New switching methods will be available, both electrooptical and purely optical. They will be needed for the much higher bandwidths and digital bit rates. See Appendix 11.6. By AD 2000, though, they are unlikely to be more expensive than present types, in some circumstances being in fact cheaper; so we can neglect this factor in the economics of our global planning. Some further features of future optical networks have been described recently.3 To facilitate the switching methods, and also information retrieval by subscribers from the future information processing centres, much more compact digital stores will be available, having capacities of up to at least 1012 bits per cm3 of storage material. One of the best versions is likely to arise from extensions to present work on holographic storage by the movement of small electric charges from one `trap' to another in electro-optic crystals such as lithium niobate, using laser beams for writing in and reading out, as a parallel operation, both in not more than 1 nanosecond eventually, per group of at least 1 000 bits. See Appendix 11.7.

4.1 General concepts

We shall now take a look at the world's future mobile speech equipment - the available new devices, and its network planning and operation. I say `speech' equipment not because mobile videophones are by any means impossible, but because their much greater loading on the non-optical frequency bands they will usually need, will rarely justify mobile TV equipment on an individualchannel basis. Every subscriber is often a `mobile'; the problem of wandering subscribers is a vital factor in the whole network plan. It is my experience in Britain today, even during normal working hours at each end, that when I call someone therein about a 40 per cent chance that the person I wands not there. If it is urgent he has to be searched for, usually a lengthy and often an almost impossible process, a frustrating and mainly unnecessary obstacle that our next generation will not tolerate.

Because of much better telecommunication their future wanderings will be far more often for leisure and pleasure than for their work, but faster and cheaper air travel will take many of them into almost any part of the earth. The problem of immediate access to subscribers, therefore cannot be truly solved without the use of completely mobile world-wide personal telephone numbers. We shall now explain how the vastly cheaper and smaller, fast data storage-and-retrieval devices that are on the way, together with the much wider-band, cheap transmission channels, can at last bring to reality this planners' dream.

4.2 Global personal telephone numbers

Assume that in AD 2020 the world population is about 16 x 10⁹. Even if a quarter of this number became subscribers, 32 binary digits would be enough to identify each. We may decide to reduce to about one-thousandth the chance of getting and disturbing a wrong number due to a single false digit, by adding ten more. Each personal number would then contain 42 digits, recorded in permanent and very compact form within an attachment to a miniature pocket-phone that the subscriber would always carry with him, be supplied by his local telephone administration, and be used automatically for call-costing and all other identification purposes.

The attachment could be as in Fig 1, and contain 42 printed-circuit connections about 0.1 mm apart, the code group being made by now-standard laser-beam cutting techniques.' A `cut' could be a `0' digit, with a `no-cut' as a `1'. When power is supplied each connector that is uncut could change the bias and the bistable state of one of 42 integrated-circuitry flipflops, all monitored in sequence at a convenient rate by conventional techniques. The code group would then go into a temporary store in the sub-set or pocket-phone for automatic transmission when required at a second convenient rate. The attachment would also contain a variation on now-available devices - a narrow magnetic tape up to about 10 cm long, on which up to about 90 other global telephone numbers could be erasably stored, each in the lateral direction. This number would be the maximum that most people would need to record in advance, pre-call selection being by moving two small wheels each in decimal steps. For both write-in and read-out the magnetic head could be quickly moved laterally across the desired code group by a small spring released electrically at the right moments, again to or from a temporary store in the pocket-phone, or directly into the sub-set when a pocket-phone is plugged in.

In the subscribers' nominal homes and offices they would plug in the pocket-phone to their fixed sub-sets on arrival. When in any vehicle, they could plug their pocket-phones into a sub-set socket in that vehicle, connected to an appropriate mobile radio network. When just walking they could use and be called by their pocket-phones alone. When a pocket-phone is plugged into a fixed sub-set or into a vehicle, its do power supply would be automatically transferred from its miniature storage cells to the central battery system of the network or vehicle, the pocket-phone cells being then on charge instead of discharge, thus ensuring that these cells are almost never allowed to run down. When the pocket-phone is plugged in to a fixed (optically operated) sub-set, the operation would also cause its stored data to be written-in to an optical-type store in the sub-set having the much faster read-out time appropriate to the optical network. At all times except when a subscriber has his pocket-phone plugged into a fixed sub-set at his nominal home or office, this pocket-phone, either via the fixed networks or by free-space paths, will cause his location to be automatically recorded at suitable fixed points, at intervals appropriate to his network, unless he deliberately presses a bistable button to cut this operation off. When plugged into someone else's fixed sub-set, the record will show the personal number of its nominal user. At other times it will show the wanderer's areanumber, channel group and channel number in the mobile network he is in - but usually only in terms of changes in this information.

In spite of much better telecommunication, some firms will prefer to go on using some large office blocks even after AD 2020. In these cases most of the movement of staff in working hours will be within these buildings. Within each, all that will be needed is a fairly conventional audible calling system, picked up by an additional miniaturised part of his pocket-phone, by which any one of up to about 1 000 people in the building can be called to the nearest fixed sub-set, only a few feet away when he is a guest in someone else's office. Appendix 11.8 illustrates a simple way of minimising mutual interference, both within any one and between adjacent such buildings. It is based on the use of a fairly large number of low-power calling transmitters, together with an inverse-cube law with distance from each. For this application the wandering subscriber's nearest sub-set or small group of sub-sets will be constantly recorded on a channel of a temporary store entirely in the building, together with his global telephone number, only these sub-sets being used if he receives a call. When he leaves the building, as long as he remains in his nominal local area his new location will be stored, in terms of only the appropriate last few digits in his global location code, entirely at new fixed points associated with fixed or mobile sub-networks within that local area. When he leaves his nominal local area all his digits associated with that area will usually change, to conform also with his then new sub-area, channel group, and channel number coding. As he moves further away a similar process takes place, all his digits relative to these larger areas and his location within them being changed, but still recorded in stages as before.

When he has moved more than an average of about 100 km away, however, the process changes; it will then usually be found more economic to up-date automatically his entire personal global code, in world capacity stores (see Appendix 11.9), in developed areas about 400 km apart, except for ocean spacings. After this has happened the world-wide store concerned, by a similar process in stages back to its own location, keeps track of the wanderer's further movements in every relevant detail, until he moves into the next 400 km square which then takes over. When the man is called on his nominal number, then via this particular worldwide store, a switching centre associated with it completes the call's establishment by the quickest free routes, just as if it were the original caller, and at the same time automatically up-dates the called person's then location code in the calling subscriber's pocket-phone store.

By the then high-speed optical methods, the total delay in ringing the called subscriber, due to the opticalswitch delays alone in a tandem chain of up to about 20, should not exceed about 50 nanoseconds. On a transglobal call between fixed subsets of route length of say 20 000 km, an additional and more important cause of delay, will be the propagation time (about 150 milliseconds) after the call button on the pocket-phone has been pressed. A third delay, the time for automatic dialling of the 42-digit code, may amount to 5 milliseconds. These figures assume no waiting for free channels all the way, which will usually be true, even via ocean routes, as in future optical sub-ocean paths a total of at least a 10 gigabits/sec rate through each fibre of a 30-fibre bundle will become common practice - sufficient for 100 000 improved-quality PCM speech circuits in both directions, if a pair of fibres in the cable were reserved for such non-TV demands.

The world-capacity, global personal-number stores, 400 km apart, will also be used as the telephone directories of the future, having the added advantage that they can be constantly up-dated. Their needed access rates, though, will depend on the type of subscriber to be called, the main division being between private, and business or professional people having direct access to the public. The need for private numbers will be quite rare. For any information a man receives that leads him to need a new individual private number will practically always contain this telephone number as well, being added automatically, at least for a time, to his own small store in a free place within it. An average of only on such need per year per subscriber is probably a fair estimate. He may well need indiscriminate access to business or listed professional numbers, though, in a given category about three times as often, on account of the advertising he electronically receives. Particularly when he is in an area strange to him, the event leading to his need will not usually give him automatic access as well to the corresponding listed numbers. So these requirements must be catered for by additional means, leading to a total average directory demand per subscriber of about four per year.

Much the quickest and most practical way for the average subscriber to get the telephone number of a particular name and address, or of any one at random in a listed business or professional group, will be to say what he wants through his microphone, especially when he ie using his pocket-phone alone. Speech recognition machines present many challenging problems, and do not yet exist in really satisfactory forms. But the present state of this art is such that I believe we can confidently predict their arrival, for directory-enquiry purposes and for stylised general information retrieval, within twenty years from now, and at economic prices. So an access point in such a machine in his local area would receive his demand; it would then transmit it to the nearest world-capacity store at a rate of 10 gigabits/ sec. In a local area containing 50 000 subscribers the average total directory demand would thus amount to about one every three minutes, handled by about 6 machine access points, with a normal maximum waiting time that would be acceptable for this very occasional service. When a subscriber is at a fixed subset the directory-enquiry service will be improved if suitable parts of the automatic information feedback to the caller are in visual rather than audible forms, displayed on his TV screen. Confirmation that the computer has interpreted correctly the spoken name and address of the person demanded, for example, will then take appreciably less time, and be always unambiguous.

At the world-capacity storage centres, we have seen already that if the recording is in hologram form in, say, a lithium niobate crystal, both the write-in and read-out times per number could probably be not more than 1 nanosecond; and that at least 10¹² information bits could be stored in 1 cm³ of crystal material. It is interesting and important to estimate how the needed access arrangements for up-dating the global numbers among the mobiles could be provided. We can see from Appendix 11.9 that all that is needed at the crystal is a single 10 gigabits/sec optical fibre.

The access arrangements at world-capacity stores for directory enquiries could possibly be as follows: The subscriber capacity of each, by a previous assumption, will be about 4 X 10⁹, leading to a bit-capacity need of about 1.6 x 10¹¹.Even assuming that the average rate of changing a nominal global personal telephone number is as high as 1 per year, this leads to an average up-dating rate of only 120 per second - many orders smaller than could be dealt with quite easily by a single movable laser beam on to a crystal of 1 cm3 in volume, allowing 10 per cent extra space for the listed business and professional people. The read-out access would be to its own area containing not more than 50 million subscribers, at the total average demand rate we have already assumed of 4 per year per subscriber, thus totalling only 6 per second. This is still more easily obtainable - it is in fact so much below the limiting speed that it is almost inelegant to use it! But it will almost certainly prove in on economics. In practice, of course, we shall at least triplicate each type of largearea store in case of occasional breakdowns, in separate locations not too near together.

4.3 Devices and principles in future mobile networks

We shall now look briefly at the extreme limits of the mobile density that could be handled. It is clear that this possible number of mobiles per km², by new methods, could be many times greater than has so far been accepted, well above any easily foreseeable demands, and by means quite within cheap and sound engineering tolerances.

The internal calling networks within a building has already been dealt with. Next, what about cars on motorways and ordinary roads; walkers on golf courses, hills, beaches; and inland and near-shore boats? To make interference impossible between mobile areas, the oxygen absorption band centred on 60 GHz will have to be used more-and-more frequently. Fig 2 shows that there is a central band about 6 GHz wide where the extra (exponential) attenuation is nowhere less than 12 dB/km at seal level. With digital systems, receivers will usually be immune at 3 km away as they will have low-level threshold clippers, and any normal type of receiver will be immune at 6 km.

Miniature solid-state oscillators suitable to fit into the pocket-phones, for both transmitting and receiving, will almost certainly be available within less than 10 years, drawing up to about 1 watt do mean power at 10 to 15 volts, and radiating up to about the needed 0.15 watts on the peaks. It looks now as if these will be based on Gunn-effect diodes, probably within the LSA sub-group (Limited Space-change Accumulation).¹⁶ In a c.w. oscillator of the LSA type in gallium arsenide, the theoretical power efficiency is limited to about 30 per cent, on account of the rather small peak-to-valley current ratios available in their N-shaped current-tovoltage characteristics. I believe that most future mobile applications will use a pulse modulation on the carriers, normally lasting about 1 nanosecond per pulse, produced internally within the oscillator in the wellknown `squegging' mode. To prevent further waste of power during the non-oscillating periods, a suitable highfrequency transistor can be used to cut off the drive current during these parts of the modulation cycles. A resulting over-all pulsed-oscillator power efficiency of about 15 per cent is likely to be achievable, giving about 150 milliwatt RF output from 1 watt input - quite adequate for most applications, as the normal maximum path length will be only 1 km. For higher efficiencies, it may well be necessary to find or make, if possible, a semiconductor material otherwise as suitable as gallium arsenide at 60 GHz but that has a larger peakto-valley current ratio.

No quartz crystals will be used in the mobiles for local frequency control. All accurate timings for channel selection will be derived from the ground stations, tolerances on resonant-circuit tunings being as wide as about ± 0.3 per cent. This will eliminate the much more accurate quartz crystals that otherwise our future mobile networks would need.

Mobile networks by FDM

With FDM(Frequency-Division-Multiplex) systems, in the 6 GHz-wide oxygen-absorption band there is room, of course, for about 40 000 PCM channels all in use at once, each 100 kHz wide, (to cater for the higher quality speech circuits of the future). On double sideband, probably essential in practice, there could be 20 000 channels. To ensure that all parts of the signal band are always at least 36 dB down in adjacent areas, these 20 000 mobile channels would have to be spread out over 25 kmz, giving a maximum mobile density of 800 per km², having an average linear spacing of about 28 metres. This would be quite adequate for most purposes, allowing for, say, a 30 per cent simultaneous usage from all cars in two 8-lane motorways in any one area, with a 40 per cent reserve space for other mobiles, even if the cars have an average longitudinal spacing of only 10 metres. This shows that the useful, limitedrange frequency band in the 60 GHz region could deal with most mobile-traffic densities by conventional FDM methods alone. But though equipments exist already called 'synthesisers', for handling the frequent wavelength changes at any one user point that such an FDM-PCM mobile network would need, it is very doubtful if even by future miniature integrated circuits, terminating in the 60 GHz band, the apparatus could be either small or cheap enough to fit into a pocketphone, for adequate protection against cross-modulation among signals of widely different field strengths. By using instead a rather revolutionary new principle, though, we can handle if need be still higher mobile densities, and do it more cheaply.

This different method, that I have named `SYNSOL' (the SYNthetic SOLid-state principle)⁸ is a very recent concept. Both more rigorous analysis and a fair amount of computer simulation will be needed before all its difficulties and limitations are apparent. It is a new and powerful non-linear network-planning concept. It is obvious that not only its possible snags but also its full latent advantages, together with applications in other fields, can be appredated only after a great deal more thought.

I shall now describe some of the aspects of what it can do, not how it does it.

Without contravening orthodox information theory, it can handle a theoretically vastly greater number of users without mutual interference, all in the same frequency band and in a given area or volume. It is self-adaptive to the number of users, and to wide variation in their location densities. The needed channel changes, too, as a mobile progresses through a network can be automatic by simple means. The required equipment, even for an output in the 60 MHz band, can be cheap, have wide tolerances, and by integrated circuit methods be small enough for pocket-phones; and where the mobiles and ground stations have the same designs, reception reliability of the signals can be satisfactory and nearly constant up to a predetermined distance, e.g. the point where the channel concerned is transduced into the land network, and then fall substantially to zero quite abruptly.

It is unusual to publish an idea while it is still in an initial, rather `woolly' stage. In speaking about the future, though, I believe it is sometimes necessary, as otherwise the picture painted could have a vital feature left out. The main idea behind the method is easy to grasp. First, a line of mobiles between the calling unit and the ground station, A and B in Fig 3, form a chain, normally in single file. Each link gives a 2-way speech channel between the mobiles or ground station at its ends, becoming when established immune to everything else. The result is something like a solid-state longchain molecule, if its cross-section were monatomic. Similar chains then form successively among the remaining, non-aligned mobiles. In this way we produce separate paths between the mobiles and the ground stations feeding the land network, each becoming immune to signals from other paths just as if insulated copper conductors had been grown between the desired points. Each link in the chain is bistable, in the sense that once broken it remains so unless specifically reestablished - as a broken soldered connection has to be soldered again.

Each mobile in a given chain, as well as being able itself to have 2-way conversation with the ground station that temporarily terminates it, can also act as a repeater towards a ground station for all other mobiles in that chain. Thus the limit to the mobile density is set almost entirely by the ground stations; but in the same way they themselves can use the same frequency bands, or by pulse methods the same time-slots within them, so if there are enough of them the true limit can be put almost indefinitely further away. If while a new chain is being formed it finds no unaligned mobiles immediately round it, these are automatically by-passed until free, alignable units are found. The same thing happens if it finds any mobile in its normal chain to have a radio that is out of action.

To use an analogy, in the solid state, the adhesive forces between atoms and molecules - due to the interactions between the nuclear and electronic electric fields, the magnetic fields due to the electron spins and orbital movements, and the quantum 'exchange forces'have very short ranges of action, the electron clouds becoming predominant and causing only repulsion between atoms when two metal surfaces (say) are pressed together after a mechanical break. It is by analogous processes that in SYNSOL the mobiles in a chain adhere together firmly as a signal-carrying group up to a certain value of tension, while repelling away the causes of forces from other chains. It is still another fascinating example of the multitude of useful things that can be done by delving into the so-far largely untapped storehouse that non-linear methods provide. I find that such more sophisticated non-linear circuits are often liable to frighten away the average electronics designers. But I assure you that the reason is merely that they are not yet sufficiently familiar with them. When correctly made they can be just as reliable as a good watch, which already uses such circuits, in mechanical form. By an AGC (automatic gain control) at each receiver, in combination with an APC (automatic power control) at each transmitter, after a link in the chain has been established the receivable range of the chain-locking signals at each end of that link is made to be nearly linearly proportional to the link length. In this way the same frequency band can be used by every mobile, even when some of them are only a few metres apart, still maintaining negligible mutual interference between different-chain members. In the preferred form, the chain-locking signals will be pulses of about I nanosecond duration, 100 per centmodulating the 60 GHz carriers. It is after a link has been established that it can start to convey speech information to and from the appropriate ground station on a multi-channel basis, connecting all the mobiles within it. Some further details of the SYNSOL method are given in Appendix 11.10. Any approach to a complete description or explanation though, even on some of its more basic features, must wait for later publications.

An alternative has been called by me the SYNSPEC scheme (the SYNthetic SPECular reflection method).⁹ Instead of relying on a grouping of the mobiles into long chains, it operates by the formation of temporary holograms, within larger groups of mobile users. Study of this alternative has hardly yet been started. It is of course impossible to predict now which of the two will prove the better, and in what circumstances.

To cope with long links in the SYNSOL chains under heavy rain conditions, a reserve carrier frequency of, say, 1 GHz will automatically be brought into action, as at 60 GHz the rain attenuation itself can occasionally amount to about 90 dB/km, while being less than 1 dB at 1 GHz. The circuit plan to achieve this is particularly simple if the chain links when at 1 GHz consist of dipulses (Fig 4) containing mostly 1 GHz components. See Appendix 11.11. The 1 GHz carrier will also come into use sometimes without rain, when a link needs a degree of diffraction round an obstacle that is impossible at 60 GHz but obtainable at I GHz.

5 Information retrieval centres
We shall now examine briefly the design and operation of the future equipments that will replace the central libraries that we now have. If such an IRC in AD 2020 contained all the information then considered worth recording for immediate access, at least at one place and at one time, how many information bits would it need to store? There are several reasonable bases on which to form an estimate. Consider an extreme estimate of the storage capacity needed for immediate access, assuming that in AD 2020 the man-made written information considered worth storing by the IRCs, assessed by the economics of the probable demands to extract it or in any other way, has been generated by 10⁹ people, with an average of 10 000 words, i.e. about 300 000 bits, from each. Assume we also want to store in PCM form, at any one time and place, 500 films each running for an average of one hour. Suppose we decide we need to store, in addition, 20 per cent of the above total number of bits for machine-made information, e.g. computer results, photographs, etc. The total capacity then needed is about 1015 bits.

This number of bits will almost certainly be storable, in hologram form, by trapped charges in not more than about 1 000 CM³ of suitable electro-optical material. If we multiply this volume by 3 000 to allow for laserbeam access to each crystal, and for the needed optical switch-gear to and from the user wideband optical channels in the local areas that each IRC serves, we get 3 cubic metres. The main space requirement, though, would be for the man-machine interface; for though both fault-finding and fault correction will become more and more automatic, some manual monitoring and servicing will still be required. To this extent the relatively enormous size of the human body, and particularly of its arms and fingers, will be the over-riding factor. However, the whole future IRC should go into a building smaller than many present telephone exchanges. The storage could be for at least 100 years, without regeneration where the number of read-outs per year is low, and at ordinary temperatures. To achieve our aim we must learn how to dope the material to ensure that almost all the traps are of a fairly uniform depth of, say, 1 electron-volt in energy level. The resulting bit capacity of about 10¹² per cm³ is enough for one good-quality film in PCM form running for one hour. The write-in can be slow and use a variety of methods. The direct read-out, though, causes all the bits in one frame to arrive in parallel. To feed into the user's lines, a parallel to-serial converter is needed. One means of access to future crystal stores, suitable for the film-storing types, and using future optically pumped integrated circuit lasers instead of the nowusual junction types, is described in Appendix 11.12. If the user requires not a film but only a digital word contained in a fraction of a film frame, the same technique can be used, the required bits being extracted by the then-normal TDM pulse-selection methods. The stores for global telephone numbers, perhaps about 400 km apart will usually be sections of the more general information retrieval centres. The needed much faster write-in, though, will require the different crystal-store lay-out that is discussed in a section of Appendix 11.7.

Each general-purpose IRC will rarely need an information store of truly world capacity, but be partly selective and be to some extent complementary to others. The best access method at the subscriber end is a very thought-provoking problem. On balance I think that a necessity for a subscriber to look up the nearest information sub-group title in a very large and complicated list, even when displayable quite quickly when he is at his desk, will not be accepted by the general public. I believe that his demand for general information will eventually have to be dealt with via an access point in a multichannel speech-recognition computer, these units being in each local area. They will have to be more sophisticated than the speech recognisers for directory enquiries. Research to date has already shown us the main lines on which they can be made to be quickly self-adaptive to individual voices, e.g. as to accent and speech formant frequencies.1° Each machine will need, too, audible feedback from itself to each user, via a channel in a stored-voice mechanism, sent in PCM form. Here again we already know fairly well how to do it.

The subscribers will have to learn to use a stylised, nonredundant syntax. Multi-lingual machines will present added difficulties, at least quantitatively. But I think that eventual success is assured - but only by people clear-sighted enough not to be appalled at the apparent magnitude of the task!

It is necessary to say a few words about rapid access. Suppose the local area serves 50 X 10³ subscribers and that, except for films, we choose the extreme aim of never allowing any appreciable waiting time even when all the subscribers are using their speech-recognition machine at once. The total rate then needed from each local area to its IRC is 5 gigabits/sec (half the capacity of a single wideband fibre). Access time to longer and shorter films, and to words or word-groups occupying fractions of one frame, is discussed in the following paragraph. There seems to be no difficulty in making the waiting times quite acceptable.

Suppose that at any one IRC there are 250 64-minute films stored at any one time, and an equal number of 32-minute, 16-minute, and 8-minute films. The 64minute films would occupy 250 cm slices of crystal, one for each; the remainder could be stored in the form of 2, 4, and 8 per similar crystal slice respectively. If we cater for groups of simultaneous users for all these films, each kind having the same spacing between optional starting times of 2 minutes, the number of simultaneous user groups needed is 32, I6, 8, and 4 respectively, and the read-out techniques already described can apply. When the demand is for a bit group occupying one frame, (5 X 10⁵ bits) or less, stored in other crystals at one or more per frame with as rapid as possible a random access in these cases, the waiting period would in general be limited merely by the time needed to transmit and process the binary-coded address. In the case of a demand for a 1 000-bit word, the address would need 27 bits, if there were (say) 10⁸ of them stored. This would take 2.7 nanoseconds to transmit from the local-area processor, using a single wideband fibre - and about another 5 nanoseconds to complete the total access to the IRC crystal.

6 Isolated mobiles

We have not yet dealt with the communication needs of aircraft and ships when out of range of the future land-based stations, nor of land mobiles in areas so sparsely populated or undeveloped that they do not justify such networks. Our future ocean-going aircraft, public or private, when out of visual-distance range of land stations, will certainly get their communications by Cosmat-type satellites, with HIP equipment in reserve for emergencies only. While they are taking off or landing they can use the land stations, probably at the lowerfrequency end of the 10-to-0.9 GHz band. Before they are too far away to continue on this band they will almost always be above cloud; they will then use a stationary satellite, on an optical or 10 GHz radio band, the satellite completing the call to ground at 10-to-0.9 GHz. In the gallium arsenide optical band, the mobile would in practice have to produce a 0.004° wide beam, e.g. by using a paraboloid 1.5 em in diameter at each end, to produce enough photons to give an adequate signal-to-noise ratio from a mean transmitted power of 300 milliwatts. This is likely to become feasible, and also the corresponding auto angular lock-follow, but not at or too near the earth's surface, on account of directional shimmering due to air turbulence. Above 40 000 ft altitude, with certain types of airframe, it might conceivably be possible.

Existing optical techniques, when combined, could achieve this by an ititial rough manual locking, with a defocussed beam and a narrow received band-width, followed by an automatic electro-mechanical finer adjustment, and a final non-mechanical, precise adjustment from a miniature controller that changes the positions of virtual images of the lasers by electro-optic crystals. In addition, the time-constant and the angular range of the final process could easily be made fast and large enough to compensate against aircraft vibration.

Beam-splitting methods have been devised by which a single satellite could serve a number of aircraft at the same time, with auto lock-follow to each. As to the ocean-going larger ships, they will use satellites via radio, preferably near the 10 GHz end of the cloudpenetrating band so as to minimise antenna size. Beams 0.1° wide with auto lock-follow will be possible, as will beam-splitting at a given satellite, if justified; but 3-metre-diameter dishes at each end, or the equivalent in total antenna gain, would then be needed. A 2-metre dish at the satellite, with a 1-metre version on the ship would usually be more practical, 0.2 watts transmitted power then being used at each end. At the 10 GHz end 6.5 per cent of the frequency band could cater for at least 2 500 PCM audio channels per satellite, with a total radiated power of 0.5 kW from the satellite. Currently planned satellite solar cells, with solid-state transmitters of 20 per cent efficiency, could thus cater for double this number per satellite. Satellite nuclear power packs, too, are now being considered. Air turbulence would in general prohibit the use of sufficiently narrow laser beams. Lone yachtsmen, etc., who sail the oceans in small boats, will in general still use the 2-12 MHz radio band.

We are left now with the cars in undeveloped areas, out of range of the future visual-distance land networks. They too will be unable to contact a satellite efficiently by laser beams. So they will have to make their calls, rather expensively, by using part of the 10-to-1 GHz band. They will get into trouble if they switch on their HF equipments except in a real emergency! To do so they can use a 1-metre dish, with auto lock-on and lockfollow to a I°-wide beam-splitting satellite, with 0.2 watts transmitted from each end, like the ships in similar circumstances.

7 Space communication

There will again be the problem of getting through the earth's cloud. In the cloud-penetrating band of 10-to-1 GHz, though, it will probably never be reasonably possible to achieve enough field strength at the earth's surface on really long-distance paths, since for good audio channels the paraboloids would have to be about 100 metres in diameter at each end (or the equivalent), from a space vehicle about 2 000 X 10⁶ km away, if the spacecraft's transmitted power is limited to 100 watts. But there is a feasible answer: to transmit by laser beam between the vehicle and a satellite, and then to change frequency to the 10 GHz band for thesatelliteearth path. The limiting factors would then be the sizes and tolerances at the laser paraboloid. If the spacecraft had a 1-metre-diameter paraboloid and the satellite (then probably a space platform - or the moon) had a 2-metre paraboloid good 2-way audio contact up to about 6 000 X 10⁶ km would be possible, with a delay, of course, of over five hours in each direction. This would cater for manned spacecraft at the distance of the planet Pluto.

The speed problems in the orientation of both dishes by automatic lock-follow are basically no worse as the spacecraft gets further away. The maximum speed of the feedback-loop controller to achieve it is reduced linearly with distance; but for given values of the unpredictable changes in the vehicle's position, so are the parallax changes that these unpredictables cause. The lock-on orientations of both dishes, of course, would need precisions varying linearly with distance, as the reciprocal of this distance determines the needed beam widths. But this is not a basic problem, and in practice it would be quite soluble at 6 000 X 10⁶ km. For intercommunicating entirely outside planet atmospheres that absorb in the useful laser-beam band, suitable laser-beam methods alone will of course suffice.

8 Conclusion We have taken a brief look at many aspects of the communication revolution that is undoubtedly coming. We have seen that mere distance can be virtually eliminated, and by means that even now have outlines that at least to my mind are clear. I suggest that those in the communication field who believe the picture I have painted, at least in the main, should first decide if they like it and if they want it. If they do want it they should take all the necessary steps, to ensure that the countries and areas that concern them do not leave it too late to reap their shares of the fruits it will bring. The goal may not be the single most vital thing that the world now needs, but a world-wide `nerve system' of an altogether new order, streamlining the way to all other aims, will be no mean achievement. The effort will be hard and long enough to bring creative fun and pleasure to some of the very best minds that we have!

9 Acknowledgements

So many of my STL colleagues have helped me in the preparation of this lecture that it is impossible to mention them all individually. I ask them to accept my sincere thanks. Though all of them would have put in an equal effort if I had asked them, the following have in fact contributed most. They are: C. F. Drake, C. C. Eaglesfield, V. G. Herrington, K. C. Kao, S. F. Laurence, A. E. Mounter, M. M. Ramsay, J. C. de Rivaz, I. F. Scanlan, P. R. Selway, G. B. H. Thompson, and R. J. Whelan. I hold none of them responsible for any predictions or opinions that I have expressed! I am indebted as well to the University of Strathclyde, Glasgow, for their permission to reproduce certain extracts from their 1969 `John Logic Baird Memorial' Lecture, which is published by them. I am similarly indebted to the Elsevier Publishing Company, London and Amsterdam, concerning Fig 5.

10 References

1. PANISH, M. B. et al, IEEE Journal of Quantum Electronics, Apri1,1969.pp.210-212.
2. THOMPSON, G. H. B. Unpublished report at STL,* predicting the results of reference 1.
3. REEVES, A. H. 1969, John Logie Baird Memorial Lecture. Published by the University of Strathclyde, Glasgow.
4. JACKSON, T. M. and Rowe, T. J. Laser cuts masks to size eliminating most errors, Electronics, Vol. 42, No. 2, 20 January, 1969, pp. 81-84.
5. GILMOUR, W. D. Wireless World, Jan. 1969, pp. 5-7; and Feb. 1969, pp.76-80.
6. STRATTON, A. W. Progress in Radio Science, 1960-1963,Vol. 2, p. 278.
7. MARTIN, D. J. IEE Conference Publication No. 14, Paper 15,Sept. 1965.
8. REEVES, A. H. and DE RIVAZ, J. C. Studies at STL to be published later.
9. REEVES, A. H. Unpublished study at STL.
10. FRAIPONT, D. Report 3, Electronic Associates Inc. Mar. 1966.
11. REEVES, A. H. Unpublished study at STL.
12. RAMSAY, M. M., REEVES, H. and THOMPSON, G. H. B. Unpublished studies in progress at STL
13. BASS, J. C. A Review of Electro-optic beam deflection techniques. Radio and Electronic Engineer, Dec. 1967.
14. BROOM, R. F. and SCHULZ-Du Bois, E. O. Demultiplexing of fast optical pulse code modulation, IBM Research Laboratory, Switzerland. Private Communication.
15. RAMSAY, M. M. and REEVES, A. H. Unpublished studies in progress at STL.
16. COPELAND, JOHN A. LSA Oscillator-Diode Theory, Journal of Appl. Phys., Vol. 38, No. 8, 1967.
17. JONES, M. and KAO, K. C. Spectrophotometric Studies of ultra low-loss glass - II Double beam method'. J.Sci.Inst.Feb. 1969.
18. KAO, K. C. and HOCKHAM, G. A. Dielectric-fibre surface waveguides for optical frequencies'. Proc.IEE. Vol 113, No. 7, July, 1966.
19. TIE N, P. K., GORDON, J. P. and WHINNERY, J. R. Focussing of a light beam of gaussian field distribution in continuous and periodic lons-like media. Proc. IEEE. Feb. 1965.
20. REEVES, A. H. and MOUNTER, A. E. Unpublished study in progress at STL.
21. Leading light in Japan. IEE News, No. 87,p. 7. Jan. 27th,1969.
22. DRAKE, C. F. and REEVES, A. H. British Patent applied for.
23. MOUNTER, A. E., RAMSAY, M. M. and REEVES, A. H. Unpublished study in progress at STL.
24. REEVES, A. H. 1969 John Logie Baird Memorial Lecture: Future prospects for optical communication, at Strathclyde University, Glasgow. Published by this University.
25. The Magic of Holography. Electronic Design II, May 24th,
26. CHEN, F. S., LAMACOHIA, J. T. and FRASER, D. B. Applied Physics Letters, Vol. 13, No. 7, 10 Oct., 1968, pp. 223-224.
27. REEVES, A. H. Unpublished study at STL.
28. THOMPSON, G. H. B. The double heterojunction. Discussion on gallium arsenide lasers, IEE (London), Dec. 15th., 1969.
29. REEVES, A. H. Discussion on gallium arsenide lasers, IEE (London), Dec. 15th, 1969.