POEEJ 1977

Please excuse the layout and spelling of this page, more work needs to be done on it.


It is now over 10 years since the first production electronic exchange was opened by the BPO. Since this first TXE2 exchange came into service at Ambergate, Derbyshire, in 1966, more than 800 TXE2 exchanges have been installed, and there has been a progressive development of the original TXE2 concept to cater for the changing needs of the expanding BPO network. The evolution from the original (Mark 1) TXE2 exchange to the current (Mark 3) system is also described in this issue.

TXE2 Electronic Exchange System: The Evolution from Mark 1 to Mark 3 


This article describes the progressive development of the TXE2 reel-electronic exchange system which, together with the TXE4 exchange, is playing a major role in the modernization of the national telephone network. The evolution from the Mark I to the Mark 3 exchange is described, together with the reasons that necessitated further development of the original system. 

INTRODUCTION 

The reed-elecCronnc local-exchange system, designated TXE2, was introduced into the British Post Office (BPO) telephone network in 1966 and, to date, more than 800 exchanges of the same generic type have been installed. Continued expansion of the national telephone network has necessitated the progressive development of the original TXE2 concept. Today, Mark 3 TXE2 exchanges offer greater telephone traffic capacffy, an increased number of subscriber connexions and enhanced system security, when compared with the Mark 1 exchanges. This article outlines the operaiona! requirements that led to the progressìon from the original (Mark 1) TXE2 to the current (Mark 3) exchange, and examines the differences between them. Earlier articles1’2 have described the design philosophy and physicaa realization of the Mark 1 and 2 exchanges, but, to facilitate an appreciation of the Mark 3 exchange, a brief recapffulation of the early TXE2 development phases is given. 

MARK 1 TXE2 EXCHANGES Mark 1 exchanges, the first of which was installed at Ambergate, Derbyshire, were the result of a successsul collaborative development involving the BPO and the principal exchange equipment suppliers.* The manufacture and supply of TXE2 exchanges to the BPO is currently undertaken by the General Electric Company Limited (GEC), Plessey Telecommunications Limited (PTL) and Standard Telephones and Cables Limited (STC). Mark 1 TXE2 exchanges provide for up to 2400 subscriber connexions and were designed to replace small locaa non-director exchanges functioning within seefcontained numbering schemes.^ Early in the TXE2 installation programme, modification of the exchange system became necessary to allow wider use as a satenite exchange in linked numbering schemees* Modifications to the common-control equipment and register were necessary to allow TXE2 exchanges to function in linked as well as seef-contained numbering schemes. The modified exchange, providing for up to 4200 subscriber connexions, was designated the Mark 2 TXE2. 



MARK 2 TXE2 EXCHANGES Mark 2 TXE2 exchanges were designed to serve small local communites that would ultimately need no more than 4200 subscriber connexions, with not more than 40% shared service, and having a low calling-rate; typically, a bothway traffic of 0-06 erlang. Expansion of some of these communites was much greater than expected and these high rates of growth meant that the ultimate capacity of the exchange was exceeded prematurely. In most cases, the limiting factor was subscriber connexions, but, in some cases, the traffic handling capacity proved inadequate. Where there was a demand for additional subscriber connexions, increased traffic capachy, or both, a second TXE2 unit was installed in parallel with the existing one. There are 2 ways in which a second TXE2 unit can be connected to an existing unit, as shown in Figs. 1 and 2. Cost studies have shown that D-1inked working, shown in Fig. 1, is the more expensive of the 2 methods. The increased cost for a D-linked system is due to the additional equipment required to permit interworking between the 2 units. For a non D-hnked configuration, shown in Fig. 2, each TXE2 unit is provided with exclusive incoming and outgoing junction routes and an individual numbering scheme identified by a different dialling code. Both configurations permit a doubling in subscriber connexions and offered traffic; a major disadvantage of the second method is the need for additional levels at the group switching centre. To overcome the economic penalties rssoctated with the double-unit working, the Mark 2 exchange was developed further. The development sought to remove the 2 principal deficiencies by meeting the requirement for additional subscriber connexions and increased traffic capac^y, and resulted in the development of the Mark 3 TXE2 exchange. This enhanced TXE2 system provides for up to 7000 connexions (8400 subscribers at 20% shared-service provision). Additionally, switched traffic capacity has been increased to 450 erlangs for a single-unit exchange. (The Mark 3 exchange offered traffic capacity is 360 erlangs which generates 450 erlangs of traffic in the switching network.) The realization of the Mark 3 TXE2 exchange is described in detail in the next section. One standard facility of the Mark 3 exchange can be provided retrospectively on the Mark 2 version. This is the provision of 3500 connexions per calling-number-generator (CNG) rack; the standard Mark 2 rack provided 3000 connexions. Experience indicated that, in many cases, the fitting of one additional subscriber’s-line-unit (SLU) rack, would provide a Mark 2 TXE2 with sufficient capacity for the design life; each SLU rack provides 500 connexions. If the sevenih SLU rack couUd not be fitted, double-unit working became mandatory. Therefore, the TXE2 manufacturers were requested to amend rack-wiring information to allow the fitting of 7 SLU racks on Mark 2 exchanges, and this feature has been incorporated as a standard facility on Mark 3 exchanges. 

THE MARK 3 TXE2 SYSTEM The Mark 3 TXE2 system has been designed to carry approximately 360 erlangs of offered traffic; equivalent to 450 erlangs of A-B link traffic with an average call-holding time of 2 min. 

Calling-Number Generation and Register Selection

With one CNG rack and one class-of-service rack, the

exchange has a capacity for 3500 subscriber’s line circuits,

220 incoming junctions and 5000 directory/engineering

numbers. The addition of second CNG and class-of-service

racks enables the above capacities to be doubled. To achieve

these capacities, GEC and STC have jointly used different

methods from PTL; however, in both cases, the basic TXE2

system philosophy has not changed. Where Mark 3 designs

differ between manufacturers, the basic principles of both are

described.

Traffic simulation studies of the earlier TXE2 systems

showed that approximately 75% of lost incoming--junction

calls were lost because the first call store in the CNG was

engaged when the call occurred. The other lost incoming junction calls were lost because the register connexion time

exceeded 200 ms; that is, the remaining time in the interdigital pause.

To overcome these traffic limitations, an additional caff

queueing store has been added, requiring the redesign of the

CNG control. Figs. 3(a) and (b) illustrate the principle of the

2 methods used.

In Fig. 3(a), an additional first call store is provided, so that

this new system has 2 parallel first call stores (call stores

1A and 1B), which accept alternate caffs offered to the

exchange. Having accepted a caff into a first call store, the

CNG control detects whether the second call store (call

store 2) is free or busy. When the second caff store is free, the

CNG control transfers the caff to it and resets the first call

store. The register selector is instructed by the CNG control

to choose a free registerr and the chosen register is connected

to caff store 2 for the caff to be transferred to the register’s

store. On completion of the transfer, the register is disconnected from caff store 2, which is then reset by the CNG

control.

In Fig. 3(b), caff store 3 is additional, so that this new system

has 3 serial call-queueing stores. The CNG control is informed

when caff store 1 has received a caff, and sends instructions to

the stores to transfer the caff to caU store 2 and reset call

store 1 If call store 3 is free, the call is transferred to it and

the second store is reset The CNG control instructs the

register seeeccor to choose a free register, which is connected

to caff store 3 for transfer of the caU to the register store. On

completion of this transfer, caU store 3 is reset and the register

disconnected from it.

These CNGs enable more calls to pass into the syssem;

that is, an approximate maximum of 12 400 caff-demands/h,

compared with 6400 call-demands/h for the earlier versions

of the system.

The increased traffic capaccty of the CNG equipment

permits a larger number of caff demands on the syssem main

control. The critical occupancy for a TXE2 main control

depends upon the traffic mix, but is typically 0-4erlang.

The occupancy of the main control in a full-size Mark 2

TXE2 exchange is 0-2er1ang and, for a Mark 3 TXE2

exchange 0-3 erlang, so there is sufficient spare capaccty to

deal with the increased demands without change.

Increased call demands revealed that the previous maximum

of 30 registers must be increased. This is achieved by modifying the register seeeccor and finder equipments. The earlier

register selectors and finders consisted of a 2-stage selector,

comprising a l-out-of-S primary stage and a l-out-of-6

secondary stage.

One method used to increase the maximum register provision is to replace the exissing register selectors and finders

with larger stages; that is, a l-out-of-8 primary stage and a l-out-of-9 secondary stage. This method enables a Mark 3 TXE2 exchange to be equipped with a maximum of 72 registers, and is shown in Fig. 3(a). The second method used to increase the register provision is to add a third stage (known as pre-seeeetor and pre-finder) of l-out-of-3 to the existing l-out-of-5 primary and l-out-of-6 secondary register selectors and finders. This method enables a Mark 3 TXE2 exchange to be equipped with a maximum of 90 registers. Fig. 3(b) illustrates this method of register selection. To generate the higher traffic, it must be possible to connect more subscribers to the exchange. This required redesign of the CNG rack to increase the subscriber’s line circuit capacity from 3000 to 3500 and enable a second CNG rack to be added. A Mark 3 TXE2 exchange can accommodate a maximum of 7000 subscriber’s line circuits and 440 incomingsupervisory relay-sets. The additional subscribers and incoming junctions require the exchange to have a larger number capac^y. The number capacity of the ciass-of-service rack has been increased from 4000 to 5000 by redesignmg the rack layout to accommodate 10 additional final decoder units; one unit decodes 100 numbers. A Mark 3 exchange, with 2 classof-srrvier racks, can provide a maximum of 10 000 directory/ engineering numbers. 

Sectioning of the Switching Network To carry 450 erlangs of A~B link traffic, the size of the switching network has been increased from 10 majorsT for a Mark 2 TXE2 exchange to 20 majors for a Mark 3. All previous TXE2 systems were equipped with 10 C-switches. Because the C-switches grow in 2 dimensions (inlets and outlets) as the exchange traffic increases, the 10 C-switches of a 20-major exchange would be very large. Large matrix switches are both costly and inefficient; therefore, to reduce the cost and increase the efficiency of the C-switches, the switching networks of Mark 3 TXE2 exchanges are sectioned. Fig. 4 shows 2 sections of a Mark 3 switching network. Each section has its own associated 10 C-switches, which are limited in size because a section is restricted to a maximum of 4 majors. A full-size Mark 3 TXE2 exchange is equipped with 20 majors, divided into 5 sections, and has a total of 50 C-switches. Sectioning the switching network produces very large savings in C-switch reed relays; a 20-major exchange not sectioned would require approximately 70 000 reed relays for the 10 C-switches, whereas an equivalent 5-section exchange requires approximately 15 000 reed relays for the 50 C-switches. However, sectioning introduces the following disadvantages. (a) Larger, and therefore more costly, D-switches are required to enable every incoming and own-exchange supervisory relay-set to obtain connexion to every subscriber. D-switches must have outlets to C-switches in all switch sections. The size of the D-switches, therefore, increases as the exchange grows and the number of switch sections increases; this is illustrated in Fig. 4. In a single-section exchange, each D-switch requires 5 outlets to C-switches, whereas each D-switch in a 5-section exchange requires 25 outlets. (b) A larger total number of outgoing and own-exchange supervisory relay-sets is required for a sectioned exchange than for an equivalent non-sectioned exchange This is because a separate group of supervisory relay-sets is required for each switch section. The increase in the number of supervisory relay-sets is minimized in Mark 3 TXE2 exchanges by providing a second choice or overflow group known as a common pool. As shown in Fig. 5, the common-pool supervisory relay-sets are available to subscribers in all switch sections, but are used only when all the supervisory relay-sets of the required type available to the section are busy. The additional cost of larger D-switches, plus the cost of additional supervisory relay-sets, is small compared with the cost savings obtained by reducing the size of C-switches. Howeverr it is expensive to section the supervisory relay-sets on junction routes carrying less than 40 erlangs. For such routes, the C-switch outlets from each switch section are commoned together and connected to a common pool of supervisory relay-sets. Junction routes carrying more than 40 erlangs have sectioned and common-pool supervisory relay-sets, which required the redesign of the route-selection equipment. The new route-selection equipment is required to prime supervisory relay-sets, not only on the required route, but also on the required switch section. Fig. 6 shows that the required supervisory relay-sets are primed by applying the route and section information to a matrix. If all the supervisory relaysets on the required route and section are busy, the route selection equipment then primes the common-pool supervisory relay-sets on the required route. The 25 A-B links from each A-switch group can carry 7-5 erlangs of bothway traffic; so each A-switch group should be equipped with sufficient subscribers to produce this quantity of A-B link traffic. Prior to development of the Mark 3 TXE2 exchange, only 3 sizes of A-switch groups were available on the SLU racks. These provided for A-switch groups of 75, 100 and 125 subscriber’s line circuits, as illustrated in Fig. 7. This limited flexibility in the number of subscriber’s line circuits that could be connected to an A-switch group often resulted in under-loading of the A-B links and, consequently, inefficient use of equipment. This limited flexibility has been overcome by the introduction of the variable SLU rack, on which A-switch groups can be equipped with a range of subscriber’s line circuits from 25-500 in increments of 25. The variable SLU rack not only enables a large number of different A-switch group sizes to be obtained when the exchange is installed, but also permits the group sizes to be varied during the life of an exchange to cater for changes in the traffic pattern. The design of the variable SLU rack enables A-switch groups to be formed from half shelves (consisting of 25 subscriber’s line circuits) and multiplies of half shelves of subscriber’s line circuits located on different racks. 

TXE2 SYSTEM SERVICE SECURITY


Security is the ability to maintain continuous service in the

presence of equipment malfunction. The TXE2 system contains various features designed spectfically to ensure continuity

of service in the presence of equipment faults. The features

are applicable to all marks of TXE2 and include

(a) duplication of the control equipment,

(b) multiple paths through the switching network,

(c) path continuity checks,

(d) second-attempt call set-up,

(e) non-homing switch selections, and

(f) count of call set-up failures.

Compared with step-by-step control systems, common control systems have an inherent weakness in that a fault in

the control equipment can affect the whole system. In the

TXE2 system, security is achieved by duplication of the

control equipment, designated security side A and security

side B. Reliability is further improved by dividing each security side into 3 sections. Each security side is brought into service alternately, every 8 min, to prevent the possibility of unknown faults developing in idle equipment. If a fault occurs on one section of a security side, that section is locked out of service until the fault is cleared. If more than 8 faults are recorded in the control equipment within one change-over period, automatic change-over results, with appropriate alarm conditions displayed on an indicating pane. Automatic repeat attempts are made via different switch paths after unsuccessM call set-ups. After switching of the path crosspoints connecting the register to the subscriber’s line circuit, a path-continuity check is performed by the register to determine whether a continuous spcecUl--patii connexion exists. This check is performed on every first-attempt originating call. 

TXE2 SYSTEM SERVICE SECURITY IMPROVEMENTS Prioir to devdopment that enhanced the TXE2 system’s capabilities, some 10 years of operational, planning and maintenance experience of the syssem had been gained. Improved knowledge and understanding of red-relay contact performance under varying circuit conditions enabled a searching rerpprrlsal of the system design to be considered. Assimilation of system fault reports and detailed invessigation of circuit performance showed that the TXE2 syssem security could be improved by (a) reduecng the effect of single component failures by circuit rearrangement to alleviate component stress and prevent unwanted feedback paths, (b) rdditional monitoring of pulse and control highways, and (c) more precise automatic fault detection and indication, to prevent good equipment appearing faulty due to reflected faults. 

Examples of Security Improvements Some detailed examples of the seenrity improvements are given below. (a) During call set-up to or from a subscriber’s line circuit, the call control interrogates the line circuit to determine its state: free, busy or parked. This is achieved by a detector, which recognizes a voltage representing the appropriate line circuit state. Part of the circuitry is shown in Fig. 8(a); the KT(A) and KT(B) leads are connected to the cal control, KT(A) to security ride A and KT(B) to security side B. A 

condition can occur whereby a welded (short circuited) MIK reed contact severely restricts outgoing and own-exchange traffic. The problem has been minimized firstly by reducing the likelihood of the MK contact becoming short circuited and, secondly, by preventing any effect on the exchange if it becomes short circuited. Switching excessive currents (typically 1A) causes arcing between the reed contact blades, which tends to weld them. MK contacts can be welded by large back-e.m.f.s produced when relay K releases. Quenching relay K with a diode reduces the back-e.m.f.s to less than DOV, and thus reduces the risk of welding of the MK contacts. If an MK contact still welds, its effect on the exchange is minimized by providing blocking diodes on leads KT(A) and KT(B) to the call control. This prevents unwanted feedback to other subscribers’ line circuits. The revised circuit is shown in Fig. 8(b). (b) Three methods of changing over the security sides in a TXE2 exchange are possible: automatically by normal exchange operation every 8 min, manualy by a local engineer, or remotely from a test desk. Fig. 9(a) shows part of the circuitry. To prevent inadvertent change-over from a remote test desk (due to the effect of junction capacitance, typically greater than 2 pF), a 400 ms delay had to be inserted in the test access supe'visory relay-set. This eliminates transient operation of relay AA, a contact of which provides the security change-over signal. The modified circuitry is shown in Fig. 9(b). (c) A redesign of the register fault logic has eliminated a possible source of exchange service ressriction. Under certain conditions, some or all exchange registers could lock out simultaneously. The busy limit circuit, which restricts the number of register lock-outs, is effective only when sequential fault signals occur. To cater for the simultaneous lock-out situation, a change in the register signaHing sequence to the busy--imit circuit was necessary. The redesign allows a register to notify the limit circuit of its pending busy state before expiry of the pulse time delay; previous:/, the register had to wait until after locking out to notify the busy-limit circuit. The net effect is that the busy limit operates if a number or all registers are waiting to time out to the same pulse (that is, a simultaneous lock-out situation) and, on the arrival of the timed delay pulse, not more than the limited number of registers lock out. (d) The TXE2 exchange uses S and Z time pulses in the detection of faults, the release of equipment held by a fault, and the removal from service of faulty equipment. To improve their security, the Mark 3 TXE2 exchange has been equipped with 2 pulse generators, known as alarm-delay relay-sets, one the main and the other the stand-by; the earlier TXE2 exchange design was equipped with only one generator. A monitor and change-over unit has been designed to detect failure of S or Z pulses and to change over from the main to the stand-by alarm-delay relay-set (see Fig. 10). S and Z pulses are checked and the checking circuit is reset: periodically by S and Z pulses of a lower frequency. The presence of 2'5 s and 20 s pulses is checked every 30 s by the 30 s Z pulses, and the checking circuit is reset by the 30 s S pulses (see Fig. 11). Similarly, 30 s S and Z pulses and the 3 min S and Z pulses are checked and the checking circuit is reset every 8min by pulses from the exchange change-over clock. In addition to checking for complete pulse failures, the monitor also detects the following types of pulse faults: (I) S and Z pulses of the same frequency overlapping by more than 400 ms, (ii) continuous S or Z pulses, and (Hi) short S or Z pulses; pusses must persist for longer than 10 ms. The security modifications, some of which have been mentioned above, have greatly improved the reliability of the TXE2 system and reduced the posssbility of exchange service restrictions. It is not practicable to apply fully all of the modifications retrospectively to all marks of TXE2 exchanges; however all of the improvements will be applied to Mark 3 exchanges. Development will be continued to improve further the TXE2 system. 

References 1 Long, R. C., and Gorringe, G. E. Electronic Telephone Exchanges: TXE2—A Small Electronic Exchange System. POEEJ, Vol. 62, p. 12, Apr. 1969. 2 Stacey, R. R. Teletraffic Studies of the TXE2 Electronic Exchange. POEEJ, Vol. 67, p. 73, July 1974.