PBXs, most stations have almost as much power as a console, and the electronic key telephone sets presently available for SL-1 and Dimension and soon to appear on other systems have visual displays that are almost as complete.

Displays are important in any event, and not just because of what they can tell the attendant. The important question to ask is what technology provides the display: LED or incandescent lamps. LEDs have very long lifetimes and thus do not have to be replaced so often. Thus, the console designer can use more LEDs with the same reliability and make the system easier to learn by replacing one lamp having several flash rates with several suitably labeled lamps. Dimension’s console, in particular, takes full advantage of this philosophy.

If a console has conventional lamps, some means for making periodic tests is vital since failure of a lamp to light may be due either to no activity or else to malfunction. Usually there is a test button to push that lights all lamps and operates the buzzer or other audible indicator. Be sure such tests are possible on the system of your choice.

A major distinction between console types is "key-per-trunk" and "switched-loop," as illustrated in Fig. 2. Usually older systems, particularly those using electromechanical switching, used key-per-trunk. The console looked very much like a large call director, and to answer any trunk, the attendant would observe the blinking lamp that indicated ringing and would then depress the trunk button. This connected her directly to the trunk; after obtaining the called extension, the attendant could then instruct the system as to how to complete the call. Key-per-trunk permits observation of traffic very easily; you can observe directly which trunks are busy and which aren't, and when they are all busy, you know it.

Switched-loop consoles, by far the more common these days, have only a few "loops" available for connecting to lines, trunks, etc. When a trunk call comes in, it is switched to one of the switched loops and the loop's lamp gives an indication. The attendant then proceeds as before. With a switched-loop console, the number of trunks permitted is not limited by the number of console buttons; however, failure of a specific trunk may not be discovered for some time.

One advantage among many in switched-loop operation is a great reduction of wires to the console. Most of the newer systems run 25 pairs or fewer between the switch and the console while older systems might run 400 pairs or more. The smaller cable permits greater flexibility in the location of the console relative to the switch, greater ease in moving the console if required, and easier maintenance.

Extension status lamps and direct station selection are features commonly found. DSS was more desirable prior to the availability of DTMF signaling than it is now; when the console attendant had to dial up all connections with a rotary dial, a lot of time was wasted.

Thus, the ability to complete the call by simply depressing the extension button was highly efficient. With DTMF, however, the extension number can be keyed in almost as easily as a particular extension number can be located among 100 or 200 buttons. This is an advantage when line hunting is used; most modern PBXs do not require consecutive numbering for hunt groups, but DSS suggests that hunt groups should be so numbered. If all you have to do is key in the requesting extension and let the system do the hunting, full advantage of flexible numbering can be taken.

A console should display the identity and class-mark of the calling line or trunk, and the identity and status of the called trunk or extension. Further, the nature of the call should be identified: new call, recall by station, ring-no-answer time-out, etc. Such displays speed call handling and eliminate the need for per-line displays which, as in DSS, give only a limited amount of information.

Note that some PBXs, unsure of the effectiveness of stored program control, hold a call on a switched loop until it is answered. This makes it easier for the system designer to return a camp-on or a ring-no answer call to the attendant, but can completely tie up the console during lunch time or some other period when many people are not answering phone calls. More advanced PBXs use "released loop operation," and free the switched loop immediately upon instituting ringing or camp-on. If the called party, camped-on or ringing, does not answer within a specified time, a software link re-connects the call to a switched loop to the attendant. When it is important for the console attendant to identify the particular caller, a feature called "serial call" can suspend released loop operation and hold the console loop until released by the attendant. Serial call also facilitates additional connections to other internal extensions when the outside caller wishes to contact more than one person.

Many current systems have additional special purpose consoles for various applications. The most common is the special maintenance console that allows testing and monitoring of the system. However, consoles for indicating the calling extension are popular for "room service" applications in hotels, and hotel message registers have been replaced in many cases by electronic memory and a special readout console.

The most general capabilities of consoles have hardly been explored as yet. Northern Telecom pioneered the use of the standard cathode ray tube data terminal for a console in their TOPS system intended for toll operators. With a full alphanumeric capability to and from the system control, considerable flexibility is possible. One voice and one TTY channel connect the TOPS position to the switch; the TOPS positions can be located individually or in groups, anywhere that voice and TTY signals can be sent.

In the PBX area, the Danray is the only one so far to use a CRT console for the attendant (many use a CRT console for maintenance access). Taking advantage of full alphanumeric capability, directory assistance is built into the system. Even at the largest sizes, the caller can ask for an employee by name: the attendant keys in the first two letters of the name and gets back the extension number almost immediately.

CO trunks

A 5,000-line PBX, for reasons known only to the telephone company, is a piece of "station apparatus" like a Princess phone. As such, it must, with rare exceptions, connect to the public telephone network just as a telephone does. That is, it comes in as one or more

lines at the central office (CO). Because the design specifications of the CO switch require each per-line appearance to be kept to minimum cost, the complex burden of the interface lies at the PBX end of the circuit. A line at the CO becomes a trunk at the PBX and, in the old days, PBX trunk circuits might have as many as 50 relays in them.

Today's PBX trunk circuits are somewhat simpler, with most of the 50-relay complexity built into the computer program. And the creation of interfaces to "protect" the public telephone network from evil interconnected equipment has changed and, to some extent clarified, signaling requirements at the expense of a vast multiplication of wires at the meeting point. In any event, there are several things to be known about the CO trunks themselves. We'll consider interfaces a little later.

CO trunks can be one-way or two-way; that is, they can be seized from only one end, either the CO or the PBX, or both ends. A two-way trunk group can, of course, carry more traffic for the same grade of service than two one-way trunk groups with the same total number of trunks. But with two-way circuits, there is always the possibility that both the PBX and the CO will grab the same trunk simultaneously and "glare" at each other to the confusion of both connected callers.

When any trunk is seized, it must be immediately made busy so that it will not be seized again for another call. With a one-way trunk, where only one switch can make the seizure, the problem is relatively simple: the switch notifies itself that it is using a given trunk and to go on to the next trunk in the group for another call. With two-way trunks, both the switch that has seized the trunk and the switch at the distant end must make the circuit busy to further use. If the far end doesn't know that the trunk has been seized, trouble will follow.

Seizure from the PBX end causes a circuit to be completed and DC current to flow. Seizure from the CO end, to match seizure of a line to a Princess phone, consists of the application of ringing, an AC signal not unlike what comes out of a wall plug to run a toaster. This signal is on for two seconds and off for four in a conventional ringing cycle; if ringing is connected to the line during the "silent interval," the PBX may not know it for as much as four seconds. In a busy PBX, four seconds is an eternity and half a dozen calls may come up desiring to use the circuit seized so long ago by the CO.

There are just two ways to prevent double seizure and glare. The easiest is to use only one-way trunks: the PBX has its group to use outgoing, which cannot complete incoming calls from the CO, and the CO has one-way circuits toward the PBX which the PBX cannot seize. The more difficult approach is to use signaling between PBX and CO to minimize the "unguarded interval" when one end can grab a trunk already seized by the other. In this approach we can have "combination trunks" at the PBX which the CO can seize to reach the PBX attendant, the attendant can use for outgoing calls, and which station users can access outgoing by dialing 9.

To make combination trunks possible, it is necessary to add a different kind of line circuit at the CO. In keeping with the standard telephone industry policy of making per-line circuits as inexpensive as possible, a very slight modification is made: the CO line circuit is designed to expect one wire to be connected to ground at the PBX rather than to the other wire to signal a seizure. Thus we have a "ground start" trunk rather than a "loop start" trunk. For those who like circuit diagrams, Fig. 3 shows the general idea.

The alert reader will now doubtless be wondering how a ground at the PBX, when the PBX originates a call toward the CO, will help the PBX know that the CO has seized the trunk for a terminating call and will soon apply ringing. The answer is simple. Ground start works both ways. From the CO to the PBX, it simply uses the other wire. On a terminating call, the CO disconnects the line circuit, connects through a metallic switching matrix to a connector, trunk circuit or ringing circuit, and returns ringing on the lower wire in the diagram and, more to the point, applies ground to the upper wire. This is shown in Fig. 4. Ground is applied immediately, telling the PBX to make the trunk busy and to expect ringing eventually.

Note that ground on the top wire is also applied on originating calls when the line is connected through to a dial pulse or tone receiver for dial-tone. Thus the PBX, when it originates a call, can use this ground as a start dialing signal. On either originating or terminating calls, however, grounds at the PBX are removed during conversation to minimize noise pick-up and a connection is made from one wire to the other.

All of this is important for two reasons. First, it shows that a PBX must be able to operate with both "loop start" and "ground start" trunks. Ground start is required for two-way (or combination) trunks, and also for outgoing trunks when the PBX is "senderized," usually in connection with automatic route selection, because senders usually can't hear dial tone and must not unload until the CO is ready.

The second reason loop versus ground start is important is that it shows the limited number of signals available to convey information between the two switches. Note in particular that a loop-start trunk will not tell the PBX when the called party has answered or hung up* and, if the call is incoming to the PBX, the PBX will not know when the far end is off hook. Thus, the PBX can only monitor its internal extension to know when to end a connection via a loop trunk, and when it should stop timing for message accounting on an outgoing call. With a ground start trunk, it will know when the called party hangs up, but not when he answers. Most message accounting systems thus assume the called party answers some number of seconds after outpulsing has been completed. When the outside party hangs up, a ground start trunk removes ground from the top wire in the diagram to let the PBX know the call is over. This is some help, but not really enough.

[* FOOTNOTE: Step-by-step central offices do send such signals, but relatively few large PBXs are served by SXS COs these days]

It should be noted that some PBXs claim to be able to test CO trunks. In general, the test routine causes the PBX to seize a CO trunk and detect ground which the CO's dialing detector connects to the line at the same time the central office returns dial tone. This form of test leaves much to be desired because (a) it does not work with loop start trunks, which always have the ground present, and (b), by not detecting dial tone, it cannot tell if the CO has chosen to make the right connection, or if the CO dial tone distribution is working. Lack of dial tone happens only occasionally, but even if everything is working perfectly, a shorted-out dial tone bus can block user calls if the user won't dial without the familiar reassurance. To be really useful, a test should dial a digit and observe dial tone removal.

Trunks for direct inward dialing (DID) are a relatively recent innovation. They require a major upheaval in the telephone central office, however, because central offices just aren't set up to send dial pulses to Princess telephones. After all, what would a Princess phone do with a train of dial pulses?

DID trunks are, in general, used one-way only, from the CO toward the PBX. They can be arranged to work with or without wink start. Wink start is a signal returned from one switch when it learns that the far end of a trunk has been seized. The "wink" says, "I know you have seized this trunk-, and I now have my digit detector connected. You are free to send me digits." Note that a wink performs the same function as dial tone, but the central office can't detect dial tone. The wink is actually an "off-hook" signal of about a fifth of a second duration.

There are two points to remember here. Step-by-step systems, being ready to receive dial pulses on a trunk at any time, do not have to hold off the CO until they get a digit receiver attached. Thus, they do not have to send a wink. Similarly, modern PBXs which extract dial pulses directly from trunk circuits as they come in, and which have enough processing power so that they are always ready for a new call, no matter how many are in progress at the time, also need not bother with a wink. But if the PBX, for any reason, may not be ready to take pulses when the CO is ready to send, wink start is required. If, in the future, DTMF to the PBX is permitted, all DTMF trunks will have to use wink start to hold off the CO until the DTMF receiver is attached.

The second point is that dial tone detectors would be preferable for many reasons, for both originating and terminating calls, but the Touch-A-Matic repertory dialer is one of the few machines to realize this.

In any event, the DID trunk passes information back to the CO to tell the latter to send digits; the CO then sends the two, three or four digits required to identify the extension. The PBX rings the called extension (or some other in a hunt group), detects answer (or a call pick-up signal), and makes the trunk-to-extension connection. Since the CO does not start charging the calling party until the extension answers, the answer signal must be transmitted to the CO.

On interconnected systems, the interface device has a one-way amplifier that lets the calling party hear the audible ringing (or the busy tone) from the PBX. When the called party answers and answer supervision is returned to the CO, the amplifier is shorted out and the called party can hear the calling party. But if no answer supervision is returned, as might happen in a malfunction or if the customer has tricked up his system to give free calls to people trying to reach him, the one-way amplifier stays in the circuit and the PBX user cannot hear any incoming conversation. This clever arrangement leads to a major problem which will be discussed subsequently.

DID is a great advantage over Centrex in that you need only provide DID phones for people who get large numbers of calls; you don't have to make every extension a DID number. Further, since DID numbers can hunt to non-DID numbers and can be picked up by non-DID extensions when pick-up is available as a feature, the effectiveness of DID is greatly increased. Note that some vendors do not understand this interaction with hunting and pick-up and, as a result, insist that DID calls can only go to DID extensions. Don't be fooled.

Tie trunks

Tie trunks connect one PBX to another. They are usually used two-way, in spite of the "glare" problem mentioned in connection with CO trunks. But to work two-way, more signaling information must be exchanged. In particular, each end of the trunk must be able to tell the other whether the local user is on-hook or off-hook. Thus, E&M signaling is used.

"E" and "M" refer to wires coming out of a trunk circuit in a PBX or CO and going to the signaling equipment of the facility to the distant office. The M lead tells the trunk and, subsequently, the distant switch, that the local user is off-hook, while the E lead receives information from the trunk (and distant switch) with regard to the status of things off yonder.

A mnemonic frequently used to remember which way the signals are going is E for Ear (listen to the far end) and M for Mouth (talk to the far end). Some writers will even tell you that this is what E and M actually stand for. Although useful as a memory jogger, the device simply stems from lead designations on Bell System trunk circuits which were assigned in alphabetical order. It just happened (some time around 1900) that the signaling leads got the names they have retained all over the world to this day.

Just for the record, the M lead is grounded for on-hook, and transmits -50 volts for off-hook. The E lead, on the other hand, is an open circuit for on-hook and a ground for off-hook. This all refers to the leads coming out of the trunk circuit going toward the signaling equipment of the trunk.*

[* FOOTNOTE: This Type I The new Type II sends a closure from PBX to trunk for M off hook, and receives a closure from trunk to PBX for E off hook]

In older systems, particularly step-by-step, all dial pulsing on trunks is under the direct control of the user. The user is "cut through" to the trunk and sends digits directly into the next switch. If a user causes a switch to seize a trunk at the same time that someone else causes the other switch to seize the same trunk from the far end, the two users are connected together and, after a few seconds of confusion, they figure out what happened, get off, and try again. With senderized systems such as those using automatic route selection, however, the problem is more difficult. When two senders grab the trunk from opposite ends, they usually aren't smart enough to figure out what is going on.

The reason for this is that senderized tie trunks use "delay dial," similar to wink start on DID (one-way) trunks. But since they are two-way facilities, and since the delay dial signal, like the wink, looks just like a seizure (an off-hook signal), confusion can get pretty thick. Since dial tone does not look like a seizure, dial tone detectors for spotting start-sending signals would again be highly desirable. Why nobody uses them beats the heck out of me.*

[* FOOTNOTE: Danray does, and apparently Stromberg-Carlson does too in some systems]

A tie trunk terminal on a PBX must be able to respond to delay dial or wink-start signals from the far end when the far end is ready to receive and, when the same PBX is ready to receive, it must be able to send delay dial or wink start unless, like step-by-step, it is fast enough to accept digits whenever the far end wants to send them. Few electronic switches are as fast as step-by-step in this regard.

The reader may be wondering why DTMF (Touch Tone) signaling is not used instead of dial pulses on DID CO trunks and tie trunks. It is tempting to say there is no reason for it, it's just policy, but there is a reason. With DID trunks, dial pulsing, not being a voice frequency signal, can bypass the one-way amplifier in the interconnect device. DTMF, however, would be blocked. Then, too, many of the existing PBXs of older vintage find it just as hard to connect a DTMF receiver (intended for extensions) to a trunk as the CO finds it to connect dial pulses to a Princess telephone.

With regard to tie trunks, it turns out that most existing networks were implemented with step-by-step switches which can accept only dial pulses. Add an electronic PBX to an existing network and it has to put out dial pulses to function. With CCSA, it is possible to DTMF into the tie trunk switch, but internal to the network, a different form of voice frequency signaling (called MF) is used. Coming from CCSA to an end PBX, only dial pulsing or MF is available since CCSA works like the toll network. DTMF is not used in the toll network to replace MF, it should be noted, because it would give every user with a Touch Tone phone the equivalent of a "blue box."

If you're building your own tie trunk network with your own switches and circuits from a specialized common carrier, it is quite likely that you can set up the system with DTMF pulsing throughout, and can even, with some arm twisting, obtain dial-tone detectors for start-pulsing signals. But you'll have to know exactly what to ask for, and be willing to fight every step of the way.

Perhaps the most important reason for dial-tone detectors and DTMF outpulsing is in connection with "off-net calls." Ideally, a user at one PBX should be able to call another via tie trunks, insert a 9 (or some other code), get CO dial tone, and make a local call. This increases the utility of the tie trunks, helps justify their cost, and saves money. With step-by-step, this sort of thing was done all the time

With the new equipment, particularly when senderized automatic route selection schemes are present, this is not always possible. The problem comes from the sender not being able to tell when the distant CO has been reached. Even with a ground-start trunk from the distant PBX to its CO, the ground signal cannot be returned to the calling PBX and the sender has only time-out as a way to tell when the far end is ready. Usually a three second delay is required; if outpulsing is via DTMF, it takes only one second. Clearly, this is absurd. A dial-tone detector just can't be beaten.

With tie trunk to off-net connections desired, there are many instances when dial pulses should be sent to the dial tandem network, and DTMF used into the CO when reached. Few designers are even aware of this requirement, but keep asking.*

[*FOOTNOTE: Access to Execunet Sprint and City Call is waking the design community up]

All things considered, transmission is probably more important than signaling. There's not much point in setting up a connection if you can't talk over it. But the most important factor to determine about a PBX, if it is intended to switch one tie trunk to another, is whether or not it can switch four-wire facilities on a four-wire basis. Only a few can.

Since combined extension and tandem switching is a good way to save money and justify the expense and risk of buying your own system, it is important to understand what is needed here. It must be possible to connect tie trunks to each other with the transmit and receive channels completely separate to minimize echo, but tie trunks must be able to connect to extensions, usually two-wire, in such a way that echo is again minimized. The transmission rules, described in detail in AT&T's Notes on Distance Dialing,* require an additional 2 dB of attenuation to be placed in a tie trunk-to-extension connection, while this 2 dB must not be present on a tie trunk to tie trunk connection (terminal net loss versus via net loss). Thus, the switch must be able to make 4 to 4 and 4 to 2-wire connections, and must be able to insert or remove an attenuator, depending on whether the call terminates locally or goes on through.

[* FOOTNOTE: The new edition dated 1980 is called Notes on the Network]

A PBX which is not intended to be a "hub" in a tie trunk network need not worry about all this, or about four-wire switching. It is just as easy to convert from two-wire to four-wire on the trunk side as the line side, and the 2 dB pad can be left in the trunk circuit at all times. However, if the PBX is four-wire internally, as most of the modern electronic PBXs are, the 2/4 wire conversion should only be made once, at the line side.

Although all of the above may sound like Sandbox One to the sophisticated communications manager, I have been shocked to find some system designers who are unaware of these several points. This worries me, since transmission is relatively important.

There is one more transmission topic to mention here. In the all-digital world of the future, digital switches will switch digital signals from digital trunks (whether CO trunks or tie trunks) without ever decoding back to analog. This will save a great deal of money in that the "channel banks" that normally do the analog to 'digital and reverse coding can be completely eliminated, and it will improve transmission in that there will be no attenuation or phase shift and very little noise as long as the signal stays within the all-digital mode. However, all digital signals have to be compatible for this to work and, as we have seen above, there is a lot of variety. It will take a while for the all-digital revolution to get here, so this isn't a vital factor in evaluating proposals in 1978. But as time goes on, the importance will increase.

With regard to all-digital systems, it is interesting to note that level adjustments can be made on a digital basis. A random access memory is used, and any coded input, representing the amplitude of the audio signal at the instant of sampling, can call out another coded word that is 2 dB (or whatever else you like) less. This "look-up table" form of pad is interesting and shows one of the ways digital signals can be processed that have no analog in analog transmission. But there is a booby trap here.

Suppose we have an all-digital system: PCM coded PBXs tied together by T-carrier span lines. If we now have some extensions arranged for data, allowing them to enter the digital world directly without going through an A/D converter, we can eliminate modems and handle 50 Kbps data over voice channels. But we have to stay clear of digital pads. If we take a digital word, fresh from our computer or terminal, and pass it through a conversion process where it comes out "2 dB lower," we have translated information into garbage.

Before leaving the fascinating world of tie trunks, there are two more related design requirements that must be mentioned. Tie trunks must be capable of facilitating satellite and centralized attendant systems (CAS). In a satellite system, one PBX is custodian of the "front door" number and answers for itself and all the other PBXs the customer has in a given area. This permits a single telephone number in the directory, and reduces the number of attendants required to handle the incoming traffic. The calls all come in to one point, are answered by the attendants there, and are completed via tie trunks to other PBXs when necessary. Combined with DID at each location, the reduced directory number traffic can be handled quite economically and excellent service maintained.

It should be noted that many customers want to be able to complete calls to any PBX on their tie trunk network, whether such calls come in via the directory number and console attendant, or via DID direct to a particular individual. Other customers, however, do not want such operation since it can lead to abuse. The Dimension, for instance, cannot transfer a call over a tie trunk, allegedly because "customers don't want this." Note that with message accounting, abuse can be contained fairly well.

CAS is slightly different. Here, each PBX has its own number, and users call the particular branch of the company they want. However, the "switched loop" to the console extends all the way to another location where all the attendants are centralized. They answer the directory number calls, key in the desired extension, and release.

In CAS service, tie trunks may or may not be used, but "released links" are needed to connect the attendant consoles to the CO trunk circuits at the distant PBXs where calls enter the system. If tie trunks can also serve as "released links," then it becomes possible for a large company with a tie trunk network to centralize its attendants, at least after hours, allowing the people on the East Coast to handle early-bird calls to the California locations and the West Coast attendants to respond to the evening calls to the offices and plants in New York and New Jersey.

It is straightforward to program a PBX to seize a tie trunk, and set up a tandem connection to a distant console (or even an extension with the proper features and class marks), while observing ringing seen on an incoming CO trunk. When the console or phone answers, the CO trunk is connected through to the tie trunk and, consequently, to the distant attendant. It is a little harder for the attendant at that PBX to flash the switchhook to call in the features needed to complete the call (station dial transfer, in particular) in the PBX with the incoming CO trunk, but with a little care, it can be done. Once the DTMF digit receiver in the incoming PBX is connected, tones can be keyed into it to identify the desired extension and the attendant can release (Reference 1).

Features such as satellite and centralized attendant service greatly increase the utility of a business communication system and go a long way toward reducing costs. However, once again I am startled to find system designers who have never heard of the sort of thing that step-by-step has been able to do for years.

Trunk interfaces

Some of the more common trunk interfaces are listed in Table 1. These devices degrade signaling and transmission, increase the number of wires (and cross-connect frame space) at the interface between telco and interconnect by a factor of seven or eight, and materially reduce the reliability of communications. That is, they protect the telephone industry from interconnected terminal equipment and the vendors thereof. But, as indicated briefly above, they do make possible precise, clear specifications at the interface.

When they depart, however, some problems are going to surface.*

[*FOOTNOTE: CO trunk interfaces were eliminated after this article was first published but as of 1980, interfaces are still required for tie trunks.]

What has to be known about trunk circuits and their relation to interfaces is this: What happens when the interfaces are removed? Some companies have designed their trunk circuits to interface directly with interfaces. Thus, such circuits will have to be replaced before interfaces can be removed. Other companies have designed an "anti-interface" that can be removed when the interface is taken out, leaving nothing but a single pair of copper wires.

There is one interface that will probably remain, however, and that is the good old C22 for DID trunks. It, or something analogous, will be necessary to insure answer supervision before the trunk is put into the talking state. But a problem, hinted at briefly above, is quite serious.

Suppose you have ten DID trunks, and the little one-way C22 amplifiers in the two most used circuits are dead. Everything goes well except that calling parties do not hear audible ringing or busy tone as the case may be. They go "high and dry." If the called party answers very quickly, little harm is done except for some surprise that the phone was answered before it was rung. However, if two or three rings do not produce an answer, the calling party, hearing nothing, will abandon.

Now comes the fun. You report the trouble to the phone company, they send somebody out to check, and the two bad trunks are fixed (maybe). But there were eight good trunks, and they want to charge you for testing them when no problem was found. Don't pay! Scream to the local PUC. And be sure that your potential vendor knows how to make tests on specific incoming DID trunks periodically. (You have to busy out all trunk circuits at the PBX but one, and then dial any DID number. If all works properly, busy out that trunk circuit and free-up another. Don't do this during the busy hour.)

There is one other little point about interfaces that should be kept in mind. Don't put them in your switch room. Put them somewhere else, with completely separate access. This is extra floor space you must allow for, but keeping interco and telco personnel and equipment separate is worth the effort. Do I have to explain why?

Telephone Instruments

Telephone instruments will be discussed in more detail in connection with features. Here, we simply want to see what should be in a general system description concerning the equipment itself.

Rotary dial versus DTMF is an important choice. Almost everybody wants pushbuttons, but the cost differential should be spelled out. DTMF sets cost almost exactly twice as much as sets with rotary dials, but the installed price, including labor, tends to mask the difference, since it costs no more to run wires for one phone than the other. Some PBXs won't work with a rotary dial; all require a digit receiver for DTMF, since DTMF requires elaborate circuitry to discriminate between tone signals, accepting valid tones and rejecting speech or other signals that look like tones.

One point that is frequently overlooked in the rotary versus DTMF discussion is the ease with which features may be invoked. The Womack was originally designed for rotary dial, and features can be activated by dialing a single digit at any time during the progress of a call. No switch-hook flash. However, all systems with DTMF, including the Womack, require a flash to get the DTMF receiver on the line. Then the feature code or extension address can be dialed. In other writing (see Reference 3), I have called this stupid/smart supervision, indicating that a relatively dumb monitor must always be present, and be able to call in a smart monitor. Only the smart monitor can cause the system to take appropriate action.

Most conventional sets use power ringing: 20 Hz at 86 volts. This method of calling the customer to the phone, invented by Bell's sidekick, Watson, is hard to beat. But it poses problems for electronic switching. In particular, ringing must be applied at each line circuit, which makes the circuit unduly complex. Tone ringing has been used experimentally for years, and is available on some specialized sets. Being a low-level voice frequency tone, it can be switched through an electronic matrix from a common circuit.

In the hotel-motel version of most PBXs, message waiting lamps are provided as standard features. When a message is being held for a particular extension at some central point, the lamp is lit and remains lit, even if the user makes a phone call, until the message center turns it out. This seems to me to be a feature that might be used more widely in business. But the ability of the PBX to light the lamp is something to ask about.

If a switch can work with conventional instruments, it can usually work with standard key systems. There are a few problems to watch out for, however. Standard key telephone units are completely independent of the PBX; that is, they provide their features and the PBX provides its features. If a key telephone is on hold and call waiting tone is connected to it, the PBX has no way of knowing that nobody heard the tone. There are some places where PBX features simply will not do the job, and when key phones are used, be sure you know what interaction to expect.

One interaction that seems to have slipped by everybody with an electronic PBX is the inability of the PBX to release a line on hold when the other party has hung up. In step-by-step and crossbar systems, the power to the phone is interrupted momentarily when the connection is taken down, and the hold relay on the held phone can release. Not so with electronic systems. With a line circuit that is in the connection at all times, there is no momentary open and a phone (or data set) can appear to be in use forever, even when the other party has long since gone home.

A sort of half-way house between conventional key equipment and modern electronics is the built-in KTU. The Womack, for instance, and the ITT TD-100 both work this way. The KTU is built into the PBX. You still have to run a 25 pair cable to a six-button set (a bigger cable for a call director), but you don't have a separate KTU, and the system, using its electronic control, handles all the key features. Thus, it knows when a line is on hold and acts accordingly.

Rolm offers an electronic KTU that greatly reduces the number of pairs to a six-button phone or a call director, keeps the key phone operation under control of the stored program system, and renders excellent key telephone service were needed. As the reader can see, there is a progression here from standard to built-in to remote electronic KTUs, and each has advantages. However, all three use conventional instruments, available from a number of manufacturers.

The next step, of course, is special instruments designed to work with the particular system. These come in a variety of types, and reflect many different design concepts. The simplest replace the switch-hook flash with a "recall" button, often labeled "hold."

Depression of this button sends a timed open-circuit to the line circuit, making it easy for the switch to recognize a flash instead of a dial pulse or a hang-up. The user then dials or keys in the feature code or the desired extension number. Chestel and the Tele/Resources System 32 work this way, and Executone has built sets to work with their Digital PBXs.

Some telephone sets have added features that are not obvious. For instance, just by looking, you can't tell that the TR System 32 has four-wire transmission from subset to subset. Danray telephone sets, although filled with electronics, look like regular single-line or multi-button telephones with a call-waiting lamp. They, too, have four-wire transmission.

The Danray sets use digital signaling directly to the PBX, independent of the voice channel. Thus, although they look like regular pushbutton phones, they are not. There is no need to flash the switch-hook; feature codes can be keyed in directly, any time during a call. You hit *4 to put a line on hold (H is associated with 4), *R (for Recall) if you want a busy extension to call you back when it comes free, etc. The call-waiting lamp performs a variety of functions to let you know you have interacted properly with the system. Tone ringing is used rather than power ringing.

In passing, it should be noted that Rolm and Womack, on their conventional pushbutton telephone sets, have a panel that identifies feature codes for the user. Tricks of this sort, or the Danray approach or using letters of the alphabet, make feature use on single-line instruments much easier for the caller.

The Danray telephone has a single visual display which helps considerably in letting the user know what is going on. But the new electronic key sets for Northern Telecom's SL-1 or Bell's Dimension increase the visual displays to the point where they rival or exceed a conventional key telephone. SL-1 telephone sets have keys and lamps for sending and receiving information, and the Dimension Custom Telephone Sets have keys for sending and two lamps per key for receiving. Since buttons can be used to pick up lines as in a key phone, or to invoke features, as in modern PBXs, but without the need to memorize a number of feature codes, the electronic sets offer many interesting opportunities to improve telecommunications. Now that Danray has been taken over by Northern Telecom, it seems reasonable to expect SL-1 sets to be used on the Danray switch, and the Danray software to help out the SL-1.

In any event, the electronic sets (and Rolm's electronic KTU) greatly simplify the wiring when a new PBX is installed. This will be covered below.

There are a great many other gizmos and gadgets that are available: speaker phones, voice announce systems, use of feature buttons for abbreviated dialing signals, etc. The SL-1 sets are particularly nice in that extra modules can be plugged in to add these extra buttons or features without changing the station wiring.

Station wiring

In most metropolitan areas and, indeed, almost everywhere else since the cost of labor is going up rapidly everywhere, it is becoming standard to simply wire business locations with 25-pair cable to all telephones. In this way, six-button sets can be installed with minimum effort, and no further wire needs to be pulled. Unfortunately, call directors require larger cables, and many older office buildings are now stuffed with telephone cables of various sizes.

This is all very well for the telephone company. The cable is in the rate base, and everybody is happy. But if you own your own system, and will have to keep track of moves, changes, etc., the shoe is on the other foot. The cost, bother and the difficulty of getting expensive KTU equipment wired correctly is a headache nobody needs.

Thus, with your own system, you need uniform wiring. Once the wiring is in, it should handle any kind of telephone you want to put at a given point, from a single line set to a 30-button call director. This is the main advantage of the new electronic key telephone sets. If you have two, three or four pairs, (depending on the set) to every point, you can handle everything a station user might ever want. All phones plug into the standard connector, and the features that match the phones (part of the PBX) do all the work.

If you could use single line instruments everywhere, you would, of course, have uniform wiring at one pair per set. However, there are places where you will not be able to provide good service with just single line instruments. Thus, you cannot count on single pair wiring to remain uniform. The electronic sets offer a much greater chance to save on moves and changes over the life of the system. What you need to know from the proposal is the minimum number of pairs to provide universal wiring, and if this is the kind of wiring that the vendor is planning to install. Single pair wiring is almost certainly going to turn out to be a bad bargain in the long run, but you shouldn't have to go to 25 pair or larger cable to be universal.

Summary

This completes our look at what you ought to know about PBX equipment that you are considering, whether from the telco or from an interconnect vendor. Actually, we have only just scratched the surface but, with luck, most of the important points have been mentioned. If your proposals don't provide you with even this minimum, go into the attack mode and demand the necessary information. It's better to find out before you buy than after. A PBX is expensive, and you have a right to known what you're getting. If enough of us keep asking, it is possible that descriptive information in proposals will be greatly improved. But if we don't ask, nothing will happen.

Next time, we'll take a look at features. We won't bother with the hundreds in the alphabetical lists, but we'll pick out the ones that count and take a good look at them. You may be surprised, but it is better to be surprised reading a book than when you are trying to give a demonstration of your new PBX to your company's assembled vice-presidents.

References

1. "Multi-location service," Bell Laboratories Record, June, 1975, p. 264.

2. The following Bell System tie trunk interface technical descriptions are of general interest: 11-TTEG, 12/11-TT-MB, 31-TT-EG, 31-TT-MB, 32-TT-MB, all dated March, 1979.

3. Design Background for Telephone Switching, Chapter 2. Lee's ABC of the Telephone, 1977.

[ Top ] [ Next ] [ Table of Contents ]

Copyright 2006 Lee Goeller. All Rights Reserved.