GB2043402A - Improvements in or relating to methods of, and apparatus for, demodulating frequency-division multiplex (FDM) signals - Google Patents

Abstract

A method of, and apparatus for, demodulating FDM signals in which the input FDM signal is time compressed followed by conversion into time-division multiplex (TDM) format enabling high speed serial demodulation to be effected. Time-compression is achieved by means of a buffer store 2 into which samples of the input signal are written at a first rate R1 and then cyclically read out at a second faster rate R2. The contents read from the buffer store during each reading cycle occupy a separate time slot of the TDM signal which has a number of channels corresponding to the number of sub-channels of the FDM signal from which information is to be derived. The time-division multiplexing of the time-compressed FDM signal involves down-conversion of the respective sub-channels to a common carrier frequency, after which the TDM signal can be serially demodulated by a common demodulation unit 7. The output of the demodulation unit is then de-multiplexed at 8 to produce a plurality of parallel output signals each containing data from a respective sub-channel of the original FDM signal. The technique obviates the need for separate frequency selection and demodulation circuits for each channel resulting in a substantial economy of hardware. The apparatus described can be programmed by appropriate selection of certain parameters to deal with a variety of signal formats having different numbers of channels, baud rates, carrier frequency spacings or bandwidths. <IMAGE>

Description

SPECIFICATION Improvements in or relating to methods of, and apparatus for, demodulating frequency-division multiplex signals This invention relates to methods of, and apparatus for demodulating frequency-division multiplex (FDM) signals.

Frequency division multiplex uses a separate subcarrier for each channel of information with spaced subcarrier frequencies. The subcarriers are modulated in accordance with the information signal carried by each channel, and are linearly added into a single multiplex signal which is transmitted along a single path. On reception of the transmitted multiplex signal, the individual modulated subcarrier frequencies are segregated by frequency selection, and then individually demodulated to extract the original information signals.

In many digital communications systems, such as telegraphy systems, the original information signals each comprise a digitally encoded sequence of 0's and l's, or 'spaces' and 'marks', and these are used to amplitude-, frequency-, or phase-modulate the respective subcarriers in accordance with any one of a number of well established modes such as frequency shift keying (FSK) and phase shift keying (PSK).

As discussed above, in order to extract the original information signals from the transmitted FDM signal, the receiving station must first segregate the individual subcarriers before they can be demodulated. This process of segregation followed by demodulation of the individual channel signals will hereinafter be referred to simply as demodulation of the multiplex signal.

Conventional apparatus for carrying out this demodulation of FDM signals usually comprises a number of parallel frequency selective circuits or band-pass filters, one for each channel, each arranged to segregate a respective one of the channel signals from the multiplex signal, followed by a corresponding number of parallel demodulator circuits arranged to demodulate the individual outputs from the filters to thereby reproduce the information signals in their original digitally encoded form.

This parallel demodulation process thus requires a large amount of hardware employing separate frequency-selective and demodulator circuits for each channel.

According to the present invention, a method of demodulating a frequency-division multiplex (FDM) signal to extract signal information carried by two or more channels of the signal, comprising: time-compressing the FDM signal by writing it into a buffer store at a first rate and then reading the contents of the store at a second, faster rate in a cyclic manner such that any given portion of the input signal is read from the store at least once for each of the channels from which signal information is to be extracted; converting the output of the buffer store into time-division multiplex (TDM) format so that signal information contained in the time-compressed portion of the input signal derived from each cyclic reading of the store contents occupies a separate time slot ofthe TDM signal, and each said time slot contains only signal information from one channel of the input FDM signal;serially demodulating the TDM signal; and demultiplexing the demodulated TDM signal.

Conveniently samples of the input FDM signal are continuously written into the buffer store at a first sample rate R1, and are cyclically read at a second sample rate R2 so that the FDM signal is timecompressed by a factor R2/R1 = F 3 N, the number of channels from which signal information is to be extracted. F will be referred to as the timecompression factor.

The cyclic reading of the buffer store contents may be effected by cyclically reading the same set of F stored sample values representing a given portion of the input signal, F times, before updating this set with a new set of samples representing the next successive portion of the input signal. The output from the store for each portion of the input signal would thus comprise F identical sets of F samples of that portion of the signal, each time-compressed by a factor F relative to the corresponding set of input samples. However, this method would require a buffer store having a storage capacity of at least 2F samples.

Preferably therefore the set of F stored samples read from the store during successive reading cycles is updated by one new sample of the input signal for each reading cycle, so that during the course of F reading cycles, the set of samples is completely replaced by a new set of samples of the input signal.

This method of time compression may be implemented using a buffer store having a storage capacity of only F samples.

In the application of the invention to the demodulation of telegraphic FDM signals in which there is no synchronisation between the telegraphic elements of the individual channels, the time compression factor F in such an application is preferably set, in relation to the length of the cyclically read portions of the input signal, such that each channel of the TDM signal contains a plurality of time slots for each telegraphic element of the corresponding channel of the input FDM signal. Effectively this means that the condition or mark/space state of each telegraphic element of each channel is sampled a plurality of times, preferably more than six times, thereby reducing the timing ambiguity in such data bit decisions (multiple-point sampling).

In such applications, the conversion of the timecompressed FDM signal into TDM format may be achieved by mixing it with an N-channel frequencystepping reference signal the inter-channel switching rate of which is synchronised with the cyclic reading rate of the buffer store, and each channel of which contains a signal at a frequency corresponding to the time-compressed sub-carrier frequency of a respective channel of the time-compressed FDM siganal.

The invention may also be applied to the demodulation of FDM signals, eg multi-tone phase-shift keyed (PSK) signals, in which the telegraphic ele ments of individual sub-channels are synchronised with one another. In such applications, the demod ulation of the time-compressed input signal in its TDM format is preferably synchronised with the individual telegraphic elements of each channel, thus enabling unambiguous data bit decisions (marWspace state) to be made on the basis of a single sampling of each telegraphic element in each sub-channel.For non-differential PSK signals, the time compression factor F may be made equal to the number of sub-channels from which information is to be extracted, while in the case of differential phase-shift keyed (DPSK) signals, the time compression factor must be made equal to at least twice the number of sub-channels, and the cyclically read portion of the store contents must correspond in length to at least two telegraphic elements or bits. This latter requirement arises from the necessary differential demodulation of DPSK signals, in which the phase of the telegraphic element on which a mark/space decision is being made, is compared with the phase of the preceding telegraphic element to determine whether a transition has occurred.

The invention also extends to apparatus for carrying out the demodulation of FDM signals by a method as aforesaid, the apparatus comprising buffer storage means into which the input FDM signal is written at a first rate, and read therefrom at a second, faster rate in a cyclic manner such that any given portion of the input signal is read from the store at least once for each of the channels of the input signal from which signal information is to be extracted; time division multiplexing means for converting the output signal from the buffer storage means into time-division multiplex (TDM) format in which signal information contained in the portion of the output signal derived from each cyclic reading of the contents of the buffer storage means occupies a separate time slot of the TDM signal, and each said time slot contains only signal information from one channel of the input FDM signal; demodulation means for serially demodulating successive time slots of the TDM signal; and demultiplexing means for converting the demodulated TDM signal from the demodulation means into a plurality of parallel signals each containing signal information from a respective one of the channels of the input FDM signal.

The invention will now be described in greater detail, by way of example only, with reference to the accompanying drawings, in which Figure 1 is a block schematic diagram of one form of demodulator apparatus in accordance with the present invention; Figures 2 and 3 show in more detail different parts of the apparatus shown in Figure 1; Figure 4 is a block schematic of a second form of demodulator apparatus in accordance with the present invention; and Figures 5 and 6 shown in more detail different parts of the apparatus shown in Figure 4.

Referring to the drawings Figure 1 shows apparatus in accordance with the invention for demodulating frequency-division multiplex voice frequency telegraph (FDM-VFT) signals.

Most present day telegraph systems employ a train of current pulses (marks) and no-current intervals (spaces) to encode each of the letters, figures or symbols to be transmitted. Each mark or space is referred to as a telegraphic element, and the number of telegraphic elements transmitted per second in any channel is known as the baud rate.

Frequency division multiplex (FDM) is a technique by which a plurality of telegraphic channels can be transmitted in parallel along a common path. The information in each channel modulates a separate subcarrier frequency of the FDM signal, the subcarrier frequencies being spaced apart at nonoverlapping frequency intervals. A FDM-VFT signal is one in which the su b-carrier frequencies all lie within the voice frequency band, nominally 300 to 3400 Hz.

The apparatus shown in Figure 1 comprises an analogue-to-digital converter 1 which converts samples of the input FDM-VFT signal into digital words at a speech-band sample rate R1. The output of the analogue-to-digital-converter 1 is written into a buffer store 2 at the input signal rate R1, and is cyclically read out at a faster sample rate R2, via a high speed digital-to-analogue converter 3. The ratio between read rate R2 and the write rate R1 of the buffer store 2 will be referred to as the time-compression factor, F, and is controlled by a time-compression control unit 4.

The output of the digital-to-analogue converter3 consisting of a time-compressed representation of the FDM-VFT input signal is converted into timedivision multiplex (TDM) format by a multiplexer unit 5 controlled in synchronism with the read-out cycle of the buffer store 2 by a timing signal provided along path 6 by the time compression control unit 4.

Each channel of the TDM signal corresponds to a channel of the original FDM signal.

The TDM signal emerging from the multiplexer 5 is demodulated in conventional manner by a highspeed demodulator7 which samples the state (mark or space) of each time slot of the TDM signal, and is also controlled in synchronism with the buffer store read-out cycle by the timing signal along path 6.

The output of the demodulator is then applied to a time-division demultiplexer 8 which separates the demodulated TDM signal into a plurality of parallel output streams each containing the information from a respective one of the channels of the original FDM signal at the original signalling rate.

The construction and operation of the above form of apparatus, adapted to demodulate a 24-channel 50 baud frequency-shift-keyed (FSK) FDM-VFT signal, using a 60Hz frequency slot scheme, the lowersubcarrierfrequencyft being420Hz(7 x 60Hz) the highest frequency f,, being 3180 Hz (53 x 60Hz), and having a tone-shift frequency of +30Hz. with respect to the subcarrier centre frequency, will now be described in more detail, with reference to Figures 1,2 and 3.

In Fig 1, the analogue-to-digital converter 1 converts the analogue FVT input signal into digital 8-bit word samples at a sampling rate R1 of 8kHz (being above the Nyquist rate of the input signal) controlled by write pulses from the time compression control unit 4. The same write pulses control the rate at which these samples are continuously written into the buffer store 2, which comprises a 256 x 8-bit word bi-po!ar TTL random access memory of standard commercially available form.

The samples stored in the buffer store 2 are cyclically read into the digital-to-analogue converter 3 at a rate R2 of 1.92 MHz controlled by read pulses from the time compression control unit 4, this representing a time-compression factor F = R21R1 of 240. Thus 240 samples are read out of the buffer store for each new sample written in.

The manner in which the buffer store is controlled will now be described with reference to Figure 2 which shows the time-compression control unit 4 in greater detail. The unit includes a clock pulse generators which produces clock pulses at the buffer store read-out rate R2 of 1.92 MHz.

These pulses are applied directly to the digital-toanalogue converter 3 and an 8-bit read address register 10 to control the read-out rate of the buffer store 2; and via a divide-by-F circuit 11 to the analogue-to-digital converter 1 and an 8-bit write address register 12 to control the write-in rate of the buffer store. The output of the read address register 10 and the write address register 12 each comprise an 8-bit digital word which represents respectively the read-out and write-in addresses (1 to 256) of the buffer store. The write-in address is cyclically advanced at a rate of 8kHz, while the read-out address is cyclically advanced 240 times as fast, at 1.92 kHz.Thus, the 240-sample output batch from the buffer store produced by each successive readout cycle, representing a time-compressed portion of the input signal, is updated by only one new sample value, so that complete updating of the 240 signal samples contained in these batches takes place over a period of 240 cycles.

The time-compression control unit4 also includes an 8-bit digital comparator circuit 14 which compares the instantaneous read-out and write-in address codes produced by the respective address registers 10,12, and produces a timing pulse along path 6 each time the two address location numbers coincide. These timing pulses are used to synchroniso the later stages of the demodulation process with the beginning of each read-out cycle of the buffer store 2.

The signal applied to the time-division multiplexer unit 5 thus comprises a serial stream of data batches each representing a time-compressed portion of the original input signal 1.5 telegraphic elements in length, with successive batches containing 239/240ths of signal information in common with its adjacent batches.

The TDM unit 5 is shown in greater detail in Figure 3 and comprises a master oscillator 20, and a bank of 24 phase-lock loop frequency-synthesised oscillator circuits 21, (only one of which is shown) each producing from the master oscillator frequency, an output signal frequency corresponding to the centre frequency of a respective one of the 24 timecompressed telegraphic channels of the original TDM-VFT signal.

The master oscillator frequency is selected to be a common factor of all the centre frequencies of the time-compressed VFT signal channels. In the present example, since the baseband sub-carrier frequencies are all multiples of half the channel spacing, ie 120/2 = 60Hz, then the centre frequencies of the timecompressed signal channels will all be multiples of Fx 60 = 14.4kHz. The master oscillator frequency may thus be set at 14.4 kHz, with the option of applying the output of the master oscillator 12 either directly to the oscillator circuits 21, or via a divideby-two circuit26 underthe control of a reference select gate 27, giving a choice of two reference frequencies 14.4kHz and 7.2kHz.

Each of the phase-lock loop frequency synthesized oscillator circuits 21 is of conventional form and comprises a phase comparator 30, a loop filter 31, a voltage controlled oscillator 32 and a frequency division circuit 33. The reference frequency derived from the master oscillator 20 is applied to one input of the phase comparator, the other input of which receives a feedback signal representing the output frequency of the VCO 32 divided by a factor D in division circuit 33. The phase comparator compares these two input frequencies and produces a DC output voltage representing the frequency difference between them which is applied via the loop filter 31 to control the VCO output frequency. The output frequency of the VCO is thus stabilised at a value equal to the reference frequency multiplied by the division factor D of the division circuit 33.

The output frequency of each of the 24 frequency synthesised oscillator circuits 21 may thus be set at the time-compressed centre frequency of a respective one of the 24-channels of the time-compressed VFT signal by appropriate selection of the respective division factors D, to D24. These time-compressed centre frequencies are equal to the centre frequencies (f1 to f24) of the original VFT signal multiplied by the time-compression factor F which is equal to 240 in this case. Thus the lowesttime-compressed subcarrier frequency or centre frequency is Fxf1 = 240 x 420 Hz = 100800 Hz and the highest timecompressed centre frequency is Fxf24 = 240 x 3180Hz = 763.2kHz.

To produce the lowest time-compressed centre frequency from the master oscillator frequency of 14.4kHz, the oscillator circuit 21 associated with this channel must multiply the master oscillator frequency by a factor Fxf1114400 = 100800/14400 = 7, so that the division factor D1 for the division circuit 33 of this oscillator circuit must be set at7. Similarly, the division factor D2 for the oscillator circuit associated with the channel having the next highest frequency is set at 9 and soon up to D24 = 53 forthe highest time compressed centre frequency.

The 24 output signals from the oscillator circuits 21 are applied to respective inputs of a 24-channel time division multiplexing module 35 which timedivision multiplexes the 24 input signals from the oscillator circuits in synchronism with the timing pulses from the time compression control unit 4 under the control of a cyclic 1-24 channel-selection counter 34. Each of the channels of the TDM signal emerging from the module 35 is thus occupied by a different one of the 24 time-compressed centre frequencies produced by the oscillator circuits 21, the time-slots of the signal coinciding in time with successive ones of the data batches emerging from the digital-to-analogue converter 3.

The TDM signal from the TDM module 35 is then applied to one input of a double-balanced modulator 36, the other output of which receives the output signal from the digital-to-analogue converter 3.

Thus, successive ones of each group of 24 data batches of time-compressed VFT signal are mixed with a respective one of the 24 time-compressed centre frequencies produced by the oscillator circuits 21, thereby converting the time compressed VFT signal into a 24 channel time-division multiplex signal, each channel of which contains only signal information from a respective one of the channels of the original FDM-VFT input signal.

To remove undesirable harmonics, the TDM output signal of the balanced modulator 36 is then passed through a band-pass filter 38 having a bandwidth of 20kHz, corresponding approximately to the individual bandwidth (80Hz) of the original VFT input signal, multiplied by the time compression factor F.

The filtered output of the time-division multiplexer unit 5, which essentially consists of the original FDM-VFT input signal in TDM format is then applied to the high-speed demodulator 7 which samples the frequency of each element of the TDM input signal, under the synchronising control of the timing pulses from the time-compression control unit 4, to determine its state (mark or space). The demodulator employs in known manner a coincidence-gate to determine whether the instantaneous signal fre quency is above or below the TDM signal centre fre quency which is equal to the centre frequency of the band-pass filter 30 to produce a DC output indicating either a '1' or a '0' as appropriate.Because of the discontinuities in the input to the demodulator caused by the cyclic reading of the buffer store contents, and the settling time of the band-pass filter 38 at each switching of the time-division multiplexer unit 5, a settling time of 33% of the buffer store cyclic reading period is allowed before each element of the TDM signal is sampled. It is for this reason that the duration of each portion of the input signal 'captured' in each cyclically read 240-sample batch of the store content is arranged to correspond in length to 1.5 telegraphic elements at the 50 baud signalling rate.

The demodulated TDM signal is then applied to the demultiplexer unit8 to convert, in conventional manner, the demodulated samples of each telegraphic channel into a respective one of 24 parallel output streams, each at a sample rate of 62/3 times the original signal baud rate of 50. The apparatus thus in effect samples each telegraphic element of each channel on average 62/3 times, this multiplepoint sampling being necessary to remove ambiguities in the telegraphic data bit decisions because of the lack of synchronism between the telegraphic elements of individual channels.It will be appreciated that if the telegraphic elements of the individual channels were synchronised with one another, the TDM demodulation could be synchronised with the telegraphic elements of all channels enabling data bit decisions to be made on the basis of only one output sample per telegraphic element in each of the parallel output streams.

Each of these parallel output streams containing an average of 62/3 pulses per telegraphic bit of the original input signal, may then be applied to a respective sample-and hold circuit (not shown) to convert it back into a telegraphic signal at the original baud rate suitable for operating a teleprinter or like apparatus.

In general the time compression factor F, applied to the input signal is determined by three considerations, namely the number of subchannels (N) in the input FDM signal to be demodulated, the number of samples (n) per telegraphic element (multiple point sampling) required for unambiguous data bit decisions, and the number of telegraphic elements per channel (d) that must be captured in each timecompressed batch of the FDM signal to allow for the settling time of the time-division multiplexer unit 5.

Thus F = Nnd, nd usually being set constant.

The apparatus may be adapted to handle a wide range of standard forms of FDM-VFT signals having different numbers of channels, different channel spacings and different baud rates by providing a number of modifications based on the following analysis.

Consider a VFT signal having the following characteristics: Number of channels = N Channel frequency spacing = fsp Baud rate of telegraphic data = fk As discussed above, where the telegraphic data in each of the individual channels is asynchronous, then multiple point sampling of each telegraphic element is necessary, this being represented by the number of samples per telegraphic element (n) appearing in each of the parallel output streams from the de-multiplexer unit 8.

Thus, the sample rate of each of these output streams is nfk and the time-division multiplex rate of the TDM unit 5 is Nnfk. The read-out rate R2 of the buffer store 2 is then L.Nnfk, where L is the number of cyclically read samples of the store content (240 samples in the above embodiment).

The duration of the signal contained in the L samples of the store content = UR 1 = d/fk, and so the time-compression factor F = R2/R1 = L.NnfklLfk.'/d = Nnd.

Normally nd will be set constant (10 in the present embodiment) as will the input signal sampling rate R1 (8kHz in the present embodiment) and so the time compression factor must be varied in accordance with the number of channels in the VFT signal to be demodulated. The only parameters of the time-compression stage of the apparatus, comprising the analogue-to-digital converter 1, the buffer store 2, the digital-to-analogue converter3 and the time compression control unit4, that require adjustment to cater for input signals having different numbers of channels N, and different baud rates fk are the frequency of the clock pulse generator 9 and the division factor of the divide-by-F circuit 11, these being adjusted in such a way as to produce the required write-in rate R1 of 8kHz.

The other parameters which require adjustment to accommodate different forms of VFT signal are the output frequencies of the oscillator circuits 21 and the cyclic count of the channel selection counter 34, the latter being set to the appropriate number of channels N.

Examination of the standard FDM-VFT types of signals, eg types R31, R35, R37, R38A, R39 recommended by CCITT, shows that the centre frequencies of the channels in each signal are all multiples of half the characteristic channel spacing of the signal type, being either even or odd multiples of the frequency spacing, or using alternate values of odd multiples of one half the channel frequency spacing. Thus by providing a choice of master oscillators 20 having frequencies of 14.4kHz and 13.6kHz, the same set of oscillator circuits 21 can be used for all of the above standard VFT types by appropriate adjustment of the division factors D, of the division circuits 33 in combination with the selection of the appropriate master oscillator frequency and appropriate switching-in, or-out, of the divide-by-two circuit 26.

Conveniently, the apparatus may be controlled by a programmable control unit in which the above parameters for the various VFT types of signal are stored and switched in automatically upon selection of the appropriate mode.

Although the invention has been described in its application to the demodulation of FSK-FDM-VFT signals, the principle may equally be applied to other forms of FDM signal, such as amplitude-modulated or phase-modulated FDM signals.

In this connection a suitable form of apparatus in accordance with the invention for demodulating PSK-FDM signals will now be described with reference to Figures 4 to 6 of the accompanying drawings. A characteristic of PSK signals is that the telegraph elements ofthe individual channels are synchronised with one another, the mark-space condition of each element being indicated by a phase-shift (usually 1800) in the carrier frequency, rather than a frequency-shift as in the case of FSK signals. As discussed earlier, because the individual data channels are synchronised with one another, multiple-point sampling is not required to make unambiguous data bit decisions because the time-division multiplexing of the FDM-PSK input signal can be synchronised with the individual telegraphic elements.The apparatus is adapted to demodulate a differential phase-shift-keyed (DPSK) signal, ie in which the encoder produces, in known manner, a transition in the input data sequence for a '0' input, and no transition in the signal sequence for a '1' input (or vice versa).

Referring now to figure 4 of the drawings, the demodulator apparatus comprises an analogue-todigital converter 40, a buffer store 41 and a timecompression control unit 42 which are substantially the same as, and perform the same functions as the corresponding components of the Figure 1 apparatus, ie conversion of the input FDM signal into time-compressed digital form. The digital output of the buffer store is then applied to a digital timedivision multiplexer (TDM) and demodulation unit 44 which performs substantially the same function as the TDM unit 5 and demodulator unit 7 of the Figure 1 apparatus, converting the time-compressed FDM signal into demodulated TDM format which is then demultiplexed in a demultiplexer unit 46.The multiplex rate of the TDM and demodulation unit 44 is controlled by timing pulses from the timecompression control unit 42 applied along path 45, these pulses being generated in the same manner as the timing signal applied along path 6 in Figure 1, while the demodulation of the TDM signal is synchronised with the baud rate of the time-compressed signal from the buffer store underthe control of a baud synchroniser 47 and a timing logic circuit 48.

Synchronisation of the demodulation process with the baud rate in this way enables the timecompression factor F applied by the timecompression control unit 42 to be reduced to 2N, ie twice the number of channels in the FDM signal, the factor of two arising from the differential operation of the demodulator as will be explained below.

The buffer store length must correspond to the length of at least two telegraphic elements, ie it must contain at least two bits of the signal at any time in order that their relative phases with respect to the sub-channel frequency can be determined by the demodulator.

In all other respects the considerations affecting the choice of parameters in the time compression control unit 42 are substantially the same as those for the time-compression control unit4 of Figure 1.

The baud rate synchroniser 47 is shown in more detail in the block schematic diagram of Figure 5, and comprises a digital shift register delay 50, a subtraction (or addition) circuit 51, a peak amplitude detector 52, a threshold level detector 54 and a phase-lock-loop circuit 55. The digital output samples from the buffer store 41 pass through the shift register delay 50 at the read-out rate of the buffer store underthe control of the buffer store read-out pulses from the time-compression control unit 42.

The length of the shift register 50 is such as to introduce a delay T, which is equal to 1/2Fsp, where Fsp is the frequency spacing of the subchannel tones of the input signal. The shift register is cleared at the end of each read-out cycle of the buffer store 41 by the timing pulse applied along path 45 from the timecompression control unit 42, so that at any given time the shift register delay contains samples from only one read out cycle of the buffer store.

The subtraction (or addition) circuit 51 subtracts (or adds) delayed samples of the buffer store output signal from (to) undelayed samples of this signal, so as in effect, to implement a digital filter having a transfer function which is zero at all the timecompressed subchannel tone frequencies of the DPSK FDM input signal. The frequencies at which these 'zeros' in the transfer function occur depend on the delay time T, and upon whether the circuit 51 operates in a subtraction or an addition mode.

Where the subtraction mode is used, 'zeros' in the transfer function will occur at even harmonics of the subchannel frequency spacing Fsp (where T = 1/2asp) while if the addition mode is used these 'zeros' will occur at odd harmonics of the frequency spacing.

The operation of the synchroniser is thus dependent upon a fixed frequency separation between the sub channel frequencies, and on the subchannel tone frequencies being harmonics of this frequency separation which is the case in most PSK modes recommended by CCITT.

The selection of the subtraction or addition mode will depend upon whether these tone frequencies are even or odd harmonics respectively of the channel spacing.

Thus, if there is no change in the phase of any one of the subchannel frequencies, the filter will produce a zero output; but if a 180 phase change in any one of the subchannel frequencies does occur, then the filter will produce an output signal the peak amplitude of which will depend on the number of subchannels in which such a phase change has occurred. There is thus a high probability that the amplitude level of the filter output signal will change at each telegraph element boundary.The peak amplitude detector 52 and the threshold level detector 54 are arranged to detect such changes, the peak amplitude detector operating to sample and hold the peak value of the output of the subtraction (or addition) circuit 51, while the threshold level detector effectively compares the held value with the instantaneous amplitude of the output of the circuit 51.

As the approximate data rate of any incoming signal will normally be known, the output signal from the threshold detector 54 is applied to the phase-lock loop circuit 55 which phase-locks the output of a clock signal generator 56 having a nominal frequency equal to the baud rate of the time compressed FDM signal with the data element transitions.

The operation ofthetime-division multiplexer and demodulation unit 44 will now be described with reference to Figure 6. The TDM section of the unit comprises a pair of digital multipliers 60, 61 and a read only memory (ROM) sine look-up table 62 which is arranged to produce a pair of time division multiplex digital outputs representing the sine and cosine respectively of frequencies corresponding to the subchannel tones of the time compressed FDM signal, the multiplex rate of these sine and cosine signals being controlled in synchronism with the read-out cycle of the buffer store by the timing signal from the time compression control unit 42 applied along path 45.Each of these digital multiplexed outputs from the ROM 62 is multiplied with the buffer store output samples in a respective one of the two multipliers 60,61 to produce a pair of TDM output signals, one of which contains, in each channel, the in-phase component of a respective one of the subchannels of input signal, and the other of which contains in each channel the quadrature component of a respective one of the subchannels of the input signal.

The two TDM output signals from the multiplexer section of the unit 44 are then applied to the demod ulatorsection which operates in synchronism with the time-compressed baud rate under the control of the baud synchroniser 47 and the timing logic control circuit 48. The demodulation section in essence comprises a digital correlator, and includes a separate integrator 65, 66, storage register 67, 68 and multiplier 69,70 for each of the two input TDM signals, a summation circuit 71 and a threshold detector 72.

During each time slot of the two phase quadrature TDM input signals, each of the integrators 65,66 is actuated by a signal from the timing logic control circuit 48, and towards the end of the current signalling interval (or time compressed data element) the contents of each integrator are passed into the associated storage register 67,68, again under the control of the timing logic control circuit 48.During the next succeeding signalling interval or data element contained in the same TDM time slot, each integrator is again actuated by the timing logic control unit 48, and towards the end of this interval, the contents of each storage register 67, 68 are multiplied in the respective multipliers 69,70 with the contents of the associated integrator, whereby to produce an output signal representing the correlation coefficient between the two successive telegraph elements in the same TDM time slot. The two outputs from the multiplexer 69,70 are then combined in the summation circuit 71, the output of which is applied to the threshold detector 72, to provide an indication as to whether or not the adjacent telegraph elements are in phase with one another.The technique described is in principle the same as conventional DPSK signal detection techniques with special adaptation for dealing with the signal in TDM format.

The above demodulation cycle is repeated on each time slot of the phase quadrature TDM signals from the TDM section of the unit, it being noted that in successive TDM time slots associated with the same subchannel, the second data element of one time slot appears as the first element of the next succeeding time slot.

The output of the TDM-demodulation unit 44 thus represents the time-compressed FDM signal in demodulated TDM format, and this is applied to the demultiplexer unit 46 operating in synchronism with the buffer store read-out cycle under the control of the timing signal applied along path 45. The demultiplexer unit converts the demodulated TDM signal into a plurality of parallel output signals each containing the signal information from a respective one of the subchannels of the original input DPSK FDM signal.

It will be appreciated that although the invention has been described in its particular application to the demodulation of FDM telegraphic signals, it may also be applied with similar advantage to the demodulation of FDM signals in which the carrier frequencies are modulated by other forms of data signal.

Claims (24)

1. A method of demodulating a frequencydivision multiplex (FDM) signal to extract signal information carried by two or more sub-channels of the signal, comprises: time-compressing the FDM signal by writing it into a buffer store at a first rate and then reading it out at a second, faster rate in a cyclic manner such that any given portion of the input signal is read from the store at least once for each of the sub-channels from which signal information is to be extracted; converting the output signal from the buffer store into time-division multiplex (TDM) format so that signal information contained in the time-compressed portion of the input signal derived from each cyclic reading of the store contents occupies a separate time slot of the TDM signal, and each said time slot contains only signal information from one sub-channel of the input FDM signal; serially demodulating the TDM signal; and demultiplexing the demodulated TDM signal.

2. A method as claimed in Claim 1, wherein samples of the input FDM signal are continuously written into the buffer store at a first sample rate R1, and are cyclically read out at a second sample rate R2 so that the FDM signal is effectively time-compressed by a time-compression factor F = R2/R1 > N, N being the number of sub-channels from which signal information is to be extracted.

3. A method as claimed in Claim 2, wherein F samples of the input FDM signal are read from the buffer store during each reading cycle, and the set of F samples read from the store during successive cycles is updated by one new sample of the input signal for each reading cycle, so that during the course of F reading cycles, the set of F samples is completely replaced by a new set of samples of the input signal.

4. A method as claimed in any one of Claims 1 to 3, for demodulating telegraphic FDM signals in which there is no synchronisation between the telegraphic elements of the individual sub-channels, wherein the time-compression factor applied to the input FDM signal is set, in relation to the length of the portions of the input signal cyclically read from the buffer store, such that each sub-channel of the TDM signal contains a plurality of time slots for each telegraphic element of the corresponding subchannel of the input FDM signal.

5. A method as claimed in Claim 4, wherein each sub-channel of the demodulated TDM signal contains more than six time slots for each telegraphic element of the corresponding sub-channel of the input FDM signal.

6. A method as claimed in Claim 4 or Claim 5, wherein conversion of the time-compressed FDM signal from the buffer-store into TDM format is effected by mixing it with a multi-channel frequency-stepping reference signal the interchannel switching rate of which is synchronised with the cyclic reading rate of the buffer store, and each channel of which contains a signal at a respective frequency corresponding to the time-compressed sub-carrier frequency of a respective one of the sub-channels of the time-compressed FDM input signal read from the buffer store.

7. A method as claimed in any one of Claims 1,2 or 3, for demodulating telegraphic FDM signals in which the telegraphic elements of individual subchannels are synchronised with one another, wherein demodulation of the time-compressed input signal in its TDM format is performed in synch ron- ism with the common timing of the individual telegraphic elements of the sub-channels, each telegraphic element of each sub-channel from which information isto be extracted being sampled only once to determine its mark/space state.

8. A method as claimed in Claim 7, for demodulating differential phase-shift keyed (DPSK) signals, wherein the input signal is time-compressed by a factor substantially equal to twice the number of sub-channels from which information is to be extracted, the cyclically read portion of the buffer store contents corresponding in length to at least two telegraphic elements of the input signal, so that each time slot of the time-compressed input signal in TDM format, prior to demodulation, contains a pair of consecutive telegraphic elements from a respective sub-channel, and the second telegraphic element of the pair in any time slot is the same as the first element of the pair in the next successive time slot associated with the same sub-channel, and wherein demodulation of the time-compressed TDM signal is performed by comparing the relative phases of the two telegraphic elements of each pair within each time slot of the TDM signal whereby to determine the mark/space state of the second ele mentofthe pair.

9. A method as claimed in Claim 7 or Claim 8, for demodulating telegraphic FDM signals having a fixed frequency separation between the individual sub-channel frequencies, and the sub-channel frequencies of which are all harmonics of this fixed frequency separation, wherein the baud rate of the time-compressed FDM input signal is determined by applying it to a filter having a transfer function which is zero at all the sub-channel frequencies of the time-compressed input signal whereby to produce an output signal the amplitude of which varies at each telegraphic element boundary, and applying the filter output signal to produce a pulsed timing signal synchronised with the timing of the telegraphic element boundaries for controlling said demodulation of the time-compressed input signal in its TDM format.

10. A method of demodulating a FDM signal substantially as hereinbefore described with reference to Figures 1 to 3, or to Figure 4 to 6 of the accompanying drawings.

11. Apparatus for demodulating a frequency divison multiplex signal to extract signal information carried by two or more sub-channels thereof comprising; buffer storage means; means for writing the input FDM signal into the buffer storage means at a first rate, and for reading it out at a second faster rate in a cyclic manner such that any given portion of the input signal is read from the store at least once for each of the sub-channels of the input signal from which signal information is to be extracted; time division multiplexing means for converting the time-compressed output signal from the buffer storage means into time-division multiplex (TDM) format in which signal information contained in the portion of the output signal derived from each cyclic reading of the contents of the buffer storage means occupies a separate time slot of the TDM signal, and each time slot contains only signal information from one channel of the input FDM signal; demodulation means for serially demodulating successive time slots oftheTDM signal; and demultiplexing means for converting the demodulated TDM signal from the demodulation means into a plurality of parallel out put signals each containing signal information from a respective one of the channels of the input FDM signal.

12. Apparatus as claimed in Claim 11, wherein the means for writing the input signal into the buffer storage means and for reading it therefrom is arranged to periodically sample the input FDM signal and to continuously write these samples into the buffer storage means at a first sample rate R1 and to cyclically read them from the buffer store at a second rate R2, so that the FDM signal is timecompressed by a factor F = R2/R1 3 N, where N is the number of sub-channels from which signal information is to be extracted.

13. Apparatus as claimed in Claim 12, wherein the buffer storage means is capable of holding a set of at least F samples of the input signal, and the means for reading and writing the samples into and out of the buffer storage means is arranged to read a set of F stored samples from the buffer storage means during each reading cycle, and to update the set of F stored samples read from the buffer storage means during successive reading cycles by one new sample for each reading cycle, so that during the course of F reading cycles, the set of samples is completely replaced by a new set of samples of the input signal.

14. Apparatus as claimed in Claim 11,12 or 13, for demodulating telegraphic FDM signals in which there is no synchronisation between the telegraphic elements of the individual sub-channels, wherein the time-compression factor, being the ratio of the read-out rate to the write-in rate of the buffer store, is set in relation to the length of the cyclically read portions of the input signal, such that each channel of the time-compressed FDM signal in TDM format contains a plurality of time slots for each telegraphic element of the corresponding sub-channel of the input FDM signal, so that during serial demodulation of the TDM signal, the mark/space state of each telegraphic element of each sub-channel from which information is to be extracted is sampled a plurality of times.

15. Apparatus as claimed in Claim 14, wherein the time-compression factor is set such that each telegraphic element of each sub-channel of the FDM signal is sampled more than six times to determine its mark/space state.

16. Apparatus as claimed in any one of Claims 11 to 15, wherein the time division multiplexing means comprises means for generating a multi-channel frequency-stepping reference signal, the interchannel switching rate of which is synchronised with the cyclic reading rate of the buffer storage means, and each channel of which contains a signal at a respective frequency corresponding to the time compressed sub-carrier frequency of a respective one of the sub-channels of the time-compressed FDM input signal, and means for mixing this frequency-stepping reference signal with the time compressed FDM input signal read from the buffer storage means.

17. Apparatus as claimed in Claim 16, including means for varying the time-compression factor applied to the input FDM signal in accordance with the number of sub-channels from which signal information is to be extracted, and means for varying the number of channels, and the reference signal frequencies of the respective channels, of the frequency-stepping reference signal in accordance with variations in the time-compression factor applied to the input signal and with differences in the parameters of the input FDM signal.

18. Apparatus as claimed in Claim 16 or Claim 17, wherein each of the reference signal frequencies of the frequency-stepping reference signal is generated from a common frequency source by a respective phase-lock loop frequency synthesiser each comprising a phase comparator connected to receive at one input a master frequency signal from the common frequency source, and the output of the phase comparator being connected to control the frequency of a voltage-controlled oscillator (VCO), the output of which is connected via a feed-back loop containing a divider ci rcuit to the other input of the phase comparator, whereby the reference output signal produced by the VCO is equal to the master frequency multiplied by the dividion factor of the divider circuit.

19. Apparatus according to Claim 18, wherein the division factor of the divider circuit of each of the frequency synthesisers is variable to control the respective reference frequencies of the frequencystepping reference signal.

20. Apparatus as claimed in Claim 19, wherein the reference frequency output signal from each of the frequency synthesisers is applied to a timedivision multiplexer unit arranged to time division multiplex a pre-selected number of the reference frequency output signals in synchronism with the reading cycle of the buffer storage means to produce said multi-channel frequency-stepping reference signal, the selected number of channels being variable in dependence upon the number of subchannels of the FDM input signal from which signal information is to be extracted.

21. Apparatus as claimed in any one of Claims 11 to 13, for demodulating telegraphic FDM signals in which the telegraphic elements of individual subchannels are synchronised with one another, wherein the operation of the demodulation means for serially demodulating the time-compressed FDM signal in TDM format is synchronised with the common timing of the individual telegraphic elements of the sub-channels, whereby each telegraphic element of each sub-channel from which information is to be extracted is sampled only once to determine its mark/space state.

22. Apparatus as claimed in Claim 21, for demodulating differential phase shift-keyed (DPSK) signals, wherein the write-in and read-out rates of the buffer storage means are selected such that the input signal is time-compressed by a factor substantially equal to twice the number of sub-channels from which information is to be extracted, the cyclically read portion of the buffer store contents being set to correspond in length to at least two telegraphic elements of the input signal, so that, in use, each time slot of the time-compressed TDM signal, prior to demodulation, contains a pair of consecutive telegraphic elements from a respective sub-channel, and the second telegraphic element of each pair in any time slot is the same as the first telegraphic element of the pair contained in the next successive time slot associated with the same sub-channel, and wherein the demodulation means is arranged to compare the relative phases of the two telegraphic elements of each pair contained in each time slot of the TDM signal, whereby to determine the mark/space state of the second telegraphic element of the pair.

23. Apparatus as claimed in Claim 21 or 22, for demodulating telegraphic FDM signals having a fixed frequency separation between the individual sub-channels frequencies, and the sub-channel frequencies of which are all harmonics of this fixed frequency separation, including a filter connected to receive the time-compressed FDM signal read from the buffer store, the filter having a transfer function which is zero at all the sub-channel frequencies of the time-compressed FDM signal, such that in use, the amplitude of the output of the filter varies at each telegraphic element boundary of the timecompressed FDM signal, the apparatus further including means arranged to apply the output of the filter to generate a pulsed timing signal synchronised with the timing of the telegraphic element boundaries, for synchronising the operation of the demodulation means with the baud rate of the time-compressed FDM signal.

24. Apparatus for demodulating FDM signals substantially as shown in and as hereinbefore described with reference to Figures 1 to 3, orto Fig urns 4 to 6 of the accompanying drawings.