Speech multiplex allows a number of speech channels to be carried over the same physical circuit. This is a description of the analogue multiplex that was current during the cold war era, nowadays this is done digitally.
The Human Voice
Our voice consists of a spectrum of sound frequencies from a few hertz to many tens of kilohertz. The Hertz is the unit of measurement for the number of times per second, an object or electric signal oscillates. However the human brain can still decipher words and music even when a large part of this spectrum is missing. CD quality music restricts the highest frequency to around 24 kilohertz and we find this perfectly acceptable. The telephone network makes use of this human ability, by further restricting the bandwidth from 300 Hertz to 3400 Hertz. Whilst this is not acceptable for music an individual's voice can still be recognised.
If the average sound energy in a passage of speech is measured and plotted so the horizontal line represents the frequency and the height represents the energy at that frequency. The curve for a typical telephone circuit will look something like this drawing. Although most of the energy is contained in the lower frequencies, it is the higher frequencies that convey the intelligibility.
Bandwidth of Copper Wires and Radio Transmitters
Pairs of copper wires used to carry telephone calls have a much greater bandwidth than is necessary to carry human speech. As the length of wires increases the available bandwidth decreases due to absorption of the higher frequencies. The line absorption losses may be overcome by amplifying the signal at regular intervals, there is a trade off between usable bandwidth and the cost of amplifying the line.
Radio transmitters can be designed for any bandwidth required. There are a number of factors influencing the choice of bandwidth mostly related to the operating frequency and channel spacing. In densely populated countries, the available radio channels can be at a premium, restricting the actual bandwidth available.
Utilising the Bandwidth
The extra bandwidth available on copper wires and radio circuits may be exploited using a speech multiplex to send more than one speech channel down a single circuit. The actual number of channels combined together depends on a number of factors and may be from two to thousands. The early warning receivers can be considered as a 1+1 system. Nowadays, Broadband Internet delivered over telephone lines (but not Cable TV) uses this extra bandwidth to carry the modem signals into the house.
The principle of a 1+6 speech multiplex is shown in the diagram below (A-End equipment). Seven unidirectional speech channels enter on the left. Each is filtered to restrict its bandwidth to 300-3400Hz. Channels 2 to 7 are translated into higher frequencies then filtered again to remove signals outside the allocated bandwidth. All seven channels are combined and amplified before being sent down the line or radio transmitter. As an example, on channel No.2 voice frequencies of 300 to 3400Hz are shifted by 12,000Hz into the 12,300Hz to 15,400Hz range. Whereas channel No.7 the same voice frequencies are shifted by 32,000Hz into the 32,300 to 35,400Hz range.
At the distant end the seven channels have to be separated by a reversal of the process. The diagram above (B-End equipment) shows the demultiplexer arrangement. The combined signal enters from the left and is separated into the individual channel frequency ranges. Channels 2 to 7 are translated and shifted back to the original frequency range that entered the system. For example channel No.7 comprising of 32,300 to 35,400Hz are returned to 300 to 3400Hz range. Channel No.1, just passes through the system without being translated. Quite often, this untranslated audio channel, is used as an Engineering Order Wire (EOW), an orderwire is a circuit that engineers use to speak to their colleagues between multiplex stations for fault finding and circuit alignment purposes.
As speech circuits are bi-directional, there will also be a multiplexer at the B-End and a corresponding de-multiplexer at the A-End for the return speech path. Steps have to be taken to ensure both ends of the multiplexed / demultiplexed circuit shift each speech channel by the same amount otherwise frequency distortion will occur. Even a small difference of a few Hertz is noticeable, so a high stability master oscillator is used to derive the channel translating frequencies, supplied to the modulators. It is usual to send a pilot signal which is used to automatically adjust the gain of the amplifiers and to detect a circuit breakdown.
Practical Uses in Communications
As telephone lines are bidirectional 2-Wire circuits they must be converted to 4-Wire circuits having a separate pair of wires for transmit and another pair for receive. For signalling between the two ends for dialling or calling the switchboard operator, there will be signalling converters too (not shown). The signalling converter changes the direct current signalling into a tone within the speech band for sending over the multiplex and this is converted back to direct current by another converter at the distant end. In the case of GPO / BT circuits this may be 2280Hz or the older standard of 500 / 20Hz.
Signalling conversion may be included within the multiplex equipment, but instead of the signalling tone being within the speech bandwidth, it is transmitted as an internationally agreed out-of-band tone frequency of 3825Hz. The automatic exchange or switchboard uses E&M direct current signalling, on an additional pair of wires to the multiplex equipment.
To create a long distance circuit a number of hops may be joined end to end. The 4-Wire circuit emerging from one Mux/Demux is cross connected to another Mux/Demux going off in another direction. This applies to both radio and line based systems. The conversion from four to two wires only taking place at the terminal ends.
Prior to modernisation to a fully digital network in 1993 the use of frequency division multiplexed speech was widespread throughout the BT network. 24 circuits per pair of twisted wires were widespread on specially designed cables. All the Go direction pairs in one cable and the Return direction in another. Some special cables were used to carry 60 channels per pair of wires. Coaxial cable has a much greater bandwidth than twisted copper wires so this was used to carry large scale multiplexed lines between telephone exchanges in major towns. The most common form of multiplex was 960 circuits carried on a pair of coaxial tubes operating at 4 MHz, with one cable containing many pairs of tubes. This was later extended to 12 MHz giving 2,700 circuits. The long haul microwave network used multiplexed channels too.
On systems multiplexing hundreds of circuits, it would be very costly to build a single high capacity multiplex. Instead, 12 circuits were multiplexed into a 'Group'. Five Groups entered a translator to produce a Supergroup of 60 circuits. These were further translated, combining 16 Super-Groups into a Hypergroup of 960 circuits transmitted on 4 MHz coaxial cables. This could be taken further, translating three Hypergroups for transmission along a 12 MHz coaxial line system. The huge number of multiplexers was referred to as the 'Mux Mountain'. Each end of a 960 channel system would require 80 Group multiplex, 16 Supergroup translators and one Hypergroup translator, 97 in all. For a 12 MHz coaxial system this would be a whopping 291 per end.
UHF Radio Multiplex
In the UHF radio spectrum used by the GCN during the sixties and seventies, the bandwidth was limited by the spacing of the individual radio channels at 25kHz intervals, restricting it to 1+6 multiplexes using thermionic valves.
The RN1 and RN2 modernisation of the eighties introduced the 1600MHz bands to the network. These bands had wider channel separation allowing more speech circuits to be multiplexed together. Typical types of multiplexer used on RN1 and RN2 are the Pye L700 and PRD1100. These could be arranged to multiplex 8, 12, 24 and 36 circuits.
Individual multiplexed circuits are cross connected at radio nodes as required to provide longer distance connections. Smaller capacity Multiplex would typically be used on spurs with the larger sizes providing a spine network. Economics dictates that a non-multiplexed single circuit per radio channel would be used to feed sites requiring only a few links. The diagram below is a hypothetical case to illustrate the principle involved but does not represent any real life example.
Something like this arrangement was used in the ECN radio network spine RN1 and RN2 known also as the 'Highway Network'.