EP1000478A1

EP1000478A1 - Optical data processing using electroabsorption modulators

Info

Publication number: EP1000478A1
Application number: EP98935210A
Authority: EP
Inventors: David Graham Moodie; Andrew John Gloag; Andrew David Ellis
Original assignee: British Telecommunications PLC
Current assignee: British Telecommunications PLC
Priority date: 1997-07-28
Filing date: 1998-07-24
Publication date: 2000-05-17
Also published as: AU8455898A; CA2295825A1; GB9715936D0; WO1999005812A1

Abstract

A method of processing an optical data stream modulated with a return to zero formatis described which comprises splitting the optical data stream into first and second data streams; coupling the first data stream into a first input port of an electroabsorption modulator (EAM); processing the first data stream in the EAM; coupling the processed first data stream out of a first output port of the EAM; coupling the second data stream into a second input port of the EAM; and processing the second data stream in the EAM. OTDM transmission systems and nodes for such systems are described in which an electroabsorption modulator is used as a gating element for optical data passing in opposite directions through the modulator. Demultiplexing, Drop and Insert, clock extraction/synchronization are all suggested as options for the process accomplished by the modulator.

Description

OPTICAL DATA PROCESS ING US ING ELECTROABSORBTION MODULATORS

The present invention relates to optical communications and in particular, but not exclusively, to optical time division multiplexing (OTDM) . To fulfil the demand for increasingly sophisticated and bandwidth intensive services, telecommunications operators will require reconfigurable optical transmission networks which operate at a higher data rate than those installed to date. It appears likely that data rates substantially higher than 10Gbιt/s will need to be used at least in the core networks. It is currently believed that in place of conventional "electronic" multiplexing, optical wavelength division (WDM) or time division (OTDM) multiplexing techniques will be used to achieve such higher data rates. Both multiplexing techniques could be used in conjunction with switching in either of the wavelength or time domains to allow increased network flexibility through, for example, the drop-and-insert function. For moderate line rates requiring only a small number of wavelengths, WDM provides the simplest implementation of optical multiplexing. However, the provision of very high capacity links using WDM would require a large number of closely-spaced channels. With large scale networks this implies the use of dispersion-shifted transmission fibre to minimise dispersion penalties. It has been reported that the operation of WDM networks with more than a few channels over fibre with a low dispersion can lead to significant system degradation due to four-wave mixing and other non-linearities, even over fibre spans of less than 50 kilometres. An alternative, longer term approach to ultra-high-speed transmission is to use OTDM, requiring a transmitter configuration based on short (approximately 1 pico second) optical pulses, implying the use of return-to-zero (RZ) format optical data. Again, it is necessary to use dispersion-shifted fibre as the transmission medium to minimise linear dispersion penalties, or to use some other form of dispersion compensation such as mid-span spectral inversion, or soliton transmission can be used over non- dispersion shifted fibre. In a typical OTDM system n individual x Gbit/s optical data streams are interleaved to form an n by x Gbit/s data stream. The multiplexing in OTDM systems is sequential in that single bits from each channel are interleaved in the same order, as opposed to packetised data, where blocks of bits are interleaved sequentially. In OTDM networks the optical pulses produced by any transmitter are significantly shorter than the bit period at the highest multiplexed line rate in the network. This means that modulated data streams may be interleaved and demultiplexed with negligible cross talk between channels. A generalised OTDM network node consists of a set of building blocks: traffic enters the node through a demultiplexing unit, which drops out every jth bit. The dropped out channel may be detected at the network node or may be transported further to a remote location. The remaining channels of data are forwarded to the multiplexing unit, where one or more channels may be inserted to replace the dropped optical data streams. The inserted channel may originate from a transmitter local to the node or from a remote location. Clock recovery is necessary to provide a signal to drive both the demultiplexer and sychronisation unit to ensure that inserted channels go into a vacant timeslot in the multiplexed data stream.

The present invention is particularly concerned with clock recovery, synchronisation, demultiplexing and multiplexing, wherever these functions are carried out.

According to a first aspect, the present invention provides a method of processing an optical data stream modulated with a return to zero format, the method comprising: (i) splitting the optical data stream into first and second data streams; (u) coupling said first data stream into a first input port of an electro absorption modulator (EAM); (in) processing said first data stream in the EAM; (iv) coupling the processed first data stream out of a first output port of the EAM;(v) coupling said second data stream into a second input port of the EAM; and (vi) processing said second data stream in the EAM.

According to a second aspect, the invention provides a method of processing an optical time division multiplexed (OTDM) data stream, the method compπsιng:(ι)splιttιng the OTDM data stream into first and second data streams;(ιι)coup!ιng said first data stream into a first input port of an electro absorption modulator (EAM);(ιιι) processing said first data stream in the EAM;(ιv)couplιng the processed first data stream out of a first output port of the EAM;(v) coupling said second data stream into a second input port of the EAM;(vι)processιng said second data stream in the EAM; and (vn) coupling the processed second data stream out of a second output port of the EAM According to a third aspect, the present invention provides a method of processing an optical data stream modulated with a return to zero format, the method comprising: (i) coupling said first data stream into a first input port of an electro absorption modulator (EAM); (n) processing said optical data stream in the EAM; (11)1 coupling the processed optical data stream out of a first output port of the EAM;(ιv) splitting the processed optical data stream into first and second data streams;(v) coupling said second data stream into a second input port of the EAM;and (vi) processing said second data stream in the EAM. According to a further aspect, the invention provides a node for an OTDM transmission system in which an electroabsorption modulator is used as a gating element for optical data passing in opposite directions through the modulator.

According to a further aspect, the present invention provides use of an electroabsorption modulator as a gating element in the processing of optical signals, the arrangement being such that optical signals are passed bidirectionally through the modulator.

Embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings in which:

Fig. 1 shows schematically an experimental set up of a two channel demultiplexer;

Fig. 2 shows schematically an experimental set up for simultaneous clock recovery and demultiplexing;

Fig. 3 shows schematically an experimental set up of a three-node OTDM network, demonstrating "drop and insert" multiplexing; Fig. 4a shows a 40Gbιt/s OTDM input data stream used in conjunction with the Fig. 1 arrangement;

Fig. 4b shows a 10Gbιt/s demultiplexed channel, channel 1 , from port A of the arrangement shown in Fig. 1 ;

Fig. 4c shows another demultiplexed channel, channel 3, from port B of the arrangement shown in Fig. 1 ;

Fig. 5 shows the bit error rate of the two 10Gbιt/s demultiplexed channels shown in Figs 4b and 4c;

Fig. 6a shows the 40Gbιt/s OTDM input data stream used with the Fig. 2 arrangement; Fig. 6b shows a 1 0Gbιt/s demultiplexed channel, channel 1 , from port A of the Fig. 2 arrangement;

Fig. 6c shows the 10Ghz recovered electrical clock from the Fig. 2 arrangement; Fig. 6d shows the RF spectrum of the recovered clock from the Figure 2 arrangement;

Fig.7 shows the bit error rate of the 10 Gbit/s demultiplexed channel with and without clock recovery;

Fig. 8a shows another 40Gbιt per second OTDM input data stream used in the Fig. 3 arrangement;

Fig. 8b shows a transmission window, indicating that channel 3 is to be dropped, from the Fig. 3 arrangement;

Fig. 8c shows a 40Gbιt/s OTDM data stream with channel 3 dropped in accordance with the Fig. 3 arrangement; Fig. 8d shows a 40Gbιt/s OTDM data stream with channel 3 reinserted according to the Fig. 3 arrangement;

Fig. 8e shows a 10Gbιt/s demultiplexed data stream as present at the port D in the Fig. 3 arrangement;

Fig. 9 shows the bit error rate for each 10Gbιt/s demultiplexed channel from the Fig. 3 arrangement;

Fig. 10 shows schematically an arrangement to perform a simultaneous channel drop and photodetection of the dropped channel; and

Fig. 1 1 shows schematically an arrangement for demultiplexing several channels of a bitstream and for simultaneously photodetecting the dropped channels.

First Embodiment

In OTDM systems it is known to use electro absorption modulators for demultiplexing. Electro absorption modulators (EAM) have the advantage of being compact devices which can be driven by simple, low voltage sinusoidal drive electronics.

For demultiplexing applications the transmission window must be narrow enough and the extinction ratio high enough to select the required channel and to sufficiently reject any remaining channels. A 10 Gbit/s channel can be selected from a 40 Gbit/s data stream, by smusoidally driving an EA modulator at the channel clock frequency of 10 GHz to create the desired transmission window

(typically ^" 20 ps). For full demultiplexing, we would normally use a separate EA modulator to select each channel, therefore requiring 4 modulators for a 40- 1 0 Gbιt/s system, or 8 modulators for a 80-10 Gbit/s OTDM system, see, for example, the paper by Marcenac et al, "80 Gbit/s OTDM using electroabsorption modulators", ECOC 1997. In this embodiment we show that two independent 10

Gbit/s channels can be demultiplexed simultaneously from a 40 Gbit/s OTDM data stream, using a single EA modulator. This will improve network reliability and simplify its management, whilst also representing a substantial cost saving.

Experimental Details

The experimental set-up is shown in Figure 1 . A 10 GHz pulse stream was generated using a CW-DFB laser, with a wavelength of 1 557 nm, incident on an EA modulator. The modulator was harmonically driven, enabling the duty cycle to be reduced, producing 5.5 ps pulses, as described in the paper by Marcenac et al, "80 Gbit/s OTDM using electroabsorption modulators", ECOC 1 997. The output was then amplified and coupled into a second EA-modulator, allowing the extinction ratio to be increased to greater than 45 dB. Dispersion compensating

2 fibre was used to remove residual chirp, producing sech transform limited pulses, compressing the pulse width to 3.7 ps. A 2 -1 pseudo-random data stream was applied, using a lithium niobate modulator downstream of the second EA - modulator, the word length of which was limited by the low frequency response of the lithium niobate (LιNb0 ) amplitude modulator. The resulting 10 Gbit/s data stream was passively multiplexed up to 40 Gbit/s, before being passed through a polaπser to equalise each OTDM channel. Of course in practice the OTDM data stream would carry real data , could be generated quite differently, and could be at a higher or lower bit rate. In particular it could contain more ( or possibly fewer) channels. For example, the OTDM data stream could be 10 channel, with a bit rate of 100Gbιt/s .

At the input to the bi-directional electro-absorption modulator the 40 Gbit/s data stream, which is shown in Figure 4, was split using a 3 dB coupler 10 , 50% being fed in a clockwise direction through the EA modulator 1 2 , and detected at port A. The remaining 50% was fed in an anti-clockwise direction back through the same EA modulator, and monitored at port B Isolators 14 and 1 6 were used to prevent any back reflections from creating interference effects. Alternatively a 3-port circulator could be used in place of the second coupler 1 7, decreasing the overall loss. A high modulation depth packaged EA modulator of the type described in the paper by Moodie et al, "discrete electroabsorption modulators with enhanced modulation depth", Journal of Lightwave Technology, 1 996, 14(9), pp2035-2043, was used, which had a fibre to fibre insertion loss of 7.2 dB and a 3 dB electrical bandwidth of ^" 14 GHz. The device was sinusoidaily driven through bias- T 1 8 at 10 GHz from source 20, to produce a 22.5 ps approximately rectangular transmission window, enabling a single 10 Gbιt/s channel to be demultiplexed at port A The demultiplexed channel is shown in Figure 4b. Demultiplexing of a second 10 Gbit/s channel was simultaneously achieved at port B (Figure 4c), by adjusting the phase of the 40 Gbit/s signal fed in the anti-clockwise direction using the variable fibre stretcher 22.

Each facet of the EA-modulator had an anti-reflective coating, and a measured residual reflectivity of -30.5 dB. This was due to uncoated lens-ended fibre pigtails, and resulted in incoherent interference, as discussed by Gimlett et al in Journal of Lightwave Technology, 1 898, 7(6), pp888-895, between the reflected and required signals, causing an additional crosstalk penalty at each of the output ports. This interference effect can be minimised using the polarisation controllers, by ensuring that the reflected wave is orthogonally polarised to the demultiplexed signal.

Results

Figure 5 shows the results of BER measurements made on the two 10 Gbit/s demultiplexed channels. A 10 Gbit/s optimised data signal is used as a back-to-back reference of the receiver sensitivity. With the polarisation states set to minimise incoherent interference, no additional penalty was observed when compared to the same modulator operating as a traditional uni-direction demultiplexer. This was true for both ports A and B. With the signals aligned for maximum incoherent interference the system still operated error free, albeit with a small (0.8 dB) penalty. Conclusions The example of this embodiment has shown that simultaneous demultiplexing of two 1 0 Gbit/s channels from a 40 Gbit/s OTDM data stream is posssible by using a single EA modulator, here bi-directionally. By appropriate arrangement of signal polarisation the system operated with no penalty on both demultiplexed channels. Even at the point of maximum incoherent interference the system still operated error free, with less than 1 dB penalty. This technique therefore represents a straightforward method of improving network reliability, whilst also representing a substantial cost saving, by reducing the required number of modulators by up to a factor of 2.

Although the arrangement shown in Figure 1 has the advantage of a low device count and extreme simplicity, other configurations could be used. As already indicated, the coupler 1 7 could be replaced with a 3-port circulator. Coupler 1 0 could have a split of other than 50/50.

The embodiment has been described in the context of a fibre based system. It would be possible, albeit generally not desirable, to construct using bulk / free space optics. Subject in particular to the achievement of sufficient delay, the system could be realised using an integrated optical approach, in indium phosphide for example, perhaps in combination with an optical fibre delay line.

The illustrated embodiment splits the data stream 8 before the fist pass through the modulator 1 2. While this approach has many advantages, it is not essential to configure the system in this way. It would be feasible to pass the datastream 8 through the modulator 1 2 before splitting it and returning the split part back through the modulator. For this approach to be worthwhile, the initial pass through the modulator would select several channels ( which in unreturned form would provide output 22) from which a further selection of one or more channels would be made on the return pass, the selected channel or channels forming output 24. Second Embodiment

High-speed demultiplexing and clock recovery will be essential parts of future optical time division multiplexed (OTDM) networks. Clock recovery is necessary for synchronisation of the demultiplexing and detection processes Low jitter between the incoming data stream and the recovered clock is necessary to prevent degradation of the bit error rate (BER) performance of the system. Many approaches to this problem have been investigated. All-optical clock recovery methods typically involve the modulation of a laser cavity by the transmitted data stream to achieve mode-locking Another technique is that of injection locking in which an incoming data stream seeds a self-pulsing semiconductor laser. Whilst many of these methods are promising, the most favourable in terms of flexibility and stability are those based on the electronic or optical phase locked loops (PLL). PLL techniques rely on a phase detector to determine the frequency mismatch between the incoming data stream and a locally generated clock, and appropriate optical phase detection methods include, loop mirrors, gain modulation and four- wave-mixing (FWM) within semiconductor optical amplifiers (SOAs) . Recently the latter achieved outstanding performance by recovering a 6.3 GHz clock from a 400 Gbit/s OTDM data stream, as described by Kamatani et al, IEEE Photonic Technology letters 1 996, 8 (8) pp 1094-1096 .

In order to separate out the base-rate channels from an OTDM data stream, some form of active demultiplexing is required. A wide variety of optical demultiplexers have been demonstrated, one of the most promising of which is the electroabsorption (EA) modulator. Traditionally, clock recovery and demultiplexing have been performed independently More recently however, colleagues have demonstrated these functions simultaneously using a semiconductor optical amplifier based non-linear optical loop mirror (SOA-NOLM) as a combined demultiplexer and phase detector. In order to lock the peak of the switching window of the device, it is necessary to dither the local clock to obtain a differential error signal. This gives the direction and magnitude of any phase drift, allowing active control of the drive frequency. This in turn imposes a degree of phase modulation on the recovered clock, giving rise to a residual timing variance. In this embodiment we show simultaneous 'dither free' clock recovery and demultiplexing using an EA modulator in the configuration used in the first embodiment. Experimental Details:

The experimental set-up is shown in Figure 2. A 40 Gbit/s OTDM data stream was generated as before by passively multiplexing a high extinction ( > 45 dB), low duty cycle (4 ps), 10 Gbit/s data stream, realised from two cascaded harmonically driven EA modulators and a lithium niobate amplitude modulator which had a low frequency response that limited the word length to a

2 -1 pseudo-random data pattern. The 40 Gbit/s data stream was split using a 3 dB coupler 10, 50% being fed in a clockwise direction through the EA modulator 1 2 and detected at port A, after passing through a second 3 dB coupler 17. The remaining 50% was fed in an anti-clockwise direction using a fibre delay line, back through the same EA modulator, and detected with photodetector 30 at port B. Rather than synchronizing the gating function of the EAM with the peak of a channel so that a channel is cleanly gated through the device, which is the aim for the EAM in its demultiplexing mode ( e.g. as essentially happens to a channel of the data stream passing in the clockwise direction), the gating function is timed so that the leading (or trailing ) edge of a pulse is "cut" by the EAM window. The idea is to choose a point on the pulse edge where the slope is great ( eg the greatest slope) so that a small difference in the phase of the pulse relative to the window makes a significant change in the power of the gated part of the pulse. The effect of this change in power is detected by the photodiode 30. A small phase change in one sense will cause the time-averaged power detected by the photodiode to drop, while a similar phase change of the opposite sense will cause the time averaged power to rise. In the transmission system, the power entering the demultiplexer varies, so the phase locked loop will need to make allowances for this variation. Thus the photodiode output can be used in a control loop to ensure clock synchronization between the gating pulses applied to the EAM and the incoming data stream. The sychronized clock can of course be used for other purposes in the node or elsewhere. The EA modulator was the same as that used in the first embodiment. The device was again sinusoidally driven by a voltage controlled oscillator (VCO) 20 at 10 GHz to produce a 22.5 ps rectangular transmission window, enabling a single 10 Gbit/s channel to be optimally demultiplexed at port A.

To achieve clock recovery, the output of port B was detected on a low speed photodiode 30 and used as an error signal. The error signal was fed through standard PLL controller electronics which were used to drive the VCO 20, thus creating a closed loop. The EA modulator 1 2 acts as an electro-optic sampler, the time averaged intensity of whose output is a function determined by the phase / frequency difference between the 40 Gbit/s OTDM data stream and the locally generated 1 0 GHz electrical clock. By adjusting the phase of the 40 Gbit/s signal fed in the anti-clockwise direction using the variable fibre stretcher 22, it is possible to offset the error signal by a fraction of the clock period, to enable locking to a point in the data stream (here a point of high slope, such as half way up the leading or trailing edge of a pulse) therefore making the system extremely stable. This removes the requirement for a dither signal by achieving independent control between the error signal used in clock recovery, and the demultiplexed signal.

The photodiode 30 was a pin diode with an InGaAs absorber layer, which provided good gain and low noise. The diode had a bandwidth of 1 25MHz, chosen to be considerably higher than the bandwidth of the electrical filters in the PLL electronics ( 1 -2 MHz) 32 and lower than that required to "see" the individual pulses of the data stream ( i.e. slow relative to the bit rate of the data stream).

Minor incoherent interference effects between the required 10 Gbit/s demultiplexed signal at port A, and residual reflections (-30.5 dBm) from the EA modulator did not degrade the performance significantly but could be eliminated by ensuring that the signals were in orthogonal polarisations. Results

Figure 6a shows the 40 Gbit/s OTDM data stream. Figure 6b shows a 10 Gbit/s demultiplexed channel, and Figure 6c shows the 10 GHz recovered electrical clock. Jitter analysis of the recovered clock was performed using an RF spectrum analyser, by measuring the noise spectral density and the total power of each signal harmonic as described in the paper by Taylor et al, Applied Physics Letters, 1 986, 49 (1 2) ,pp681 - The RF spectrum of the recovered clock is shown in Figure 6d; the phase noise pedestal is 50 dB down from the peak. The rms timing jitter was calculated to be ~1 50 fs, making this technique suitable for use in an OTDM network at > 100 Gbit/s as in the paper by Kamatani referenced above. Figure 7 shows the results of BER measurements made on the 10 Gbit/s demultiplexed channels, with and without clock recovery. A 10 Gbit/s optimised data signal is used as a back-to-back reference to indicate receiver sensitivity The system operated error free, with minimal penalty and no indication of an error floor.

Conclusions

The example of this embodimentshows that clock recovery and demultiplexing of a 10 Gbit/s channels from a 40 Gbit/s OTDM data stream can be achieved simultaneously using a single electroabsorption modulator. The system showed excellent stability and used reliable low speed inexpensive components to achieve clock recovery. The recovered clock had a phase noise-mduced timing jitter of < 1 50 fs, indicating operation up to and beyond 100 Gbιt/s is possible. This technique reduces the number of high-speed components necessary in an OTDM network node, therefore increasing reliability and simplifying management, whilst also reducing costs.

Third Embodiment

As indicated in the introduction, demultiplexing, 'drop and insert' (D&l) multiplexing, clock recovery and synchronisation are the key functions required within an OTDM network node. To perform effective D&l functionality an optical switch must completely remove a single data stream (the 'drop'), whilst leaving the remaining channels undisturbed. It is then a simple matter to insert a data stream into the vacant time-slot (the 'insert'). It is necessary that clock recovery is achieved to ensure correct synchronisation of all data channels. To date a wide variety of optical demultiplexers and 'drop and insert' multiplexers have been demonstrated, including loop mirrors, Mach-Zehnder intergrated interferometers, four-wave mixing in semiconductor optical amplifiers, FWM in SOAs, and non-linear optical loop mirrors including semiconductor optical amplifiers. However, in our opinion the most promising, is the Electroabsorption (EA) modulator.

In this embodiment we deal with the use of EA modulators as the core element in a 3-node OTDM network. Each node is self contained and performs its own clock recovery. Node 1 uses two EA modulators to generate a 40 Gbit/s OTDM data stream, node 2 performs the D&l function on a single channel, whilst simultaneously recovering the 10 GHz clock by exploiting the bi-directiona ty of the EA modulator in the phase locked loop configuration of the second embodiment. Node 3 uses an EA-Modulator based demultiplexer with electrical clock recovery to extract a 10 Gbit/s channel.

Experimental Details

The experimental set-up is shown in Figure 3. At node 1 a 40 Gbit/s

OTDM data stream was generated as before by passively multiplexing a high extinction ( > 45 dB), low duty cycle (4 ps), 1 0 Gbιt/s data stream, realised from two cascaded harmonically driven EA modulators and a lithium niobate (L1N1O₃) amplitude modulator This was driven by a 2 -1 pseudo-random 10 Gbit/s pattern generator, the word length of which was limited by the modulator's poor low frequency response. The resulting OTDM data stream 8 was amplified and fed into the input of the 'drop' section of the OTDM node 2. Node 2 is based on the arrangement of the second embodiment. The 40 Gbit/s input data stream was split using a 3 dB coupler 1 0, 50% being fed in a clockwise direction through the EA modulator, and detected at port A via a 3 dB coupler 1 7. ( Again, a 3 port circulator would be preferred at this point in place of coupler 1 7.) The remaining 50% was fed in an anti-clockwise direction using a fibre delay line, back through the same EA modulator, and detected at port B. Of course, there is no need to use a symmetrical coupler at this point: it will often be more attractive to pass more than 50% of the input signal to port A ,e.g. 60, 70 or 80 %. Of course, the coupler provides loss to the signal passing in the anticlockwise direction, the loss depending on the splitter ratio, but even with a coupler splitting 80/20, there should be ample signal reaching port B. ( The use of an asymmetric coupler at this point will also be of interest in other of the embodiments described herein, as those skilled in the art will appreciate.) The EA modulator 1 2 is the same as that used in the first and second embodiments. The device was reverse biased at 1 V and sinusoidally driven using a voltage controlled oscillator (VCO) at 1 0 GHz, producing a 75 ps rectangular transmission window. This enabled three 10 Gbit/s channels to pass through the device unaffected, while completely removing (dropping) the remaining 10 Gbit/s channel. This is illustrated in Figure 8, (a) shows the 40 Gbit/s OTDM input data stream, (b) shows the transmission window for the 'drop' process, (c) shows the result after channel 3 has been dropped.

Simultaneous clock recovery was achieved by detecting the output at port B using a slow photodiode 30, as described in relation to the second embodiment. The resulting electrical signal is a time averaged intensity which is related to the phase difference between the 40 Gbit/s OTDM data stream and the locally generated 10 GHz electrical clock. This was fed through standard PLL electronics and used to control the VCO driving the modulator. In this way a closed loop is formed, with any fluctuations in phase being constantly corrected for. By locking to the leading edge of the data stream, it was possible to extract an extremely clean and stable clock signal with an rms. timing jitter of < 1 50 fs.

A different 10 Gbit/s channel, which had been amplified in an erbium doped fibre amplιfιer42, was inserted into the vacant time slot using a 3 dB coupler 40and a variable fibre stretcher 41 to adjust its phase. The resulting 40 Gbit/s data stream (see Figure 8d) was fed into node 3. This consisted of demultiplexer based on an EA modulator 50 driven by a 10 GHz VCO. Clock recovery was achieved electrically, using a 32 GHz high-speed pin diode 52, an electrical mixer 53 and the appropriate phase locked loop control circuitry 54 as described in the paper by Ellis et al, Electron. Lett. 29 ( 1 1 ) pp990-992, 1 993. Figure 8e shows an example of a demultiplexed channel. Different channels could be selected for demultiplexing using an electrical phase shifter. Results

Figure 9 shows the results of BER measurements made on each of the 1 0 Gbιt/s demultiplexed channels. An optimised 10 Gbit/s channel is used as a receiver sensitivity reference. Each of the undropped 10 Gbit/s channels operated error free with no indication of an error floor, but incurred a 1 dB penalty due to an imperfect transmission window at the 'drop' stage. The inserted channel however suffered from no or minimal penalty. To assess the extinction ratio of the channel 'drop' function, a high extinction 10 GHz pulse, pulse duration 5ps, was passed through the EA modulator and detected on an optical power meter. The pulse was tuned through the transmission window of the EA modulator using a variable electrical delay line The results are shown in the inset of Figure 9, which indicates an extinction ratio of ^" 14 dB. Conclusions

The example of this embodiment shows that the main OTDM processing functions of nodes for an OTDM network can be realised using electroabsorption modulators as the core elements. Each node can be self contained, performing its own clock recovery and synchronisation. Drop and Insert functionality has been demonstrated for the first time with an EA modulator by completely removing a 10 Gbit/s channel from a 40 Gbit/s OTDM data stream and then subsequently inserting a different 10 Gbit/s channel into the vacant time slot.

The un-dropped channels suffered from a 1 dB penalty as a result of an imperfect transmission window associated with the 'drop' process and because they had passed through more processing stages.

For full Drop and Insert functionality two synchronised modulators may be necessary at each node, one to perform the drop function, and the other to perform the demultiplexing operation.

Further Functionality

Where a channel is to be dropped or where there is to be demultiplexing, the EAM which performs that processing can also, in many instances, provide a further useful function as a photodetector. This further function can, in general, be achieved whether or not the EAM is being used to process signals which enter it from different directions.

As an example, consider the arrangement shown in Figure 10. Here the EAM is driven so that it transmits all but one of the incoming OTDM channels and the EAM is in a highly absorbing state for the remaining dropped channel. The absorbed channel is converted into a photocurrent by the electro absorption process within the modulator. Provided that the amount of background photocurrent due to the relatively low absorption of the transmitted channels is sufficiently low and an appropriate filtering arrangement is used (as shown in Fig. 10), the dropped channel may be detected as an electrical output from the EAM. This technique avoids the need for a further EAM (connected in parallel) to demultiplex the dropped channel This technique could be combined with the clock recovery technique used in node 2 of the Fig. 3 third embodiment (counter propagating clock recovery technique) to give simultaneous clock recovery, channel drop and photodetection of the dropped channel. The optical set up could be left as in Fig. 3, but an arrangement of electrical filters like that shown in Fig. 10 would be required.

The counter-propagating signal used for clock recovery will also generate a photocurrent in the EAM and thus could potentially degrade the quality of the electrical dropped channel signal. To minimise this effect one could (i) reduce the intensity of the counter-propagating optical signal (perhaps using asymmetric couplers) so that it gives rise to a relatively weak photocurrent in the EAM, or (u) adjust the fibre stretcher so that the detected counter propagating channel in the EAM is the same as the dropped channel synchronised to the same bit.

The process described in relation to Fig. 10 could also be viewed as demultiplexing. Demultiplexing could be achieved using an array of EAMs optically connected in parallel (as is conventional but without the need for any separate photodiodes) or in series. In either case a saving in the number of high speed opto electronic components could be achieved.

If a number of EAMs were connected in series as shown in Fig. 1 1 and driven with a similar filtering arrangement as in Fig. 10, then it is easy to see how demultiplexing of all of the channels could be achieved.

Electro absorption modulators have been used for in line control, i.e. cleaning up digital (not necessarily OTDM) signals. An example of this is described in the paper by G Aubin et al, "40Gbιt/s OTDM sohton transmission over transoceanic distances," El. Lett. Vol.32, 24, pp 21 88-21 89, 1 996, where an EAM was driven at 40GHz to provide in-line control ( e.g. reshaping and re-timing) in a so ton transmission system. Clock recovery is needed to achieve this in line control. The counter propagating clock recovery technique described here in relation to the second and third embodiments could be used in such an application to reduce the number of high speed opto electronic components required.

If required, it may be possible to detect the signal in the same EAM that is used to provide in line control and clock recovery (although the signal photocurrent is likely to be weak as a lot of the light in all the pulses will be transmitted): making possible simultaneous in line control (e.g. reshape, retime in a soliton system), clock recovery and photodetection

Wavelength conversion has been demonstrated using an EAM, see the paper by N Edagawa, et al, "Novel wavelength converter using an electro absorption modulator: conversion experiments at up to 40Gbit/s", OFC '97 Technical Digest, pp 77-78, 1 997[2]. One incoming high power optical signal modulates the absorption of the DC biased modulator due to carrier saturation, thus causing a weaker CW incoming optical signal to be modulated effectively converting the wavelength of the signal. If required, the signal could be monitored by using the photocurrent generated in the EAM.

Claims

1 . An optical processing arrangement comprising an electroabsorption modulator ( 1 2) having first and second optical ports, an optical splitter ( 1 0) having an input and two outputs, a first output being optically coupled to the first optical port of the modulator, and a second port being optically coupled to the second optical port of the modulator, the arrangement being such that an optical signal provided at the input of the optical splitter is split and fed, via the first and second outputs of the splitter, into the first and second optical ports of the modulator for processing by the modulator.

2. An arrangement as claimed in claim 1 , wherein the optical splitter is a four-port device, the arrangement being such that the fourth port of the device passes optical signals which have passed through the modulator from one of the outputs of the splitter.

3. An arrangement as claimed in claim 2, wherein each one of the outputs of the splitter ( 1 0) receives signals passed through the modulator from the other one of the outputs of the splitter.

4. An arrangement as claimed in any one of claims 1 to 3, configured as a node of an optical time division multiplexed system.

5. A node for an OTDM transmission system in which an electroabsorption modulator is used as a gating element for optical data passing in opposite directions through the modulator.

6. Use of an electroabsorption modulator as a gating element in the processing of optical signals, the arrangement being such that optical signals are passed bidirectionally through the modulator.

7. A method of processing an optical data stream modulated with a return to zero format, the method comprising:

(i) splitting the optical data stream into first and second data streams;

(n) coupling said first data stream into a first input port of an electro absorption modulator (EAM);

(in) processing said first data stream in the EAM;

(iv) coupling the processed first data stream out of a first output port of the EAM;

(v) coupling said second data stream into a second input port of the EAM; and

(vi) processing said second data stream in the EAM.

8. A method as claimed in claim 7, comprising the further step of

(vn) coupling the processed second data stream out of a second output port of the EAM.

9. A method as claimed in claim 7 or claim 8, wherein the second input port and the first output port are common.

10. A method as claimed in claim 8, wherein the first input port and the second output port are common

1 1 . A method as claimed in any one of claims 7 to 10, wherein the split between the first and second data streams is unequal.

1 2. A method as claimed in any one of claims 7 to 1 1 , wherein the processing carried out in step (in) is demultiplexing.

1 3. A method as claimed in any one of claims 7 to 1 1 , wherein the processing carried out in step (in) is channel dropping.

14. A method as claimed in any one of claims 7 to 1 3, wherein the EAM is used as a photodetector.

1 5. A method as claimed in claim 14 as dependent on claim 1 2, wherein the EAM is used for photodetection of a demultiplexed channel.

1 6. A method as claimed in any one of claims 7 to 1 5 , wherein the processing carried out in step (vi) is demultiplexing.

1 7. A method as claimed in any one of claims 7 to 1 6, wherein the processed second data stream is used for clock synchronisation.

1 8. A method as claimed in claim 17, wherein the processed second data stream is detected by a photodetector having a frequency response significantly slower than the modulation rate of the optical data stream.

19. A method as claimed in claim 17 or 1 8, wherein clock synchronisation produced according to claim 8 is used to synchronise operation of the EAM.

20. A method as claimed in claim 19, wherein in respect of the second data stream the EAM acts as an electro-optic sampler, the time averaged intensity of the output of the sampler being determined by the phase difference between the second data stream and a locally generated electrical clock applied to the EAM.

21 . A method as claimed in claim 20 as dependent on claim 1 8, wherein the EAM is synchronized by the clock to sample through an edge of pulses of the second data stream, so that small differences in phase between the clock and the second data stream lead to relatively large changes in the electrical output of the photodetector.

22. A method as claimed in any one of claims 7 to 21 , wherein the data stream comprises optical sohtons.

23. A method as claimed in any one of claims 7 to 22, wherein the optical data stream is an OTDM data stream.

24. A method of processing an optical data stream modulated with a return to zero format, the method comprising:

(i) coupling said first data stream into a first input port of an electro absorption modulator (EAM) ; (ii) processing said optical data stream in the EAM;

(ii)i coupling the processed optical data stream out of a first output port of the EAM;

(iv) splitting the processed optical data stream into first and second data streams; (v) coupling said second data stream into a second input port of the EAM;and

(vi) processing said second data stream in the EAM.

25. A method as claimed in claim 24, further comprising the step of coupling the processed second data stream out of a second output port of the EAM.