GB2187365A - Pulse density modulation fibre-optic system - Google Patents

Abstract

In a multi-channel fibre optic transmission system in which a plurality of analogue channels are each pulse density modulated (PDM) and then time division multiplexed, channel identity is conferred by superimposing an out-of-band pilot tone on one analogue channel. At the receiver the pilot tone is filtered out in the relevant decoded analogue channel and used to effect channel synchronisation of the demultiplexed channels. <IMAGE>

Description

SPECIFICATION Pulse density modulation fibre-optic system This invention relates to a multi-channel fibre-optic transmission system suitable for video links.

Currently British Telecom is utilising the so-called "Switched Star" system providing 4-channel video links. However it is desirable to provide an improved system with an increased number of channels, the benefits of which include reduced costs (equipment, installation and maintenance) and/or improved range of services for subscribers.

Several techniques have been considered for increasing the number of video channels per optical fibre. These have included a 6 or 8-channel version of the existing 'FM' system. Other modulation schemes such as VSB have been considered including semi-digital techniques such as PFM. While many modulation schemes can increase the number of video channels per fibre they suffer from difficulties with fibre optic transmission or cost disadvantages in achieving simple interfacing with switches, especially at shorter distances. Our investigations have concluded that true digital transmission of video signals over an optical fibre will provide the highest number of channels per fibre. However, conventional methods of encoding and multiplexing the digital signals would not appear to provide a satisfactory cost solution.

The standard method of encoding video signals digitally normally requires a 7 or 8-bit encoding process plus an additional bit for word synchronisation. It would be likely, therefore, that up to 140 Mbit/s rate per video channel would be required to meet Switched Star System requirements. The encoders and decoders required are also relatively expensive and power consuming, disadvantages which will continue for some time. Further, standard methods of multiplexing a number of channels require a frame word with corresponding logic circuitry of some complexity.

Finally, the optical transmission system would require suitable line coding circuitry and a corresponding increase in the line transmission rate. In conclusion, an 8-channel system of this design would require a line transmission rate over 1.2 GBit/s. it would also be comparatively large, expensive and power consuming and therefore not an appropriate choice for the Switched Star System.

According to the present invention there is provided a multi-channel fibre-optic transmission system including for each channel respective pulse density modulation (PDM) analogue-to-digital encoding and decoding means, means for multiplexing and demultiplexing a plurality of PDM channels, means for converting the multiplexed PDM channels into optical signals for transmission via an optical fibre, means for reconverting the transmitted optical signals to electrical signals for demultiplexing, and means for inserting in at least one of the channels channel identity signals whereby the demultiplexing means is synchronised with the multiplexing means, characterised in that the channel identity signals comprise a 'pilot tone' analogue signal superimposed on the analogue input to the encoding means of the one channel, said pilot tone having a frequency outside the signal baseband of the channel, the decoding means for the one channel including pilot tone extraction means and channel synchronising means dependent thereon to generate demultiplexing control signals.

Embodiments of the invention will now be described with reference to the accompanying drawings, in which: Figure 1 illustrates the principle of PDM coding and decoding, Figure 2 illustrates signal-to-noise (S/N) characteristics of PDM compared with PCM, Figure 3 illustrates an example of a multiplexed PDM system for television, Figure 4 illustrates the application of a multiplexed PDM system in a switched star system, Figure 5 illustrates a PDM encoder in block diagram form, Figure 6 illustrates a PDM decoder in block diagram form.

Figure 7 illustrates a typical optical fibre implementation of a multiplexed PDM system, The principle of PDM can be understood by referring to Fig. 1. The decoder regenerates the pulse in D flip-flop 10 and then low pass filters it in L1, C2 and C3 so producing a baseband signal which is then amplified to a normal TV level.

Pulse Density Modulation is seen to have a running average which matches the baseband TV signal it represents. This can be generated at the encoder by means of a clocked comparator (e.g. a D flip-flop 11) whose input consists of the difference between the signal being encoded and the running average of its own digital output. The low pass filter R1, C1 which determines this average has to be simple to ensure loop stability, whereas the decoder filter L1, C2, C3 can cut off much more sharply.

The higher the loop gain of the encoder the more correct will be the code, and therefore the lower the noise content when decoded by L1, C2, C3. In order to meet the Bode stability criterion for feedback systems the feedback should not be rolled-off faster than 12 dB/octave up to half the bit rate (e.g. to 50 MHz). The higher the bit rate the more feedback can be applied in a given baseband and the better the signal/noise that should be achieved. Increasing the bit rate also spreads the quantising noise over a wider bandwidth improving the signal by a further 3 dB/octave.

In practice, there are always around the feedback loop due in part to the finite width of practical pulses and this makes the achievement of 15 dB/octave unlikely to be realised. In this context it is assumed that 12 dB/octave is feasible. Another practical problem is that the pulses ahead of the low-pass filters must have an area which is completely independent of the pulse pattern. In our implementation this has required the introduction of pulse shaping buffers at the positions marked 'X'.

A useful comparison between PDM and PCM codecs is their base-bandwidth and signal/noise behaviour as shown in Fig. 2. A PCM code is often implemented at 13.3 MHz (3 times colour sub-carrier) and its base-bandwidth is therefore limited to 13.3/2=6.65 MHz. The line rate will be 8x 13.3=106.4 Mbit/s for a 7 bit code plus one frame bit. It will have a flat noise spectrum up to the 6 MHz cut-off frequency of the filter which also limits the baseband to approximately 6 MHz. This gives Signal and Noise spectra as Fig. 2(b).

The PDM coder, by contrast, need only have a gentle roll-off beyond 6 MHz so that signals can be sent up to say 10 MHz with lower quality. The noise spectrum generally rises with frequency at up to 12 dB/octave consistent with Bode's stability criterion, so that much better noise is achieved in the luminance band (0 to 1.5 MHz say) and worse noise from 5 to 10 MHz.

This gives PDM a triangular noise spectrum like FM. It is nevertheless possible to send FM or FSK modulation subcarriers above 6 MHz. The worst TV performance will generally be found in the chrominance band. This however can be improved by providing some extra feedback in this band. Indeed, it is possible to shape the noise spectrum of a PDM modulator to the least subjectively disturbing spectrum by shaping the loop again in the baseband as in Fig. 2(a). These benefits in terms- of weighted signal/noise amount to approximately 10 dB at 100 Mbit/s data rate This is roughly equivalent to 2 more bits PCM word length and can provide superior S/N for PDM encoding at this rate.

The behaviour of the systems in the presence of errors is also different. In PCM an error is likely to disrupt a very significant bit. In PDM no bit is more significant than any other. Thus errors are less disturbing. This independence from bit significance can also ease the switching of digital TV. PCM has to be reframed if switching (or re-routing) disturbs the frame position but PDM, which has no frame, is free from this problem.

PDM tends to have a soft overload characteristic, which eliminates the need to provide clamping at its input. It is generally operated at approximately 5 dB below gross overload, and this implies that long runs of like symbols do not occur. The code therefore ensures that clock regeneration will not be a problem under normal conditions.

A useful feature of PDM is its ease of interfacing to VSB equipment. A PDM encoded signal in return-to-zero format is equivalent to an amplitude modulated signal with the clock frequency as carrier. It can readily be converted to VSB by means of a filter, and can then be frequency translated elsewhere if desired. Conversion to VSB of the received 100 Mbit/s signal using a bandpass filter 12 is also shown in Fig. 1.

The comparative characteristics of PDM and PCM are summarised in Table 1.

PCM PDM For PDM TV Weighted Signal/Noise Good Better Framing System Essential None Affected by Switching Fragile Rugged Clock Content Erratic High Conversion to VSB Elaborate Simple Filter Quality Required Elaborate Simple Power Consumption (Serialised) 2.5W 0.7W (Typical) Subjective Effect of Error Intolerant Minor Ability to Carry Data or Sound Subcarriers Poor Good Teletext or MAC Signals Questionable Good Against Product Available Yes No Standards Emerging None? TABLE 1 COMPARISON OF VIDEO ENCODING TECHNIQUES An example of how PD can be used to carry several digital TV signals over a large (optical fibre) network into local cable TV networks is shown in Fig. 3. Up to 24 channels ought to be feasible on single mode fibre. It is assumed that the channels are combined in a simple synchronous time domain multiplex 30. At least one of the channels is flagged with some form of identification 'ID'. This can be a baseband pilot in the frequency domain which can be identified on reception by means of a ceramic filter 31. Its purpose is to ensure that the TDM demultiplexer 32 is in the correct phase to deliver the channels to the correct outputs. Other methods of inserting the necessary identification are also available.

The demultiplexed PDM streams can be low-pass filtered to a baseband signal for switching at baseband (to interface with, for example, the existing Switched Star System). An alternative option is to use SAW filters 33 to produce VSB signals which are then translated to an FDM multiplexer f, to fn by n conventional frequency changers 34 for connection to a 'tree and branch' co-axial cable TV network (perhaps employing tuner switches) as shown in Fig. 3.

The most likely initial use for PDM (digital) transmission in the Switched Star System is for primary links 40 from the Head End 41 through Hub Sites 42 to the Wideband Switch Points (WSP's) 43 as shown in Fig. 4. This can provide cost and operational advantages in its own right. However, PDM can be integrated further into the system to provide further advantages.

This would allow the whole system to become digital and provide a low cost interface at subscriber's premises.

Plan 'A' is the current analogue system employing the 4-channel links, a baseband video switch and coaxial drops to subscribers. Plan 'B' shows the introduction of PDM into the primary links where the digital signals can be distributed to the WSP's in a similar manner to the existing 4-channel systems. For interim systems the PDM signals are decoded to baseband signals at the WSP to interface with the switch. This and the secondary link to the subscribers would remain unchanged.

Although plan 'B' provides a significant upgrading to the Switched Star System itself, there are nevertheless further improvements to be gained by allowing the digital signals to penetrate further to the subscriber. These are shown, in steps, as plans C to E.

In plan 'C' the 12 PDM channels are transmitted to the WSP's as before. Instead of being decoded down to baseband signals the 12 channels are demultiplexed into the 12 100 Mbit/s (PDM encoded) streams. These 100 Mbit/s streams are then passed through the switch in lieu of the conventional baseband signals. The current switch itself is expected to be able to handle the digital signals without causing digital errors although special line filtering and repeatering would be required. The transmitting of digital signals through the switch will overcome any reduction in signal quality to analogue signals due to crosstalk.

Plan 'D' is an extension of 'C' and provides for the PDM encoded digital signals to be transmitted over the secondary link directly to the subscriber. Here, the 2 or 4 channels would be demultiplexed into the 100 Mbit/s streams. These would be VSB filtered and each upconverted into the UHF band for direct connection to subscribers' TV sets. The number of PDM channels may be varied according to the transmission needs of the secondary link. The present coaxial links are expected to support at least 2 PDM channels.

Plan 'E' is the final example of PDM penetration into the Switched Star System. Here the PDM channels at the subscribers' premises are decoded into baseband signals for connection to television sets which have baseband inputs. The transmission system will then have full digital transmission and would await a direct digital interface to television sets (when available).

The PDM encoder basic block diagram is given in Fig. 5. It includes a decision circuit consisting of a 'D' flip-flop 50 and a pulse shaping circuit 51 producing return to zero (RZ) pulses. These are connected in a first (inner) feedback loop with R and C as loop integrator. At the 100 Mbit/s rate the choice of flip-flop, circuit lay-out and frequency response around the loop have been critical. The pulse shaper has been developed from a 5 GHz transistor array to ensure that the pulses have equal area regardless of position in the pulse stream.

The second (outer) feedback loop consists of an amplifier 52 which takes the difference between the input signal and a feed-back signal which is partially filtered by a 'pre-emphasis' network 53. The loop gain around the outer loop is further shaped and controlled by a 'noise shaping' network 54 which together with the amplifiers, inner loop and pre-emphasis network determine the spectral distribution of quantising noise in the base-band.

We have shown that noise can be reduced in the chrominance band of a PAL TV signal by including special filtering in the 'noise shaping' network. The relative noise levels of luminance and chrominance band can also be altered by the pre-emphasis network. This simultaneously alters the signal and overload frequency responses in such a way as to match the energy, subjective noise criteria and signal handling capability of the coder. Pre-emphasis also reduces the effect the 'average picture level' (d.c. content) has on video parameters, such as differential gain and overload.

A video clamp 54 and/or clipper could be placed ahead of the coder to ensure that overload by an abnormally large signal does not cause catastrophic overload.

The decoder shown in Fig. 6 consists of a 'D' flip-flop 60 and RZ pulse shaper 61 to reproduce the RZ pulses in the coder, followed by a de-emphasis network 62 whose output matches that applied to the difference amplifier in the coder. The remainder of the decoder consists of a video low-pass filter 63 and video clamp 64 as in the 4-channel FM links.

We have shown that by passing the RZ pulses of the PDM decoder through a bandpass-filter a double sideband AM signal results. Use of a narrower band filter produces a Vestigial Sideband (VSB) signal. This is potentially of greater importance because of the widespread use of VSB for present generation television equipment, especially consumer television receivers.

The carrier level of the VSB signal can be controlled by simple digital processing of the RZ pulses ahead of the band-pass filter, and the de-emphasis can be corrected by careful design of the filter frequency response. A Surface Acoustic Wave (SAW) filter would be a convenient method of implementing such a filter.

A number of VSB channels can be frequency translated to other frequencies by conventional means to form a frequency division multiplex of channels. These could be sent on existing coaxial cable networks in tree and branch or switched systems.

Noise is normally introduced in switching as well as transmission. For example, the switch in the Switched Star System is limited by capacitive cross-talk between channels especially at the higher video frequencies. If PDM encoded video is passed through the switch the cross-talk may be removed by a regenerator in the same way that a regenerator removes transmission noise in a digital system.

It has been estimated that the amount of cross-talk capacitance present in a MOS cross-point video baseband switch will be low enough to permit satisfactory regeneration of the 100 Mbit/s PDM digital stream.

This permits the gradual evolution of television networks such as the Switched Star System so that the switch would become part of the digital network, eventually taking such digital signals to subscriber's terminals in digital form and regenerating there to maintain the original noise level of the PDM encoder.

Several PDM encoded TV channels may be multiplexed in the time domain and carried over a fibre optic system.

A laser would be used as a transmitter because of its wide modulation bandwidth. If the system is to cover substantial distances 1.3 micron single mode fibre would be used, with either a Ge or GaAsP avalanche photodiode or possibly a PIN-FET receiver. The number of channels feasible depends on the laser bandwidth. A few types could handle fifty channels, but others suffer penalties with as few as five channels. A reasonable target for proven lasers is 12 channels.

A sending terminal can therefore consist of 12 PDM video encoders 70 each running at 100 Mbit/s. These may be grouped into 3 groups of 4 channels as illustrated in Fig. 7. Each group of 4 are interleaved by a first order multiplexer 71 to form three streams of 400 Mbit/s which could be an optimum arrangement to allow using high speed versions of ECL. Finally, the three 400 Mbit/s streams are serialised in a second order multiplexer 12 to form one 1200 Mbit/s stream, using discrete device principles borrowed from microwave experience. A number of laboratories have described digital systems from 1 to 5 Gbit/s using these principles which include step recovery diodes, PIN diode switches, GaAs transistors etc.

The low frequency content of PDM encoded video signals could cause difficulty with capacitive coupling and maintenance of adequate clock content. We have shown that this can be overcome for PDM coded video signals by differential encoding. The 400 Mbit/s streams can be differentially encoded 73 to improve the line transmission in this way.

In order to identify the channels one of them has an additional signal added. With PDM a pilot 'identification' tone may easily be added above the video band in, for example, the 12th channel.

In the receive terminal a narrow band filter 74 (e.g. 10.7 MHz) is connected to the 12th channel. When the ID signal is present in the 12th channel all the channels are in the correct phases. The phase of the demultiplexed channels is established by repeatedly interrupting the frequency dividers 75,76 generating the clock structure and testing for the presence of the ID tone in channel 12. Once the ID tone is found the phase of the dividers is held. An alarm lamp can indicate the absence of the ID tone in channel 12.

The receive terminal contains an avalanche photodiode 77, optical head amplifier and limiting amplifier 78 to achieve good sensitivity and wide dynamic range. The photodiode current could be monitored as part of system supervision. If AGC instead of limiting amplifiers is used it is possible to monitor the AGC control voltage also.

The output from the limiting amplifier 78 is passed to a phase detector forming part of a phase locked loop timing recovery circuit 79. The other clock frequencies are derived by frequency division from this oscillator.

The 1200 Mbit/s stream from the limiting amplifier is applied in gates 80 at 400 MHz to form three parallel 400 Mbit/s streams which are passed to differential decoders 81. Then the three 400 Mbit/s streams are each passed to a respective group of four sampling gates 82 to give a total of 12 streams of 100 Mbit/s. Finally each 100 Mbit/s stream is fed to a PDM decoder 83 to form a video signal and suitably amplified and clamped to suit the customer's requirement.

Other combinations of channels and multiplexing arrangements are, of course, also possible based on this description.

The primary objective in developing this multiplex structure will be to achieve economies of power consumption and size to make the system competitive with multi-channel FM transmission. Other objectives include the possibility for extension into an extended digital architecture to handle other forms of traffic than video at a later date and flexible distribution of these signals.

It would be possible to use any 100 Mbit/s channel (except perhaps the 12th) for digital services other than video so long as they are made synchronous with the multiplex. ECL interfaces would be made avilable to permit this.

One such application might be for digital error monitoring which is not provided for in the simplest structure so far described. An example of this is the transmission of a parity stream sent in one channel created by modulo 2 addition (exclusive OR operation) of the other 11 channels. Other possibilities include the transmission of a standard 140 Mbit/s channel in two 100 Mbit/s tributaries by adding redundant bit(s) in such a way as to bring the total to synchronism. In a similar way many standard 2 Mbit/s streams could be combined to form a 100 Mbit/s channel.

In future systems there may be a need to send a digital signal (e.g. 2 Mbit/s) with a video signal.

One means for doing so could be to send it on carrier above the video band at approximately 7.5 MHz, as can be done on FM systems.

There is also the possibility of impressing the data into the PDM coder by modifying its structure. A variation of this principle would be to impress a known data pattern, (e.g. all marks) into the coder, and then replace these data positions with the intended data elsewhere in the transmission path. Both these possibilities make use of the capability of the PDM coder principle to compensate for corrupted bits occuring within the normal video encoding process.

Claims (13)

1. A multi-channel fibre-optic transmission system including for each channel respective pulse density modulation (PDM) analogue-to-digital encoding and decoding means, means for multiplexing and demultiplexing a plurality of PDM channels, means for converting the multiplexed PDM channels into optical signals for transmission via an optical fibre, means for reconverting the transmitted optical signals to electrical signals for demultiplexing, and means for inserting in at least one of the channels channel identity signals whereby the demultiplexing means is synchronised with the multiplexing means, characterised in that the channel identity signals comprise a 'pilot tone' analogue signal superimposed on the analogue input to the encoding means of the one channel, said pilot tone having a frequency outside the signal baseband of the channel, the decoding means for the one channel including pilot tone extraction means and channel synchronising means dependent thereon to generate demultiplexing control signals.

2. A system according to claim 1 characterised in that at least one channel PDM encoding means comprises a clocked comparator whose input consists of the difference between the analogue signal being encoded and the running average of its own digital output, said running average being determined by a low pass filter network.

3. A system according to claim 1 or 2 characterised in that each or the PDM encoding means has a loop gain in the baseband shaped to control the noise spectrum of the encoder to a predetermined shape.

4. A system according to any preceding claim characterised in that each or the PDM encoder includes a pulse shaping network in the encoder feedback path.

5. A system according to any preceding claim characterised in that each or the PDM encoder includes a further feedback path incorporating a pre-emphasis network and each corresponding decoder includes a complementary de-emphasis network.

6. A system according to any preceding claim characterised in that at least one PDM decoder comprises a clocked comparator the output of which is fed via a bandpass filter to an output amplifier stage.

7. A system according to any one of claims 1-5 characterised in that each of the demultiplexed and decoded PDM channels further includes a bandpass filter means followed by frequency translation means, the output frequency of each channel being different to that of every other channel, the system including means for combining the different output frequencies in a frequency division multiplexed channel.

8. A system according to any preceding claim characterised in that the PDM signals are differentially encoded before transmission of the multiplexed channels and the transmitted signals are differentially decoded before being PDM decoded.

9. A system according to any preceding claim characterised in that the plurality of PDM channels are first multiplexed in groups, each group multiplex being differentially encoded and the groups then being further multiplexed onto a single optical fibre transmission medium, the transmitted signals being first demultiplexed into groups, the groups being differentially decoded and then further demultiplexed into the plurality of PDM channels.

10. A system according to any preceding claim characterised in that the transmitted digital signals are, at a receiver, fed to a phase locked loop to extract clock signals, that the pilot tone extraction means comprises a bandpass filter and that the extracted pilot tone is applied to alignment means to synchronise the clock signals to effect correct channel synchronisation of the demultiplexed channels.

11. A multi-channel fibre-optic transmission system substantially as described with reference to the accompanying drawings.

12. A multi-channel fibre-optic transmission system including for each channel respective pulse density modulation (PDM) analogue-to-digital encoding and decoding means, means for multiplexing and demultiplexing a plurality of PDM channels, means for converting the multiplexed PDM channels into optical signals for transmission via an optical fibre, means for reconverting the transmitted optical signals to electrical signals for demultiplexing, and means for inserting in at least one of the channels channel identity signals whereby the demultiplexing means is synchronised with the multiplexing means, characterised in that the PDM signals are differentially encoded before transmission of the multiplexed channels and the transmitted signals are differentially decoded before being PDM decoded.

13. A multi-channel fibre-optic transmission system including a plurality of channels wherein some at least of the channels each include respective pulse density modulation (PDM) analogueto-digital encoding and decoding means and at least one other channel carries digital data signals, wherein all the channels are synchronous, means for time division multiplexing the PDM and digital data channels, means for converting the multiplexed channels into optical signals for transmission via an optical fibre, means for reconverting the transmitted optical signals into electrical signals for demultiplexing, and means for inserting in at least one of the PDM channels channel identity signals whereby the demultiplexing means is synchronised with the multiplexing means.