Field of the Invention

This invention relates to an optical amplifier, particularly a semiconductor optical amplifier having non-linear characteristics.

Background

It is known that semiconductors can act as optical amplifiers. When certain semiconductors are subject to an injected electric current, an incident photon causes an eleαron to traverse semiconductor's gap band with the result that an additional photon is generated thereby producing light amplification. Semiconductor optical amplifiers which operate in this way are well known and reference is directed to" Long Wavelength Semiconductor Lasers" G. P. Agrawal and N. K. Dutta; Van Nostrand, Chapters 1 to 6.

The semiconductor material which is used as the active amplification region of the device may comprise a bulk material or for example, a stack of multiple quantum wells (MQW). A problem with the semiconductor material that is used for the amplification region is that it suffers from a gain-saturation effect which may be produced by amplified spontaneous emission (ASE). For this reason, typical devices that use bulk semiconductor material in a parallel sided channel, usually have a length of 500 microns or less and a width of 1-2 microns, because if the device were to be made longer, there would be no improvement in gain. For MQW devices, the gain per unit length is slightly lower than for bulk material devices, and amplifiers of length up to 1 mm have been produced hitherto but it has been considered that devices of longer length would suffer from ASE, with no improvement in gain. Longer, tapered devices have been reported, of length l-3mm, in which tapering of the amplifier is arranged to offset partially the onset of gain saturation. Reference is directed to S. El Yumin et al, " Taper Shape Dependence of Tapered- Waveguide Travelling Wave Semiconductor Amplifier", IEICE Transactions on Electronics, Vol. e77, No 4, April 1994, Tokyo Japan. Reference is also

directed to D. Mehuys et al, " 11.6 W Peak power diffraction limited diode-to- diode optical amplifier" Appl. Phy. Lett. Vol. 62, No 6, 8th February 1993, pp544-546, which discloses a broad area travelling wave amplifier of width 600 μm and length 2200μm.

Another disadvantage of longer devices is that they are harder to mount. Conventionally, SOAs are mounted using headers designed for laser diodes, which tend to be shorter and as a result, it is not straightforward to package longer devices.

Also, longer devices consume more power, so that it has been considered disadvantageous for the device to be longer than that at which optical saturation occurs.

Semiconductor optical amplifier devices can be used for a number of different purposes and a review is given in K. E. Stubkjaer et al,"Optical Wavelength Converters", Proc. ECOC '94, pp 635-642. SOAs can be used as modulators, in which an optical signal, modulated at a given bit rate, is fed into the amplifier, together with a separate target wave. The modulated signal produces gain-saturation for successive bits and as a result, the target wave is modulated with the input bit pattern. This is known as cross gain modulation (XGM). The modulation may also produce a phase shift in the target wave and this is known as cross phase modulation (XPM). Both of these processes may produce wavelength conversion. For example, the target wave source may be at a different wavelength to the modulated input source so that the bit modulation is transferred from the input optical source at a first wavelength to the target wave at a second different wavelength.

In order for the modulator to be effective, for example in an optical data transmission network, it is desirable that the amplifier exhibit uniform amplification characteristics over a wide range of bit modulation frequencies. For example, Stubkjaer supra suggests a bit rate transparency to more than 5-

10 G-bit/s. A bit rate of 20 G-bit/s has been reported by J. M. Wiesenfeld, J. S. Perino, A. H. Gnauk and B. Glance, "Bit Error Rate Performance for Wavelength Conversion at 20 G-Bit/s", Electron. Lett. 30, pp 720-721 (1994) although it is not clear from Wiesenfeld et al whether the modulator was operating within a 3 db bandwidth.

Hitherto, it had been considered that the bandwidth was limited by the differential carrier recombination rate in the amplifier, this rate including spontaneous emission and stimulated emission.

However, in accordance with the present invention, it has been found that the -3 db bandwidth of the gain of the amplifier in respect of the bit modulation rate, is a function of the length of the path through the amplifier. Thus, in accordance with the invention, it has been appreciated that by increasing the length of the path, this bandwidth can be increased.

Semiconduαor amplifiers can also be used to produce wavelength conversion by a different process known as four wave mixing. This is discussed in Stubkjaer supra and a fuller theoretical discussion is given in " Population pulsations and nondegenerate four-wave mixing in semiconduαor lasers and amplifiers" G. P. Agrawal, J. Opt. Soc. Am. B, Vol 5, No 1, January 1988 ppl47-159. In four wave mixing, pump radiation at a pump wavelength λ _p is fed into a semiconduαor amplifier, together with an input signal of a different wavelength to the pump signal. In a typical example, the pump waveform has an energy of lOmw whereas the input signal has an energy of lmw. The wavelength of the input signal is close to that of the pump, typically with a wavelength difference of = 2nm. The two beams are of the same polarisation and consequently beat coherently, with a beat frequency in this example of =■ 100 GHz. The resultant beat waveform causes the carrier density in the amplifier to oscillate. This produces a non-linear effeα on the gain, which lags the input waveform and beats with it. It can be shown that this produces a wavelength converted signal λ _c , with a wavelength

- 2λ _p - λ,. The converted signal λ _c and the input signal λ, are equally spaced in terms of wavelength above and below the pump wavelength λ _p .

Four wave mixing has the advantage that the conversion process is extremely fast as it does not rely on carrier recombination as in XGM and XPM. Furthermore, there is less distortion but four wave mixing suffers from the disadvantage that the converted signal is of low power and the signal to noise ratio can be a problem in respeα of the converted signal.

However, in accordance with the invention, it has been found that the conversion efficiency for four wave mixing is funαion of the length of the path through the amplifier. Thus, in accordance with the invention, it has been appreciated that by increasing the length of the path, the four wave mixing efficiency can be increased.

Summary of the Invention

In accordance with the invention from a first aspeα, there is provided a semiconduαor optical amplifier apparatus including an optical path from an input to an output through an optical amplification region of substantially constant width of less than 50μm along its length, the path having a length through the region that exceeds 1 mm.

In another aspeα, the invention provides optical apparatus comprising a target wave source, a modulated optical source that is modulated at a given bit rate, and an optical amplifier to receive radiation from the sources and operative to produce modulation of the target wave according to the modulation of the modulated optical source whereby to produce a modulated target wave output, wherein the amplifier has an optical path from an input to an output through an optical amplification region and the length of the path is seleαed 0 to be longer than that at which optical saturation in said region occurs.

Also, in accordance with the invention there is provided optical modulation

apparatus comprising a target wave source, a modulated optical source that is modulated at a given bit rate, an optical amplifier to receive radiation from the sources and operative to produce modulation of the target wave according to the modulation of the modulated optical source whereby to produce an 5 modulated target wave output, wherein the amplifier has an optical path from an input to an output through an optical amplification region and the length of the path is seleαed so that the gain of the amplifier for the modulated target wave output exhibits a -3 db bandwidth that extends beyond 20 GHz in respeα of the modulation bit rate. w

The modulated optical source may produce cross gain modulation in the amplifier. Furthermore, cross phase modulation may be produced which may alternatively be used when modulating the target wave. The target wave source may have a different wavelength charaαeristic from the modulated is optical source with the result that wavelength conversion occurs. This may be used to provide an all-optical switch, in which the amplifier is conneαed to a plurality of optical output channels each with their respeαive optical filter tuned to a different wavelength, so that, by controlling the wavelength of the target wave source, the modulated output can be direαed to the different

20 channels seleαively.

The path through the optical amplification region of the amplifier is seleαed so as to optimise the bandwidth as aforesaid but may be seleαed not to be sufficiently long as to cause undue problems with ASE. Typically, the path 2J length does not exceed 5 mm. Convenient choices for the path length are that it should exceed 1.0, 1.125, 1.25, 1.5, 1.75, 2.0, 2.2, 2.25, 2.50 or 2.75 nun.

The amplifier may comprise at least first and second amplifier elements 30 conneαed in cascade, with the aggregate length of the paths through the elements being seleαed so as to achieve the desired bandwidth.

In praαical examples of the invention, the -3 db bandwidth may extend beyond 28, 35 or 40 GHz by suitable choice of device parameters.

The invention is also applicable to four wave mixing and in a further aspeα, the invention provides a semiconduαor optical amplifier that includes an optical path from an input to an output through an optical amplification region, first and second sources of different optical radiation for the amplifier, whereby the amplifier produces an amplified optical output in dependence upon the optical radiation from the first and second source, wherein the path has a length through the region that exceeds 1 mm.

The first source may comprise a source of pumping radiation for the amplifier at a wavelength λ _p and the second source may comprise a signal source at a wavelength , such that the amplifier is operative to produce an output by four wave mixing at a wavelength λ _c , with an improved amplitude and signal to noise ratio than hitherto.

In another aspeα, the invention provides semiconduαor optical amplifier apparatus including an optical path from an input to an output through an optical amplification region wherein the path has a length L through the region that exceeds 3 mm.

Brief Description of the Drawings

In order that the invention may be more fully understood examples will now be described with reference to the accompanying drawings in which:

Figure 1 is a top plan view of a semiconduαor optical amplifier; Figure 2 is a seαional view taken on I-F of Figure 1; Figure 3 is a schematic diagram for illustrating modulation effeαs in the o amplifier of Figures 1 and 2;

Figures 4A-4C are schematic diagrams for explaining XGM effeαs in the amplifier of Figures 1 and 2 by a superposition of SGM effeαs;

Figure 5 is a graph of the real and imaginary parts of a funαion (T(ω)-1);

Figure 6 is a schematic diagram of an experimental set-up for testing the bandwidth dependence upon path length for the amplifier;

Figure 7 is a graph of the bandwidth for two amplifiers in cascade and a corresponding one of the amplifiers alone, for a drive current into each amplifier of 140 mA;

Figure 8 illustrates the effeα of increasing the drive currents to the amplifiers to 400 and 500 mA respeαively;

Figure 9 is a graph illustrating the linearity of the conversion bandwidth with amplifier length L, and also the corresponding linearity of the peak wavelength for the wavelength conversion;

Figure 10 illustrates a network switch incorporating an amplifier according to the invention;

Figure 11 illustrates schematically a demultiplexer; Figure 12 illustrates how an amplifier according to the invention can be used in a Mach-Zehnder interferometer, configures as a demultiplexer;

Figures 13 to 15 illustrate alternative examples of how amplifiers according to the invention can be configures in the Mach-Zehnder interferometer; and

Figure 16 illustrates the results for the apparatus of Figure 6, when operated by four wave mixing.

Detailed Description

A typical semiconduαor optical amplifier Al is shown in Figures 1 and 2 which is fabricated in the InGaAsP material system for use in telecommunication systems, with wavelengths centred on 1.55 μm. The device consists of an elongate amplification region 1 formed on a substrate 2.

The amplification region 1 has an input 3 at one end for optical radiation and an optical output 4 at the other end each formed with anti-refleαion coatings.

A typical refleαivity is of the order of 10 ^"3 ~ 10 ^"4 . The input 3 and output 4 are configured to couple into single mode optical waveguides e.g. optical fibres or alternatively other structures (not shown) integrated onto the substrate.

The region 1, as explained hereinafter is elongate with a length L and a

and a width W defined by a lateral confinement structure. As will be explained hereinafter, the length L of the amplifier region 1 is seleαed to be longer than hitherto in order to improve the amplifier charaαeristics. The width W of the region 1 is seleαed for guided travelling wave operation and is typically of the order of 1-2 μm, less than 50 μm and usually less than 5 μm.

The aαive amplification region 1 can be fabricated in a number of different ways and an example that comprises a buried heterostructure is shown in Figures 1 and 2. Referring to Figure 2, the aαive region 1, which produces amplification, comprises a stack of multiple-quantum-wells comprising a plurality of alternate layers of InGaAsP-InP 5 formed on a n-InP substrate 6. The aαive region 5 is overlaid by a p-InP region 7 formed with a conduαive over-contaα 8 formed of p-InGaAsP material, provided with a metallic contaα 9. A metallisation layer 12 is formed in the underside of the substrate 2.

When forming the device, regions to the side of the elongate aαive region 1 are removed by seleαive etching in a manner known per se and layers 10 and 11 of p-InP and n-InP material are grown back.

Thus, the MQW stack 5 is sandwiched between p-InP region 7 and n-InP layer 6 so that when a voltage is applied between the metallic contaα 9 and the metallisation layer 12, an eleαric current passes through the stack 5. The arrangement of the n and p doped layers 10 and 11 forms a reverse-biased junαion so that the current produced by the voltage applied to contaα 9 is direαed seleαively through the aαive region 1 rather than to each side, to provide lateral confinement. Thus, photons incident on input 3 (Figure 1) cause eleαrons to traverse the bandgap of the MQW stack 5 so as to generate additional photons, thereby producing optical amplification. For further details of the structure of the device and various modifications thereof, 0 reference is direαed to "Long Wavelength Semiconduαor Laser" Agrawal and Dutta, supra, from which it will be appreciated that a bulk device can be used as an alternative.

A semiconduαor optical amplifier can be used as an aαive component in a modulator as will be explained with reference to Figure 3. An essentially continuous target wave source 13 e.g. a laser operating in the aforesaid optical telecommunications wavelength band, is direαed into the input 3 of the aαive region 1 of the amplifier. Also, an optical source 14, which has been modulated at a given bit rate is direαed to the input 3 of the amplifier. The "1" bits of the modulation produce saturation effeαs in the aαive region with a result that the target wave becomes modulated by the signals from the source 14 and the resulting output at output 4 comprises the target wave modulated with the modulation pattern from source 14. The target wave source 13 and the modulated source 14 can operate at different frequencies so that a wavelength conversion takes place. This process is known as cross gain modulation. As previously explained, the modulation may also produce a phase change, which can be deteαed by comparing the phase of the modulated output target wave with the phase of the input target wave from source 13 (by means not shown).

In order to operate as a praαical device, the modulator needs to be able to operate over a wide range of bit rates for the modulated source 14. Thus, the amplifier gain should not vary significantly over the desired operating range of bit rate modulation.

In accordance with the invention, it has been appreciated that this gain bandwidth of the amplifier in respeα of the modulation bit rate is a funαion of the length L of the optical aαive region shown in Figure 1. A derivation of the wavelength conversion frequency response will now be given in order to show the length dependency of the bandwidth charaαeristics. This is carried out by firstly considering the response of the amplifier to a single input beam and then extending the analysis to two input beams by superposition.

Considering firstly the case of a saturated amplifier with a single optical input

beam, the carrier density N in the amplifier evolves with time as:

where R(N) is the spontaneous recombination rate, / the injeαed current density, d the aαive layer thickness, e the eleαronic charge, A the aαive cross seαion, T the mode confinement, hv is the photon energy, a is the material gain, N _t is the transparency carrier density and P is the optical power.

This is known as self gain modulation and a fuller explanation is given in: G. P. Agrawal & N. A. Olsson, "Self-phase modulation and speαral broadening of optical pulses in semiconduαor laser amplifiers" J. Quantum Eleαron., 25, pp 2297-2306, (1989).

It can be shown that the travelling wave equation for the optical power is

where _sc is the waveguide loss per unit length. In the small signal regime, the zero-mean time-varying parts of P and N are denoted by δP and 5N respeαively. The carrier density is then given by:

-aT (N-N)bPIAhv δ_V= P) iω+(l/τ +aTP ₀ jAhv)

For a saturated amplifier, the gain becomes close to the waveguide loss, and the optical power becomes constant at a saturation value. The propagation equation (2) then becomes, ignoring the phase faαor:

Integrating this over the length L of the amplifier gives :

(aT ) N-N)P ₀ /Ahv

-δP=- -δP (*) d z iω+(llτ _c +aT P ₀ I Ahv)

BPβ.) = T (ω) δP(0) where T(ω) is a transfer funαion defines as follows:

Writing the stimulated carrier lifetime as:

Ahv aT P _Q

and writing the optical gain as g = T (N-NJ it follows that:

The analysis for a single input optical beam for the amplifier can be extended by superposition to a situation where two beams of different wavelengths are fed into the amplifier, in order to charaαerise cross modulation effeαs. This will now be explained with reference to Figure 4.

In Figure 4, two different wavelength input optical beams I, II of respeαive powers P _j and P ₂ are shown for optical amplifier A, with the beam I constituting the modulating beam and beam II constituting the target beam.

In Figure 4A, the situation is considered where each input beam is imparted with a modulation pulse δ so that input pulses δ and δ /r are applied to the amplifier, as shown. The amplifier is operative so that both inputs are changed in proportion and so the inputs are amplified according to the transfer funαion T(ω) of the amplifier, which results in outputs δ.T(ω) and δ.T(ω)/r for the output beams.

In Figure 4B, a different situation is shown in which the input beams are modulated to produce inputs +δ /r and -δ /r respeαively. The amplifier is operated so that the input (and output) power is kept constant so that the outputs are the same as the inputs i.e. +δ /r and -δ/r.

Figure 4C shows a superposition of the inputs and outputs of Figures 3A and 3B. By considering the relationship between the values of the input and the outputs shown in Figure 4C and substituting for r, it can be shown that the cross gain modulation response T _XGJ ^ω) is given by:

^τ xω ° <T (ω)-l) (*)

P _l + P _i

The bandwidth of wavelength conversion through cross gain modulation as given by equation (8) is found, in accordance with the invention, to increase almost linearly with amplifier length. The 3db limit for the bandwidth of this conversion can be understood by further analysis of terms of equation (8). It will be understood that for a given operational condition the term P P _j +P ₂ is a constant but that the term (T(ω)-1) is variable in both frequency and phase, and has both real and imaginary parts. Figure 5 shows a plot of the real and imaginary parts of the funαion F - T(ω)-1) on orthogonal z and y axes. For increasing frequency, the funαion F describes a widening spiral 15 beginning at near z--l for ω-0, and finishing at z-0 for ω- oo. Stated differently, the funαion T(ω) spirals outwardly from the point (-1,0) to the

point (0,0). The 3db wavelength conversion bandwidth limit for funαion F is described by the locus 16 of a veαor of modulus 0.5 centred on the origin and is encountered when the widening spiral 15 crosses the arc 16. By plotting the real and imaginary parts of F for different values of gL , it has been found that when the spiral 15 crosses the arc 16, the phase angle for the expression T(ω) is always close to π/6. Thus by equating the phase of T(ω) in equation (7) to 7r/6 permits a frequency condition F _idb for the 3db bandwidth to be to be expressed as follows, making the assumption that the amplifier is operating in a saturated condition i.e. the stimulated carrier lifetime τ _s is smaller than the spontaneous lifetime τ _c :

_ 3g

3db P)

Thus, it can be seen from equation (9) that the 3 dB bandwidth is approximately linearly dependent on to the length L of the amplifier when other faαors such as injeαion current density are maintained constant.

This effeα can be seen from the experimental set up that will now be described with reference to Figure 6. The target wave source 13 comprises a DFB laser operating at 1.555 μm, and the modulated source 14 comprises a tunable laser 14a operating at 1.560 μm, that feeds radiation to a Mach- Zehnder (M-Z) modulator 14b. The outputs of both amplifiers are fed through respeαive erbium doped fibre amplifiers 17, 18 and associated polarisation controllers 19, 20 attenuators 21, 22 and filters 23, 24. The resulting modulated signal and target wave are combined by a 3 db coupler 25 and fed into a first semiconduαor amplifier Al having an aαive region of length L, which is cascaded to a second semiconduαor optical amplifier A2 with an aαive region, also of length L. The amplifiers are conneαed in series with optical isolators 26A, 26B, 26C for suppressing ASE. The resulting output is fed through a further filter 27 and an attenuator 28 to a deteαor diode 29, which feeds an eleαrical signal through a eleαrical amplifier 30 to a

network analyser 31. The analyser 31 also provides eleαrical control signals to the modulator 14b. The modulator 14b was swept through a bit rate of 300 Mhz to 40 GHz. The responses were measured for three different cases: with only the amplifier Al, with only amplifier A2 and also with both amplifiers cascaded. The amplifiers used were both of length L - 1.125 mm with a constant width W of the order of 1 micron, and a gain peak at 1.550 μm. The signal and target powers injeαed into the amplifiers were +2.5 and -1.8 dBm respeαively which resulted in the gain being saturated and the Fabry- Perot ripple being reduced to below .5dB. The drive current to each amplifier was set at 140 mA and the single amplifier XGM bandwidth was below 10 GHz.

The results obtained are shown in Figure 7. Trace 32 shows the gain of the amplifier as a funαion of frequency for a single amplifier, whereas trace 33 shows a result for the two amplifiers aαive regions IA, IB cascaded. It can be seen from Figure 5 the -3db bandwidth for a single amplifier, of 6.5 GHz is increased to over 13 GHz when two amplifiers are cascaded. This is attributed to the faα that the length L through the cascaded amplifiers is increased, in this example to 2.25 mm by conneαing the two amplifiers in series.

Figure 8 shows an arrangement in which the cascaded amplifiers have their drive current increased from 140 mA as described above, to 400 and 500 mA respeαively. The resulting bandwidth exhibits a -3db roll off at 28 GHz, substantially greater than hitherto.

It will be appreciated that the cascading of two amplifiers produces substantial losses at the junαions between them and so by forming a single amplifier of length that exceeds 1 mm, a substantially improved bandwidth can be 0 provided, that extends beyond 20 GHz bit modulation rate for amplifier drive currents ~ 400 mA.

Typical examples of the length L of a single semiconduαor optical amplifier are in excess of 1.00 mm, 1.125 mm, 1.25 mm, 1.5 mm, 1.75 mm, 2.00 mm, 2.20 mm, 2.25 mm, 2.50 mm or 2.75 mm, with a drive current adjusted to achieve a current density in the aαive region of typically 50kA/cm ² . The s drive current may be adjusted to achieve a -3 db wavelength conversion bandwidth that extends beyond 28 GHz or 35 GHz or 40 GHz. the width W is typically of the order of 1-2 μm, less than 50 μm and usually less than 5 μm

Alternatively, a number of devices may be cascaded as described in order to o achieve the desired path length, with isolators between them in order to suppress ASE.

In these examples, an amplifier with a MQW structure is used, in which case a path length at least in excess of 1 mm is provided for a channel width W of 5 less than 5μm. However, as previously explained, bulk devices can alternatively be used, in which case the path length can be shorter, typically in excess of 800 microns, because of the higher efficiency of bulk devices compared with MQW devices.

0 Referring now to Figure 9, this shows the -3db bandwidth (plot 34) and peak frequency (plot 35) for the wavelength conversion as a funαion of amplifier length. It can be seen that both charaαeristics have a linear relationship with the length of the aαive region of the amplifier. The amplifiers used to obtain the data were construαed as described with reference to Figs. 1 and 2 and 5 were of the same construαion apart from the length of their aαive region. They were operated under the same conditions.

A typical praαical example of the device may be construαed as described with reference to Figs 1 and 2 with an aαive region 1 of length L - 3.5 mm and 0 width W - 1 - 2 μm,with the depth of the layers in the confined aαive region being of the order of 0.1 - 0.2 μm. This can be operated with a wavelength conversion bandwidth of the order of 40Gb/s, with a typical drive

current in the range of 500 mA to 2000 mA.

A praαical example of the invention will now be described with reference to Figure 10. A semiconduαor optical amplifier in accordance with the invention, as previously described, is used as a wavelength converter. Optical digital communication signals from a modulated source 14 at wavelength are applied to the converter, which includes an optical amplifier with an aαive region 1, as previously described, with a length L » 3.5 mm. A target wave source 13, which is tunable in terms of wavelength, is also applied to the wavelength converter. A number of output channels are conneαed to the output 4 of the amplifier, which are typically constituted by optical fibres Fl- F4 conneαed to different geographic destinations. Each fibre has an associated band pass filter 35, 36, 37, 38 tuned to a particular wavelength λ,-)^. The target wave source 13 is tunable to the individual wavelengths λ,-λ ₄ so that by setting the target wave source wavelength, the modulation from source 14 can be converted to the wavelength of the target wave source and consequently direαed to a seleαed individual one of the communication channels.

An alternative use of an amplifier according to the invention is as a time demultiplexer, in which data configured in interleaved time slots are separated into different channels. Referring to Figure 11, successive data time slots S1...SN contain data denoted by the presence or absence of optical pulses. The time slots are interleaved and are to be direαed to different channels Chi, Ch2 respeαively by a demultiplexer DM1. A suitable struαure for the demultiplexer DM1 is shown in Figure 12.

The demultiplexer DM1 comprises a Mach-Zehnder loop struαure including first and second optical fibres 39,40 with two coupling regions 41, 42. As well known in the art, for a Mach-Zehnder loop structure, an optical output is produced at the output OP1 or OP2 depending on the relative phase of the signals travelling in the portions 43,44 of the fibre loop. In accordance with

the invention, an optical amplifier Al, construαed as previously described, with an aαive region of length > 1.25mm, is conneαed in one of the loops so as to control the relative phases of signals in the portions 43,44 of the loop and direα signals to the outputs OPl, OP2 seleαively.

The interleaved optical data stream for both Channels 1 and 2 from source 14 is fed to input LP1, whereas a control optical pulse stream is fed to the input LP2 from source 13. The control pulse stream includes pulses which delineate in time, data slots for Channel 1. The signal streams fed to the inputs LP1,2 are of respeαive different wavelengths or polarisations, or both and are mixed together by the coupler so that the mixed stream travels along both fibre portions 43,44. The amplifier Al produces a phase shift in the signal stream travelling in fibre portion 43 relative to portion 44 as a result of cross phase modulation that occurs in the amplifier during the occurrence of the control pulses that occur for the data slots for Channel 1, and as a result, optical data pulses for Channel 1 are direαed seleαively to OPl whereas data pulses for Channel 2 pass to output OP2. It will be understood that either Channel 1 or 2 may itself include a plurality of further channels that can be separated by means of additional demultiplexers.

Whilst the described Mach-Zehnder device has been described for use as a demultipexer, it can also be used as a switch for other applications e.g. for routing packets in packet switched networks, for gating signals for other purposes and for improving the extinαion ratio of a digitally modulated optical signal.

Alternative Mach-Zehnder loop struαures will now be described, that incorporate optical amplifiers Al, A2 in both of the portions 43, 44 of the fibre loop, with both of the amplifiers being construαed as previously described e.g. each with an aαive region of a length greater than 1.75 mm.

In Figure 13, the optical path length from the splitter 41 to each of the

amplifiers Al, A2 differ by an offset d. Thus, considering the inputs LPl and LP2, the input signals reach the amplifier Al before the amplifier A2, with the result that a very narrow switching window is produced, useful for demultiplexing. The use of amplifiers Al, A2 in accordance with the invention, sharpens the definition of the switching window and so enables shorter switching windows to be achieved than hitherto. In this configuration, the width of the switching window is fixed by the spatial offset d, during the fabrication process for the device.

Referring to Figure 14, an alternative configuration is shown in which the amplifiers Al and A2 are disposed symmetrically in the interferometer but the control signal LP2 is fed into the portion 43 of the loop through a separate coupler 46 so that for example, demultiplexing can be carried out in the manner described with reference to Figure 12. However, the presence of the symmetrical disposition of the amplifiers Al, A2 enables matching of losses in both portions 43, 44 of the fibre loop. The use of the longer amplifiers Al, A2 in accordance with the invention enables the phase changes that occur in amplifier Al to be speeded up as compared with the prior art.

In Figure 15, the amplifiers Al, A2 are disposed symmαrically in the interferometer and control signals can be injeαed into the amplifier inputs individually through respeαive optical couplers 46, 47. In the example shown, the input LP2 is fed into both of the couplers 46, 47, with the input to coupler 47 being delayed in time relative to the input for coupler 46. As a result, a very narrow switching window is produced in the manner described with reference to Figure 13. The use of amplifiers in accordance with the invention can sharpen the switching window. In the configuration of Figure 15, the delay between the two injeαed control signals LP2, IP2 + Δ can be varied so as to adjust the length of the switching window.

The invention also has application to semiconduαor optical amplifiers which are used in modulators that operate by four wave mixing. It has been found

according to the invention that by increasing the length of the amplifier, the efficiency of four wave mixing is improved and the signal to noise ratio is increased. This will now be explained by way of example with reference to Figures 6 and 16. The apparatus of Figure 6 was driven such that the target s wave source 13 operated at relatively high power and provided a pump for the amplifiers Al, A2, and a pumping wavelength λj - λ _p - 1.555 μm. The tunable laser 14a was operated at a wavelength of λ ₂ = λ; - 1.560 μm. The outputs of the lasers were applied with the same polarisation to the amplifiers Al, A2. The amplitude of the output of laser 14a was significantly lower o than that of laser 13 with the result that four wave mixing occurred as can be seen from Figure 16. A wavelength converted signal λ _c was produced. The wavelength relationship between the signals is as follows:

λ _c - 2 λ _p - λ _; . 5

Thus, the converted signal and the input signal are equally spaced in terms of wavelength on opposite sides of the pump wavelength λ _p as shown in Figure 16.

0 A non-limiting, qualitative explanation of the four wave mixing process will now be given. Due to the faα that λ _p and are of the same polarisation, they form a beat frequency of the order of 100 GHz which causes the carrier density in the amplifiers to oscillate. This produces a non-linear effeα in respeα of the gain, which produces further beating with the pump wavelength 5 so as to produce the wavelength converted signal λ _c . For further details, reference is direαed to Stubkjaer et al supra.

The effeα of operating the apparatus of Figure 6 with only one of the amplifiers Al, A2, and then with both of the amplifiers is shown by a thick 0 trace 48 and a thin trace 49 respeαively. It will be seen that the four wave mixing conversion efficiency improves with increased amplifier length L.

amplitude of the converted signal λ. increases. Also, the signal to noise ratio is improved. It will be seen that the noise floor 48a of trace 49 is suppressed downwardly as compared with corresponding floor 48a when a single amplifier is used so that the signal to noise ratio is increased, in respeα of the converted signal λ _c . It is to be noted that the suppression of the noise floor occurs asymmetrically, and the wavelength converted signal λ _c is positioned on the appropriate side of the pump wavelength λ _p to take advantage of the asymmetrical, downward changes that occur to the noise floor.

As previously mentioned, four wave mixing has the advantage that the wavelength conversion occurs rapidly as compared with cross gain modulation. The converted signal λ _c can be imparted with a modulation, by modulating the input signal λ, . This can be achieved in the configuration of Figure 6 by operating the modulator 14b.

An amplifier that uses four wave mixing, can be used to correα dispersion in a signal travelling along an optical communication path e.g. a signal travelling over long distances. The signal travelling along the path may be subjeα to a frequency shift due to the dispersive effeαs of the path e.g. an optical fibre. An amplifier which operates by four wave mixing can be used to change the wavelength of the input signal λ, to λ _c to achieve an inversion of the signal wavelength relative to λ _p , so as to compensate for chirp.

It has been found that for four wave mixing, the efficiency of the amplifier is approximately proportional to the square of its path length i.e. E=L ² .