First-Hand:Philips Telephone Exchanges and Denmark 1970-1980
Contributed by: Swenn Poulsen
Carrier Frequency Request
Around 1970 something happened that was close to a catastrophe for Philips Telecommunication in Copenhagen. The telecom administrations made an open request for carrier frequency equipment.
Why was this so serious? Philips had since the middle of the 1950ies been sole supplier of carrier frequency transmission systems to the administrations, and when something was required they just telephoned us and got a delivery. Due to the increased use of transistors the factory costs had been reduced considerably through the 1960ies, but Max Hansen had not reduced the prices correspondingly. The result was both a large turn-over (one year the turn-over of the Telecommunications Group was the second-largest among all groups in Copenhagen although we were only 10 employees) and a large profit. Max Hansen was also the hero one day when the Danish krone was devalued, as Telecommunication was the only group, which had tied its prices to the Dutch Guilder and could therefore immediately increase the prices in kroner.
But new people were employed in the administrations and thus carrier frequency equipment was asked for in an open request. Max Hansen and Svend Erik Andersen worked of all their might to make an advantageous offer, but the prices at which we had sold equipment prevented us from reducing the new prices too much, as this would cause trouble with the customers. Had they really paid so much too much for the goods?
The result was that Philips lost the request due to the prices. Another supplier came in and our major turn-over during the next years was in extension of the already delivered systems. At the new prices, so it was not the good business it had been.
Autumn 1970: PRX
In the spring of 1970 I got an invitation from Hilversum to be a pupil at the first course in the PRX exchange (Processor Reed eXchange). The course would be in Dutch, however with the documentation in English. It was a course for the Dutch installers and technicians, who should handle the first installations a couple of years later. I was invited as we still expected to enter the Danish market for public exchanges and as the language was no problem.
This fitted well in with both my jobs. ETS3 was kept in operation by the technicians of Jutland Telephone with only a rare assistance by me. As to that, ETS3 had proved that it was not necessary to employ an engineer for each electronic exchange. As to the job as Egon Hansen’s personal assistant there was a sensible replacement. For me it was also fine, my main interest was switching technique and this was an opportunity to concentrate on that again.
My replacement was J. Skjold Sørensen. He worked in Philips-BOFA, the professional audio-visual equipment firm we had together with Bang & Olufsen. There was trouble here between Skjold and Niels Øberg. It was so bad that their opposition to each other might lead to one of them leaving Philips. They were both clever, so it would have been a loss. I suggested to Egon Hansen to remove Skjold from Philips-BOFA and employ him as his personal assistant. There was at the same time a reasonable second job for him as Philips’ contact to Telstar. This became the result and I could spend the autumn in Hilversum (with a weekend in Copenhagen occasionally).
Telstar was started to make video-cassettes with news and entertainment for distribution to i.a. the ships of the Maersk lines. Cassettes were used according to a system designed by Philips. It was a long time before the battle between our VCR 2000, JVC’s VHS and Sony’s Betamax. But possibly because video recorded material for other than studios was so new, Telstar was closed down after a few years.
PRX was computer controlled, but the switching of connections was done with hard contacts in reed-relays. In opposition to the thyristors in ETS3 these contacts could handle ringing voltage and DC current to normal telephone sets.
Already when ETS3 was officially opened in 1967, de Kroes had pointed to this being the right way when Philips should go on from ETS3 to a saleable series-produced system.
A reed contact consists of two thin iron pins inserted in a glass tube from either end. When a magnetic field is applied the two free ends, which are close to each other, will be resp. a North- and a South-pole. They attract each other and make the contact. The Philips reed contacts were about 2 cm long. There was a major development behind the contact, i.a. of the surface layer of the iron pins where they touched each other, such that the contact resistance would remain low, of how to ensure that the pins when the contact was made would touch over a certain area and of the hermetic sealing of the glass tube, such that humidity from the surroundings could not enter the tube. This last point was helped by iron and glass expanding in nearly the same degree when heated. Also in radio valves iron is used to lead wires through the glass, something that Philips were experts in.
One or more reed contacts were assembled into a relay with a common coil around them, placed in a plastic box which was then filled up with araldite. Break contacts were made by replacing one of the reed contacts with a permanent magnet, the effect of which was counteracted by current in the coil. More relays could be placed on a printed circuit board. Their weight and mounting method corresponded to the electronic components, such that no special construction was required for the selector network.
This was an advantage compared with e.g. crossbar selectors. On the other hand each cross point had its own coil, where a crossbar selector needed only 20 coils for a 10 by 10 selector.
The computer (or processor) had been developed considerably since the development of ETS3 had been frozen. Now pure transistor logic was used and more tasks were done in the central computer, as this technique was much faster than using core amplifiers. However, all real-time operations, i.e. operations which have to be done within defined short intervals, took place in peripheral computers. These used also transistor logic.
It was even so, that signals from the lines were filtered in the line circuits. This meant that when a point scanned by the computer changed its state it was a true change to which the computer could react. Accidental noise pulses and transients were removed.
In Bell’s Morris, Illinois, exchange ESS1, the first truly computer controlled exchange (ETS3 was except for the PG controlled by wired logic just as a relay exchange), which started operation in 1967, the main computer was used for signal filtering and real time jobs. It had to look at each point in the line circuits 100 or 200 times a second, day and night, whether there was activity in the circuit or not. This took up a considerable part of the time, so the exchange could not serve a lot of line circuits. L. M. Ericsson’s AKE exchange, where the first was to go in operation in 1967 as a trunk exchange (connecting exchanges in Copenhagen to exchanges in the rest of Denmark and abroad) was based on the same “modern” idea. T. Albrectsen from Copenhagen Telephone once asked me about the signal filtering in PRX and did not like that it was done peripherally.
But the idea of Philips was the right one, and not only because the Dutch administration required it. For them it was not sufficient if signal filtering was done by looking at the state in some discrete moments. ESS1 soon got a signal processor for the real-time jobs added as a front-end to the main computer and this doubled the capacity of the exchange. AKE was changed in a similar way before it at last, in 1974, went into operation.
The main computer of PRX was built as a general purpose computer and not specifically for control of an exchange. The idea at that time was that it should be possible to assemble the program for a computer at that computer itself, i.e. when the program units were written in a mnemonic code, the computer should be able to translate this code to the final program units for the computer, catch errors made by the programmers, assemble program units to complete programs, control that the modules fitted together etc. This required a general purpose computer which could also afterwards handle the telephone traffic.
J. Brouwer ran the course. Shortly after it he emigrated to a position in Philips Australia. This was quite a change, as he was about 55 years of age at that time.
The first PRX went into operation in Overvecht near Utrecht in 1974 and then there was a fast introduction of PRX in Holland. Holland and Saudi Arabia were during a period the two countries where the largest percentage of the subscribers were connected to computer controlled exchanges. In both countries this was under PRX exchanges. The order for exchanges to Saudi Arabia in the end of the 1970ies, obtained in a joint venture with L. M. Ericsson, was a peak point in the history of PRX.
“We could also accept such an order”
In reality Philips was the front runner among all European telephone exchange manufacturers in the development of computer controlled public exchanges. This in spite of that L. M. Ericsson already early in the 1960ies had received the order for a trunk exchange of the AKE type for Copenhagen. As said above it was to go in operation in 1967, but was only ready in the spring of 1974. In the meanwhile Ericsson filled up all available space in the exchange building with older equipment to cover the need for connections. I remember A. H. Kvist from the State Telephone saying when the AKE had begun operation that now it would work with no problems for the next 40 years, as it had already used all the down-time allowed in the contract for this period!
As to the State Telephone, we had in Denmark 4 administrations, 3 local for Sealand (Copenhagen Telephone), Funen and Jutland, and 1 (state-owned) over them all for mutual connections and connections abroad. And for local telephony in Southern Jutland! The latter was a remnant of the situation from 1864 to 1920 when this part of Denmark had been under German rule. After the First World War when the people in Southern Jutland had voted for returning to Denmark the telephone system was broken down and none of the local administrations would take over. So the state had to do it. My perception was in the early 1970ies that Ericsson was far in front of us. I said so to de Kroes, who answered that such an order Philips could also have accepted. I persisted with a question if such an order might also have a condition about a fine if delivery was late? Oh sure, said de Kroes. Again I asked if the fine might have been unlimited in stead of the maximum 10% which was usual? Then de Kroes said definitely no.
As said the first PRX went into operation in 1974 and many were set up in Holland over the next years. There was a funny consequence of the administration being a state institution with a monopoly and at the same time being required to treat all subscribers equally. The facilities already built into the first PRX and made possible by the computer control were not offered to the subscribers, not even at an additional price. They could at that time not be offered to all users in Holland, and then it was not allowed to offer them to anybody! Only in the 1980ies when PRX was spread all over Holland the additional services came up for sale. Until then PRX was preferred only because it gave operational advantages for the Dutch administration.
Test with keytone dialling in Skanderborg
The keypad of ETS3 was standardised as to the place of the keys (however with 7 at the upper left in Denmark as required by Copenhagen and Jutland Telephone), but the electrical signals were not. The Bell Laboratories had in 1962 described a signalling system in the Bell System Technical Journal and it was agreed that this system should be used everywhere. At that time ETS3 was so far in its development that it was not used in it – and it would also have been more expensive as it required two tone generators in the telephone sets.
This is an early example of standardisation within the national networks. The dials were not standardised, except that they all gave from 1 to 10 short interruptions (pulses) of the current to the set when they turned back to their rest position. But the digits were not standardised, most countries used 1 pulse to indicate the digit 1 up to 10 pulses indicating 0, i.e. the dial counted 1, 2,…0. But in Sweden it counted 0, 1, 2,…9. And on New Zealand it counted 0, 9, 8,…1. This latter dial was also used in Oslo, Norway, while the rest of Norway used 1, 2,…0! Quite some adaptation when you should deliver exchanges in many countries.
Bell’s system has the characteristic that it uses two sets of 4 frequencies each, one set from 697 to 941 Hz (cycles per second) and one set from 1209 to 1633 Hz. Each signal consists of one frequency from each set, sent while the telephone set is off hook (i.e. the handset is lifted up and there is current to the set and its microphone). The higher frequencies were not harmonics of the lower, and this made it difficult to simulate a digit by whistling in the microphone. 4 by 4 gives 16 combinations, used for the digits 1 to 0, *, # and four signals called A, B, C and D. These four are the only ones using 1633 Hz, so normal telephone sets, which shall only send digits can do with 3 tunings of the oscillator for the high set of frequencies and 10 or 12 keys.
Originally the system was for signalling from a set to the nearest exchange, it would thus only have to cope with the attenuation in a subscriber line between sender and receiver. The receiver could therefore be rather insensitive. This protected it also from simulations through the microphone.
But down the line the system is also used for signalling from subscriber to subscriber. Thus it has to cope with the attenuation in two subscriber lines and the telephone network in between. The receiver must be more sensitive and is easier to fool. This has not prevented a widespread use of the system, only think of voice-response, where you get spoken directions and answer them by keying digits, or in-dialling, where you call a PABX, get dial tone from it and keys the wanted extension number.
But back in the early 1970ies there were no key telephone sets (except under ETS3, but this was also the only exchange able to receive the signals), and although Bell’s system looked promising nobody knew if it would work in real life. Thus the administrations started testing it.
This was easier said than done. To test you must have equipment to work on. Telephone sets could be had from the US. Receivers would be too much of an effort to develop, but luckily they found suitable receivers in a catalogue from Philips in Canada. Jutland Telephone wanted to make tests around the exchange in Skanderborg (a town some 40 km from their headquarters in Århus) and called on us. We got the necessary receivers with their specifications etc.
Jutland Telephone also modified the registers (receivers of dial pulses) in the Skanderborg exchange, such that they could be controlled by the output of the tone receivers. This made the test realistic: Users of key telephone sets went off hook as usually, dialled by pushing keys and got the wanted number.
The test went so well that the Danish administrations could after a few years introduce keytone receivers in all their exchanges and introduce key telephone sets as the standard set you got when you entered a subscription. Already in 1976 the F76 phone from the Kirk Telephone Works came, the model called “Comet” designed by Jacob Jensen. In the start it did not live up to the quality requirements of Copenhagen Telephone, but in 1980 it was improved and introduced in the whole of Denmark as DK80. In the meanwhile Telephone Factory Automatic had developed its F78 model, later called danMark.
When telephone sets were later liberalised (after a type approval they could be sold directly to customers) it was a requirement that they signalled with keytones. In 1995 it was so far that all other terminals in order to pass a type approval should be able to do the same (if they were to signal, an answering machine should not comply with this requirement). But there were still people maintaining that they could not grasp how to dial with a keypad, only the rotating dial was good enough. They also said that it is too bad that new services were introduced which require keypads, such as in-dialling.
Quite to the contrary our children hardly knew how to use a dial. I used the possibilities I had in connection with equipment from Pye-TMC (see later) to test key dialling at home even before the exchange could receive keytones. Pye-TMC had designed key-sets which converted each push of a key into the corresponding number of dial pulses!
1971: Remote reading of electricity meters
Early in the 1970ies the idea came from TeKaDe in Nuremburg that they could expand their program within data transmission with modems for remote reading of the electricity meters in private residences.
The power companies paid much to send people out every quarter year to read the meters. If a modem was set up in connection with the meter and the telephone line the modem could be called in the middle of the night at almost no cost. The problem is of course that the devices are not free of charge and can break down. The whole service organisation behind the new form of reading must also be in place.
There had just been an article in the IEEE Transactions on Communications on remote reading, and I took this article with me to a meeting at TeKaDe. But I brought something more: A description of what the Danish power companies did to reduce the cost.
They had replaced the quarterly visit with a post card. The customer filled it in himself and mailed it. The bill was then made up on the basis of this card.
4 cards a year (later just one with intermediate bills made up based on the expected use of current), 2 USD for the mailings (the power company paid this) and maybe 5 USD for other costs. Even one visit would cost more!
Some people would fear that people were dishonest. But the power company could just visit those with a funny development of the electricity use and those moving house. This system is still in use after 2000, now with the use also of voice-response or internet.
A yearly cost of some 7 USD was what remote reading should compete with. This was simply not possible.
I do not know if it was my arguments and the reference to the praxis of the Danish power companies which resulted in TeKaDe giving up remote reading. Only after 2000 remote reading might come in as an extra function when meters shall adjust to a variable price for electricity over 24 hours and control when (at low cost) e.g. a washing machine may start.
1972: Leaving Philips?
PRX never made it to Denmark. We tried to introduce the system, but as Hilversum would not support it seriously, i.a. by adapting to the Danish signalling protocols without an agreement about delivery of exchanges with thousands of lines, and as the administrations would not agree to anything before exchanges in Denmark had proved their worth, there was not much to discuss.
We knew that Copenhagen Telephone should decide on a replacement of the semi-automatic exchanges (from the 1930ies) around 1972. They were exchanges with a total of about 200.000 lines, so it was somewhat better than delivering exchanges to cope with the yearly expansion of about 2% of their 1.000.000 lines and about the same number of replacements. I found that our system was so good that we should aim at this bunch of exchanges, adapt to the network and supply an exchange without being promised later deliveries. We would anyhow be the natural choice for the replacement of the old exchanges due to our quality.
I went so far that I wrote a draft about this, threatening to leave the company if nothing was done in this matter. However, this was a rather crucial step, so I would not go ahead with it without some good advice. I telephoned a Saturday to Henning Drejer, an engineer in the Telecommunications Group, who then handled deliveries to the Danish defence. I had confidence in him and asked if I might come along with “something”. He said OK and I visited him and showed him my draft.
His reaction was that my threat would not influence Philips, this big machine would not alter course due to my arguments. We did also not know all the considerations behind the decision to pursue a certain policy. In brief, if I was generally satisfied with the company and thought that also without delivery of the replacements of the semi-automatic exchanges, there would be interesting tasks, I should tear the draft in very small pieces and go on.
I followed his advice and did never regret it. Activity with public exchanges never materialised, but I soon got so busy with private exchanges (PABXes) that it took up all my energy.
1973: Data communication course in Montreux
Data communication was coming to the fore. Earlier it was about telegraphy and telex, but new uses appeared and the first modems came on the market.
There came invitations to courses on the new possibilities and I was allowed to participate in one of them, in Montreux, Switzerland, in April 1973.
It was a good introduction to the possibilities as seen by the supplier organising the course. It was very useful later on!
The course lasted a week and I seized the opportunity once again to also have a little holiday. My wife was with me and we took the train through Germany and Switzerland. From our room we could look over the Lac Leman (the lake of Geneva) to the French Alps on the other side.
Our hope was that we were coming down to the spring, but we woke up several days to snow on the hillsides. Thus, one day we decided that now we would see the spring and took the train up through the Rhone valley and through the Simplon tunnel to Stresa, Italy. Her it was spring and we enjoyed a day’s warmth before we returned to the cold Switzerland.
Another day we went by train into the area behind Montreux, to Gstaad. There were gearwheels under the train to cope with the steep gradients to the winter sports place of the rich and beautiful. We did not see them, the season was probably over, but walked a good long way before we again took the train back.
We paid of course also a visit to Chillon, a small castle close to the coast of the lake near Montreux, and walked to neighbouring small towns like Vevey. One should always take the time to be a tourist the first time you are in a new place. And bring the family if the stay lasts a week!
1973: Ericsson DC-signalling over PCM systems
PCM (Pulse Code Modulation) transmission systems were so far developed in 1973 that it seemed they might be useful (i.e. economic) between exchanges within one town. In Copenhagen it meant that they should be able to transfer the DC-signalling used between the L. M. Ericsson exchanges. ETS3 had adapted to this signalling, so I was put on the task.
Signalling can be divided in two types: Line signalling over each link (trunk line) from exchange to exchange to seize and release the link and register signalling end to end from the register in the caller’s exchange to a marker in a remote exchange such as the called subscriber’s exchange. The DC-signalling contained both types of signalling and was therefore limited to use within one town. If you called outside Copenhagen a code converter in the trunk exchange would adapt between the signalling systems.
Hilversum had developed circuits for translation between the 2-wire interfaces of the exchanges and the signalling interface of the transmission equipment, and these circuits were to be tested in cooperation with Copenhagen Telephone. From them, Poul Friis participated and from Hilversum van Deursen came from the development laboratories. The tests took place in one of the crossbar exchange buildings.
We did not put in transmission equipment, the two test circuits were connected back to back as a transmission line. We compared the function via the test circuits and without them.
It all worked fine, but there was seemingly a problem: The signal sender detected if there was current in the line (to see if the receiver was ready) before sending each digit. Without the signalling converters we saw that there was current immediately, such that the signalling started with a short pulse. With signalling converters the receiver was never switched in at the start, such that the first pulse was longer. Why?
The explanation was rather simple but it took time to find it: Before the sender in the A-end was switched to the line there was no polarity on it and consequently – when our circuits were used - no polarity from our receiving circuit to the receiver in the B-end. Therefore our circuit in the B-end could not detect if the receiver was connected and the signal back to our circuit in the A-end was that the receiver in the B-exchange was disconnected. Thus our circuit showed a large resistance to the A-exchange. When the A-exchange switched on polarity there was (almost) no current, but our circuit could now send information about the polarity to the other end. Our circuit in the B-end would then connect the same polarity to the receiver in the exchange. Now there was a current and this was signalled back to our circuit in the A-end which changed to a low resistance. Now the sender in the exchange could see the current and send the rest of the pulses.
Thus there was an unavoidable delay through our equipment, aggravated by the need for a certain filtering against noise pulses. Our circuits must not react immediately on signals from the exchanges. This meant a little longer time for the establishment of connections, and as the time for this, when registers, code senders, code receivers and markers were busy in a call, was very short, this time was significant and could mean a need in some places for more code senders and receivers. The users would not feel a thing, the delay was after all too short for this.
We never delivered equipment for the DC-signalling via transmission systems in Copenhagen. I do not think it was the delay and the consequent requirement for expansion of certain exchanges which was the cause. It was rather that the signalling converters were complicated and that there was no use for a large number of them. Thus they would be rather costly, making PCM between the exchanges in Copenhagen too expensive. The exchanges were later modified so it was only the “slow” line signalling about seizure and release, which was sent over the signalling channels in the PCM systems. The fast register signalling was then transferred over the speech channels as MFC-signals (Multi Frequency Code) before the call was switched through. At that time Philips almost-monopoly on delivery of transmission systems was broken.
I was also mixed up in another future project in the early 1970ies, in mobile telephony. We had in Denmark at that time a manual system, but in cooperation with the other Nordic administrations the Danish State Telephone started the development of a common automatic system, later to be called NMT (Nordic Mobile Telephony), the first system in the world with roaming (transfer of calls from base station to base station) across borders.
Within Philips the development at that time was a responsibility of TeKaDe in Nuremburg (incidentally they had a real telecommunication address, Thurn und Taxis Strasse, named after the first organisers of a postal service in Europe). In Germany there was already an automatic system in operation, called öbl-B, meaning “public mobile landradio-B”. Their manual system was called öbl-A. They later went on with C (an analogue system like NMT), D (GSM) and E (DCS 1800). We would like to sell öbl-B in the Nordic countries, as TeKaDe was supplier of both base stations and mobile telephones to it, so we had meetings with the State Telephone, whose delegation was led by Marius Jakobsen.
Our proposal was however too far from what the Nordic administrations had in mind. There was especially two points: In Germany you had to know approximately where the mobile telephone was in order to reach it. You dialled an area number and the directory number and the call was only made in that area. NMT should have an automatic follow-up on where the mobile telephones were, so it was only necessary to dial the directory number.
The other point was the capacity as to directory numbers. In the German system it was far too little, only 100.000 numbers in the whole of Germany. NMT should be able to have 1.000.000 directory numbers in each of the Nordic countries. This was due to a quite different philosophy regarding the marketing and the price for the service. In Germany mobile telephones were only leased from the State Telephony and the price was as I remember around 125 USD per month. This resulted of course in that there were only a few customers.
In the Nordic countries the policy would be followed which had already made the manual service popular. Mobile telephones could be sold on the open market after a type approval and the State Telephone should only be paid for the traffic. That is: be paid what it took to deliver the service. Therefore it could be anticipated that prices would be low and there would be many customers.
Thus TeKaDe was unable to sell the German system. But they returned with a proposal for exchanges, which was very advanced for that time: The traffic circuits of the exchange were assembled in bundles of the same type of circuits, each bundle controlled by its own computer. These bundles were connected by a high-speed bus, over which the computers could communicate. This was the first time I saw a draft of a totally decentralised system. But this did not result in an order, when NMT started operation early in the 1980ies the exchanges were from L. M. Ericsson.
But Philips played its part in mobile telephony in Denmark. For one thing the company took over AP Radiotelephones in the late 1970ies, a firm making mobile telephones for the manual system and active in the preparations for NMT. Here at last Philips had found a replacement for the making of mechanical channel selectors for TV-sets. The production was moved to Philips’ factory in Copenhagen and for the next 15 years the NMT telephones were developed and produced there.
In addition Philips Telecommunications was busy selling base stations to the State Telephone. There were two distinct groups of customers: The sellers of mobile telephones, which AP Radiotelephones (the name was kept) took care of, and the providers of base stations and exchanges in the State Telephone. The latter was a typical Telecommunications customer and thus we sold equipment to them. K. B. Mortensen joined us from the development group in the factory and he got the products of Philips adapted to the NMT specifications. Our prices were also right in many cases, so we succeeded as a supplier to the NMT. This continued later with base stations for GSM, the common European system.
After giving up production in Denmark in the early 1990ies the still smaller production of analogue mobile telephones moved to Austria. Development and production of equipment for GSM took place in France. TeKaDe was almost closed after considerable losses during several years.
Automatic Key Boxes, Key and Lamp Units
In order to improve the telephone service of the different groups in Philips Copenhagen I had earlier designed the “Meadow Road Arrangement” for our neon factory and introduced group hunting for Philips Light in our UB49 PABX. Now more groups wanted a better service where a secretary could answer calls to employees who were away from their office.
In the meanwhile the care for the UB49 PABX had been taken over by Max Andreassen, who had worked in Philips Telecommunications with the installation of the many carrier frequency stations supplied to the administrations since the middle of the 1950ies. He was a clever and ingenious technician, but on the other hand it was clear that a repetition of the arrangement to the neon factory was out of the question. This assembly of individual keys would be too large and too expensive.
We considered the possibilities and found at first a component with 10 mechanically coupled keys from the Telephone Factory Automatic. When a key was pushed down, any other pushed key would release. Max Andreassen designed and made a box with this component and 10 neon lamps, one for each key. The extension lines were led in parallel to the offices and to this box. If an extension was rung its lamp would go on, and the secretary could connect to the caller by pushing the corresponding key.
The secretary should not answer immediately in all cases. If she thought somebody was in his place, she had to give him a chance to answer himself. In groups where people were thought to remember well, we provided as an alternative switches in the offices, so the extension line was either led to the telephone set or to the key box. People had to remember the switch when they left or entered their offices, and the secretary should in that case answer every ringing immediately. Everybody liked this flexibility.
Max Andreassen made several boxes of this kind but they required a wooden box for their assembly and were therefore still a little too expensive. When we later had contact with Pye-TMC (see later) we introduced their KLU (Key and Lamp Unit), where keys and lamps were already mounted in a plastic box.
However, we could not use the KLU as it was when we only wanted the lamp to go on during ringing. Thus the lamps were replaced with other types of lamps and ringing detectors. The keys were wired such that each had the positions “up” without a connection, “mid” with a connection from that extension line to the telephone set of the secretary and “down” where the extension line was held. The wiring was of course also done such that if more keys were in their “mid” position only one line was connected to the secretary, the others were held.
KLU came in rather general use at Philips as it was well adapted to the wishes of the users for easy handling and provided a good service towards the calling customers. We showed it of course to the PABX people of Copenhagen Telephone and they found it to be a solution suited for one of their customers. Thus a KLU was installed in the emergency oil stock in Hedehusene, where the oil companies had a common stock of oil for several months (it was after the 1973 oil crisis).
In 1968 Max Hansen had visited Pye-TMC in Southern London, a company which had become part of Philips when the whole of the Pye-concern was bought. In England they said of course that it was Pye who had bought Philips!
TMC meant Telephone Manufacturing Company and they supplied many telephone sets to British Telephone. But Max Hansen had looked more to their development and came home enthusiastic over new developments he had seen in their laboratories, led by John Rhodes, one of whose ancestors gave name to Rhodesia. They had developed a new transistor technique, MOS for Metal Oxide Semiconductor. With this technique one could with a metal electrode, isolated with a thin layer of silicon oxide from the semiconductor, control the current in the semiconductor. The control electrode was thus totally isolated from the controlled circuit, except for charges and discharges of the small capacitor formed by the electrode and the semiconductor. This was so to say a return to the good old valve technique but at the low voltages of the transistors. It was now possible to make circuits using very little current. MOS opened for the epoch where a pocket calculator for a few USD has more power than a house full of computers in the early 1950ies and could deliver this power for years from an in-built battery!
They had especially developed a four phase logic where the state of the circuit was controlled from step to step, where states could be combined for the next step and they did this with a minimum of supply current because there was never a conducting path from plus to minus, current was only used to charge the MOS transistors.
In these years liberalisation of the telecommunications networks began. An opening was made for type approval and sale on the open market of auxiliary equipment to connect to a telephone set. Speech amplifiers were such a product. Another one, interesting for the MOS technique, was digit senders, i.e. devices which could store telephone numbers and send them with the push of a single button. The Danish administrations issued “Circular 27” with rules for such auxiliary equipment and we had it translated (if it was not already in English, the administrations made an early official translation of it) and sent to England.
The first product was Prestafone. It was a complete telephone set (and could therefore not be approved). It was a key telephone, but when a key was pushed the corresponding number of dial pulses was sent out. Dial pulses were the only language the public exchanges could understand then. Prestafone could do with the little current the telephone line could provide when idle (when ideally the set should interrupt the line completely).
We later got proper digit senders which were approved such as Sphericall for up to 10 numbers and Multicall for up to 62 numbers. Both remembered also the latest number keyed digit by digit. They were sold through Max Manus, a company which also sold the dictation machines etc. from Philips.
From Philips Telecommunications we sold directly to the State Telephone shop in the airport the Operator’s Key Sender. In this shop the operators made international calls for their customers and connected the calls to telephone boxes. It was a major improvement for the operators when they should not any more dial hundreds of calls each day but could just push keys!
A couple of Prestafones were sold to the Soviet trade representative in Copenhagen. They addressed us and I demonstrated the device. They wanted to buy but this was delayed a bit due to the Cocom rules determining what might be sold to the Soviet Union. Electronics was mostly blacklisted. I asked Pye-TMC, who answered back that this MOS-equipment was not prohibited by the rules and we delivered the telephone sets.
The concern language within Philips was English. In principle all communication between the national organisations should take place in this language. It did not always work. We received e.g. manuals for data transmission equipment from TeKaDe in German and might get manuals for equipment from France in French. But mostly the letters from Paris were in English and I boasted of this once when visiting Pye-TMC: Even the French wrote us in English, what a discipline within Philips!
I got a typical British answer: “Yes, they had experienced the same – when the French wanted to obtain something!”
Data circuits at the Copenhagen University
Liberalisation was also significant for data circuits. “Circular 27” was aimed at data modems for switched telephone connections, but it was not always sufficient when you wanted to send many bit per second. At the time one could send 300, 1200 or 2400 bit per second over a telephone channel (and only the first speed in duplex, i.e. in both directions simultaneously).
One thing was the possible speed via switched or leased circuits, where transmission systems limited the bandwidth to 300 to 3400 Hz. Another was transmission over short distance leased lines, where the line was just a pair of wires in a subscriber cable, also called leased lines of local quality. They had no sharp limits to the bandwidth, there was only an increasing attenuation with frequency. For leased lines one should have an approval of the terminals to be connected to the line according to the relevant section of “Circular 12” which had different rules for each type of leased lines.
In Nuremburg TeKaDe made a DC-modem for data transmission over wire pairs, which we got approved. In short it translated the levels from the V.24 interface (with levels according to V.28 (two CCITT recommendations)) of the terminal, which were rather high and with large input and output impedances, to the low levels and impedances characteristic for a wire pair. It could work as fast as 48 kbit per second even if the V.28 interface was only specified for up to 20 kbit per second. However, if one did not load the V.28 interface with the allowed capacitance it could work at higher speeds. They required a 2-wire line in each direction.
The University of Copenhagen had a computer centre in a campus a little to the North of central Copenhagen. About 500 m away at an institute a front-end processor for the main computer was situated. Programs and data could be loaded into this processor and the results could be printed just as on the terminals in the computer room. The connection between computer and front-end processor had to be high-speed. Thus the university bought two of the DC-modems from us and bought a couple of leased line of local quality between the two addresses from Copenhagen Telephone and set up the equipment. It did not work.
The problem turned out to be the leased line, for which also the State Telephone (who cared for all data transmission) was called in. The leased line went from the computer address to the nearest public exchange in a subscriber cable. From here it went in a trunk line to the centre of telecommunications in Copenhagen (where the L. M. Ericsson AKE exchange was being installed). It was then coupled to another trunk line to the public exchange nearest to the front-end processor. The total resistance in this route was rather high, but in addition the trunk lines were loaded lines. This means that there were inductances inserted which reduce the attenuation of the speech frequencies, but increase the attenuation of higher frequencies considerably. Exactly the frequencies we needed to transmit!
Both the State and Copenhagen Telephone could see this and contrary to all what is said about monopolies and bureaucracy in these dinosaurs they found a solution. The leased line was led a much shorter route and not via loaded lines, and then the data transmission worked as it should.
This was not to be my last visit to the computer or the front-end processor. Some time later I was called upon – there was no connection from computer to front-end processor! So I went along on trouble shooting. I brought my home-made universal meter as a tool. It was a Heathkit and it worked fine. I could see from the computer room that there was no DC passage in the leased line, meaning that the fault was in the connection. Then I went to the front-end room. There was no DC to the modem in that end from the plug in the wall, but there was a connection in the modem itself. By examination of the plug it was found that one of the wires was interrupted in it, probably because somebody had walked along the wall behind the front-end processor and stepped on the cable. The wire was soldered in place, the cable connected and the path behind the processor blocked somewhat better. Then back to the computer room and see that now there was a DC passage the whole way. And the data transmission worked again.
Data circuits at SAS
Within Philips we had also another supplier of modems for leased lines of local quality (or base band modems as they were also called). It was the French subsidiary TRT (Telecommunications Radioelectronique et Telefonique) who made the Sematrans modems. They worked according to a very different principle than the modems from TeKaDe, as they translated the 1s and 0s on the terminal interface to patterns of positive and negative pulses, which should balance each other (come in equal numbers such that the absence of a DC component made it possible to have transformers in the line) and come as few as possible (to transmit at the lowest possible frequency having the least attenuation). Finally they should for an arbitrary pattern of bits ensure a certain number of changes on the line between plus and minus (such that the receiver could adapt its clock to the sender).
I seem to remember that the pattern on the line had a transition between plus and minus or the other way in the middle of each 1 and a transition of the same kind in the start of every 0 if the previous bit was also a 0. If it was a 1 there was no transition. Thus, if a series of 0s or 1s were sent, the line pattern was just a square wave with each plus or minus lasting one bit time. In that case the receiver could not see if it got 0s or 1s, but such a sequence is also devoid of information. As soon as there was a change between 0s and 1s from the terminal, the pattern was such that the receiver could detect the bits correctly. It was e.g. only possible to receive a plus or a minus lasting two bit times if the sender had delivered the sequence …101…
While the modem from TeKaDe could follow any speed of the transmitted signal, the change into pulses in Sematrans necessitated that it was adapted to the clock of the sending terminal. This could only take place in steps, but they covered all expected data speeds. There was also sufficient tolerance in the receiving clock to adapt to a received signal if it was within the speeds allowed in the recommendations from CCITT. The V.24 interface provide at the sending side for a clock from either the terminal (telling the modem when to read the bit) or from the modem (telling the terminal when to forward the next bit). At the receiving side the clock must come from the modem, telling the terminal when to read the bit.
There were two kinds of Sematrans. Their difference was in the V.24 interfaces. One had the levels recommended in V.28, i.e. each data, clock and control circuit had one pin, with a common return for all of them in another pin. The other Sematrans had levels according to V.35. This means that the data and clock circuits each had two pins in the interface with a balanced transmission, i.e. the same current in opposite directions over the two pins. This ensured if the wires behind each pair of pins were a twisted pair that the circuits would neither disturb other circuits or be disturbed by them. The impedance of these circuits was also low such that a capacitive line would not influence the signal much. All other interface circuits used levels according to V.28 (they changed more slowly).
SAS (Scandinavian Airlines System) had bought a couple of Sematrans modems with V.35 interfaces for use between their data centre at Meadow Road (Philips’ former neon factory) and the airport. They should work at 48 kbit per second, a high speed then. But the test done by SAS showed that the connection did not work, although it was within the range of the modems. There was even a peculiarity in the test result: They had tested with two different test equipments, one a Sematest from TRT, with test pattern generator, receiver and comparator built into an attaché case, the other from another supplier. The Sematest showed only a few bit errors, the other showed lots of bit errors.
A technician came up from TRT and for a whole morning he evaluated the set-up but could only confirm the results reported by SAS. He had no explanation why equipment from TRT acquitted the modems while the other denounced them. We went for lunch and there I got an idea: Could it be the relation between data and clock signal? The clock signal from the sending side of the V.35 interface is plus in the first half of each bit time, minus in the other. The receiving side of the V.35 interface shall read the data bit when the clock changes from plus to minus. If the clock wires were changed on one side of the interface, the order to read the bit would come just as it changed to the next bit. This might also explain why Sematest acquitted the modems. If it measured the bit a little postponed from the clock bit change, the data bit had reached its final value.
We went back and checked the drawings in relation to the modems. And behold! One of the clock circuits had its wires crossed when soldered in the plug. It was corrected and both test equipments now showed all to be OK.
1972: UH 900
In the office building of Philips Copenhagen there was originally several factory rooms. About 1970 there were by and by so many offices in the building requiring telephones that the UB49 exchange was filled up. There was no space for more equipment in the exchange room. The exchange also served the factory site some 200 meters down the road (we had a cable laid with the necessary wire pairs for the connections to the telephone sets). A logical solution would be to install a separate PABX for the factory site to free extension numbers in the UB49. There had to be some lines between the two exchanges, but this traffic was not great so a few lines would do.
However there was no place for an exchange at the factory site. No problem, there was room in the office building and we had the necessary cable pairs between the two sites. Thus the PABX was installed in the office building with a common battery room with the old PABX. We choose the latest system from Hilversum, a UH 900 with 300 extension numbers.
UH 900 was an electromechanical system with relays and rotating selectors. The selectors had 50 outlets and could move over 150 contacts per second. Thus they operated as fast as the selectors in UB49 where both figures are twice as high. They were considerably simpler. There was only one magnet in the selector. It pressed on a thin flexible plate when the magnet was energised. This plate was a gear wheel which either fitted into a fixed stop controlling the position of the selector wipers on the fixed contacts or fitted into a gear wheel on a common shaft when pressed by the magnet. The tolerances were such that any selector mechanism fitted into any fixed selector part without further adaptation. In UB49 one had to adjust the mechanism after mounting a selector in a new place.
The exchange was of course supplied with power via a rectifier and a battery. The common shafts of the selectors were driven by AC motors and an electronic circuit generated the AC. The motor was very simple, its principle was used in many household items from Philips: The rotor was a toothed block of iron, the stator had two poles, about 90 degrees apart. One had a thick copper wire around it. This delayed the magnetic field such that its maximum at the two poles did not occur simultaneously, and this gave the momentum to turn the rotor and the shaft. It was an improvement from UB49, which used DC motors with their brushes, commutators and wired rotors, i.e. parts subjected to wear, while the motors of UH 900 only had their bearings to worry about!
The electronics in the AC generator were also not subject to wear. This generator did double duty as it also supplied the ringing voltage to called telephone sets.
There was more electronics in UH 900. Where a diode could be used in stead of a relay contact it was done. The exchange was also born with a memory of ferrite cores to store extension numbers, so the number appeared for the operator if a call returned to her. We had already this function working in UB49 since 1954. In UH 900 the displays used rear projection. There were 12 lamps in a space of about 1 by 2 cm deep inside the display. In front of the lamps was a dark film with the digits as clear areas. Each digit (one at a time) was projected through a lens to a matt glass plate in the front.
But the operators’ desks were not ergonomic enough for the physiotherapist of Philips, Anette Prætorius. The operators should stretch too much to press the keys, their upper arms could not hang down vertically during this. The carpenter of the company was called in and made a semicircular cut in the desks so the operators came nearer to the keys. This caused a new problem: Under the desk there was a row of soldering tags where the cables from the exchange were connected to the desk. They were screened by an iron plate with an iron bar to support the plate. Bad enough for the operators’ shin bones before the cut, catastrophic after! So we replaced the bar and plate with a larger softly rounded plate which did not harm the shin bones.
We asked for a special installation in the exchange: There was not space enough in the cabinets for the number of rerouting circuits we would need. Such a circuit will pass a call on to another extension if the one called first does not answer within some 15 seconds. So we ordered a whole cabinet filled with rerouting circuits and with cables to an addition to the main distribution frame (where the cables from the exchange meet the cables from the extensions). This made it easy to change allocations.
UH 900 was a good system, very reliable and stable because it was rather simply built. It was also installed in a professional way as we hired two installers from Hilversum to do the job. Krauwel and de Voogd did a good piece of work.
We invited the people from Copenhagen Telephone, who dealt with PABXes, to see the new exchange which was now connected to their network. They came with their manager, Arne Oksholm, in front. It must have impressed them, for they started discussions with us about introducing the system to their customers. These discussions took place up to the middle of 1973, when they choose another type of PABX (Minimat 800 from ITT Standard Electric). But it opened for discussions about the EBX 8000, the first computer controlled PABX from Philips. More about that later!
Spring 1973, the Rubber Band Exchange
Jutland Telephone had taken over ETS3 in 1969 as it worked so well that they wanted to keep it in operation after the agreed test time of 5 years.
They did, however, not have the exchange in operation much longer than originally agreed. Something quite unexpected happened. The electronic cross points in the exchange consisted of a Germanium thyristor, a Germanium diode and a resistor of 27 kΩ held together by a rubber band. In 1972 it happened more and more often that a connection through the exchange could not be made. The in-built programs for trouble shooting told that the faults occurred in the cross points and that they happened at random over the network. The common cause was that the resistor was interrupted and then the marking voltage could of course not switch the cross point on.
Further analysis showed that there were tiny holes in the lacquer of the resistors under the rubber band. The softening agent in the rubber had in some way attacked the lacquer and the resistive material under it. In Hilversum they could not understand how this was possible, they had made great efforts to find the right combination of materials and i.a. used synthetic rubber. However, no other explanation could be found.
It was a very awkward fault as it occurred at random, and as it would be quite impossible to replace all the cross points, quite apart from the fact that the thyristors were a series made to very special specifications several years before at the transistor factory in Belgium and were not more available. Jutland Telephone found that the exchange room could be used for other purposes and stopped the test. The subscribers were connected to “old fashioned” exchanges again.
15 years later we presented our then newest PABX, Sopho-S, for Jutland Telephone in Århus. Rob Mulder, who was in charge of the development, could in his introduction not refrain from mentioning the old ties between Jutland Telephone and Philips, referring to the cooperation around the “rubber band exchange”.
June 1973, EBX 8000
In June 1973 we had a two days meeting with the PABX people of Copenhagen Telephony, led by Oksholm. During the first day we continued our talks on UH 900 and on the second day they had asked us to tell about our future program within PABXes. This was the first time we (and they) were to hear about EBX 8000, and the head of its development, Rom van der Schaaf, was in Copenhagen to do the presentation.
He explained the system and this made obviously quite an impression. He answered many questions, of which I asked the most. My background from PRX was invaluable. Oksholm left the meeting around 2 pm due to another meeting, thanking for the presentation, and the rest of us continued with questions and answers.
Around 3.30 pm the telephone in the meeting room rang and I answered it. It was Oksholm who was in a meeting with the CEO of Copenhagen Telephony, Johs. Rosbæk. Oksholm had told him about EBX 8000 and now he would like to know when such an exchange might be delivered to serve the headquarters of Copenhagen Telephony? Would the sales manager from Hilversum, Hugo Simons, give an estimate – but in secret, the other people from the administration should not be informed. I promised to return soonest possible and took it up with Simons (and Max Hansen) after the meeting. Spring 1976 was the answer from Simons. We passed on the answer to Oksholm the next day and he visited us later that day with his secretary, Anja Krumins, to give us more details about the required size and the wanted facilities. This information was immediately sent to Hilversum.
June 1974, Offer of EBX 8000 to Copenhagen Telephony
Discussions about the requirements to the new PABX for their headquarters continued all through the autumn and winter. In April 1974 we were asked to provide a definite offer. The request went to Hilversum, as the system was not yet revealed for the national organisations - except for the few chosen like us. But rules had not yet been agreed such that we could ourselves make an offer.
In the middle of May we at last received an offer in a telex from Hilversum. There were many reservations and it was not very clear, so the only way for us was to translate it as best as we could and send it to Copenhagen Telephone.
It was Whitsun on June 1st and on the 2nd I was due to go to Holland for a fortnight where EBX 8000 should be presented for the most important national organisations. The Friday before, May 30th, Oksholm phoned me and asked me to come to a meeting the same afternoon concerning our offer.
I came of course and there were a number of people from the PABX group around the table. From the end of the table Oksholm started by saying that they were unable to treat our offer, it was too mixed up and unclear. It could only be scrapped. It would be necessary for us to provide a new offer which fulfilled their need for clarity and precision.
He continued by telling exactly how such an offer should be set up. How it ought to be divided in sections on traffic capacity, amount of equipment, reliability, signalling on extension and exchange lines, prices and other crucial matters. I wrote down all that was said.
I felt almost as if I had been through a wringer when the meeting ended. What now? Well, when the TV program of that evening ended and the rest of the family went to bed I started writing a proper offer at our dining table. The raw material was of course the mixed-up offer from Hilversum, but many things from the discussions during the winter had to be added to comply with the customer’s wishes, and I had taken them with me home on paper or in my head. At certain points I also improvised statements, e.g. concerning the signalling on the exchange lines (lines to the public exchange). We had not earlier discussed this item as seriously as it was now required. I was halfway through at 2 am and went to bed.
Saturday evening I continued and finished the other half. I also wrote a letter to Max Hansen reporting the meeting and referring to the draft of the offer I had now made. I wrote that he should only send it to Copenhagen Telephone when I had discussed the contents with Hugo Simons and others who would be at the meeting in Holland the next fortnight.
Sunday I went to a shopping centre where I knew there was a copying machine. Such devices were also to be found at Philips, but at that time only in our printing office. I only had a key to enter the telecommunication offices, so I had to find other means. I copied all the pages I had written over the two last nights and visited of course also the baker’s shop to get something nice for our coffee!
Monday I went to Holland. On my way to the airport the taxi drove along Philips so I could put the draft of the offer and the letter on the desk of Max Hansen. And having arrived in Lage Vuurse, in the conference centre, Ernst Sillem Hoeve, owned by the Dutch YMCA, where we were to have our meeting, I immediately went to Simons to arrange a discussion of the offer.
We succeeded in having this discussion on Thursday. There were Simons, de Raaff, van der Schaaf, Veldkamp, the whole brain trust behind EBX 8000, and me. I read the offer for them, simultaneously translating it into Dutch (or English, I do not remember which). They had remarks which I noted in my copy. I remember especially one, where I said to them that this would mean a goodbye to the order. It had to be included anyway. It was that in the program package they could provide in an EBX 8000 in the middle of 1976, the exchange could only send dial pulses to the public exchange at a speed of 10 pulses per second. EBX 8000 could accommodate key telephone sets, but their signals would be converted to dial pulses towards the public exchange.
After the discussion I phoned Max Hansen and gave him the corrections. And then a week went by with more information about EBX 8000 (and a weekend in Amsterdam in between). Thursday I phoned again to Copenhagen and talked with Max Hansen. I said that tomorrow all the Dutchmen would spread out and be unavailable. It was the last chance to get an answer if there were still questions. Max said there had been one and how he had answered it (an answer I could confirm). And then he went on to say that Oksholm had been very satisfied with the offer, which now, as the PABX group sent their recommendation further up in the company was on top of the heap. The other offers were from L. M. Ericsson and Siemens.
It was like a bombshell at the meeting when I could report this. Most of the participants had only heard about EBX 8000 during these two weeks and here we told about a complete offer which was even the preferred one among the specialists in such a well established company as Copenhagen Telephony!
It became better still when we two weeks later were phoned by our contacts. Rosbæk phoned Max and Oksholm phoned me to tell that we would get the order. After a short discussion we phoned in our turn, he to Simons, I to Veldkamp, to bring them the good news. Then the work started seriously in Hilversum with a deadline: To deliver the first EBX 8000 to serve a whole office of about 2000 extensions such that it could start operation “in the middle of 1976” as we had written in the offer. It actually went into operation on July 1st, 1976.
When Rosbæk and Oksholm phoned us that the order was ours it was of course not only in Hilversum they were glad. For us in Philips Copenhagen it was a major breakthrough. It was our first switching order. For me personally it was also crucial. I had been employed for 15 years and finally I got an order in my main field!
Some celebration was called for, and as Max did not look like finding the Champagne and it was a really hot summer day I went to the nearest ice shop and bought ice cream for the whole group.
ISS 74, EWS becomes EWSD
In the autumn of 1974 the ISS, International Switching Symposium, was held in Munich. Especially Siemens was active in its organisation. I think the intention was to show their latest development, the EWS public exchange, for all the experts.
I participated with my wife and we had a nice time in and about Munich. She went on trips with the other wives (it was still called “ladies’ programme”, only later it was dubbed “spouses’ programme”), I listened at the meetings to the new ideas.
There were also trips for the participants to one of the new exchanges installed by Siemens. And during the ISS came the message showing how big Siemens is and how close their relations are to their main customer: Delivery of EWS exchanges to German Telephone would not begin. EMD-exchanges would in stead be delivered also the next 10 years while Siemens developed EWS into EWSD, a digital exchange system (meaning that it would switch PCM-coded speech in time division multiplex). It also said something about the development process and the long time planning within telecommunications. It reminds of Kennedy’s words in 1961: “A man on the moon in this decade”.
We were also on a trip to Salzburg in Austria with ITT, who also wanted to show some new equipment. It was on Saturday, after the ISS, and we enjoyed being tourists.
Jørgen Lindegaard’s first job
After getting the order for the EBX 8000 to the main office of Copenhagen Telephone I thought there would be too much to do for just one person. Max Hansen agreed to employ another engineer, but whom?
We took it up with the telecommunications professor at the Technical University, A. Kjerbye Nielsen: Who would be a suitable candidate? Many years later Tage Fox Maule, with whom I shared an office at Copenhagen Telephone Business Division, remembered that I had also discussed this with him during the ISS 74. He worked at the TU then. The result was that they recommended one of the new engineers, who would finish his studies ultimo 1974, to us. It was Jørgen Lindegaard, whose father was head of the Southern District within Jutland Telephone and who was himself interested in a career within telephony.
We contacted him and got his application for the job. An interview was arranged. On the day I looked through the papers and saw that it was his birthday. So I could greet him with a “Congratulations” when he came and we went to talk with Max Hansen.
The result was that Jørgen was employed from January 1st, 1975. It turned out later that his interest was not the technique, but management. So he stayed only for two years with us before he went on, first to be head of business telephones at Funen Telephone, then a series of CEO positions, at Funen Telephone, Copenhagen Telephone, GN Great Northern, Scandinavian Airlines System and International Sanitary Systems.
But in 1975 he was to spend most of the year in Hilversum to get to know EBX 8000. First there was a course from April to June, in which we both participated, and to which Copenhagen Telephone also sent two of their engineers, Jørgen Michaelsen and Niels Hansted Jørgensen. After the course, until November, the three others were stationed in the laboratory to study the system in depth.
The course was international with pupils from many countries. One of the pupils was from South Africa and he liked to boast about how big everything was down there. One day when we discussed the length of the extension lines under EBX 8000 he expressed concern. They were not long enough in his opinion. Our teacher, R. van der Veen, asked how long they needed to be, and then I could not hold myself back but drawled “well, you know the distance from Johannesburg to Pretoria”. He turned on me playing angry while the rest of the class laughed out loud.
He also came with the best definition of a pessimist I have ever heard: “A pessimist is an optimist with experience”.
As to the Danes, except for Niels we had our families with us and lived in rented houses with furniture in Hilversum. Lindegaard’s wife had a lot of books with her from her study and was busy with them while the men were at Philips. She had a problem: They had brought their son, Morten, with them and needed a kindergarten for him. That was difficult, as it was not usual in Holland that married women do other than household work.
I think the people in Hilversum felt somewhat pressed and would rather have had the course and our stay later. However, that would not suit the rest of the programme for teaching of technicians at Copenhagen Telephone and the installation preparations. I had the impression that van der Veen, our teacher, wrote the manuals from day to day. It was hard to keep the customer people from asking so many questions that it delayed the development. However, they had an office in the laboratory allocated to them so Hilversum had to live with that risk!
Autumn 1974, through-dialling
Copenhagen Telephone had required that their new PABX should have through-dialling, such that subscribers in Copenhagen could reach its extensions with 6 digits, just as they could reach subscribers on public exchanges in Copenhagen.
At that time the public exchanges in Copenhagen were a mix of semi-automatic exchanges reachable with 2 digits giving access to one of their operators, who then put the call through to the wanted subscriber, and automatic exchanges where you reached a subscriber by dialling 6 digits. The semi-automatic exchanges had names and there were letters on the dials, 1-C, 2-ABD, 3-EFG etc. You dialled e.g. 32 and got a recorder saying “Fasan” and then you told the operator the wanted subscriber number. By the way, it was a standing joke that if you wanted a proposal for the meal of the day you just dialled 32 and listened. “Fasan” means pheasant in Danish!
With 6 digits there could be approximately 1 million subscribers in Copenhagen. Later that was not enough and the area was first split in two groups. Still later 8 digits were introduced for calling an arbitrary subscriber in Denmark. With 0 and 1 as first digits used for other purposes there are then 80 million subscriber numbers in Denmark and for the time being it is sufficient also for mobile numbers.
The outgoing number analysis in a PABX must of course adapt to these changes. Thus, if it had sent 32 to town it should immediately switch through the speech connection. If an extension had chosen 30 as the first digits to town the PABX had to know that there would follow 4 more digits, as this was a call to an automatic exchange in Valby.
As said, Copenhagen Telephone wanted their extensions to be reached with 6 digits. In that respect the lines from public exchanges to EBX 8000 were not subscriber lines but equal to the trunk lines between the public exchanges.
In our offer we had promised this facility to be available in the middle of 1977. The development up to the first delivery in 1976 was frozen and additions to it would only delay the delivery. This delivery date was accepted and during the autumn of 1974 I often travelled back and forth between Copenhagen and Hilversum to get the exact specification of the whole protocol for through dialling from the company and translate it to a language which the developers could understand. There had to be no doubt about what the exchange should do in any situation. Luckily I had done something like it 12 years before (for a different protocol, now only line signalling was loop signals on the lines, while the register signalling was with MFC (Multi Frequency Code) signals) so I could finish the work in the spring 1975 in time to be a pupil at the first EBX 8000 course. During the next 2 years I often asked the developers if there were any doubtful points. There never were and in the summer of 1977 through-dialling went into operation without any fuss.
Register signals are only active while a connection is being build up, then they have played their role. Therefore we only needed a few receivers for the MFC signals, switched through to the incoming through-dialling lines when an incoming call begins. We used the selector network for this switching and connected the MFC receivers to the extension side of the network although they belonged to the PCTs. This was not new, from the beginning we had adapted to the requirement that outgoing dialling must not take place before a dial tone had been heard. We had a few dial tone receivers (they were only needed for about 0,5 seconds of every outgoing call) connected to the extension side already in the first generation of EBX 8000, switched through the selector network at the start of every outgoing call. These receivers were also controlled by a PCU.
The OCTs, operator circuits, were connected to both the extension and the external line side of EBX 8000 for their switching to resp. an external line and to an extension. In this way a special selector between OCTs and external line circuits was avoided.
I mentioned trunk lines between public exchanges in Copenhagen above. They are normally 2-wire and the signalling protocol was also based on the use of only 2 wires. But the first EBX 8000 was to be placed in the same building as the public exchange and it was therefore a requirement that it should be connected as a group selector in the local trunk exchange. Thus there had to be 2 more wires, one to mark the trunk busy and one to mark when the trunk should be taken in use. For the latter there even had to be a special common earth from the trunk exchange, different from the earth to the rest of EBX 8000. I realised that if this type of through-dialling should be used in other EBX 8000 exchanges in Copenhagen it had to work with only the two first wires, so I specified it such that this was possible.
This was lucky as the system for through-dialling was introduced later in three EBX 8000 exchanges for the state administration, in the new Foreign Ministry and two other places. We were also interested in using it in our own EBX 8000 when it came, but Copenhagen Telephone would not allow this. They had had a similar request from Siemens and had also refused that. The reason was the number capacity in Copenhagen of less than 1 million numbers. Soon many would want this type of through-dialling and there would not be numbers enough for it, so the first-comers would have a commercial advantage. This was not permissible and therefore they could only introduce the system in their own and in the state-owned PABXes.
Later when 8 digits had to be used for all calls within Denmark and the number capacity in Copenhagen was increased 10 times, this type of through-dialling was introduced generally. But not on 2-wire lines, only on PCM lines where the PABX was connected to the public exchange via 30 speech channels in a 2 Mbit per second connection. This meant that this type of through dialling was only suitable for large PABXes. Small PABXes had to do with so-called in-dialling, where one first choose 8 digits (as to a subscriber) and then got a speech connection in the public network and a dial tone from the distant PABX. One could then dial the extension number on one’s keypad.
5% in the country, 5% in industry
In several years I was on the board of the personnel club PAP and was its representative in the committee for the personnel newspaper “Philiskopet”. It consisted of people from all parts of the company, including the worker representative Svend Loft from the factory. Due to the cooperation between him and the CEO, S. A. Windelin, there was a good relation between workers and company since a major strike in 1958.
There was of course much recognition of his significance for the company when there was an occasion. One occasion was when the workers club had its 25th anniversary. From the paper committee they got a wooden statue with three Chinese wearing broad hats and symbolising cooperation. To deliver it the committee met with three people (I was one) who with corresponding hats and a long cloak sat behind a curtain when the reception began. During it (when we were already tired of sitting still) some well-chosen words from the committee were spoken, we were revealed and the gift handed over.
I will mention another happening, I think from the middle of the 1970ies, during a meeting of the committee. I do not remember how we came to talk about the future employment, but I expressed the opinion that as 5% of the Danes could make all that we could eat and even produce for export, it had to be possible that another 5% was all that was needed to produce all the other things we needed. Svend Loft did not like the idea.
At that time there were about 800 people in the offices and 800 in the production of Philips Copenhagen. The production stopped (apart from some minor items) in 1993 and much work has been outsourced, so around 2000 there were only some 300 people left.
Teaching at Copenhagen Telephony
It was not enough that two engineers from Copenhagen Telephone and two from Philips knew something about EBX 8000. The knowledge should also be spread to the installers and technicians who should set up the exchange and keep it running. There was still at that time a distinction between these groups. One should take care of everything from the equipment came in its transport packages until it worked, the other took over from the day the exchange was delivered to the customer. Later the borders were less prominent, from the day current is switched on the tests are very much like the procedures necessary during operation. Also during expansions of an exchange (which happens more often for a PABX than for a public exchange) both groups must have access.
This teaching was my task after having returned from the course in Hilversum. The first job was to make Danish documentation, another requirement in the contract. Rumour had it that this was a question about the workers’ agreements, the salaries to installers and technicians had to be higher if they were required to use documentation in English!
Thus, based on the documentation from the course I wrote during the summer months 1975 the Danish manuals. One thing was clear for me from the beginning: There should be only one set of manuals, not like in Hilversum one set used during courses and one set used during installation and operation. It took ten years more before Hilversum for the Sopho-S exchanges cut costs by having just one set of manuals.
In October the teaching started. I still felt that I knew almost nothing about the exchange and now I had to teach others about it! All the pupils were very keen on learning, much more than I could live up to, and there was ample time to treat all aspects of the exchange in depth. The point was that while the technicians in electromechanical exchanges should be able to adjust individual circuits, they should in EBX 8000 generally spoken just localise a fault to a circuit board and replace it. Repair should then take place centrally. A deep knowledge about the function of the exchange was not required, if only they could read the indications printed on the service terminal (a teleprinter) and find further directions in the manual it was sufficient.
But as said there was ample time to treat all aspects of the exchange and I felt badly each morning when I went to the class room.
In the middle of November the three other engineers came home from Hilversum with the newest and most comprehensive knowledge available about the exchange. I did something which I have regretted since then. It was not nice. They were present one day at the course and with the justification that now they knew much more I announced – without consulting him beforehand - that from the next day Jørgen Lindegaard would take over the teaching. Jørgen lived up to the challenge, nodded yes, he would, and took over from then on. I could again concentrate on the documentation, there was still a lot to do before it was ready with its four main parts: Description of the exchange, Installation and drawings for this, Administrative procedures for e.g. changes in the placing of extension numbers in the exchange and Operational procedures with fault codes, their significance and additional data and what to do about them. </p>
EBX 8000 arrives at Copenhagen Telephone
At last it was so far in April 1976: The EBX 8000 exchange was ready from the factory and was sent to Copenhagen Telephone.
It was packed carefully with each cabinet in a wooden box, material that soon found its way to the summer houses of the technicians. It was sent with a freight company situated next door to the factory in Hoorn on the West coast of the Zuiderzee. The 15 cabinets took up a lot of space, so they came in a large lorry with a four-wheeled and just as big trailer containing the exchange and whatever the freight company brought with it.
The lorry drove down the narrow street in central Copenhagen, through the gate and filled up most of the space in the yard. The unloading started and went reasonably fast although the access to the room was not the best one. Every box had to go through a narrow door, up a small staircase, turn 90 degrees and then through another door.
In the middle of it all Rosbæk, the CEO, was to pass in his company car, the Princess (an Austin Princess of some age). He was at first cross that his path was blocked, but hearing that it was their new exchange everything was OK.
After a concentrated effort of both the drivers from Philips Telecommunication, Svend Førsterling and Michael Petersen, movers called in for the job and the technicians of Copenhagen Telephone all goods were unloaded. I was in the exchange room for a moment and as I got out the lorry was gone! In no time the driver had with both lorry and trailer backed out through the gate, up the narrow street, out on a broader road and gone on. It still stands for me as one of the most impressive things around EBX 8000 that this large vehicle could disappear so fast! </p>
Description of EBX 8000
How was the construction of EBX 8000? It had like PRX the division between a central computer, CPU, for coordination and allocation of resources, and peripheral computers - or rather specially wired assemblies of electronic components – as special purpose processors to do the real-time jobs. Contrary to PRX there were three types of special processors, for resp. the extension lines, the switching network and the speech circuits. The latter was called PCU for Peripheral Control Unit. Before EBX 8000 went out of production this part of the exchange had been further developed to use microprocessors and was therefore called PPU, Peripheral Processor Unit.
The CPU spent its time, when it was not actually handling a call, searching for work. It polled the PCUs for new calls or new events during a call. When it had fetched data about an event it searched again. When all PCU events were fetched the first of them was treated. When this treatment was finished it sent possible commands to the PCUs, the marker (for the selectors) and the LSM (Line State Memory) or SSM (Scanner State Multiplexer) for the extension side. Then it again fetched PCU events. When all PCU events were fetched and treated it looked to the LSM or SSM for new calls from this side and treated them. Finally the CPU could look for new signals from the MMU, Man Machine Unit, which the technicians used for changing data, reading fault reports etc.
A message from a PCU held data about the peripheral circuit, PCT, which had originated the message, and what the message meant. In the CPU there was a memory table addressed with the PCT number to tell how far in the call this PCT had come. Knowing this and the message the CPU could continue from where it had come last time the PCT was treated and adapt to the new data. This might result in commands to this or other PCTs, e.g. “start busy tone”. Such a command was stored in a command buffer along with the PCT number and possibly an internal time out was started in the CPU for this PCT. Finally the CPU stored in the above mentioned memory table how far it had now come in the programme for this PCT. Then the CPU looked again for new events in the PCUs or – if there were none and no events were waiting for treatment – look to the LSM or SSM.
Strictly spoken the CPU did not search over the PCUs directly. This would take up too much time. The CPU only searched in a PMU, Peripheral Memory Unit, in the same cabinet as the CPU. The PMU scanned autonomously the PCUs until it found a message. It paused until the CPU had fetched the message and then continued its scanning. An exchange could have 4 PMUs, each scanning 32 PCUs which again each could serve 16 PCTs. Thus an EBX 8000 could maximally accommodate 2048 PCTs.
The job of a PCU was to scan the up to 16 PCTs it served for signals from the lines or from the extension side when it was connected to that side through the selector network. It treated the signals and readied a message to the CPU. The other way the PCU should pass on commands from the CPU to the PCTs. Not every change on a line (or from the extension side) caused a message to the CPU. For instance the PCU collected itself a number of interruptions and closures of the current to the extension side as dial pulses and sent only a complete digit as a message to the CPU (the current to an extension came from the PCT through the reed contacts in the selectors when an extension was busy). There were three types of PCU, one type exclusively for internal connections between two extensions, one type for external connections between an extension and an exchange line (to the public exchange), and one type for various PCTs like operator circuits, keytone receivers etc. The type for exchange lines had to be usable with a minimum of adaptations to the signalling in different countries with different signalling systems to the public exchanges. In a PCU there were some circuit boards common to all PCUs, like the board for connection to the PMU, and boards unique for each type of PCT. In the PCU for different PCTs there were more types of the latter boards and when the PCU scanned a PCT it also activated the correct one of these boards. When the PPU was introduced these boards were replaced by data in the memory of the microprocessor.
While the PMU was rather simple and the PCU was complex in order to handle a detailed real-time treatment of the PCTs, the relations on the extension side were the opposite. Here a simple LSS, Line State Scanner, placed in a shelf for extension circuits, scanned up to 64 extension circuits for new calls from the users. A complicated LSM, Line State Memory, in the CPU cabinet scanned the LSSes for data on new calls and sent them to the CPU.
The LSM scanned up to 128 LSSes for new calls, i.e. cases where an extension had gone off hook (the handset had been lifted) but there was not yet a connection to a PCT. When a new call was found the LSM stored data about it and stopped the scanning in that LSS until the data in the LSM were read by the CPU. Line blocking was also taken care of. This is the case when an extension has gone off hook and remained so for a long time, such that the handling of it is given up. It must not any more occupy a PCT. So the situation is the same as for a new call, but the LSS knew the difference and reported only new calls. Line blocking ended when the extension had again been on hook.
In small EBX 8000 exchanges (which were not introduced in Denmark) the CPU had better time and did also the LSM jobs. Between CPU and LSS there was then only an SSM, Scanner State Multiplexer.
The detection of new calls is not really a real-time job. There is a lot of time from the handset leaves the hook contact until it is at the ear of the user and dial tone shall be heard. So even with the low priority of messages from the LSM compared with messages from the PCUs there was always dial tone when the user expected it. The LSM and LSS had only one true real-time job: When there was a call to an extension the CPU wanted the latest information about its state and ordered via the LSM that the LSS checked this. The CPU could not accept any waiting time. Further there was a command carried out by the LSS, i.e. the disconnection of the DC power from the extension circuit itself to the line. When there was a connection to a PCT, current came from it and the local supply was cut off in order not to attenuate the speech too much.
There was a type of extension circuit which supplied its current through high-ohm resistances, such that they could remain in the circuit with minimum (and acceptable) attenuation. It was cheaper as it did not contain a relay per extension line, but it was not introduced in Denmark. The reason was that it was unclear how much current a terminal might drain from the exchange. Some terminals could only work if they drew current all the time (such as the MOS auxiliary devices from Pye-TMC). For safety’s sake Copenhagen Telephone would only use the extension circuit which supplied current in the idle condition through relatively small resistors.
Between PCTs and extension circuits the selector network was placed, controlled by the marker in the CPU. In opposition to the PMM and the LSM the marker could not send messages about the selectors to the CPU, it could only obey orders from the CPU. The selectors were divided in three stages, LN (Line Network), IMN (1st Mixing Network) and IIMN (2nd Mixing Network). There were 64 reed relays on a circuit board, either as one 8 by 8 selector, as two 4 by 8 selectors or as four 4 by 4 selectors. Like in the PRX exchange the power to an extension came through these relays when it was connected to a PCT. But the contacts could not stand the power spikes when this current was switched in or out, especially because their big number forbade the use of additional components to protect the contacts. Therefore the connection through the selectors was built up before power was switched on with a relay with mercury wetted contacts in the PCT. Likewise this relay interrupted the current before the relays in the selectors were released after the call.
This also ensured that – apart from the relay in the extension circuits – it was only in the PCTs that the two DC paths – to two extensions or to one extension and one external line – had to be separated from the speech connection between the two parties.
LN had two stages. The A-selector had 4 by 4 cross points, the B-selector 8 by 4. 8 A-selectors and 4 B-selectors formed an LN with 32 A-inlets (connections to extensions) and 16 B-outlets. Each A-selector had a link to each B-selector and each A-inlet had therefore one path to every outlet on the B-selectors. That is, if the link between the A- and B-selector was not occupied by another connection.
IMN and IIMN had two stages, each with eight 8 by 8 selectors, meaning that each MN had 64 inlets and 64 outlets. The stages were called C and D in IMN, E and F in IIMN. Each C- (or E-) selector had one link to each D- (or F-) selector in the same MN. Any inlet on a C- (or E-) selector could thus be connected to any outlet on a D- (or F-) selector. Again: If the link was not occupied by another connection.
This possibility of internal blocking in the selector network made it a requirement that there were always four different paths between an extension circuit on an A-inlet and a PCT on a D- or F-outlet. The links between B- and C-selectors and between D- and E-selectors were distributed such that this was realised.
128 extension circuits, 2 LSSs, 4 LNs and 1 IMN were placed in a double shelf for 128 extensions. 4 of these double shelves formed a 500-group. There were 16 outlets from each LN, but with the requirement of at least 4 connections to an arbitrary PCT, the 16 outlets were led to 4 inlets on each IMN in the 500 group.
If an exchange was larger than 500 extensions there had to be more than 4 units for 128 extensions each. Then the IIMN was installed to collect connections from more 500-groups to the PCTs. With 64 E-inlets and at least 4 paths from an arbitrary extension circuit a IIMN could be connected to a maximum of 16 500-groups or 8000 extensions, the maximum capacity of EBX 8000.
Also in exchanges with IIMNs PCTs were still connected to outlets on IMNs. When a path passed as few links as possible it reduced the risk for internal blocking. The PCTs for connection between extension circuits had mostly their A-side (towards the caller) connected to outlets on a IMN (and could only serve the relevant 500-group) while their B-side (towards the called extension) were connected to outlets on IIMN. One did not know beforehand to which extension a call was directed! A common overflow group for all 500-groups had both their A- and B-sides connected to outlets on IIMN and could serve all extensions.
PCTs for external connections were mostly connected to both IMN (for outgoing calls) and IIMN (for incoming calls). They had often two connections to IIMN to reduce the risk for internal blocking. This reflects the difference between outgoing and incoming calls. If there is an internal blocking in an outgoing call, i.e. no access to a given PCT, another can be chosen. In an incoming call the PCT can not be changed. PCTs of which only a few were installed in each exchange, like operator circuits and receivers for keytone signals, were only connected to IIMN.
Let us return to the CPU. An exchange might have up to 8192 extensions and 2048 PCTs. This did not mean that it was possible to give one fourth of the extensions a connection at any time. The dimensioning was rather that every extension would be busy in the busy hour for 14% of the time (have a traffic of 0,14 Erlang). If the average call lasted 2 minutes it meant that there would be 4 new calls per hour and per extension, or for 8000 extensions 32.000 new calls per hour. Thus the CPU was only allowed to use 0,1 second on each call (in fact less if waiting times should be avoided). This was a heavy requirement, which could not be fulfilled. We were lucky to never have this problem in Denmark, as the biggest EBX 8000 here had only about 2000 extensions and that gave ample time for the CPU. The evaluation of Copenhagen Telephone was also that the memory in the first type of CPU was not big enough to serve 8000 extensions with the heavy use of facilities they expected.
The CPU was thus a computer for handling of many simultaneous processes for many different peripheral circuits. This is the big difference between a telecommunications computer and the usual types of computer. It also has to happen with the reliability usual for telecommunications, 24 hours 7 days for 365 (or 366) days a year. Further EBX 8000 was also developed to be an economical proposition, as it had to compete with the still more developed electromechanical PABXes, which also introduced electronics for special purposes, such as number displays for the operators like in UB49 and UH 900. The CPU was for these reasons not a general-purpose computer as the CPU in PRX. The CPU in EBX 8000 could only handle telephone traffic. It could not assemble its own programs. This was done on a PRX CPU.
The programs were divided in three layers. The major memory in EBX 8000 was called DST (Data STore) and had in the 1st generation 64K (or 65.536) 16-bit words, each with a parity bit added. The two figures are linked, as you can with 16 bit address 64K places directly. This is in total 128KB (128.000 8-bit bytes) plus parity bits, far less than a PC would have just 20 years later in its cache for a fast handling of programs. But this was the early 1970ies and it was taking a chance to choose chip memories in stead of ferrite core memories as in PRX. The 1st generation was also limited by the availability of memory chips, only chips with a capacity of 1K bit were on the market, such that a 64K DST took up a whole shelf in the CPU cabinet. Only two years after the introduction of EBX 8000 it came in a new generation with 16K chips. Now there could be 1M (1.048.576) 20-bit words (20 bit to address 1M words directly) in the DST.
In the top of the DST there were 64 words (addressable with 6 bit) as a scratchpad for the actual program. Then followed a list of starting points for tables in EBX 8000, such as where extensions with call transfer could be found, up to address 4096, i.e. addressable with 12 bit. The next part of the DST contained a link table with telephony programs. When e.g. a PCT sent a message about an event a table addressed by the PCT number told where in the link table the search for the event should start. The CPU compared the message with the data in the link table and when they were equal the next word in the DST told what to do, it was the address of a master program in a PST, Program STore.
The link table thus was a mix of master program references and data. Another example is the address of the master program “send tone”. The next word in the DST told which tone.
After the link table followed the tables to which the list of starting points referred. Some tables had a word for each PCT, addressed by adding the number of the PCT, i.e. PMM, PCU and PCT number to the start address of the table. If the exchange had only 2 PMMs with 40 PCUs together, the table would only have a length of 40 by 16 words or 640 in stead of the maximum 2048 possible. This gave room for other tables, typically in Denmark for 1000 common code calling numbers, where there was otherwise only space for 100. Or it saved equipment as only the required number of circuit boards, each with 4K 16-bit words, were ordered. Remember that they were expensive! Other tables were addressed with the extension number, again only according to the installed amount of extension circuits.
As said the link table supplied the address of a master program in a PST. From this address there were a series of program steps, each consisting of 17 bit and a parity bit. The 17 bit were divided in 5 bit for the program step itself (thus, all programs were ultimately built up by using just 31 basic instructions, as 00000 was not used) and 12 bit for data. These 12 bit could be used as a whole to address the list of starting points for tables directly (12 bit and 4096 addresses fit together) or as 2 times 6 bit to address 2 places in the scratchpad (6 bit and 64 addresses fit also).
The program step was again an address in a micro program store, wired to do just this instruction. It could be the program ADS for “add and store”, where it was wired that it contained 2 times a 6 bit data word, and that the details were to fetch the contents at the first place in the scratchpad, add it to what a former program step had put into the accumulator of the CPU and store the result at the second place in the scratchpad. The next program steps in the PST might then fetch the result and send it to a table according to the list of starting points and a PCT number.
The advantage in this division in DST and PST was that many telephony programs need to go through the same steps. Several references in the DST could thus be to the same master program in the PST. It saved memory bits and that was important then!
Time for the CPU handling was also important and therefore one should as much as possible address the DST directly. This might, however, cost memory space. Take e.g. call forwarding in case of no answer. An extension is rung but does not answer. After a time-out of 15 seconds a program starts to detect if the call shall be rerouted. There are two possibilities: Either there is a table with a word for each extension. Or there is a table with only a number of words corresponding to the number of extensions, the customer wanted to have this facility, e.g. 5% of the extensions. In the first case search is fast, the table is addressed by the extension number and the word contains the alternative number or has a content meaning “no call forwarding”. In the other case one has to first search in a table of users of the facility (comparing the number of the non-answering extension with the content of each word) and – if the extension number is found – jump to another table with the alternative numbers. Searching takes time, especially if there are only a few with the facility and the table is often searched through to the end before a non-user is detected. This was later refined such that a table with one bit per extension told if the extension had the facility and the other table was only searched if this was the case. It saved searching time!
We used a thumb rule in EBX 8000: If up to 5% of the extensions should at the same time have a facility we used search tables, if more we rather saved searching time and put in the necessary memory space for all extension numbers.
We had in a data listing for each exchange noted the start addresses for every table. In addition the content of every table was defined in the exchange manuals. Thus we could on the spot make a data patch and load it into the exchange. This was used often by the technicians to change time limits in an exchange as wanted by the customer or to change the distribution of extension numbers over the extension circuits. See later.
Some words about system reliability. EBX 8000 had two CPUs working synchronously. That is they received the same messages from PCTs etc. and handled them in the same way. Orders to PCTs etc. were only sent from one of the CPUs. A small unit compared each step in the two CPUs. In case of a difference the exchange stopped a moment (so short that the customer felt nothing) and each CPU ran a program to see if it was OK. A sick CPU would be switched off and the exchange then worked on one CPU. It was of course the fastest if different content was read in the DSTs but the parity bit told which was wrong. When a stopped CPU should start work again the exchange stopped serving calls (but existing calls went on) while the full content from the PST and DST of the good CPU was transferred to the other. Then the synchronous work began again with the same starting point in both CPUs.
The whole exchange might of course fail and forget all data in both CPUs. In that case one could as during installation load one CPU from paper tape, first the PST, then the DST programs and finally the customer data into the DST. During operation one could prepare an emergency tape with customer data. Thus it was not required to load with the original data and then put later modifications in by hand. Such an emergency tape should be made regularly, once at Philips the tape was a year old when the exchange went down. With all our changes of user data it took a couple of days before the situation was up to date again!
All loading and output took place via the MMU, Man-Machine Unit. It was connected to a paper tape reader, a paper tape puncher, a teleprinter and to an MMU-panel in one of the doors of a CPU cabinet. In stead of the teleprinter a modem could be connected in order to have a remote control. The technicians made a three-way panel, such that the exchange could be connected to either modem or teleprinter, or the teleprinter could be connected to the modem. In this way they could from any EBX 8000 exchange connect to any other and read fault reports, change extension data etc. This took place via the public switched telephone network (good old PSTN!) and in principle any hacker could call an exchange and do these procedures from a teleprinter. I think it never happened, but if this had turned out to be a problem, Hilversum had made an additional program, such that a password was required to get access.
Would this be to come after the fact and what about sensitive data? Well, nothing in the exchanges could be changed from the teleprinter, which could not be rectified fast. One could e.g. close down all PCTs, such that all traffic was blocked. But they could as easily be opened again. And a PABX houses no sensitive data. If a hacker should happen to know the extension number of a certain person he could of course read where in the exchange that person was connected or to which number there was a call forwarding. But such data are neither sensitive nor of interest for other people than the user!
Finally some words on the physical construction of EBX 8000 as it was used in Denmark. It was made up of sets of two cabinets, each set having one shelf with power supplies from 48 Volts for these two cabinets. Thus, only 48 V was distributed over the whole exchange. In the first cabinet there were two shelves with power supplies, for resp. CPU0 and CPU1. Duplication of the CPUs had not had much sense if they worked from the same supply! Under these two shelves CPU0 filled up the remaining four shelves with its LSM, PMM and marker. The DST of this CPU took up the lowest shelf.
The second cabinet had in one shelf duplicated generators for ringing voltage and tones, the next shelf was empty and then CPU1 filled up the remaining four shelves.
The third and fourth cabinet housed four units for each 128 extensions forming a 500-group. The third cabinet had first a shelf with power supplies for the two cabinets, then a shelf with 16 PCTs for internal connections. The lowest four shelves contained two units for extensions, each taking up two shelves. The fourth cabinet had first two shelves with each 16 PCTs for external connections and then two units for extensions.
The fifth and sixth cabinet housed in their top two shelves the same equipment as the third and fourth cabinet, i.e. power supplies and room for 16 internal connections and 32 external connections. Under these there was in each cabinet two shelves with each one IIMN and a double shelf with room for 16 PCTs of different types. In such a double shelf there was room for two operator circuits.
The next six cabinets were to be placed back to back with the first six. Cabinets seven and eight, resp. nine and ten were like three and four, eleven and twelve were like five and six. Thus in these 12 cabinets with a “footprint” (area directly under the cabinets) of 6m² there could be equipment for 1500 extensions. An electromechanical exchange just a few years older and with the same capacity would have had a footprint of 16 m².
A still larger exchange was normally made according to a fixed pattern, such that all cables could be prefabricated, of sets of two cabinets either as cabinet three and four, or as five and six. But of course: If the available room required it, the placing of the cabinets could be adapted. This was the case in one of the state exchanges. But knowing from the parts list the length and composition of each cable, we could just order the correct prefabricated cables for the exchange also in this case.
EBX 8000 manuals
With every EBX 8000 we supplied a full set of documentation, consisting of descriptions, drawings and paper tapes with programs and data. The documentation was in English and was completed with the Danish manuals used in the daily work by the technicians. All this documentation took up as much space as two of the exchange cabinets!
The Danish manuals were written during the summer of 1975 and simultaneously with the course during the autumn 1975, as said above. They consisted of six binders taking up about 40 cm shelf space, and although the “Installation Manual” mostly consisted of drawings, copied from the Dutch material, the writing comprised several hundred pages of text and tables in A4 size, i.e. 30 by 21 cm.
We had no capacity at Philips Telecommunications to type all this. It was before the era of the PC, so as specialist I wrote a draft by hand. It went to a secretary who typed it. Then back for proof reading and if there were serious errors the whole page had to be rewritten.
On good days I could produce about 10 closely written pages of draft, which would result in about 10 typed pages. But with the capacity at Philips exhausted where then? I asked my wife, who had a past as secretary for a major industrialist, if she would type the manuals on a typewriter borrowed from Philips, and she agreed. There was no pay, with the tax rules of that time it would not pay off. The deduction per person was a deduction in the income and was moved to me when she had no income. If she had an income an equally great amount was moved away from me, so even if her income was not large enough to pay a tax, the result was that I paid tax on her income with the maximum percentage!
For several months I wrote on the draft during the day, took it at home and gave it to my wife, who gave me her work from the day for proof reading. This I read in the evening and gave it back with possible corrections. Sometimes it felt like a 24 hour per day job, but it was necessary to be relatively ready for when the course started.
We did get something, however, for our efforts. My wife was and is a fan of Roger Whittaker and when he was in Copenhagen during the autumn I suggested to Max Hansen that Philips paid for a couple of tickets to his concert. Max agreed and we enjoyed the music.
In the English manuals there were i.a. descriptions of all the master programs, from which the upper level programs in the link table were built, and listings of both master programs and the link table programs. It was part of the policy of Copenhagen Telephone: If the connection to the supplier was cut they should be able to proceed alone.
But this was only theory. If one does not absorb 100% how the people developing a program have thought, it is nearly impossible in a short time to improve or only repair on what they have made. And Copenhagen Telephone never went into such details.
They did not need to. Our agreement was clearly that if errors in a program resulted in bad service for the users (e.g. waiting time for dial tone), Philips should correct the error. And if there was a program error, which e.g. made the exchange do a restart once a year they could live with it if the connection to Philips was cut. It would have been a waste of resources if Copenhagen Telephone should study the programs at lower levels than where their effect on the users was described.
As to myself I had of course much use of the link table listings when I made the additional program for controlled interruption, see below.
But the documentation from Hilversum had a serious shortcoming at exactly a point where Copenhagen Telephone could intervene: The description of data. Based on the agreement with them and a list of information for each exchange, Hilversum prepared a data load tape for it, and it was of course followed by a listing of the content of this tape. But in the beginning there was no description of the build-up of these data, so it was difficult to read the listing.
But it was necessary to know these data, e.g. when the external number scheme should be altered (a semiautomatic exchange was closed down and the two-digit number should now give a “no number” tone, or an automatic exchange went into service and choice of its two digits should now give sending of these digits and four more before the speech connection was switched through). If the change should not each time be reported to Hilversum and result in a data patch from them, Copenhagen Telephone had to know themselves what to do so they could make the patch. Their technicians were capable of this and it was cheaper and faster than having everything made in Hilversum.
What did these data describe? As an example let us take the table for call forwarding. It was e.g. agreed that there should be this facility for 10% of the extensions. In a 1000 line exchange it meant for 100 extensions. In this case we would have a table with a length of 1000 words in the DST, each addressed by the start address of the table plus the reduced extension number, i.e. a continuous series of numbers from 0 to 999, fetched in another table with conversion from the directory numbers like 2000 to 2499, 3000 to 3299 and 4000 to 4199 to these reduced numbers. In this word one found the extension number the call should be forwarded to or data telling that this number did not have call forwarding. This was fast and gave a possibility for all to have call forwarding. But it cost memory space, 1000 words of the totally available 64K in the DST.
As already said another organisation would be to have a table with one bit per extension, addressed by the reduced extension number (for 1000 extensions 6 bit to address the word in a table of length 64 words and 4 bit to address the bit in that 16-bit word. If the bit was 0, it meant no call forwarding. If the bit was 1, a search took place in another table with a reduced extension number as content for each extension having a 1 in the first table. This table should have a length of 100 words. When the reduced extension number was found, a jump was made to an address 100 places further on. Here the number of the extension to which the call should be forwarded was found. The total requirement for memory space in the DST was in this case 264 words.
The description of data told how one could from the content of the single words and bits read the content of the data listing. Copenhagen Telephone should not have this description, Hilversum found. I disagreed completely. I succeeded in arguing for my opinion and the descriptions were supplied.
I had myself great use of these descriptions when the technicians needed assistance to serve the customers properly. They were those of the people of the administration who were in closest contact with the customers. They regularly visited each exchange, also when the control of them with EBX 8000 could be carried out remotely. The customer did not see a technician and could not understand why he had to pay so much for the service! Thus the technicians again visited the customers regularly and heard about his wishes. They then often came to me to hear about the possibilities. They were sure that if they had addressed their own sales department the result would have been an expensive offer, possibly for something quite different from what the customer wanted. But the goal of the technicians was to serve the customer in the best possible way within the amount he already paid, if possible. Of course there was quite a good deal of contempt for the sales people in this attitude, as “they did not understand a thing”.
I had to handle these requests myself. If I had gone on to Hilversum they would also ask a price (and cause a serious delay before an answer could be expected), while I saw it as a normal care for the customer. In this I had a good support in the descriptions of the data structures.
Take as an example the extension numbers. I have mentioned above the directory numbers divided in groups (three in the example) and their conversion to reduced extension numbers. There was no free numbering in the first generation of EBX 8000 (except in one exchange where it was loaded as an additional program). Free numbering means that any extension circuit could have any directory number within the range defined for the exchange (the range could well be 2000 to 7000 for a 500-line exchange). Hilversum had assumed that the directory numbers could be divided in five continuous groups, with conversion in each group to continuous reduced extension numbers equal to the extension circuit location in the exchange. Copenhagen Telephone had from the beginning required 20 groups of directory numbers to have more freedom to distribute the directory numbers within the range of e.g. 2000 to 7000. Some customers wanted to use this to have an illusion of free numbering, such that some groups could be very small, even down to a single extension. In such a case the technicians came to me, I investigated if it was possible within the limitations of the exchange and if yes wrote down what should be loaded into the tables of the exchange. The technicians converted this to a punched paper tape and loaded it into the exchange.
This had to be right the first time (we could not follow the rule “there is never time to do it right, there is always time to do it over”), as a fault would seriously hamper the function of the exchange. Thus, while I wrote the new data I checked them continually. And asked the technicians to do the same both as to my listing and when they had made a listing of the punched tape. I quoted Lenin’s words for them: “confidence is good, control is better”.
1976: Exchange name announcer adaptation
In addition to the standard execution of EBX 8000 there were special Danish requirements. They were not asked in the order for the exchange for Copenhagen Telephone itself, but that did not mean they could be neglected when they came up. One requirement was that it should be possible to use an exchange name announcer, i.e. a device that every time an operator answers an incoming call cuts in first to tell the name of the company. The customers of Copenhagen Telephone were used to this even in rather small PABXes, while it for Hilversum sounded as a very exotic wish. There was no possibility for their developing an interface to the announcer, they were busy with developments for the world market!
We had to start for ourselves in Copenhagen, meaning that I had to study both details in EBX 8000 and the announcer.
The announcer came from Kirk in Horsens and was also used in the semi-automatic public exchanges. It had a film sticking up from the edge of a plate. Just like in tone films (system Petersen and Poulsen) the exchange name was coded into the film as a wavy pattern and was read by turning the plate such that the film moved between a lamp and a photo cell. The name took up half a turn and a small magnet made a contact close between two repetitions of the name.
It was out of the question to get Hilversum to make a program to deliver a contact signal each time a given operator answered a call. We had to find a solution in hardware. We could have a signal from the answering keys A1 to A12 on each operator position. This would unequivocally tell who answered and what name should be said. Hilversum accepted this and cooperated by changing the circuit board with the keys so we had access to an extra connection for each key. With a strong limitation on the additional load we might put on the key contact.
Then I could develop a solution: Two flip-flops, controlled by the contact of the announcer and reset by the key of the operator. When the operator released the key, the flip-flops were in their 0-state. After the first signal from the announcer they were in their 1-state with a speech connection from announcer to external line. After the next announcer signal they went into their 2-state interrupting the connection. And then they interrupted further signals from the announcer when they entered their 3-state. Such a device on a small circuit board was required for every operator in an amount corresponding to the number of A-keys used in the exchange. It might be rather many.
The physical design and the production were taken care of by Erik Nielsen in the Telecommunications Group. He usually took care of the coordination of deliveries to the many addresses of the Telecom Administrations. But he was a genius with his hands and had quite a workshop in his cellar. Thus, when I had made diagrams and parts list for the content of the announcer interface, he took care of the rest. He choose a premade box, an extruded aluminium profile with slots inside in which the circuit cards could slide. Each card had two of the interface circuits and a box held six of these cards and a power supply. Each card could be used independently for two answering keys at the same or different operator desks.
The power supply was quite a headache. I intended to avoid transformers and only use capacitors to transfer the energy from 48 Volts to the electronics without galvanic coupling (no DC path). It did not work and in the end I gave up and designed a power supply according to the application notes from our component group (Miniwatt or Elcoma). And this one worked!
Erik Nielsen produced these interfaces for all the exchanges (and several for the biggest exchanges) and they worked perfectly. There has been just one fault in one of them and it was repaired quickly by replacing the card.
1976: Jørgen goes to Funen
In the spring of 1976 we had intense discussions with Copenhagen Telephone about delivery of EBX 8000 exchanges to their major PABX customers. But Oksholm seemingly did not like my way, I was maybe too much to the point (others might call it too stubborn). Anyway, he told Max Hansen that he and his whole group would rather discuss deliveries with Jørgen Lindegaard. And thus it was arranged. Max Hansen took care of Oksholm, Jørgen Lindegaard took care of Michaelsen (with whom he had been 6 months in Hilversum) and I formed the rear guard as to what EBX 8000 could do. I shall not say that I liked this degradation, but the customer rules. An expression of my feelings was that I did not come to the reception when Oksholm had been 25 years in Copenhagen Telephone.
In August Max Hansen had been 25 years with Philips and held a reception where representatives from all telecom administrations came. One was chief engineer Norman from Funen Telephone. He pulled me aside and told me that Jørgen had said to them that he was dissatisfied with Philips, he felt kept down by me and would rather like a job on Funen or in Jutland. On Funen their head of PABX sales, Stage, was very ill (he died shortly after) and this might be a job for Jørgen. It also fitted well with the end of studies for Jørgen’s wife. They had both lived in Kolding, Jutland, as children and had their families there. I could only answer that Jørgen definitely was not kept down, but I would take up with Max Hansen what Norman had told me.
Going home by bus (there had been a lunch as celebration after the reception) I stepped off it early to take a good long stroll home. I had to digest the words from Norman. I sat on a bench and noted it down as precisely as I could. Next time Max Hansen was in his office I gave him my resume of the talk. He read it and had some questions, as it was also quite new for him. And asked me to send Jørgen up to him (we had our offices one floor lower than the rest of Tele). This I did. Of course I do not know what was said, but I think Max Hansen like me found it rather disloyal to talk with others about problems inside the company before talking with us.
Well, we could not change the facts. Jørgen and his wife wanted to go to the provinces and there was a suitable job on Funen, so Jørgen gave notice and left with no hard feelings on either side. From January 1st 1977 he headed the PABX group at Funen Telephone.
Jørgen’s father phoned me that autumn. As said, he was head of operations of Jutland Telephone in Southern Jutland. He wanted to thank Philips and me in particular for everything we had meant for the development of Jørgen. We had been of much significance for it, he said. This was a compliment I could only receive thankfully and report to Max Hansen. Several years later I met him at a meeting in Århus (that was some time in the 1980ies) and he repeated the nice words.
Jørgen did not get a replacement at Philips. When I did not argue for this, part of the reason was probably that I would not like a new competitor working with PABXes. This might also be why Max Hansen did not find a replacement. But the main reason for both of us was that we now knew better what Copenhagen Telephone expected of us. They preferred to discuss both technique and prices directly with the supplier. They wanted also to have the full contact with their customers. They wanted to handle the service to their customers, both in the daily routine and if something special came up. When an exchange was delivered Philips Copenhagen only had the task to get the goods into the country, arrange for customs duty and bring the goods to the site of the installation, and for this we already had the necessary people (Børge Klemvig and the drivers Svend Førsterling and Michael Bernstrøm Petersen in the Telecommunications Group and other people in the new central store of Philips in Glostrup). As to the technique there was a need for local backing of the technicians of the administration (although everybody realised that I did not have a chance to acquire a routine like that of the technicians) and a backing of Copenhagen Telephone in discussions with Hilversum.
This last point was probably characteristic for my way of handling things. It was always the facts that spoke. Thus, Hilversum often felt like Copenhagen Telephone had an extra spokesman with them. But if the arguments of Hilversum were good, I supported them. Some times one or the other side tried to go a little too far in the arguments and then it was possible for me to cut through and get it accepted by both parties. This conduct was also the background for my warm reception in the administration in 1991, when Philips had no more use for me.
From 1976 almost ten years went by before the group for sales of PABXes was enlarged. First with Peter Miedema, a technician from Data Systems who wanted to work with telephone exchanges and took over some of my contact with the technicians of the administration. Later with Jens H. Bojsen, an engineer who had started in that part of our group which sold defence systems and later had worked much in Data Systems with data transmission. He was (and is probably yet) a sorcerer with a PC. It was also then that I moved from the Danish sales organisation to the new Philips Tele Nordic. Now the other Scandinavian countries needed some technical help! See later.
1977: Keytone to the public exchange
In 1974 I had expected we would lose the order for the PABX of Copenhagen Telephone themselves because we did not from the beginning send “dial signals” as keytone to the public exchange. But luckily this was not the case. I heard a rumour that somebody at the administration had indeed observed that we could only send dial pulses and said so to Oksholm. His answer was that this had to be a printing error in the offer and that the evaluation of it should just disregard that point. True or not, at some time in 1975 he put on a show during a meeting in Hilversum, seemingly surprised over this statement in the offer. He could not be accommodated, the development had been frozen a long time ago and we kept to the conditions in the accepted offer. But we could supply this facility with the next generation of the program in the middle of 1977. He had to accept this.
And in the middle of 1977 we delivered a new program package including this facility. The dial tone receivers were replaced by combined dial tone receivers / keytone senders on the extension side and they were connected via the selector network to the outgoing lines at the start of every call. Just as regards the circuits for dial tone or MFC signals there were only a few keytone senders in the exchange, as these signals were only active for a few seconds at the start of a call.
We got a problem, however: It happened too often that extension users did not get the outgoing connection they wanted. The reason turned out to be in the interplay between EBX 8000 and the public AKD subscriber exchange. This exchange would, when it received a loop signal from a subscriber make a line blocking while it found a register to receive the dial or keytone signals from the caller. During line blocking it sent busy tone as 250 ms tone, 250 ms pause. Our dial tone receivers would detect a dial tone after a maximum of 300 ms. That is, it would detect the busy tone as a dial tone due to the necessary tolerances and start sending keytones. Thus the first keytone signal would already have been sent before the register was ready! The solution was rather simple, we changed a capacitor in every dial tone receiver (remember we filtered signals in the PCTs) such that a tone had to last at least 300 ms before it was accepted as a dial tone.
1977: Controlled interruption
Another Danish specialty was the wish to signal recall to the exchange as a short interruption of the current to the extension, almost like a dial pulse. Recall means that the extension user wishes to consult with somebody else during an external call. During this the external line is held and the user can either return to the external line or transfer the line to the person he consulted. In Holland they only used recall by connecting earth to the extension line when the R-key on the telephone set was pressed. It gave an unbalance in the current in the two wires and this was detected by the exchange.
There were two reasons for the wish. One was that an interruption is a balanced signal, meaning minimum risk for a disturbance of other calls. The other was purely economic. Several large PABXes which EBX 8000 might replace had only two wires (a twisted pair) to each extension. There had never been a need for an earth wire. Once, in their own PABX, they could rewire the whole internal network. This was not a proposition for other customers, thus if we wanted to see an EBX 8000 exchange in these places, we had to introduce controlled interruption as a recall signal.
Seen from Hilversum it was just a hobby of the administration. When customers could do with an earth key in all other countries, then why not in Denmark? They would not spend resources on this function.
So we had to get to work again. Luckily we were still in the period where all was done to save memory space in the exchange, both for data and for programs. This meant that the whole program was written in assembler code, i.e. in a way so each line in the program listing corresponded to the code in a memory word. And the listing was in a mnemonic code, i.e. it was easy to read. Further, EBX 8000 had the possibility of additional programs which could be linked to the main program by changing some words in the main program so they referred to the additional program and of course the latter program could refer back to the main program.
An additional program was therefore needed, which would, whenever there was a signal from an extension, let a short interruption have the same effect as an earth key signal. So I went through the link table listing and each time there was an earth key signal I made a reference to the additional program which would check for an interruption. The program was made in such a way that e.g. during a conversation it waited for a signal from the PCT and the program knew how far in the link table this call was. When a signal came there was from this place in the link table a small list of the waited for signals. If one of these signals came there was a reference to where in the master programs in the PST the further handling was found. If the signal was not found the program just returned to the former place and waited for the next signal. Typically a dial pulse (or a controlled interruption) was not expected during a conversation, but now it should be treated just like an earth key signal in the additional program.
The program itself was very short, but there were a lot of linking references.
I received the necessary instruction from Hilversum so I could write the program (in assembler code) and did it. It should then be coded into a punched paper tape and we did not have a teleprinter for this at Philips. So I went to the exchange of the administration where they had a Danish made teleprinter (from GN Great Northern). The technicians helped me, but there was a problem they could not do much about: There was no colour any more in the ribbon in the teleprinter so I was unable to see what I made! This called for a high degree of carefulness. When the tape was ready we went to the main computer room of the administration, where they had a line printer. The tape was loaded and a listing of it made. It seemed to be in order.
Luckily enough as I was due to go to Hilversum the next day. Here I immediately gave the tape to van Gelder who was in charge of the programming. He also acted fast by loading the tape into a computer to change it into machine language, i.e. a tape loadable as an additional program in an EBX 8000. It changed e.g. all my symbolic names for the places to which linking should take place into real addresses in the memory. I got the new tape with me home in the evening, there had been no errors to confuse the recoding program in my tape. And the next day this tape was loaded into the EBX 8000 at the administration. We tested the function and it worked correctly.
This function was sold in other countries later (although it had started as a “Danish hobby”), and this was again lucky. We got a complaint: The operators suddenly got many returned calls but there was nobody on the line when they answered. The reason was that when you laid on hook after a call, the hook contact could bounce (i.e. interrupt, close and interrupt for good). The exchange perceived this as a recall signal followed by release and should react on this by calling the operator, who on answer was connected to the external line, which had also released then. For Oman, where the same problem had been reported (that is, they must first have introduced my program), an extra additional program had been made to prevent a call to the operator in this situation. We could also use this program and the problem was history.
1977: Which exchange is most difficult to develop?
Oksholm had shown much courage when he pressed the order of EBX 8000 through. An exchange coming from a completely new supplier of such items when the good old supplier, L. M. Ericsson, had only with much effort and after several years of delay finally got the major trunk exchange in Copenhagen up and running in the spring of 1974.
I think it was in 1977 I discussed this with de Raaff, who was in charge of development of PABXes. I meant it was also a far larger project, Ericsson had embarked upon, when they had developed a trunk exchange, where Philips with PRX and EBX had developed end user exchanges.
De Raaff rejected this. Quite to the contrary the development of end user exchanges is far more complicated, he said, although they have fewer line types to which they shall adapt. The problem is the far greater number of lines with a small traffic each, which must nevertheless be supervised all the time, and the richness of facilities especially expected in a PABX.
As to the number of lines one can only refer to a normal subscriber exchange (end office) which typically has 10.000 lines to subscribers with each line only in use for about 8% of the busy hour (10 am to 11 am on weekdays). But if a subscriber goes off hook, he expects to hear dial tone when the handset is at his ear, i.e. after just 0,3 seconds!
EBX 8000 for Philips
The telephone administrations had a monopoly on delivery of PABXes. However, three types of customers were allowed to set up their own PABXes, such that the interface between customer and administration was where the lines to the public exchange entered the premises of the customer. It was embassies, technical state organisations like the State Railways and representatives of PABX suppliers, who could have a commercial interest in installing their own PABX. It was because of this that Philips had the UB49 and UH 900 PABXes installed.
There was a cost in having ones own PABX. You had to keep it in service yourself, i.e. correct faults and make changes. It was necessary to employ a technician even if there was a long time between faults. The administration could naturally not take over the operation of equipment not on their program.
It would be different if they acquired an EBX 8000 for their own offices and offered it to customers. We could then lean on their service organisation. Already in the spring of 1974, before our offer for their exchange was forwarded, I found that it might stimulate the interest for the equipment if Philips made it clear, that we considered the system ourselves and might then buy service from Copenhagen Telephone. I wrote a draft of a letter with this content and it was approved and forwarded, signed by our head of the property administration, Arne Echart.
In 1976 while the exchange of the administration was being installed it was time to go on. We had a meeting with the administration to discuss the conditions if we acquired an EBX 8000 and bought service from them. Their participants in the meeting were Oksholm and the head of customer relations, Ludvig Edinger. It was soon clear that Edinger did not like our “if” at all. We had with the letter from 1974 committed ourselves on both points! It seemed this was the interpretation Oksholm had given of the letter internally, to help in deciding for the EBX. Echart found his copy of the letter so all could see that there was nothing definite in it and the meeting could proceed on that base. But Edinger and Oksholm did not talk with each other during the rest of the meeting.
Our aim was to have the administration take care of the exchange so there were no problems when our technician, Max Andreassen, was on vacation or was ill. In these cases the administration had several people available. But we wanted to keep the freedom we were used to on the extension side, where we had continuous moves and other changes. Due to this it could pay off to be responsible for the operation ourselves. We would divide the job between Max Andreassen and our head operator, Grete Biilmann, such that Max made the changes in the exchange which had to do with the extension network, like moves and opening of new numbers, while Grete from an operator’s desk made changes without consequences for the network, like a change in call forwarding. Generally Copenhagen Telephone had chosen that customers should order all changes with them and they carried them out in the exchanges. We choose to exploit the possibilities in the exchange as much as possible to decentralise its operation. This proved that the procedures were usable for people who were not born with a soldering iron in their hand. We had naturally to agree exactly with the administration what we were allowed to do and how it should be documented. The intention was not to make trouble for their technicians by changing data according to our whims and expect them to clear up any problems just like that. In principle the border between them and us moved to the extension side and we paid 1 USD per quarter year per extension, their lowest rate for connection of customer owned equipment. We also paid for the trouble shooting service, of course.
They also put it as a very natural condition that they should themselves install the exchange when they should provide service afterwards, but they accepted that it was delivered directly from Hilversum to Philips Copenhagen. It was not necessary to invoice it from Hilversum to Philips Copenhagen, then to the administration and finally to Philips Copenhagen again, with an addition to the price at every step, but the installation was of course to be paid separately. All agreements were made, the picture was a reasonable one for our property administration, and orders went out to Hilversum for the exchange and to the administration for installation and service.
The operator desks “as a Turkish harem”
With the decision to replace the old exchanges, UB49 from 1954 and UH 900 from 1972, with an EBX 8000 it was important to find the right place for exchange and operator desks. The exchange could be placed in the room where UH 900 was situated. Not because there was a lot of space besides this exchange for 300 extensions, but EBX 8000 was so compact that the UH 900 just needed to have its two rows of cabinets pushed a little closer with the cables between them forming a bow. In this corner of the building there was also space for the rectifier and the battery, and then it was a small problem that the extension cables became a few meters longer. The personnel office (today the HR department) could expand in the rooms formerly used for UB49 and the rectifiers and batteries for both the old exchanges.
There was a suitable room for the operators on the fourth floor just over the exchange room, so the cables would not become too long. The whole property administration was also on this floor. A short time before this I had got the book by Peter Townsend “Up the Organisation”. In a few years he had made Avis the no. 2 car rental company in the US, so there were many good ideas. One of them was that the offices of the management should be as barren as the cells of a Trappist convent, if the bosses are worth anything they do not notice their environment. But in contrast to this the working space for the telephone operators should be as lavish as a Turkish harem. Then they will meet happy at work and sound friendly and receptive when people call the company!
I made a copy of this page in the book and circulated it within Philips. I do not think I was popular with the head of the property administration, Arne Echart, but the operator room was laid out beautifully. A large circular table with 6 desks sunk into it (the physiotherapist was again active so the operators should not bend their wrists too much to press the keys while resting their arms on the table), a corner with a pair of comfortable chairs etc. Many of the new customers for EBX 8000 exchanges visited us and talked with our operators. In that way the set up was a good choice. It was directly copied for an EBX 8000 exchange at the town hall of Odense, Funen. Naturally, none of the operators would have complained openly if the set up had been worse with too little space. But our guests could hear from them that they were satisfied with the exchange, their work and all the new facilities it provided.
EBX 800 to Funen
Jørgen Lindegaard moved to Funen Telephone from January 1st, 1977, as already said and became head of their PABX group. EBX 8000 would be a too big exchange for the companies at Funen, as it was not economic for less than about 500 extensions. Thus, we did not expect his knowledge of the system to result in sales to their customers. One EBX 8000 was, however, sold to the town hall of Odense – their inspiration from our operator room was mentioned above. When I went to Funen to teach the technicians about the system I pictured it on top of the weapon of Odense with the French Lilac, its three leaves symbolising the extension side, the selector network and the traffic circuits!
Hilversum was developing a PABX, EBX 800, for the range 100 to 700 extensions. Jørgen had naturally heard about it during his stay in Hilversum but had not had an opportunity to study it closer. His stay was also back in 1975 when all efforts were concentrated on EBX 8000. He had, however, had ample opportunity to meet the people developing EBX 800 while he worked in the laboratory.
EBX 800 fitted in size to the major customers on Funen and it was a nice day when Jørgen phoned us and asked for details about the system and an offer for a specific customer. It made a great impression on us and on the people in Hilversum that Funen Telephone on the basis of the confidence, Jørgen had to the developers, as the only administration in Denmark would introduce this system (Copenhagen Telephone had settled for Minimat 800 and Jutland Telephone was not looking for a new type of PABX of this size at that time).
Thus I had to go in depth with a new type of PABX but could luckily rely on much of the information I already had about processor controlled exchanges. The technique will be described in the following. Here it suffices to say that its physical set up was very different from EBX 8000, as it would otherwise be uneconomical for the lower number of extensions. It meant a new method of calculation when dimensioning an exchange and that was an important point when we should in a short time give Funen Telephone a binding offer. If we had calculated wrongly and were corrected by Hilversum, Philips Copenhagen had to face the loss. And if we for this reason had calculated on the safe side so as to never have a loss, the prices might make the customers run away screaming!
We succeeded in presenting EBX 800 for Funen Telephone and their customers in a way so it was introduced. From 1979 to 1982 9 exchanges of this type were installed on Funen. The first one went into operation at the Hotel Nyborg Strand on January 2nd, 1979, and moved early in the 1990ies to a few years operation at another customer when Nyborg Strand got a new PABX. Another was installed at the Funen Telephone headquarters in the centre of Odense.
Taking both types of EBX together it was Philips PABXes which served the offices of two of the four telephone administrations in Denmark. The two largest banks in Denmark also had EBX 8000 exchanges in their main offices, and one of them had in addition one in their data centre. The State Administration had four EBX 8000 exchanges in operation of which three had through-dialling using the same system as in the EBX 8000 for Copenhagen Telephone. It also deserves to be said that there were EBX exchanges at four hospitals, 3 on Sealand and 1 on Funen. 15 years after the last delivery of an EBX, in 1997, there were only two still in operation, both EBX 8000 exchanges serving hospitals.
Description of EBX 800
There were three types of modules in EBX 800, the CM (Control Module), the SM (Switch Module) and the PM (Peripheral Module).
The CM took up one shelf, where each CCU (Central Control Unit) in EBX 8000 took up four. It was not only the increase in capacity of the memory chips, which was the reason, the changed organisation, without autonomous devices in the CM for collection of messages from SM and PM also played a part. This collection was done by the main program. The most significant was, however, that the whole program was now laid down in a Read Only Memory (ROM), i.e. it was burned into memory chips such that their content could be read, but not changed. The variable memory with data about installed circuits, used extension numbers, use of facilities etc. was a pure memory for data and during start up of an EBX 800 it was not necessary to load the program. It was present already.
The program was thus neither laid down in the wiring as in electromechanical exchanges (in hardware) nor in a changeable memory as in EBX 8000 (in software). This led to the name “firmware” for this type of storing programs.
A result of the reduction of the physical size of the CM to just one shelf was an increase in reliability to such an extent that doubling of the CM was unnecessary. The reliability of components as regards accidental faults is often expressed in FIT or “faults per thousand million hours” or per 114.000 years. A typical value for a component is 1 FIT, but it is of course impossible to say other than that this value applies in the start of its life. If it is true in 10.000 years even the best accelerated test can not tell. The advantage in using FIT is that the total reliability in FIT of an apparatus can be expressed as the sum of the FIT values of all its components (including the values for soldering joints and plugs). A large computer can easily contain one million components and with a value of 1 FIT for each component this corresponds to a random computer fault per 1000 hours or 8 faults per year. This is also called “the tyranny of numbers”, as it is unacceptable for most systems. In EBX 8000 the total FIT for each CCU was too big for a telephone exchange. Therefore the CCU was doubled and there was a unit (with an extra large reliability) to decide if they both worked OK, and if this was not the case, decide which was faulty, isolate it and start an alarm. In EBX 800 the total FIT for the CM corresponded to less than 1 fault per 10 years. That is why duplication was not necessary and programs to decide which side worked were also redundant.
This was expressed in a somewhat shocking way by Kaasschieter from the development group, who at a course said that all system assurance (also dubbed SAS) in EBX 800 could be left out of the system and this had simplified it enormously. This was of course an exaggeration, there were still lots of SAS programs to localise faults in the CM or the other modules.
The different parts of the CM were connected by a bus for transport of program- and data-words. The bus was only 4 bit wide and could move one hexadecimal digit at a time. With PCs around the year 2000 moving 32 or 64 bit in parallel this may seem very small, but it was fast enough for its purpose.
SM contained extension circuits and selectors for their connection to the peripheral circuits in the PMs. The peripheral circuits connected between extensions or between an extension and an external line. When idle the extension circuit connected 48 Volts to the extension line through reed relay contacts. Reed relays have in principle only make-contacts (they make contact when energised) and in EBX 8000 reed relays with a permanent magnet in stead of a contact had been used for this purpose. The magnet made the contacts be on when the relay was not energised and current in the relay coil would counteract the permanent field and make the contacts go off. In EBX 800 the relay was kept on during the idle state by a current, and as this state lasts most of the time (even in the busy hour) there were talks about superfluous energy spending. But not much current was used, the relay was more compact and had better tolerances for its function.
The selectors were built in three stages, the A-, B- and C-selectors. An SM had circuits for 96 extensions and the A- and B-selectors concentrated this number to a smaller number of inlets to the C-selector, which was in fact one big crossbar selector, with on one side the links to the B-selectors, on the other all the peripheral circuits. Thus, the C-selector was the same in all SMs and was expanded whenever another SM or PM was added to an exchange. If e.g. 1 C-selector can connect 1 SM and 1 PM, an exchange with 6 SM and 5 PM needed 5 C-selectors in each SM or 30 C-selectors. The number grew as the square of the size of the exchange and this gave one of the limitations on its size or at which stage EBX 8000 became more economic. Another limitation was of course how many extensions etc. the CM could serve.
PM contained circuits for calls between extensions and for calls between extensions and external lines. It also had circuits for operators and a P-selector for connecting operator circuits and external line circuits, as well as circuits for keytone dialling from the extensions. An operator circuit and a circuit for internal calls should have two links to the C-selectors, an external line circuit or a tone receiver only one. This put narrow limits on the placing of circuits in a PM. Likewise an operator should be able to connect to all external lines, including those with their circuits in another PM than the operator circuit. This required more P-selectors and as the total space for them was limited, the number of PMs in an exchange was also limited. But when all limits are the same during expansion to about 800 extensions (in a version for 0,15 Erlang per extension) everything is just fine!
An SM or PM took up two shelves. The first cabinet had the CM in the lowest shelf, a PM in the next two and an SM in the top two. The next cabinets mostly held an SM and a PM. Each module had its own power supply from the 48 Volts battery.
The modules were connected by prefabricated cables with a plug in each end. Funen Telephone had for all exchanges chosen to order the cables necessary for the next step of expansion. Thus, an expansion during the first years of its operation could take place by the installation of a new cabinet with a new SM and/or PM, plugging in of cables and opening of the new circuits in the data store.
EBX 800 manuals
Funen Telephone had of course required full documentation in English about EBX 800 and that all documentation necessary for their technicians should be translated into Danish. So I started on this task.
Copenhagen Telephone had received the full program listings for EBX 8000 but were not able to use them meaningfully. Funen Telephone never received the program listings for EBX 800, but they did receive a description of the firmware which told in details how connections were built up, how the facilities worked etc. Thus they could test the functions of the exchange according to this description and complain if there were deviations. A simple comparison: When you buy a car you get a description of how you start the motor, but there are no details about what happens under the hood.
Funen Telephone received however the full data listing for each of their exchanges. We had the same discussion with Hilversum as we had earlier about EBX 8000 and it ended in the same way: Funen Telephone got a description of the data structures so they could read these listings. They could also change project data in the exchanges themselves. It was both cheaper and faster than if each change should be prepared in Hilversum.
At Funen Telephone Erling Sundsted made these changes. We had met earlier in talks about public exchanges. Each time I received his description of data in the load tape (EBX 800 was also loaded from paper tape) in two copies, of which one was sent to Hilversum.
One of the changes made by Funen Telephone themselves was the change of the external number scheme. That is the changes necessitated by first introducing that all calls should start with the area number. In stead of calling with 6 digits locally and with 8 digits to other areas (by dialling 01, 02 etc. before the 6-digit local number) every call should now start with 0. Later on it was changed to dialling 31, 42, 53 etc. in stead of 01, 02, 03 etc. and this was also implemented by Funen Telephone themselves.
Funen Telephone introduced also a fast service for their customers. A user of an EBX 800 exchange could call the service centre and while he talked with the attendant the latter connected to the exchange of the customer via a modem call and made the changes asked for by the customer. Thus, during the same call the customer could test the changes and everybody was happy. It was of course only changes for which space existed in the facility tables which could be modified in this way, e.g. the making or modifying of a call forwarding if no answer (i.e. transfer to another extension if there is no answer within 3 ringing signals). If a customer wanted more call forwarding possibilities than prepared for in the data tables Sundsted had to step in and change the place and the length of the table to an idle memory space.
In the late 1970ies Jørgen Lindegaard became first CTO and then CEO for Funen Telephone. He was replaced by Bjarne Sørensen as head of the PABX group in Funen Telephone.
Visit from England and Belgium
More people than the customers of Copenhagen Telephone were interested in the already installed EBX 8000 exchanges. Also customers in other countries where EBX 8000 was offered wanted to see for themselves the working exchanges. Denmark (i.e. Copenhagen Telephone) was a pioneer in this. True enough an EBX 8000 was in operation in Leiden, Holland, in the laboratories of the Dutch Telephone Administration since February 1976, but it was rather small. The very first time 2000 extension users were let loose on an EBX 8000 was 1st July 1976 at Copenhagen Telephone.
Thus it stimulated the sales of EBX 8000 in other countries that customers could visit Copenhagen, see for themselves a working exchange and listen to the impressions of Copenhagen Telephone. They cooperated fully and with their status in the telephony world it was only fine that they were enthusiastic about the system!
Visits came especially from Great Britain. The first steps in the liberalisation were underway, after approval of a PABX type it could be sold directly to a customer, but service was still a British Telecom monopoly. The usual procedure was that the British Philips organisation asked us for a visit with a customer. We arranged the details with Copenhagen Telephone and the visit took place. Usually the guests came the evening before, were installed in a hotel and I took them to town for a good dinner. The next day we went to Copenhagen Telephone, where Michaelsen showed the exchange and told about their experiences with it. Then followed a lunch with the customer and Michaelsen, where discussions continued (this often took place in Tivoli) and then we parted. Our British colleagues were very glad for the treatment given them by Copenhagen Telephone, it helped getting several orders for EBX 8000 exchanges in Great Britain!
It might be difficult to get everything to fit in all cases. Once we had agreed to a visit while there was a fashion week in Copenhagen. All hotels were occupied, we could only get rooms in a small hotel near the airport, so the people from the customer, British-Iranian Oil, were kept awake by the noise the whole night. But mostly the visits were run almost to a fixed pattern.
Also the Belgian Philips organisation visited us with a delegation from the Commission of the EEC. They were very satisfied with the treatment, but choose in this case another type of PABX, from Siemens.
1978, Visit to Hilversum from the Technical University.
Philips Hilversum had succeeded rather well in selling PRX public exchanges in other countries than Holland, and this was why Kjerbye-Nielsen from the Technical University of Denmark wanted to visit Hilversum and other places in Holland with a group of students, just like I had been myself in München etc. with Rübner in 1956.
We got it arranged and helped finding reasonable lodging for the students in Amsterdam. Kjerbye and his wife stayed like me in Hilversum.
They visited naturally not only Philips. We paid a visit to the Dutch Telephone Administration in Utrecht, where we visited the control centre supervising all public exchanges in the district. They showed how supervision of the PRX exchanges was much more intensive than the supervision of the electro-mechanical exchanges, due to the much greater possibilities.
We were in a period of very hot summer days, but the people told us that although there was air-conditioning in the exchange rooms, it was not turned on. There was still an energy crisis and it was not required for the sake of the exchanges. PRX used silicon transistors and integrated circuits and could stand higher ambient temperatures than ETS3. But one of the people of the administration told of a function he was very glad of. He was called to a fault in the exchange in Dronten, in the middle of the newly won land in the Ijsselmeer. An hour and a half of driving in full sunshine. Before he left Utrecht he could by remote control start the air-conditioning in Dronten, so it was comfortable in the exchange room when he reached it.
Sale of UH 900 to Italy.
The EBX 8000 exchange of Philips Copenhagen should replace both the old exchanges, UB49 from 1954 was 24 years old and was scrapped. UH 900 from 1972 was only five years old and was too good for that. It was also the exchange that had started the discussions with Copenhagen Telephone about delivery of PABXes. So if only for sentimental reasons it would be a pity to scrap it. I got an OK from Arne Echart, head of the property administration, to try selling it. The buyer could get it for free as it stood, if only it would not cost us more than scrapping it.
I told this to Hilversum, who went along fine, although it possibly meant the sales of one less new exchange. They informed the other national organisations and a message came from Italy, from Mino Caviglia whom I had often met in Hilversum. They would like to get the exchange, we should just put it on a lorry and send it to Philips Milan. Mino would pay DKK 36.000 for it (roughly USD 6000).
I would not accept this. If the exchange was not packed up in a proper way all the heavy items in it would shake loose and fall to the bottom of the racks during the transportation, so the exchange would never function again. In stead I got a green light to get an offer from Hilversum covering a professional packing. One man and all the necessary material came at DKK 16.000 and this was accepted.
Thus the exchange was taken good care of. Heavy units were packed individually, other units were tied up in the racks and it was all put in cardboard boxes, ready for shipment. Both Mino and one of his colleagues were in Copenhagen to take part in the packing, but to the delight of the Dutchman, who liked to do the job himself, they were more interested in visiting Tivoli!
But they followed up sufficiently to be able to rebuild the exchange when it came to Italy. I heard later that it came to a convent school in the Vatican State and that the customer was very satisfied with it.
So was Mino. Whenever I met him during the following years he asked if I had another used exchange up for sale, he was ready to buy it!
The Repair Trunk.
EBX 8000 contained a number of supervising programs, which – in case of a fault – would isolate it and raise an alarm with data about its reason and its location in the exchange.
Repair at the site took place by a replacement of the defective circuit card. At a central workplace the card could then be examined and repaired. There were several card types, from cards almost filled up with reed relays and used in the selectors to cards tightly packed with electronic components such as integrated circuits for calculations and storage of data. How could they be examined such that one was really sure that all their functions were in order?
In the factory there was a machine for automatic test of the cards. Each card was fixed to a bed of nails, with a needle touching every path between two components. According to a program, unique for each card type, there was first applied a voltage over each component to see if it had the right value and orientation. Then other voltages were applied along with various signals and the results were measured. The advantage of the bed of nails was that also parts of a card could be examined, one was not limited to testing from the contacts at one side of the card. The results were printed and the card could then either go to an exchange or be repaired and tested again.
Such a machine would answer all prayers – except the economic! It was a fine tool for test of 50 to 100 cards per day, no question about that. But if you should only test 1 or 2 cards per week it was prohibitively expensive.
What then? Hilversum would of course have as their first answer that all defective cards could be sent to them for repair. They had the machine and their technicians would be kept in training by the number of cards they would receive.
However there were many countries where this would be impossible if only due to the customs treatment. A duty to be paid every time a card crossed the border and without a possibility to save it for cards already imported once can quickly spoil every idea about having a central repair shop in Holland. This was not a problem in Denmark, but Copenhagen Telephone wanted that their clever craftsmen should still have something reasonable to do. They should not be replaced by people only able to follow the directions in the manuals or be reduced to only doing replacements in the exchanges.
The requirements from the national organisations made Hilversum develop the repair trunk. It was a lot of electronics installed in a metal trunk, such that a card could be plugged into a socket at the trunk, voltages and signals could be applied to the contacts in the socket and to a pin, which could be applied to any point of the card. For every card type there was described which voltages and signals should be led to which contacts and how the test should be performed.
Copenhagen Telephone received such a trunk and the number of cards for repair went down considerably. It was agreed that they should report to us how many faults they had found each quarter year. Soon there were no reports and I said that then Philips could only reason that there had been no problems and that it looked fine in the statistics over the operational experience. Oh no, I should not perceive it in that way, but when the number of faults was far below the agreed number (1 fault per 100 extensions per year) they would not use time reporting.
This had, however, a negative effect when there was something seriously wrong. The technicians of the administration would go far in fighting problems themselves and in most cases they found a solution. But when they gave up it was a bad situation also in relation to their customer. They then wanted us to come up with a solution within 24 hours and this was not always possible, our people in Hilversum should also use time to get to know the problem!
1978: The National Bank of Iraq.
One day my telephone rang. It was Ole Jørgensen from Mogens Balslev, Consulting Engineers. He had a lot of questions about EBX 8000. They were advisors for the builder, the state of Iraq, on a new National Bank, and among the offers for an exchange there was also an offer for an EBX 8000 exchange. So I told about the system and the experiences already gained with it in Denmark.
I naturally told Hilversum about this. After some time Hugo Simons came back and said “forget it, we do not have an organisation in place to support it”. I wrote a letter to Mogens Balslev to tell this.
Shortly after I was very surprised to see in the Dutch personnel paper “Koerier” that Philips had received an order for equipment to “The Palace of Parliaments” in Baghdad, a splendidly equipped building, probably called “the mother of all parliaments” by Saddam Hussein. Among the equipment there was also an EBX 8000 supplied and serviced from Philips India.
This caused another letter to Mogens Balslev explaining the new situation (I did not tell Hilversum about this initiative). The contact was resumed. It resulted in that the then big German contractor Philip Holzmann got the main contract including an EBX 8000 delivered through the German Philips organisation. Thus we earned nothing. It was a funny construction: to be a consultant for the consultant of the builder concerning our product!
In the middle of the 1980es Ole Jørgensen called me again about courses in Baghdad about the system. Iraq and Iran were in a war at that time and all men were called to do military duty. The bank had to instruct a number of women to cater for the exchange. I referred to Hilversum where the necessary teachers and material for this was available.
Finally it happened during the First Golf War that the newspapers reported one day that the National Bank was hit by a rocket. Oh, I thought, now we shall hear nothing more about that exchange. I was wrong, shortly after Ole Jørgensen rang me. The exchange was not hit, but it had broken down, what could we do? At that time there was a blockade against Iraq. I mentioned this and referred once more to Hilversum. What happened after this I do not know.
Use of facilities in EBX 8000.
The basic principle in the EBX 8000 program was that the program should contain all imaginable facilities while the actual use of them was laid down in data. E.g. whether the necessary data tables for a given facility were in place and how big the tables were. In several types of public exchanges (including PRX) the program was made up of modules for the individual facilities and these modules were linked together when the program for each exchange was made.
The advantage of an approach as in EBX 8000 was, that the final program for the exchanges could be tested as a whole just once. Imagine having five additional modules which can be combined in 32 different ways, from having none of them to having all 5. In that case the program package had to be tested in all 32 set-ups.
However, this principle could not be maintained for all programs. Some functions were specific for a single market and had appeared after the program as such was finished (the program for controlled interrupt, this “Danish hobby”, was a typical case). If they should be introduced in all exchanges one had to go and reprogram them all. This was against the philosophy that the program was finished as it was supplied with each exchange. There was of course program errors needing repair by help of patches supplied to all exchanges, but it was left to the customer to decide when they should be read in. If an exchange worked without problems there was no reason to correct an error before one should anyhow work with it for an expansion or the like.
The principle was also not applied to programs used only occasionally in each exchange. Typical examples are the programs for traffic measuring. They could be read into the data store (in a place reserved for this) and linked to the basic program, i.e. some words in the basic program were replaced by others such that the exchange jumped to the additional program at suitable intervals, ran the additional program and jumped back to the basic program. When one had finished using the additional program it could be delinked, i.e. the basic program was re-established, and then the additional program could be deleted. The latter only meant that the space it had used could now be used by other additional programs.
There were two types of traffic measuring programs, Erlang-measurements and event-counting. The first type was for proper traffic measuring. At fixed intervals it was counted how many circuits of a specific type (e.g. outgoing lines to the public exchange) were busy. If a count was made every 36 seconds it meant 100 counts per hour. If in average there was found 10 busy circuits (meaning a traffic of 10 erlang) the total counted number would be 1000, or the traffic in centierlang. It will be obvious that in each single count the number could be anywhere between 0 and the total number of the relevant circuit type. Many counts were necessary before one had a figure deviating less than 1% from the true value. The number of counts depended also on the time a type of circuit was made busy during a call. A code sender occupied for a few seconds only at the start of each call needed short intervals between measurements to give the right result. Therefore the shortest interval possible was 3,6 seconds with the result of the counts of one hour given in millierlang. Such a measurement gave a 10 times heavier load on the exchange than a measurement with 36 seconds intervals, and there were rules for how many circuit types it was allowed to count with short intervals at one time.
Event-counting could e.g. count how many times the code for call-back was used during a given hour (if you had called another extension and found it busy this code would cause a call to you and on your answer a call to the other extension, after it had become free of the first call). For this purpose Hilversum had reckoned with a space for 5% of the users, 50 for a 1000 extension exchange. We had to be economical about storage space then, and I argued therefore that it was far too much. With 0,1 erlang of traffic (6 minutes per hour) for each user, equally distributed over incoming and outgoing calls and over internal and external calls, there would be a traffic of 0,025 erlang outgoing internal calls per user in the busy hour or 1,5 minutes. If the average length of a call was 2 minutes it meant less than one outgoing internal call per extension in the busy hour. With a total traffic of 0,1 erlang per extension one would at every 10th internal call meet a busy tone, or once per 10 hours. With 1000 extensions it would happen 100 times during a busy hour. Each call-back would occupy a place during half the average call duration or 1 minute, or for a total of 100 minutes in a 1000 extension exchange. As each place could provide for 60 minutes waiting time (for 60 different calls) two places for call-back or 2 promille of the extension number would be enough. These figures are all averages, so with the variability in the traffic the necessary number was rather for 1% of the users or 20% of what Hilversum had assumed. My evaluation was approved by Copenhagen Telephone, who otherwise expected a much heavier use of the facilities than foreseen by Hilversum. For this reason they did also not expect to have larger EBX 8000 exchanges of the type with 64K data words than their own with 2000 users.
Copenhagen Telephone used the traffic measuring programs in their own exchange and in 1978, when they reported some of the results to us, I found this to be a suitable basis for an article in “Philips Telecommunication Review”. EBX 8000 was supplied in several countries and it might stimulate the sales to publish that the system was already so well established that there were operational experience to draw from. In addition it was not bad that this experience was gained in Denmark!
I took it up with Jørgen Michaelsen at Copenhagen Telephone, who agreed to the idea. I used whatever could be used in the measuring results and wrote the article. Hilversum also agreed to the idea (somebody said it was nice to see an article without long Ses (integration symbols), another that its tone was too popular for this publication, well, you can’t please everyone so you’ve got to please yourself). The article was examined by Michaelsen and adjusted so Copenhagen Telephone could also accept the contents and it was published in the May 1979 issue.
What was the conclusion? As to the traffic the figures were as expected with some very intensive users and others only phoning a little. This is typical also for public exchanges: when the telephone density goes up (60 per 100 inhabitants in Denmark in the 1990es) each user uses the telephone less. In general the exchange was dimensioned reasonably. All other had also been a scandal as it was dimensioned according to measurements at the former exchange.
Regarding the use of facilities the result was disappointing. Only those facilities used often and where the users could memorise the codes were used to any greater extent. It was automatic call-back (mentioned above), abbreviated dialling (where you could choose a four-digit number (same length as an extension number) and the exchange would call an external number of maybe 15 digits) and rerouting (dialling *21 followed by an extension number, calls to your number would be rerouted to the other extension). The chief-secretary function, where two telephones could make mutual calls easily and could transfer calls mutually, was not used much. Not because there were no chiefs and secretaries at Copenhagen Telephone, but because it was too complicated to use. The functions needed other codes with a * in front and then it was easier to just call the four-digit extension number!
It was however necessary for the sales to have an impressive list of facilities. Then it was less important that many of them were only used sporadically!
At our next visit to Hilversum Michaelsen and I got a gift for our article. I choose a ship’s clock from the vast range of Philips goods. It has since then had its place in our sitting room with no other maintenance than a new battery every two years. </p>
You may judge the article yourself, see File:Telecom Review 1979.pdf.
The Supervisory Board of Philips.
During the 1970es a law was passed requiring that supervisory boards of public companies had to include members elected by the employees. As a result such members were also elected to the board of Philips Denmark. During the first two periods it was John Jensen, representative of the smiths at the factory (who was re-elected time and time again until the factory was closed down in 1993) and Alfredo Pedini, chairman of the personnel society PAP. He said in 1979 that he would stop as a board member and that he would like me as his replacement. I ran for the office and was elected.
I was not a very active member and the whole construction around Philips Denmark did not encourage activity. The company was (and is probably still) owned 100% by Philips Electronics in Holland and for the general assembly they sent a power of attorney to the CEO to represent all shares. The board members coming from outside Philips (Haldor Topsøe, Thor A. Bak to mention a few) were thus in fact elected by the management of the company. We had thorough discussions of various points but in the end it was a management decision what should happen.
This resulted naturally also in that the yearly general assembly took less and less time, in the end only the time required to read the agenda! The reason was that there was first a board assembly where i.a. the accounts for the year passed were discussed in detail after a presentation by the CEO and the accountant. Then the board signed the accounts. We then had the general assembly, which brought nothing new to any of the attendants. Finally the new board could have its first meeting.
While I was a member the CEO, Svend Aage Windelin, wanted to retire. Philips Holland selected for the first time a Dutchman as new CEO. Generally within Philips the job as CEO of a national organisation was a training job for people selected to go high up in the Dutch organisation, but until then Denmark had always had a Dane as CEO. The proposal to appoint Hoekman as CEO was discussed in the board, which had no other options than to say “please”.
Hoekman brought orders from Holland to rationalise. Work habits should be investigated and changed to reduce costs, and almost before the habits were changed reductions in the work force began. Thus, Hoekman was not a beloved person and I have kept quiet about that I had in fact had a part in his employment.
I was only member of the board during one period. At the election in 1981 Kirsten Topsøe-Jensen ran for office and was elected. She argued much about voting for a female board member and about all that she would do, and I could not do much about that. Still, it was somewhat unpleasant to become no. 2 in the election.
IEE Conference in London 1979
In 1979 there were conferences on computer controlled exchanges in Birmingham and London in two consecutive weeks and I was allowed to participate in them. This was also an opportunity to see several of the customers of the British organisation, who had visited Copenhagen to see the EBX 8000 at Copenhagen Telephone.
I took my wife with me and we had a fine weekend in Wales in between.
From the London conference I especially remember a session where two suppliers were to discuss their very different PABXes. One was van der Schaaf, who talked about EBX 8000, the other was – as far as I remember – from Micom. There was a very big difference between the two, and in a way van der Schaaf made only little impact. Here was an honest man describing a system already in use and proving its worth. And opposite him was a salesman with big rhetoric skills presenting a system still in the design phase, but of course able to do everything when it was ready, and just because it was not ready it was far more progressive and modern than EBX 8000 with its already three years of operational experience.
It was a flashing performance, but did it make an impression at the audience, which consisted of professionals able to discriminate between reality and sales talks?
1979: The Super Salesman.
There were often meetings in Hilversum for salesmen from all the countries marketing EBX 8000. At such meetings it was an important point how business was in each country. In these surveys Denmark stood out in the figure for sales per employee. In Denmark these statistics only counted me, all the persons visiting end-customers, making projects, installing and servicing PABXes were at Copenhagen and Funen Telephone. Naturally our sales price per extension were lower than if we had treated the customers directly (maybe only half), but it corresponded well to the activity at the administrations that they could double the price. Further, if we had approached the customers directly (if it had been allowed, it was still monopoly times!) we would have needed a staff of maybe 10 people on whom to spread the sales income.