Optical FDM system.

This invention relates to a frequency division multiplex (FDM) system, especially for use where the transmission medium is an optical fibre cable.

In such a system difficulties can arise when according to conventional techniques filters are used. These difficulties are due to the relatively narrow bandwidth of the optical filters needed.

It is the object of the invention to provide a frequency division multiplex (FDM) system which avoids the difficulties caused by the use of filters.

This object is attained in the way as set forth in claim 1. Further embodiments of the invention are set forth in the subclaims.

An embodiment of the invention will now be described with reference to the drawings, in which : <ul id="ul0001" list-style="none"><li>Fig. 1 is a diagram explanatory of the frequency spectrum used, Fig. 2 is a multiplexer/demultiplexer arrangement for a first F_DM system embodying the invention, while Fig. 3 is an arrangement embodying the invention in which there is an element of switching.</li></ul>

The principle on which the system is based is now described with reference to the frequency allocation diagram, illustrated in Fig. 1. The channels to be switched are modulated with double sidebands onto carrier frequencies f₁ f₂ ..... f_n and are of bandwidth _Df. Between each pair of adjacent channels is an unused section of the frequency spectrum, as indicated in Equation 1.<maths id="math0001"><img file="EP0131818A2_D0001.tif" /></maths>

The local oscillator frequencies f_lo' f_2o' etc. are so allocated as to lie in these gaps above or below the signal frequencies at a frequency difference f_IF therefrom. They are so chosen that<maths id="math0002"><img file="EP0131818A2_D0002.tif" /></maths>To ensure that f does not produce a signal at f_IF the condition must be met that :<maths id="math0003"><img file="EP0131818A2_D0003.tif" /></maths>From (1), (2) and (3), it will be seen that f_S is greater than 2f_IF and the channel spacing is given by<maths id="math0004"><img file="EP0131818A2_D0004.tif" /></maths>With a tolerance of +T on the frequency stability of each channel oscillator and with the spectral line widths of the oscillators S, the channel spacing must be increased to

^C_s⁼ 2f_IF + 2Df + 2T + 2S (5) _We assume that some form of frequency control on the local oscillator is used to lock the difference frequency constant, so it is not necessary to consider a tolerance on the local oscillator frequencies. The spectral line width is assumed to be the same as for the channel oscillator.

The minimum f_IF gives the minimum channel spacing, and the minimum f is 2Df, so that :<maths id="math0005"><img file="EP0131818A2_D0005.tif" /></maths>

With double side band modulation at least half the available channel bandwidth is wasted. Single side band modulation if usable could double the number of channels in the available spectrum but it is not clear how single side band modulation can be achieved. Double side band modulation as assumed above implies amplitude modulation of the light source. Such "Amplitude" Modulation of the guided wave is in fact a phase modulation process and for small modulation depths the double sidebands will extend many times beyond a range equal to the bandwidth of the modulating signal. This has been discussed in "Topics in Applied Physics - Integrated Optics., by J.M. Hammer (ed.T. Tamir); Vol.7, Springer-Verlag, 1982, pp 139-200. However, for large modulation depths, the phase modulation process becomes in effect an amplitude modulation one, because on extinction energy is lost by radiation, reflection or deflection from the straight-through path down the waveguide. The spectrum of the resultant amplitude modulated signal should therefore be confined largely to a frequency range equal to plus and minus its bandwidth. Regeneration of the digital signal after the receiver will in any case remove cross-talk components.

The same assumptions cannot however be made for analogue signals which demand linear modulation processes and considerably increased channel spacings are needed to accommodate the phase modulation spectrum. However, for a given type of information, digitisation normally increases bandwidth requirements so that the increased channel spacings are offset by the narrower inherent bandwidth of the information signal.

One problem with a heterodyning arrangement that requires switching (as opposed to demultiplexing) is that the local oscillator has to be able to be tuned over the complete frequency range. Semiconductor lasers currently available are unsuitable for this because their frequency of operation is related to their geometry.

To avoid this problem we use one master optical source for the entire switching or multiplexing/demultiplexing system and synthesise the required frequencies from this master oscillator using acousto-optic interactions. This has the added advantage that if the master optical source drifts in frequency all the other frequencies drift in sympathy with it, so the factor _T in equation (5) can be set to zero. By making one high quality optical source supply, one should be able to afford a relatively expensive oscillator with good spectral line width, thus minimising S.

Figs. 2 and 3 show arrangements based on the above discussed principles, _Fig. 2 for the simpler multiplexing/demultiplexing case and _Fig. 3 for a switching arrangement in which the incoming and outgoing channels do not bear a fixed relationship to each other. For illustration, we assume a specific example of five 70Mb/s digital _TV channels, an intermediate frequency of 160 MHz, and a channel spacing of 500 MHz.

In Figs. 2 and 3 the transmission (multiplexing) ends are identical.

The master oscillator output is split, with half the energy or more going to the receive end and the remainder to the transmit end. The transmit signal is fed into a splitter RNM 1 working in the Raman Nath (R_N) regime, for which see the reference quoted above. Also of interest is "Interaction Between Light and Sound" by R. Adler, IEEE Spectrum, V4, No5, pp42-54, May, 1967. This modulator also has an input via which a 500 MHz signal is applied. The phase excursion or the modulation index of the device RNM1 are such as to make the carrier and higher order sidebands have the same amplitude. For the five channel arrangement shown this means that the carrier, the first order and second order sidebands corresponding to Bessel function J_o, J₁ and J₂ must have roughly equal amplitudes. A modulation index of about 3 radians meets this criterion. When frequency modulating the input the device RNM1 also splits it spacially into five outputs of frequencies f , f_o + or - f_m and f₀⁺ or - 2f_m. Each of these outputs is then modulated in one of m the channel modulators CM1-CM5 by one of the five _TV signals, and they are then comined in a combiner _CB, whose output is applied to an optical fibre cable. Combination can be by optical means (collimation and focussing) if the functions are on separate chips or, if they are in waveguides on the same chip, by one of the standard waveguide coupling techniques.

Thus finally the combined output is transmitted along a fibre or guide to the receive end. The receive end part of the master oscillators output is also transmitted along another guide or fibre to the receive end to act as the local oscillator source.

In Fig. 2 for the multiplex/demultiplex requirement, the receive end contains a Bragg modulator BM device which shifts the local oscillator signal by an amount equal to the IF frequency, upwards in the present case. The signal is then split by another Raman Nath Modulator RNM2 into constituent frequencies; ^fo ^{+ f}_IF' ^fo ⁺ f_IF + or - f and f_o + f_IF + or - 2f

The information signal is split on reception into five equal channels using a splitter SP1 which works on a waveguide decoupling technique. Once split these channels are respectively combined with the relevant local oscillator outputs from the device RNM2. This combination can be on the chip or directly at the photodetector. Mixing takes place in the photodetector and the resultant electrical signal is filtered and amplified to give five separate _TV channels.

In Fig. 3 for the switching case it is necessary for each of the five local oscillator frequencies to be potentially available for any channel. To achieve this the receiver local oscillator output is split by a simple optical splitter OS into five equal parts, and each part is fed to a separate Bragg BMl to BM5 device. Each Bragg device can be acoustically modulated with signals at the IF frequency, at the IF frequency + or -f , or at the IF frequency + or -2f , as required. Such modulation needs several acoustic transducers only one of which is energised at any one time, and results in different Bragg angles at each frequency. The output is then collimated and focussed before application to the mixer. From that point on, operation is identical to the multiplexer case.

Although the above has taken a digitised _TV signal as an example, the principle is equally applicable to an analogue _TV signal; the major difference is that the channel bandwidth would be smaller (say 5MHz) but the channel spacing greater (say 100MHz) in terms of multiples of bandwidth. The resultant designs are not therefore greatly dissimilar from what has been described. Thus by an apparently profligate use of the optical spectrum and filtering by electrical means the short comings of filtering at optical frequencies can be overcome.

We have assumed that the application for such a system is to the switching and/or multiplexing of wide band video channels. However, it is also possible to extend the system to the 30 channels said to be needed for CATV.