WO2003013032A1

WO2003013032A1 - Optical signal regeneration

Info

Publication number: WO2003013032A1
Application number: PCT/AU2002/001018
Authority: WO
Inventors: Gregory Raymond Collecutt; Peter David Drummond
Original assignee: The University Of Queensland
Priority date: 2001-07-31
Filing date: 2002-07-31
Publication date: 2003-02-13
Also published as: AUPR673701A0

Abstract

All-optical regeneration of an optical signal is achieved by utilizing a non-degenerate parametric interaction in a one-dimensional optical waveguide (5). The optical signal (1) comprises a series of pulses of light which have been degraded. The signal pulses are combined (4) with pump pulses of light (2) of a predetermined energy, centre frequency and duration. The pump pulses have a bit rate derived from the optical signal by averaging. The pulses of the optical signal overlap with the pump pulses and are propagated through the waveguide (5) where they interact in a spatially dependent manner. The pump pulses selectively amplify the frequency components of the signal pulses that travel at the same speed as the pump pulses. The amplified signal pulses are then isolated by filtering (6). The regenerated optical signal (8) is not only re-amplified, re-timed and re-shaped, but also re-tuned.

Description

OPTICAL SIGNAL REGENERATION

This invention relates to optical signal processing, particularly (but not solely) for use in telecommunications applications. In particular, the invention is directed to a method and apparatus for ail-optically regenerating an optical binary data stream within a single mode waveguide, so that the optical pulses of the data stream are re-amplified, re-timed, re-shaped and re-tuned.

BACKGROUND ART The common method of transmitting data within an optical fibre is to convert it into binary form and send it through the fibre as a sequence of pulses of laser light. That is, a binary "1" is usually represented by the presence of a pulse, while a binary "0" is usually represented by the absence of a pulse. The pulses (or absence of pulses) are spaced at predetermined intervals. The rate at which the pulses appear (in a signal comprised solely of 1s) is known as the bit rate.

Such pulsing of laser light can be thought of as an amplitude modulation of some base carrier frequency, ω . (Note that ω is expressed in radians per unit time, so that ω = 2 f, where / is the frequency in cycles per unit time). This amplitude modulation causes the signal to be comprised of many frequency components spread over the range ω±ω_b , where 2ω_b may be considered to be the bandwidth of the signal. The bandwidth of the signal is inversely proportional to duration of the pulses. Transmitting data at a faster bit rate necessitates the use of shorter pulses, and so the bandwidth of the signal becomes proportional to the data rate. Hence data rate and signal bandwidth are commonly used interchangeably.

At some point in the fibre, each pulse of light, on an individual basis, exhibits the following properties:

• E , total energy of the pulse. • t_c , centre time (time at which half of the pulses energy has already arrived at the point in question).

• ω_c , centre or carrier frequency (frequency at the centre of the pulse's spread of frequency components).

• t_h : Half duration time or half-width, which typically is defined as the time period for which the intensity of the pulse is greater than half of its maximum intensity.

However, as these pulses of light propagate within the fibre, they degrade with distance. Primarily they lose energy due to impurities in the fibre which absorb the light and/or scatter the light into the fibre's cladding, but secondary effects include Raman interaction with sound waves within the fibre which causes additional noise on the signal. Consequently, simple propagation alone corrupts all four of the properties mentioned above to varying extents.

It is already known to use a fibre-laser to amplify signals propagating in an optical fibre, solely by optical means ("ail-optically"). However, known fibre-lasers have two significant limitations. First, they have restricted bandwidth, and are therefore not suitable for re-amplifying signals with a high bandwidth (which are commonly referred to as broadband signals). Secondly, the fibre-lasers add an additional random error to the centre frequency, ω_c , of the signal pulses. This is a problem because of a technical property of the fibre known as dispersion, or sometimes group velocity dispersion, as explained below.

Glass, like any other wave propagation medium, has what is known as a dispersion relation, which is a function k = g(ω) that defines the wavenumber of a wave as a function of its frequency ω . Wavenumber is related to its wavelength, λ , by

, _ 2π

^~Υ^'

The velocity with which the wave crests of the carrier wave travel is called the phase velocity, and is given by ω

Whereas the velocity with which the modulation (i.e. the envelope of the wave packet) of the waves travels is called the group velocity, and is given by

v = *<^

⁸ ω J_a which is the slope of the dispersion relation curve in the vicinity of the carrier frequency. This is usually not the same as the phase velocity. Group velocity dispersion is the rate at which the group velocity changes with frequency. Its most common definition is

β₂ (ω) -— = — — — . dω dω

The effect of group velocity dispersion is to cause the different frequency components within a single pulse to travel at different speeds, the net result being that after some propagation the pulse becomes chirped. This is a term used to describe a pulse where higher frequency components end up at one end of the pulse, and lower frequency components at the other, causing the duration of the pulse to increase, and the pulse shape to "slump". This effect can be negated over short distances either by using (near) zero dispersion fibre, or pre- or post-chirping the pulses with the opposite amount of chirp to that which they receive in the fibre. Similarly, pulses with different centre frequencies travel at different speeds. This in turn means that after amplification by known optical amplifiers, and some distance of further propagation, the time centres, t_c , of the pulses exhibit a random deviation, or jitter, known as Gordon-Haus jitter.

Therefore, after some distance of propagation within a fibre the pulses exhibit perturbations in all four of the above mentioned characteristics, namely: energy, timing, frequency, and duration. These perturbations must be corrected before the signal becomes too degraded or scrambled to permit the binary data to be recovered.

Traditionally the art has referred to "3-R" regeneration as being re- amplifying, re-timing, and re-shaping. Re-shaping primarily means correcting for the slump in pulse shape caused by dispersion, which can be considered as being correction of perturbations in pulse duration.

It is believed that known 3-R regeneration alone is not sufficient to enable a broadband long distance communication network. The primary cause of the need for re-timing is the error in pulse centre frequency, and if this error goes uncorrected then the collective ensemble of pulse centre frequencies will continue to diffuse causing the timing error problem to escalate with propagation. To date, while re-amplification may have been performed ail- optically, the remainder of the signal regeneration has been performed by electronic repeaters in which the signal is received using a photo-diode, processed electronically, and re-transmitted by another laser diode. The information is read electronically from the degraded data stream and re- modulated onto a new carrier wave. Thus while such electronic demodulation and re-modulation has a re-tuning effect, the effect is largely incidental and generally not recognised or explicitly stated. More importantly, electronic re- tuning has limited bandwidth and is not suited to achieving the full utilisation of the bandwidth available within optical fibres.

It is an object of this invention to overcome, or at least ameliorate, one or more of the abovedescribed disadvantages of the prior art, by providing method and apparatus for all-optical regeneration of optical signals, including effective "re-tuning".

SUMMARY OF THE INVENTION In one broad form, the invention provides a method of ail-optically regenerating an optical signal comprising a series of pulses of light of a nominal carrier frequency, the method including the steps of: providing a source of periodic pump pulses of light having a predetermined centre frequency, combining the optical signal with the pump pulses so that the pulses of the optical signal are concurrent with corresponding ones of the pump pulses, and propagating the combined signals through an optical waveguide so that the signals interact in a spatially dependent manner, wherein the pump pulses selectively amplify frequency components of the signal pulses that travel at the same speed as the pump pulses

In another form, the invention provides apparatus for Apparatus for all-optically regenerating an optical signal comprising a series of pulses of light of a nominal carrier frequency, the apparatus including: a source of periodic pump pulses of light having a predetermined centre frequency, a coupler for combining the optical signal pulses with the pump pulses so that the pulses of the optical signal are concurrent with corresponding ones of the pump pulses, and an optical waveguide connected to the output of the coupler such that, in use, the combined signals from the coupler propagate through the waveguide and interact in a spatially dependent manner, and wherein the pump pulses selectively amplify frequency components of the signal pulses that travel at the same speed as the pump pulses.

In a non-degenerate parametric interaction in the waveguide, energy is transferred from the pump pulses to the signal pulses with which they spatially overlap. Since the pump pulses selectively amplify the frequency components of the signal pulses that travel at the same speed as the pump pulses, the signal pulses are effectively re-tuned to the centre frequency with the same group velocity as the pump pulses. Moreover, the re-tuning is performed all-optically. (The term "all-optically" is intended to mean that the regeneration is performed optically, but does not exclude the use of ancillary electrical components). Typically, the source of periodic pump pulses is a mode locked pump laser that generates a uniform stream of pulses having a bit rate which is derived from the optical signal, e.g. by averaging. Such "timing recovery" ensures that, over a long term average, the pump pulses remain synchronised with the signal data stream (whose interval period may suffer longer term drift). The pump pulses have the same pulse duration (when non-chirped) as the nominal signal pulses (i.e. before degradation). The pump pulses may or may not be of the same carrier frequency as the signal pulses, and are of significantly higher energy than the signal pulses. The pump laser light is distinguishable from the signal pulse light by having a different carrier frequency and/or being orthogonally polarised to the signal light.

Typically, the signal pulses are combined with the pump pulses in a 3dB optical coupler, or similar device, so that the pulses are concurrent or otherwise aligned within the same physical space. The waveguide is typically a length of single mode optical waveguide that exhibits an interaction mechanism, so that energy may transfer from the pump pulses to the signal pulses in regions where they are spatially coexistent. Examples of such interaction mechanisms include type I and type II χ^{2) nonlinearity, four wave mixing using the self or cross ⁽³⁾ nonlinearity, and Raman interaction.

Preferably, the waveguide exhibits sufficient dispersion at the signal carrier frequency so that the length of the waveguide represents of order 1 dispersion length for the signal pulses.

Advantageously, the waveguide is constructed so that: (i) the pump signal centre frequency and the signal pulse carrier frequency have sufficiently identical phase velocity so that the relative phase between them does not shift by more than about — within the length of

the waveguide (known as phase matching), and

(ii) the pump pulse centre frequency has the same group velocity as that of the signal pulse carrier frequency to which the optical signal is being regenerated (known as group velocity matching).

Preferably, the energy of the pump pulses is sufficient high to ensure that amplification of a nominal signal pulse reaches saturation, i.e. its maximum value.

Following amplification in the waveguide, the amplified signal pulses are isolated from the combined signals, e.g. by filtering. Low level noise may suitably be removed from the filtered amplified signal, e.g. by using a saturable absorber. This enables such regeneration to be used repeatedly without substantial degradation of the signal to noise ratio.

The dispersion of the waveguide at the pump pulse centre frequency may be tailored so that the length of the waveguide also represents of order 1 dispersion length for the pump pulses.

The pump pulses may also be chirped so that, in the absence of a signal pulse, they are fully focused, i.e. they focus (or "waist") at some predetermined distance within the waveguide. This enables a greater range of timing error in the input signal to be regenerated.

Depending on the particular interaction used in the waveguide, this invention is able to extend significantly the band with of optical signals which can be amplified, as the interaction mechanism represents the primary limiting factor with regard to amplification band with.

The invention enables long distance propagation of binary optical data by providing "4-R" regeneration of the optical signal pulses, namely re- amplification, re-timing, re-shaping and re-tuning. In order that the invention may be more fully understood and put into practice, a preferred embodiment thereof will now be described by way of example only, with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a schematic block diagram of an optical signal re-generator according to one embodiment of the invention;

Fig. 2 is a diagram illustrating the relative timing and strengths of the signal and pump fields at point A in Fig. 1 ; and

Fig. 3 is a diagram illustrating the relative timing and strengths of the signal and pump fields at point B in Fig. 1.

DESCRIPTION OF THE PREFERRED EMBODIMENT The signal regenerator of the preferred embodiment of this invention employs type II parametric interaction, in which χ^{2) polarisation nonlinearity (or effective χ^{2) nonlinearity) is used to couple two orthogonally polarised light fields with frequencies 6^ and ω_} (known in the art as signal and idler fields) with a sum frequency (or pump) field at frequency ω_p = ω_s + ω_} .

Preferably, the signal and idler fields have the same frequency, i.e. ω_s = ω, and thus the pump frequency ω_p = 2ω_s , but this is not essential.

Since the signal field has a desired polarisation, this embodiment may operate in a polarisation sensitive, and a polarisation insensitive mode of operation. Further, it can be used to amplify either all of the signals in the fibre together, or (in a parallel, multiplexed configuration) just those signals occupying a specified window or frequency band. In the latter configuration, several substantially identical regenerators are normally used, and if each wavelength band was driven originally from the same clock then these parallel regenerators may be synchronised via the same timing recovery mechanism.

A schematic diagram of a signal regenerator 10 according to a preference embodiment of the invention is shown in Fig. 1. An optical waveguide, such as an optical fibre, serves as an input signal line 1 containing the optical signal pulses to be regenerated. In the simplest version illustrated schematically in Fig. 1 , the input pulses are plane polarised with a known polarisation, which is that of the signal mode of the waveguide. The input signal has a nominal carrier frequency of ω_s .

The signal regenerator 10 also includes a pump laser source 2. The pump laser produces periodic pulses of coherent, near monochromatic, plane polarised laser light at frequency ω_p . The polarisation matches that of the pump mode of the χ^{2) waveguide 5 (described below), and are output from the pump laser 2 to a fibre coupler 4 via an optical fibre or other suitable waveguide. The input signal pulses are fed by the input signal line 1 to the fibre coupler 4 also. The pump laser 2 is adjustable to vary the energy of the pump pulses. The pump laser pulses should have an approximately Gaussian temporal profile of similar duration (non-chirped) to the nominal input signal pulse duration. The pump pulses are also produced with the same pulse interval period or bit rate as the input signal, and they are synchronised, on average, with the input signal pulses so that a pump pulse and a correctly timing signal pulse are concurrent once the two are mixed at the fibre coupler 4 (as described below). The quality of the output timing is directly related to the quality of the pump pulse timing.

The pump output pulses may be chirped so that (in the absence of a signal pulse) they focus or waist by the end of propagation through the χ^{2) waveguide 5. This is further explained below. Although the performance of the signal regenerator 10 may be improved through the use of such chirped pump pulses, this is not essential. The pump laser 2 may be constructed from any number of component lasers, providing the final output pulses are as stated above.

The fibre coupler 4 is preferably a 3dB or similar coupler which mixes the output pulses of the pump laser 2 with the incoming optical signals in input signal line 1. The fibre coupler is a common device used in the telecommunications industry, and need not be described in detail in this application. The superposed signals at the output A of the fibre coupler 4 are shown in Fig. 2 (in computer simulated form) in which the field power of the signals is shown as a function of time. [For illustrative purposes, the time coordinate is normalised to units of input pulse width, while the power coordinate is expressed in arbitrary units. The signal is a binary signal, with ones represented by pulses, while zeros are represented by the absence of a pulse. The energy and time centres of the signal pulses have been varied slightly to simulate data in need of regeneration. Note also that the signal intensity has been enlarged by a factor of 1000 so that it can be easily seen alongside the pump pulses on a linear scale]. In the combined signal which is fed from the coupler 4 to the waveguide 5, the input signal pulses overlap with pump pulses.

The signal regenerator also includes means for maintaining the synchronisation between the pump pulses and the input signal pulses. If this invention were being built into an integrated photonics circuit, such synchronisation may be derived from a central oscillator. However, in a communication network, the synchronisation would normally be derived independently at each regeneration station. In the illustrated embodiment, a synchronisation feedback mechanism 3 is connected between the output of the fibre coupler 4 and the pump laser 2, and is used to adjust the pulse repetition rate of the pump laser 2 so that, on average, the output pulses from the pump laser 2 are synchronised with the signal pulses in the input signal line 1. Such synchronisation feedback mechanisms are known in the art, and need not be described in detail in this application.

The signal regenerator includes a χ ²⁾ nonlinear single mode waveguide 5 as a fundamental component thereof. The waveguide 5 may be a graded refractive index channel written within a large amount of bulk crystal. A suitable material for type II interaction is lithium niobate (LiNbO₃) carefully grown as a single crystal, or else grown as a periodically poled crystal (often referred to as PPLN, or Periodically Poled Lithium Niobate). With type II interaction, the orthogonal polarisations for the signal and idler modes are aligned with the ordinary and extraordinary modes of the waveguide, or vice-versa. If using a uniform crystal, the phase velocities of the signal, idler, and pump modes should be sufficiently matched so that their relative phasing does not change by more than about — over the length of the

4 waveguide. This usually requires maintaining the waveguide at a specified temperature. Alternatively, the poling period of a periodically poled crystal must generate quasi phase matching for the field modes used.

The dispersion of the waveguide material at the pump laser wavelength should preferably be equal to, or less than, the dispersion at the signal wavelength. If it is of similar magnitude then "chirping" of the pump pulses may be advantageous as chirped pump pulses are broader in duration to begin with, and then bunch up or focus as they propagate in the waveguide. This enables a greater range of timing error in the signal pulses to be corrected for. On the other hand, if the dispersion at the pump frequency is significantly less than the dispersion at the signal frequency, there will be little advantage in using chirped pump pulses since little focusing effect may be obtained. These dispersions are either both normal, or both anomalous. (In other words β₂(k_P) and β₂(k_s) are of the same sign).

The length of the waveguide 5 ought to be such that the signal pulses undergo of order 1 dispersion length within the waveguide, or mathematically β₂(k_s)L

1,

(2^{/ 2} where L is the length of the waveguide. The pump pulse energy is adjusted so that a nominal signal pulse causes the maximum possible conversion of energy from pump field to signal and idler fields (saturation).

The output from the waveguide 5, at point B in Fig. 1, is shown in Fig. 3. It is to be noted that the pump pulses have been depleted where they coincide with signal pulses, and the signal pulses have been regenerated in terms of both energy and timing. This regeneration process also corrects variations in signal centre frequency and pulse width. Smaller signal pulses undergo proportionally more amplification than larger signal pulses.

Since the signal pulses are amplified by energy conversion from the pump field mode, and since the energy in the pump mode is localised in space and travelling with a known group velocity, the amplification process is strongest for frequency components of the signal pulse that travel in step with the pump pulse. In other words, the pump pulses which are spatially superposed with the signal pulses and travel therewith selectively amplify the frequency components of the signal pulses that travel at the same speed on the pump pulses. Hence the signal pulses are "re-tuned" to the frequency that has the same group velocity as the pump frequency.

The group velocity of the pump pulses must be matched to that of the nominal signal pulse. Mathematically this may be expressed as

The output of the waveguide 5 is preferably passed through a colour and polarisation filter 6 to isolate the regenerated signal. Such filters are commonly used in the telecommunications industry, and need not be described in detail in this application. A saturable absorber 7 is preferably connected to the output of the filter 6 to remove inter-pulse noise from the regenerated signal and thus improve the signal to noise ratio. The saturable absorber 7 has a nonlinear response such that (unwanted) small signals are highly attenuated while the (desired) larger signals are only slightly attenuated. These are known in the art and need not be described in detail.

After passing through the saturable absorber 7, the regenerated signals which have been re-amplified, re-timed, re-shaped and re-tuned, pass into an output signal line 8 to continue propagation in the communications network (not shown). The output signals may be combined with other regenerated signals of a different frequency band if a multiplexed configuration has been used.

The connecting optical fibres or other suitable waveguides used within the signal regenerator 10 should preferably be polarisation preserving. The above described signal regenerator device performs very well in computer simulations using non-chirped pump pulses, with the dispersion in the waveguide 5 at the pump wavelength being equal to that at the signal wavelength, and the phase matching being exact. Signal pulse timing and centre frequency at output were primarily only related to their same parameters at input; signal pulse timing errors of up to +o.25t_Λ were reduced by a factor of 4, and centre frequency errors of up to five times the bit data rate were reduced by a factor of 2. The signal pulse energy and width at output were primarily sensitive only to fluctuations in input signal pulse energy and timing. For fluctuations in input signal pulse energy of ±3dB , and timing error of ±o.25t_A , the fluctuations in output signal pulse energy were about ± dB , and the fluctuations in output signal pulse width were about +0.1t_A ■

If the waveguide dispersion at pump and signal frequencies are similar in magnitude, it has been found that by chirping the pump pulses so as to waist by the end of the waveguide 5, the signal regenerator of this invention can correct timing and frequency errors or deviations over a wider range. Preliminary investigations indicate that the input ranges of both timing and frequency errors may be doubled without adversely affecting the output signal pulse energy and width.

Improved performance can also be obtained by controlling the polarisation of the input signal pulses. Although the above described signal regenerator is applied to input signals having a known plane polarisation, in current communication systems the input signals often arrive with unknown polarisation. [This may not be a necessary feature of future communication systems, as there is known technology (namely polarisation-preserving optical fibre) for ensuring a known polarisation]. However, there are several ways to deal with the problem of unknown polarisation of the input signals, even without using a different type of optical fibre.

The first involves passing the signals through a controllable polarisation rotation plate, and then a polarisation splitter which is aligned with the desired input polarisation. Since the signal polarisation drifts relatively slowly with time, an active electronic feedback mechanism may be employed to continually adjust the polarisation rotation plate to minimise the unwanted output of the polarisation splitter, with the desired output being used as the input to the signal regenerator of this invention. In this manner, the signal that is input to the waveguide 5 will have nearly all of its power in the desired polarisation mode.

In an alternative method, the polarisation rotation plate and associated electronics can be eliminated by passing the input signal directly into the same polarisation splitter as before, but this time the two outputs are then used as the inputs for two copies of the parametric signal regenerator 10. The two outputs are then recombined by using a polarisation splitter in reverse. However, one possible disadvantage of this dual regenerator approach is that the total output power now becomes sensitive to the signal polarisation, with possibly about a 2-3dB variation over the range of possible input polarisation angles.

The parametric signal regenerator of this invention has the advantage that it enables optical signals to be regenerated over a much greater bandwidth than hitherto possible. For example, a known erbium-doped fibre laser can be used to re-amplify optical signals and has a frequency bandwidth of the order of 1THz, i.e. the maximum signal data rate which this device can successfully re-amplify is around 1THz. However, the other signal parameters, such as timing, width, and tuning, are regenerated electronically. As the bandwidth limit for electronic processing is of the order of 10GHz, there is a need for wavelength division multiplexing. The duplication inherent in multiplex circuits adds to the cost of signal regeneration, and also increases transmission times. Even still, the bandwidth is limited to that of the erbium-doped fibre laser used to amplify the signal. The parametric signal regenerator of this invention on the other hand, utilises an optical non-linear effect which operates on a femtosecond time scale, and thus its bandwidth is ultimately limited only by the transmission window of the fibre, i.e. around 100THz. Thus, the optical signal regenerator of this invention can provide a bandwidth improvement of the order of 100 times greater than current signal regeneration systems.

The illustrated components can be implemented physically in any suitable manner, but most preferably in all solid-state configurations using planar integrated optics.

The foregoing describes only some embodiments of the invention, and modifications which are obvious to those skilled in the art may be made thereto without departing from the scope of the invention as defined in the claims appended hereto.

For example, in another embodiment of the invention, type I χ^{2) interaction is utilised within the waveguide 5. In such interaction, the signal and idler field modes have the same polarisation, but different (i.e. non-degenerate) frequencies. The same phase matching and group velocity matching requirements must still be met.

In general, any parametric three wave mixing or four wave mixing process may be used instead of type I or type II χ^{2) interactions. Examples include four wave mixing using the self or cross χ^{3) nonlinearity, or Raman interactions leading to three or four wave mixing.

Other variations include splitting up the input into two separate polarisations, to be amplified by two separate parametric regenerators. This would have the advantage of allowing the retrofitting of pulse regenerators to existing optical fibres which are not polarisation preserving.

In another application of the invention, the input can be split into two or more wavelength channels to allow separate amplification of different wavelengths. This may be necessary if design parameters (such as dispersion, group velocity and phase mismatch) require amplification of different parts of the input spectrum with separate regenerators. Further, a very practical modification of the signal generator 10 is to include photonic crystal technology into the waveguide 5 in order to more precisely engineer the group velocity and dispersion characteristics of the waveguide.

Claims

CLAIMS:

1. A method of all-optically regenerating an optical signal comprising a series of pulses of light of a nominal carrier frequency, the method including the steps of: providing a source of periodic pump pulses of light having a predetermined centre frequency, combining the optical signal with the pump pulses so that the pulses of the optical signal are concurrent with corresponding ones of the pump pulses, and propagating the combined signals through an optical waveguide so that the signals interact in a spatially dependent manner, wherein the pump pulses selectively amplify frequency components of the signal pulses that travel at the same speed as the pump pulses

2. A method as claimed in claim 1, wherein the pump pulses have a bit rate which is derived from the optical signal by averaging,

3. A method as claimed in claim 1 , wherein the optical signal pulses and the pump pulses interact in the waveguide in a type II χ^{{ )} interaction.

4. A method as claimed in claim 1 , wherein the frequencies of the optical signal pulses and the pump pulses in the waveguide are phase matched.

5. A method as claimed in claim 1 , wherein the waveguide is constructed from a periodically poled crystal with a poling period such that quasi phase matching of the frequencies of the optical signal pulses and the pump pulses is achieved.

6. A method as claimed in claim 1 , wherein the waveguide has sufficient dispersion at the carrier frequency of the optical signal so that the length of the waveguide represents of order 1 dispersion length for the signal pulses.

7. A method as claimed in claim 1 , wherein the pump pulses have sufficient energy so that the amplification of the signal pulses reaches saturation.

8. A method as claimed in claim 1 , wherein the pump pulses are chirped.

9. A method as claimed in claim 1 , further comprising the step of isolating the amplified signal pulses from the combined signals after propagation through the waveguide.

10. A method as claimed in claim 9, further comprising the step of removing low level noise from the isolated signal pulses.

11. Apparatus for all-optically regenerating an optical signal comprising a series of pulses of light of a nominal carrier frequency, the apparatus including: a source of periodic pump pulses of light having a predetermined centre frequency, a coupler for combining the optical signal pulses with the pump pulses so that the pulses of the optical signal are concurrent with corresponding ones of the pump pulses, and an optical waveguide connected to the output of the coupler such that, in use, the combined signals from the coupler propagate through the waveguide and interact in a spatially dependent manner, and wherein the pump pulses selectively amplify frequency components of the signal pulses that travel at the same speed as the pump pulses.

12. Apparatus as claimed in claim 11 , further comprising a timing recovery device for synchronising the pump pulses with the pulses of the optical signal, the timing recovery device deriving a bit rate for the pump pulses from the optical signal by averaging.

13. Apparatus as claimed in claim 11, wherein the optical waveguide is a single mode waveguide, and provides type II χ⁽²⁾ interaction between the optical signal pulses and the pump pulses.

14. Apparatus as claimed in claim 11 , wherein the optical waveguide is constructed such that the frequencies of the optical signal pulses and the pump pulses are phase matched.

15. Apparatus as claimed in claim 11, wherein the optical waveguide is constructed from a periodically poled crystal with a poling period such that quasi phase matching is achieved.

16. Apparatus as claimed in claim 11 , wherein the optical waveguide has sufficient dispersion at the signal carrier frequency so that the length of the waveguide represents of order 1 dispersion length for the signal pulses.

17. Apparatus as claimed in claim 11, further comprising a filter connected to the output of the optical waveguide for isolating the amplified optical signal pulses from the combined signals.

18. Apparatus as claimed in claim 17, further comprising a saturable absorber connected to the output of the filter for removing low level noise from the isolated signal.

19. Apparatus as claimed in claim 18, wherein two or more of the source of pump pulses, the coupler, the waveguide, the filter and the saturable absorber are connected by optical fibre which is polarisation preserving.

20. Apparatus as claimed in claim 11 wherein the source of periodic pump pulses is adjustable to vary the energy of the pump pulses.