WO2012172309A1

WO2012172309A1 - Optical signal processing

Info

Publication number: WO2012172309A1
Application number: PCT/GB2012/051247
Authority: WO
Inventors: David John Richardson; Periklis Petropoulos; Radan Slavik; Francesca Parmigiani; Joseph Kakande
Original assignee: University of Southampton
Current assignee: University of Southampton
Priority date: 2011-06-16
Filing date: 2012-06-01
Publication date: 2012-12-20
Anticipated expiration: 2013-12-16
Also published as: GB201110212D0; GB2492068A

Abstract

A device for outputting an optical signal of a desired output frequency based on input optical signals of first and second input frequencies. First and second input optical signals are four- wave mixed in a highly nonlinear fiber and generate a beating signal which is supplied to a frequency divider which divides the beat frequency by 1/N, where N is an integer of 2 or more. The divided down signal is used to control an optical modulator to shift the frequency of the first input optical signal by 1/N of the beat frequency. The device may be used for carrier regeneration of phase modulated data signals where the first optical signal is a pump and the second optical signal is the data and the frequency shift moves the pump frequency to the data frequency.

Description

TITLE OF THE INVENTION

Optical Signal Processing

BACKGROUND OF THE INVENTION

This invention relates to optical signal processing, and more especially but not exclusively to the processing of digital coherent signals.

The gap between the bandwidths attainable with optical signals and the bandwidth of digital or analog electronics needed to process these optical signals creates a technical challenge.

Processing optical signals with bandwidths beyond the speed of state-of-the-art electronics is therefore challenging and invites all-optical processing solutions to address the high

bandwidths. However many processing functions are difficult or impractical to achieve using existing optical methods. Techniques that open up new ways to process optical signals are therefore of significant interest, particularly in fields such as optical communications, test and measurement and sensing. Moreover, given recent improvements in laser technology, it is becoming practical to use not only the amplitude, but also phase to carry information. The preferred encoding techniques, such as CSRZ (carrier suppressed return to zero), DB

(duobinary), N-PSK (N-level phase shift keying) and -QAM (M-level quadrature amplitude modulation) remove the carrier at the transmitter, which brings benefits like suppression of undesired non-linear effects in the optical fiber transmission line. During detection at the receiver end, the highest sensitivity is obtained by the use of coherent homodyne detection, a process that involves mixing the received optical signal with a stronger beam locked in phase to its original carrier.

As an example, consider the detection of advanced modulation format signals which look set to become standard in future optical networks due to the high spectral efficiencies they offer. It is well known that optical coherent reception should provide excellent system performance both in terms of sensitivity and resilience to noise. However, to date there are no fully homodyne optical coherent receivers on the market due to issues associated with recovering a local optical reference locked to the incident carrier-less data signal. Recovering a carrier wave from a carrier-less data signal such as QPSK or 16 QAM is a significant technical challenge. The function of generating such an optical reference from an incoming data signal is referred to as carrier recovery.

As a consequence of these technical difficulties, two less optimal methods are currently used to decode such phase encoded signals. The first is based on a differential approach, in which the phase of adjacent, (temporally separated), parts of the signal are used and compared to encode/decode the data. Systems based on this method are becoming available commercially. The second, method, currently limited to research laboratories, is to use a free running CW (continuous wave) laser with a frequency sufficiently close to that of the carrier to allow the use of electronic digital signal processing to remove the resulting intermediate frequency. Although powerful, this method relies entirely on electronic processing which in most current instances needs to be performed off-line, i.e. not in real time, in a computer in the laboratory. Although dedicated DSP (Digital Signal Processing) chip sets are under development that should ultimately allow real-time operation and adoption outside of the laboratory, given the high bandwidth of modern communication systems, this electronic processing will be complex, power hungry and will in any event ultimately be slower than what would be desired owing to the inevitable limitations imposed by electronic processing speeds. A DSP approach would also be sensitive to various sources of noise, e.g., noise associated with a free running local oscillator.

If an efficient way of performing carrier recovery can be developed then it should be possible to develop fully optical homodyne coherent receivers which would provide significant advantages in terms of bandwidth, performance, and energy efficiency that can displace these DSP-based coherent receivers.

To obtain a carrier wave in order to decode a carrier-less signal several methods are known. A feed-back method is known (Ref. 1, 11). A feed-forward method is also known (Ref. Ill and IV). It is also known that the problem can be avoided completely by delivering the carrier as a separate signal (Ref. II) but which is clearly not ideal. The feedback methods lack speed which is limited by delay in the feedback loop. The existing feed-forward schemes are limited to processing signals with bandwidths less than that of electronics (e.g. , in Ref. Ill, 10 GHz signals are processed with >10 GHz electronics), or generate carriers at different carrier frequencies further away from the original carrier frequency than the signal bandwidth (e.g., in Ref. IV, 40 GHz signals are processed with a carrier generated 200 GHz away from the original carrier and which is beyond the capabilities of electronics). Even when the carrier is obtained, it is generally not straightforward to shift its frequency by more than is practical using electronically driven approaches (e.g. , >100 GHz) The existing techniques consist of using various optical combs (Ref. V).

Whilst we have focused above on a discussion of coherent reception the same issues hold for many other signal processing applications in telecommunications and in principle it would be highly advantageous to carry out some (if not all) of the processing directly in the optical domain if suitable approaches could be developed. In many of these applications all-optical recovery of the carrier will be essential. Typical examples include signal demultiplexing (to several lower- speed streams that can then more easily be processed by electronics), optical regeneration, and signal format conversion.

SUMMARY OF THE INVENTION According to a first aspect of the present invention there is provided a frequency modulating device for outputting an optical signal of a desired output frequency based on input optical signals of first and second input frequencies, the device comprising: first and second inputs for receiving first and second input optical signals of first and second input frequencies; an optical mixing path arranged to receive and propagate at least respective components of the first and second input optical signals, or signals phase synchronously derived therefrom, so that the components mix to generate a beating signal component with a beat frequency; a frequency divider having a bandwidth greater than the beat frequency arranged to receive the beating signal component at the beat frequency and generate as an output a frequency divided signal having a frequency 1/N of the beat frequency, where N is an integer of 2 or more; an optical modulator operable to shift the frequency of an optical signal input thereto by the frequency of a control signal input thereto, the modulator being arranged to receive a component of the first input optical signal, or of an optical signal phase synchronously derived therefrom, as its input optical signal and to receive the frequency divided signal output from the frequency divider as its control signal, thereby to generate an output optical signal shifted by a fraction 1/N of the beat frequency.

According to a second aspect of the present invention there is provided a frequency modulation method for outputting an optical signal of a desired output frequency based on input optical signals of first and second input frequencies, the method comprising, supplying first and second input optical signals of first and second input frequencies; optically mixing at least respective components of the first and second input optical signals, or signals phase synchronously derived therefrom, so that the components mix to generate a beating signal component with a beat frequency; frequency dividing the beating signal component into a frequency divided signal having a frequency 1/N of the beat frequency, where N is an integer of 2 or more; and modulating a component of the first input optical signal, or of an optical signal phase

synchronously derived therefrom, with the frequency divided signal to perform a frequency shift of 1/N of the beat frequency.

In the approach proposed herein, the carrier wave is generated from the carrier-less signal at the same frequency in a feed-forward fashion. In our proof-of-concept experiment, carrier waves were recovered from 56 GHz phase-encoded (carrier-less) data streams using electronics with < 10 GHz bandwidth.

The beating may be between the source signals or between FWM products derived from the source signals, or a combination of both.

In some embodiments, the beat frequency is the difference frequency between the first and second input frequencies ω_Α - ω so that the frequency divided signal has a frequency (ω_Α - ω_Β)/Ν and the frequency of the output optical signal is ω_Α + (ω_Α - ω_Β)/Ν. In some embodiments, the optical mixing path has a portion in which the respective

components of the first and second input optical signals pass through a third-order non-linear medium to induce four wave mixing (FWM) between the first and second input optical signals. This means that the FWM generates at least one idler which can be used, if desired, as one of the signals for generating the beat frequency. In one such embodiment, the idler used is at an idler frequency oj_idter = 2ω_Β - ω_Α , where ω_Α and ω_Β are the first and second input frequencies, and wherein the beating signal component is between the idler and the pump so that the beat frequency is ω_Με( - ω_Α = 2(ω_Β - ω_Α). This device may further comprise a pump source operable to generate the first input optical signal as a continuous wave and supply the pump to the first input and wherein, when the second input optical signal is a data signal from a transmission line bearing data which is phase encoded into N states separated equally in phase, the device recreates a carrier which is substantially free of phase modulation as the output optical signal, since the modulator shifts the frequency of the pump to that of the data signal by the operation ω_Α + 2(ω_Β - ω_Α)/2 = ω_Α + ω_Β - ω_Α = ω_Β, and since the idler, and hence the beating signal component, is substantially free of phase encoding as a result of optical phase erasure by the four wave mixing. The signal is also free of pump noise, because pump phase variations are canceled out in this scheme. The first and second optical input signals can have alternatively orthogonal polarizations and a polarizer can be arranged before the frequency divider to filter out the first input optica) signal (i.e. the pump) prior to generation of the frequency divided signal.

A third input may be provided for receiving a third input optical signal of a third input frequency ojc, at least a component of which is supplied to the optical mixing path, so that components of the first, second and third input optical signals pass through the third-order non-linear medium to induce degenerate four wave mixing between both the first and second input optical signals and the first and third optical signals, thereby to generate first and second idlers at respective first and second idler frequencies u½_er1 = 2ω_Β - Μω_Α and u_ldi_er2 = u)c - Μω_Α , where M is an integer and where ω_Α and ω_Β are the first and second input frequencies, and wherein the beating signal component is between the first and second idlers so that the beat frequency is

Widierl - ω,οι_6Γ2 = 2(u)_B - Jc) .

The device may further comprise first and second pump sources operable to generate the first and third input optical signals as continuous waves and supply the first and second pumps to the first and third inputs and wherein, when the second input optical signal is a data signal from a transmission line bearing data which is phase encoded into N states separated equally in phase, the device recreates a carrier which is substantially free of phase modulation as the output optical signal, since the modulator shifts the frequency of the second pump to that of the data signal by the operation u>_c + 2(ω_Β - u>_c)/2 = L _c + ω_Β - u>_c = ω_Β, and since the first and second idlers are selected to be of Mth order where M=N-1 so that the first idler is substantially free of phase encoding as a result of optical phase erasure by the four wave mixing, thereby ensuring the beating signal component, is also substantially free of phase encoding. The data signal and second pump can be directed to propagate along the optical mixing path in opposite directions, and the first pump signal is split into respective components which are directed to propagate along the optical path in opposite directions.

In some embodiments, the mixing components include a first mixing component phase synchronously derived from and shifted in frequency with respect to the first input optical signal, and a second mixing component which is a component of the second input optical signal. The mixing components may optionally further include a third mixing component phase

synchronously derived from and shifted in frequency with respect to the first input optical signal in an equal amount and opposite direction to the first mixing component.

In some embodiments, the optical modulator is connected to receive as its input optical signal a component of the first input optical signal. However, in other embodiments, the optical modulator is connected to receive as its input optical signal a component of an optical signal phase synchronously derived from and frequency shifted with respect to the first input optical signal.

The frequency divider may comprise an optoelectronic converter having a bandwidth greater than the beat frequency and arranged in the optical mixing path to receive at least a component of the mixed optical signal and generate therefrom an electronic signal at the beat frequency, and includes an electronic frequency division module arranged to receive the electronic signal at the beat frequency and generate the frequency divided signal electronically, and wherein the optical modulator is electrically actuated with the electronic frequency divided signal as its control signal. The electrically actuated optical modulator can be an electro- or acousto- optical modulator. Alternatively, the frequency divider can be an all-optical component based , e.g., on injection locking, and the control signal input to the optical modulator is an optical control signal. When an all-optical frequency divider is used, its output is already the carrier wave in which case, a frequency shifter (modulator) would not generally need to be used.

Our invention is motivated by the telecommunications applications discussed above, but may also be of use in other fields in which optical signal processing is needed, in particular in optical sensing systems when considering the processing of extremely large bandwidth digital signals, and in many test and measurement applications, e.g. phase sensitive optical sampling.

BRIEF DESCRIPTION OF THE DRAWINGS

This invention will now be further described, by way of example only, with reference to the accompanying drawings^'. Figure 1 (a) is a schematic drawing of a carrier recovery unit of a first embodiment used to recover the carrier from a carrier-less phase modulated data signal.

Figure 1 (b) is a schematic graph showing the pump, data and idler of the first embodiment as a function of frequency.

Figure 1(c) shows the block structure of a receiver 40 incorporating the carrier recovery unit of Figure 1(a).

Figure 2(a) is a schematic graph showing the frequency relationships between the four wave mixing components in a second embodiment for regenerating the carrier from a carrier-less binary phase encoded signal.

Figure 2(b) is a schematic drawing of a transmission system with a carrier recovery unit exemplifying the second embodiment.

Figures 3(a)-(d) show results from the second embodiment of the spectra) characteristics when processing data with no residual dispersion for optical (a,b,c) and RF (d) spectra.

Figure 4 shows results for data experiencing 50 km of SMF-28 residual dispersion for optical (a) and RF (b) spectra for the second embodiment.

Figure 5 shows a static homodyne measurement in the temporal domain for the second embodiment.

Figure 6 shows a static heterodyne measurement in the RF frequency domain for the second embodiment.

Figure 7 is a constellation plot (a) and phase error graph (b) for a 20 Gbit s stream with no dispersive propagation for the second embodiment.

Figure 8 is a constellation plot (a) and phase error graph (b) for 20 Gbit/s stream with 50-km of dispersive propagation through SMF-28 for the second embodiment.

Figure 9 is a schematic graph showing the frequency relationships between the four wave mixing components in a variant of the second embodiment for regenerating the carrier from a carrier-less quadruple phase encoded signal.

Figures 10(a), 10(b) and 10(c) are schematic graphs showing the frequency relationships between the four wave mixing components in a third embodiment for regenerating the carrier from a carrier-less binary phase encoded signal. Figure 10(d) is a schematic drawing of a carrier recovery unit of the third embodiment used to recover the carrier from a carrier-less phase modulated data signal.

Figure 11 is a schematic drawing of a frequency divider of low-bandwidth optical signals according to an alternative embodiment.

DETAILED DESCRIPTION

Figure 1 (a) is a schematic drawing of a carrier recovery unit used to recover the carrier from a carrier-less phase modulated data signal that has been received from a transmission line. The carrier recovery unit performs the function of shifting the frequency of a continuous wave (CW) pump signal to the carrier frequency of the data signal to recreate a carrier signal.

The data signal is at frequency ω_Β and is polarized with an s-polarization. The data signal is received on an input optical path 2 which is typically formed by an optical fiber, but could be another type of waveguide such as a semiconductor or lithium niobate planar waveguide. A CW pump source (not shown) outputs a pump signal at frequency ω_Α which is polarized with a p- polarization, i.e. polarized differently from the data signal. In view of the use of polarization specific data and pump signals, polarization maintaining (PM) fiber is used where necessary for various ones of the optical paths shown in the figure. The pump signal is received via an optical path 3 and then split by a splitter 4 into two components carried on optical paths 6 and 8. One component is supplied together with the data signal to a polarization multiplexer 10 which has an output optical path 12 which conveys both the data signal and the pump component. The optical path 12 is spliced to a portion of highly non-linear fiber (HNLF) 14 which is used in the device as a χ₃ third-order non-linear medium to induce four wave mixing (FWM) between the data signal and pump. The FWM generates an idler signal at frequency

which can be rewritten as ω_Β = uj.aier + ω_Α)/2 = ω_{Α +} (ω^ - ω_Α)/2

From this equation it is evident that a signal at the data frequency ω_Β and hence the carrier could be obtained by dividing by 2 a signal at the beat frequency between the idler and the pump and then adding it to a signal at the frequency of the pump ω_Α. This will be technically feasible so long as the beat frequency is within the bandwidth of the electronics (or optics) used to carry out the division.

Figure 1 (b) is a schematic graph showing the pump, data and idler as a function of frequency, where the data is shown smeared over a range of frequencies to indicate the data bandwidth. The transmission fiber is likely to be tens or hundreds of kilometers long in a real case. The figure also indicates a preferred range of frequency separation of the pump and idler which is 2-4 GHz, corresponding to a frequency separation of 1 -2 GHz between the pump and data. Although these precise values are not significant, what is important is that the frequency separation is sufficiently small to lie within the bandwidth of the processing electronics used for feed-forward frequency shifting which has yet to be described.

The HNLF 14 is spliced to another portion of ordinary PM fiber 16 which leads to a polarizer 18 which serves to filter out the s-polarization and hence filters out the data signal. The output of the polarizer 18 is supplied to a further portion of optical fiber 20 which terminates at a detector 22, such as a PIN detector, which converts the optical signal output from the optical fiber 20 into an electronic signal. The detected electronic signal includes a component caused by beating between the pump and idler. The beat frequency is given by:

Qbeat = Midler - ω_Α = 2(ω_Β - U)_A)

The detector 22 detects the beat Q_beai. Any parasitic signal originating in beating between the residual data signal and pump or idler d is filtered out of the detected electronic signal in a filter 24 which outputs the filtered electronic signal predominately containing the beat frequency to a divider 26. The divider 26 serves to divide the beat frequency by N where N is an integer of 2 or more. The value of N is determined to match the degree of phase encoding in the data signal. For example, with binary encoding such as BPSK (binary phase shift keying) N=2 and with four- level encoding such as QPSK (quadruple phase shift keying) N=4. In other words, the denominator in the division is set equal to the number of phase encoded states separated equally in phase.

In the general case, the divider will output a signal with frequency 2( J_B - ω_Α)/Ν

In the present example we assume BPSK so that N=2 so the frequency is

2(ω_Β - ω_Α)/2 = ω_Β - ω_Α

This electronic signal is then supplied as a control signal on an electrical line 28 to an electrically actuated optical modulator 30, such as an electro-optic or acousto-optic modulator, which is arranged to receive and modulate a component of the pump signal, namely the component not used in the FWM conveyed on optical path 8. The modulator 30 acts to shift the frequency of the pump to that of the data signal by the operation: ω_Α + 2(ω_Β - ω_Α)/Ν which in the example of N=2 is: ω_Α + 2(ω_Β - ω_Α)/2 = ω_Α + ω_Β - ω_Α = ω_Β

Not only is the output of the modulator at the correct frequency for the carrier, it also fulfils the requirement that the carrier is free of the data-related phase modulation since the beat frequency has been extracted from the idler and the pump. The pump is clearly free of phase modulation, since it was generated independently of the data signal. However, the idler is also substantially free of the phase modulation as a result of optical phase erasure by the four wave mixing. The recovered carrier is also free from any noise presented inherently in Pump (the result is independent on ω_Α).

A key advantage of the proposed scheme is that it cancels out pump noise in the process of recovering the carrier frequency. This allows the carrier to be recovered better than the linewidth of the pump. This is significant, since even very high quality state-of-the art pump lasers have a linewidth of hundreds of Hz, and standard pumps would have linewidth up to 100 KHz. Since α½_ΘΓ = 2ω_Β - oo_A this would result in the idler varying by the same amount due to the pump instability which is commonly referred to as limited linewidth. In the experimental example described below, we used pumps with ~1 kHz linewidth, but the carrier was recovered with <1 Hz precision (1000 times below that of the pump). The carrier is thus generated at its original frequency and can be recovered very precisely free from limitation by the linewidth of the pump laser(s).

The recovered carrier is then output on optical path 32

As alluded to above, the frequency separation between pump and data, and hence pump and the resultant idler generated by four wave mixing between the data and pump, is chosen in combination with the bandwidth of the electronic components, specifically the divider. In other words, the bandwidth of the divider and frequency of the pump source are jointly specified to ensure the divider has sufficient bandwidth to process the beat frequency. In summary, the present design provides a feed-forward scheme that allows relatively slow, commercially available, electronics to recover a carrier in real time from a relatively fast optical phase encoded signal.

It is noted the scheme of Figure 1 (a) can be simplified by removing the polarization separation. This will tend to increase noise at the detector as a result of the data signal so more

sophisticated filtering in the electronics prior to frequency division may be needed.

Figure 1(c) shows the block structure of a receiver 40 incorporating carrier recovery as described above. A carrier recovery unit 44 incorporates the components illustrated in Figure 1 (a) and receives a component of the data signal and the pump on respective optical paths 2 and 3 which are also shown in Figure 1 (a). The data signal received by the carrier recovery unit 44 is a component of the incoming data signal tapped off the main transmission data line 47. The pump is received on the optical path 3 also shown in Figure 1 (a) from the pump source 42. The pump source 42 is shown as being an integral part of the receiver 40. It will generally be the case that the pump source will be located generally on the receiver side of the transmission system even if it is not integral with the carrier recovery unit and other receiver components. A demodulator 46 then receives both the data signal on line 47 and the recovered carrier on line 32 and outputs the demodulated signal on output line 49. It will be understood the demodulator 46 may be of any known design and may be all-optical, optoelectronic, part optical and part electronic in any desired combination. Consequently the demodulated signal output on line 49 may be optical or electronic.

Figure 2(a) is a schematic graph showing the frequency relationships between the four wave mixing components in a second embodiment which uses two pumps P1 and P2. Compared with the first embodiment a second pump P2 at frequency u>c which is approximately coincident with the frequency of the data signal, although whether intended or not the frequency of the second pump will not be exactly coincident with the data signal frequency ω_Β. The first pump P1 at frequency ω_Α is selected to be offset from the data signal frequency ω_Β by an amount which avoids mutual overlap. The first pump P1 is used here to frequency offset both the idler and the second pump P2 signal in order to facilitate their beating detection. As both data and idler are offset by the same pump, their difference (beating) is independent of the first pump P1 .

The left hand part of the figure shows the input signals prior to FW , namely the first and second pumps 1 and P2 and the data. The right hand part of the figure shows the frequency components after FWM in a HNLF or other third-order non-linear medium. The effect of the third-order non-linear medium is to induce degenerate FWM between P1 and the data signal to generate a first idler labeled Idlerl as well as degenerate FWM between P1 and P2 to generate a second idler labeled Idler2. The first and second idlers are at respective first and second idler frequencies. Jidlen ⁼ jdler2 ⁼ 2u⁾c -

More generally, higher order idlers (not shown) are also generated and have frequencies oj^Midi_er1 = 2U>B - Μω_Α u) _idler2 = 2ω₀ - Μω_Α where is an integer. The value M=1 relates to the illustrated first order idlers and values of M≥2 relates to the higher order idlers. The higher order idlers are not exploited in the present embodiment, but are used in a later embodiment.

In this second embodiment, the beat frequency on which the division is based is beating between the first and second idlers. The beating signal component between the first and second idlers has a frequency:

D eat = Wjdierl " Uidler2 ⁼ 2(ω_Β - UJc)

It is noted the beat frequency does not depend on ω_Α the frequency of P1. Thus, any phase noise present on P1 is not transferred to Q_beat

The approach of the second embodiment differs from that of the first embodiment where the beating used by the divider is between the sole pump and the idler, the idler being produced by FWM between the sole pump and the data. In this embodiment, the idlers are generated at frequencies out of the data bandwidth and are thus easier to be extracted.

Figure 2(b) shows an implementation of the second embodiment. A transmitter 1 is shown for phase encoding an optical signal. The transmitter is based on a 200 kHz linewidth laser outputting at 1555.5 nm which is modulated by a modulator to encode data according to the binary phase shift code NRZ-DPSK (non-return-to-zero differential phase shift keying) with pseudorandom data pattern of 2³¹-1 bit length with speeds of 0, 10, 20, 40 or 56 Gbits/s. The transmitter 1 transmits the data signal into a transmission line 45. A 50 km length of single mode fiber (SMF) was used in the tests. At the receiver-end of the transmission line, the data signal was passed through a polarization tracker 52 to stabilize drift in the polarization of the received signal. The polarization stabilized received data signal output from the polarization tracker 52 is supplied to the carrier recovery unit 44 via a portion of PM fiber 55 to an erbium doped fiber amplifier (EDFA) 57 which forms part of the carrier recovery unit 44. The amplified signal output from the EDFA 57 is then supplied via a further portion of PM fiber 59 to a fiber coupler 61.

The carrier recovery unit 44 also includes first and second pump sources 54 and 56 which are CW fiber-coupled lasers which generate the first and second pump signals P1 and P2. Pump P1 has a power of 100 mW and a wavelength of 1557.3 nm. Pump P2 has a power of 40 mW and a wavelength of 1555.6 nm. The pump P1 output from the first pump source 54 is split in a 50/50 fiber coupler 58 into two components on respective output fibers 60 and 62.

One P1 component 60 is combined by the above-mentioned fiber coupler 61 with the amplified data signal from the EDFA 57. The PM fiber 64 carrying the combined data and P1 signals is connected to an optical circulator 66 which routes the combined P1 and data signal to a first end of a 500 m length of HNLF 14 with a Ge-doped silica core which is the third-order nonlinear medium for inducing FWM. Propagation of the data and P1 through the HNLF 14 induces four wave mixing to generate a first idler, labeled Idlerl . From the second end of the HNLF 14, the data and P1 signals, now with the FWM-induced idler signal Idlerl , are routed via a further optical circulator 68 to the input of a wavelength demultiplexer 70. The FWM components are shown in Figure 2(a) and also schematically illustrated in Figure 2(b) by the small inset drawing to the right and slightly above the optical circulator 68. The frequency demultiplexer 70 has three output ports separated by 200 GHz and separates out Idlerl , data and P1 into respective outputs. The optical path carrying the data signal from the demultiplexer 70 can be supplied onwards as the data signal for demodulation in the demodulator, as indicated by reference numeral 47. The first idler output Idlerl passes through a variable attenuator 91 and a first injection locked laser (IL1) 93 to remove amplitude noise and then is conveyed on an polarization controller 96. Idlerl is used to generate the beat frequency signal in the manner conceptually described with reference to Figure 1 (a) and to be described in more detail further below. The P1 output is not needed.

Going back to the pump sources 54 and 56, the other component of the first pump signal P1 is combined with the second pump P2 at an 80/20 fiber coupler 80. The fiber coupler 80 receives pump P2 after it has been amplified by an EDFA 78. A variable attenuator 76 is also arranged between the P2 pump source 56 output and the EDFA 78. From the fiber coupler 80, the combined pumps P1 and P2 are then supplied via a portion of SMF (not P fiber) 82 and the previously mentioned optical circulator 68 to the second end of the HNLF 14 where the P1/P2 signals traverse the HNLF 14 in the opposite direction to the data/P1 signals. This bi-directional set-up avoids generating four wave mixing products from P1 , P2 and data that would occupy the same bandwidth as the two idlers of interest and thus would be difficult to be filtered out. Utilizing the same HNLF allows avoiding of excessive acoustic pick-up that would be obtained if two idlers were generated in two different fibers. The FWM products of P1 and P2 are directed by the optical circulator 66 to a portion of SMF 84. The FWM products are as shown in the small inset to the right and above the optical circulator and also shown in Figure 2(a). Namely, an idler labeled Idler2 is generated by the FWM of P1 and P2. An optical add drop multiplexer (OADM) 86 is used to drop the P2 signal onto a SMF fiber 88. The transmitted P1 and Idler 2 signals are then passed to a band pass filter 90 which filters out the P1 signal, thereby transmitting only Idler2 on a fiber 92. The fiber 92 carrying Idler2 is combined at a 70/30 fiber coupler 94 with the fiber 96 carrying Idler 1. The output fiber 98 from the fiber coupler 94 then provides a medium for Idlerl and Idler2 to co-propagate and beat. The output fiber 98 leads to a photodiode detector 22 where the incoming optical signal is converted into an electrical signal containing the beat frequency. The beat frequency component is amplified in a 10GHz RF amplifier 104 and then supplied to a digital frequency divider 26 which divides the beat frequency in the signal by 2. The output from the digital frequency divider 26 is then supplied as an electrical control signal on signal line 28 to a 10 GHz electro-optic phase modulator 30 to modulate the incoming P2 signal which it receives by shifting its frequency to the frequency of the data signal and thereby regenerate the carrier through the shift: ooc + Q_beat/2 = IOB

After transit through a variable attenuator 106 and a second injection locked laser (IL2) 108, the recovered carrier is output from the carrier recovery unit 44 on a fiber line 32.

In order to test the system capability to deal with a signal carrying data, we tested it using DPSK modulated data at various data rates (up to 56 Gbit s) straight from the transmitter. To evaluate how the scheme would cope in presence of a linear impairment occurring in a typical network (chromatic dispersion), we tested it and also in the presence of high residual dispersion as introduced by a 50 km length of SMF-28 fiber.

Figures 3(a)-(d) show results from the second embodiment of the spectral characteristics when processing data with no residual dispersion for optical (a,b,c) and RF (d) spectra. Figures 3(a) to (c) are graphs of power spectral density against wavelength. Figure 3(a) shows generation of Idlerl from FWM of P1 and data. Figure 3(b) shows generation of Idler2 from FWM of P1 and P2. Figure 3(c) shows the recovered carrier directly after the modulator (heavier line of lower amplitude) and then after the injection locking laser IL2 which removes harmonics of the phase modulator used instead of a single-sideband modulator that could have been used, but was not available (lighter line at higher amplitude).

Figure 3(d) is a plot of power spectral density against frequency of the beat signal input to the divider for two modulation frequencies, i.e. 10 GHz (solid line) and 56 GHz (dotted line).

Figure 4 shows data collected after experiencing dispersion by transmission through a 50 km length of SMF-28. Figure 4(a) is a graph of power spectral density against wavelength showing generation of Idlerl from FWM of P1 and data. Figure 4(b) is a plot of power spectral density against frequency of the beat signal input to the divider for two modulation frequencies, i.e. 10 GHz (solid line) and 56 GHz (dotted line).

We also carried out two further experiments to evaluate the quality of the recovered carrier and its sensitivity to the quality of the beat detection, frequency division and subsequent modulation.

Figure 5 is a graph of signal strength against time with an inset of the experimental set up for data collection showing results of the first experiment which was a static (data modulation off) homodyne measurement in the temporal domain. We observed interference between the original and recovered carrier. Due to thermal drift, the optical intensity went slowly from a minimum to a maximum as would be expected from our current implementation. The fast noise on the curve is due to the phase error which is clearly much less than pi. Figure 6 is a graph of RF power spectral density against frequency with an inset of the experimental set up for data collection showing results of the second experiment which was a static heterodyne measurement in the RF frequency domain. We phase modulated the original carrier by 140 kHz and observed the power spectral density of the beat signal. This represents the Fourier transform of the characteristics from the previous experiment (shifted by 140 kHz). The extremely narrow beat signal observed highlights the quality of the recovered carrier. The low-frequency noise observed is due to acoustic pick up in the fibers.

Figure 7 is a constellation plot (a) and phase error graph (b) for a 20 Gbit/s stream with no dispersive propagation, i.e. direct transmission from the transmitter to the receiver. The recovered carrier was used for homodyne detection. The results for 20 GHz (limit of our instrument) are shown.

Figure 8 is a constellation plot (a) and phase error graph (b) for 20 Gbit/s stream to be compared with Figure 7(a) and (b). In this experiment, a 50 km length of dispersion- uncompensated SMF-28 fiber is inserted between transmitter and receiver to show the effect of a transmission line and associated chromatic dispersion. After homodyne detection, we removed the chromatic dispersion using electronic post-processing.

The above-described scheme can be extended to cover non-binary modulation formats having an arbitrary number of levels, e.g. for quadruple phase encoded signals (QPSK). To remove the phase encoding from the data signal, a cascaded FWM process is needed as now discussed. Following that, an electronic divider capable of dividing by four (or other integer greater than 2) is needed rather than the divide-by-two frequency divider required for BPSK signals.

Figure 9 illustrates this generalization with the example of 4-level phase encoded data (QPSK modulation). Figure 9 is comparable with Figure 2(a) in that it shows the relevant FWM components on the right hand side. As stated above, in a BPSK scheme, the first-order idler derived from FWM of the first pump P1 and the data does not contain data modulation due to the effect of optical phase erasure and thus was used for beat detection. In the case of QPSK, one needs to go to the corresponding 3^rd order idler to obtain optical phase erasure, so we use the 3^rd order idler pair for the beating. In general, for an N-level phase encoded signal, phase erasure will be possessed by the pump-data idler in the (N-1)th order idler pair.

In the above-described embodiments, we used FWM processes in which part of the energy of pump(s) and data signal is transferred into another wave (idler). Due to phase/frequency synchronization required for FWM, the frequency/phase of the idler was determined by the frequency/phase of the pump(s) and data signals involved in the FWM. To extract this information, the idler was beaten with another signal prior to detection. This scheme is commonly referred to in the art as Phase Insensitive Amplification (PIA). However, FWM also has another property which can be exploited in what is referred to in the art as Phase Sensitive amplification (PSA). The following embodiment is based on PSA.

In FW , when, at the input of the nonlinear medium, there is already an optical field present at the frequency where the idler is to be formed, the idler interferes with that optical field. This interference can be destructive or constructive depending on the relative phase between the idler and the already-present optical field. The idler is thus effectively beating with the optical field, i.e. pump, provided at that frequency, or more precisely close to that frequency. Thus, this configuration transforms the phase information of interest directly into amplitude information that can then be detected directly by a photodiode. Because in PSA, the information of interest is carried by amplitudes rather than frequencies, it is more straightforward to use analysis in the temporal rather than in frequency domain.

Figures 0(a), 10(b) and 10(c) are schematic graphs showing the frequency relationships, between the four wave mixing components in a third embodiment based on PSA for

regenerating the carrier from a carrier-less binary phase encoded signal.

As shown in Figures 10(a) and 10(b), a first pump P1 will interact with the data signal to generate an idler at a frequency close to an existing optical field which is provided by a second pump P2. In other words using the same nomenclature as for the previous embodiments

The difference between Figures 10(a) and 10(b) is that in Figure 10(a) the pump spacing is smaller than the data bandwidth, whereas in Figure 10(b) the pump spacing is larger than the data bandwidth. This is not a difference of principle, but has practical implications for the design of any implementation of this embodiment as described further below.

For correct operation of PSA, the data signal must lie exactly midway between the pump frequencies, i.e. such that:

2ω_Β - ω_Α - ωο = 0

However, in the context of a demodulator decoding a received data signal that has been transmitted perhaps over a long haul optical fiber, the frequency of the data signal cannot be exactly known. Consequently it will not be possible to satisfy the above, condition, so instead some frequency offset Ω will exist, i.e.

2ω_Β - ω_Α - ω₀ = Ω (1 ) where Ω = Cl(t) since the three signals involved are not phase synchronized. Figure 10(c) shows this situation where the data signal is offset in frequency from the midpoint - labeled as JC - by the amount Ω.

When the signals are mixed in a chi-3 nonlinear media such as a highly nonlinear optical fiber (HNLF), PSA gain g' depends on the relative phase of the signals, for example g(t) = -function [ 2<t>_B(t) - Φ_Α(0 - 4>_c(t) ] which can be re-written as: g(t) = function [ 2co_Bt - to_At - u>_ct ]

For BPSK modulation format, the modulation is stripped off due to the factor 2 in the phase and thus GOB represents its carrier frequency leading to: g(u)) = function [ 2ω_Β - ω_Α - ω₀ ] (2) As mentioned above, so that gain does not vary with time, PSA is normally operated such as

2ω_Β - ω_Α - ooc = 0 which can be rewritten as

This equation shows why it is necessary that the data signal lies exactly in the middle between the two pump frequencies for the gain to be constant. As mentioned above, this condition is difficult to obtain practically, as Pumpl , Data, and Pump2 are not phase synchronized so that in reality an offset frequency Ω = Q(t) will exist. Combining equations (1 ) and (2) we see that:. g(u>) = function [Ω] (3)

The intensity of the data signal after PSA will thus oscillate with a frequency that is directly proportional to the offset frequency Ω. The intensity of the data signal can be extracted by a simple low-speed photodetector, i.e. a photodetector having a response which is sufficiently slow to filter out higher frequency amplitude modulation components. It is thus possible to measure the offset frequency Ω. With the offset frequency Ω established, it is possible to generate the carrier using frequency shifting similar to the previous embodiments with an electrically actuated optical modulator, such as an electro-optic or an acousto-optic modulator. A suitable scheme for achieving this is now described.

Figure 10(d) is a schematic drawing of a carrier recovery unit of the third embodiment used to recover the carrier from a carrier-less phase modulated data signal. The data signal is at frequency ω_Β which in the illustrated example corresponds to a wavelength of 1550 nm and is encoded according to differential phase shift keying (DPSK). A transmitter 1 is shown schematically which generated the DPSK signal further propagated along a transmission line 45 to a receiver (not separately marked) where it is split into a component for demodulation which is passed to line 47 and a component for carrier recovery which is passed on a line 2 This is as shown in Figure 1 (c) using the same reference numerals. It is noted that Figure 1 (c) applies completely also to the present embodiment with most of the components shown in Figure 0(d) being components of the carrier recovery unit.

To generate two phase synchronous pump signals P1 and P2, a CW laser source 65 is used with a frequency u>_Cw close to ω₈, i e. a wavelength close to1550 nm in this example, but offset slightly from the DPSK carrier frequency ω_Β, either to higher or lower frequency, to ensure the frequency offset Ω does not change sign as it varies during operation. This is necessary since in the described implementation the measurement of the offset frequency is only an absolute one, so changes in sign would not be detectable. It therefore needs to be known a priori whether the CW frequency is above or below the data signal frequency. The pump signals P1 and P2 at frequencies ω_Α and uo_c are generated from the CW laser source 65 by subjecting a component of the CW source output to duobinary modulation at a frequency of 10 GHz in a modulator 67. The pump signals P1 and P2 are thus synchronized in phase to each other and the CW laser source. Another component of the CW source output is passed on for carrier recovery on line 8. The components are split by a coupler 4.

The portion of the data signal on line 2 is combined with the two pump signals output from the duobinary modulator on line 6 at a coupler 10 and after amplification in an amplifier 69 is passed through a HNLF 14 to generate FWM. The HNLF 14 produces PSA gain variations at a frequency of 2Ω since one pump is Ω closer to the data (i.e. carrier) frequency and the other pump is Ω further away, as depicted in Figure 10(c).

The frequency of the gain oscillations generated by the beating between the second pump P2 and the idler generated from FWM of the first pump and the data in the PSA are then measured by a photodetector 22 after suitable filtering in a filter 21 to remove or at least suppress the pump signals. The signal at frequency 2Ω output by the photodetector 22 is then input into a divider 26 which carries out a divide-by-two operation and thus outputs a signal at frequency Ω on line 28. This signal is then used to drive an acousto- or electro-optic modulator 30 which receives the portion of the output from the CW laser supplied by line 8 and operates to up- or down- shift the CW laser light in frequency by Ω to the carrier frequency of the data signal. Carrier recovery is thereby completed by outputting the recovered carrier on line 32. Referring to Figure 1 (c), the recovered carrier is then input to the demodulator together with the portion of data signal supplied by line 47. Similar to the previous embodiments, functioning implementations rely on the frequency to be measured with the photodetector and handled by subsequent electronics, principally the divider, to have suitable values, i.e. be within the bandwidth of the detector and processing electronics.

From a practical point of view, it is also important to be able to filter out the pumps from the data signal prior to photodetection. This is straightforward for a situation as shown in Figure 10(b), since it can be done by frequency-based optical filtering. For a situation as shown in Fig. 10(a) using vector FWM provides a simple way to separate out the pumps from the data via polarization. Scalar FWM may also be used in addition, e.g. by narrow-bandwidth filters including an optical tracking oscillator realized through injection locking.

In the implementation just described, Ω is extracted by measuring oscillations of the gain of the data. Alternatively, Q could be extracted by measuring oscillations in the pump intensity. These would occur when the PSA is in saturation, since the gain in the data signal would drain energy from the pumps which would therefore also oscillate in unison with the data signal.

In a variant of the third embodiment, the design can be adapted for higher level modulation formats such as QPSK, which is 4-level, by placing the data signal at 1/n of the distance between the pumps (plus a small offset to ensure the sign of Ω does not change in use) where n is the level of encoding, rather than 1/2 way between as in the BPSK example above.

In another variant of the third embodiment, the acousto- or electro-optic modulator 30 is connected to receive the output from the duobinary modulator 67 which is the two pump signals P1 and P2. The acousto- or electro-optic modulator 30 will then function to synchronize the two pump signals to the data signal which is needed for phase regeneration applications (Ref. IV). Referring to Figure 10(d), the output of the modulator carried on line 32 will be the two pumps P1 and P2 now synchronized to the data signal by shifting both of their frequencies by the offset Ω.

Figure 11 is a schematic drawing of a frequency divider of low-bandwidth optical signals.

Unlike the previous embodiments, this embodiment does not use a third-order non-linear medium, but rather direct beating between two input optical signals. This embodiment is not specifically directed to telecommunications applications. Specifically, it is not envisaged that one of the inputs is phase encoded.

Referring to the figure, first and second inputs 200 and 202 are arranged to receive first and second input optical signals of first and second input frequencies ω_Α and ω_Β where ω_Α < ω_Β.

An optical coupler 204 splits the first optical signals into two components, one of which is mixed with the second optical signal by a further coupler 206, and the other of which is transmitted onto an optical fiber 208. The output from the further coupler 206 provides an optical mixing path 210 on which the first and second optica! signals propagate and mix to generate a beating signal component with a beat frequency ω_Β - ω_Α. lying in the radio-frequency (RF) range. A photodetector 2 2 converts the optical signal to an electrical signal and then a bandpass filter 214 filters out all but the beat frequency component in the electrical signal. The beating signal component is then input to a frequency divider 216 having a bandwidth greater than the beat frequency. The frequency divider 216 may be analogue or digital. The frequency divider 216 generates and outputs a frequency divided signal having a frequency 1 /N of the beat frequency, i.e. (U)B - ω_Α)/Ν where N is 2 or a larger integer. An optical modulator 218 is provided which is operable to shift the frequency of an optical signal by the frequency of a control signal. The optical modulator 218 is connected to the optical fiber 208 to receive a component of first input optical signal at frequency ω_Α as well as the frequency divided electronic signal at frequency (ω_β - ω_Α)/Ν output from the frequency divider 216 as its control signal via electrical line 217. The optical modulator then acts to shift the input signal to 1/N of the beat frequency above the first . input frequency, i.e. to ω_Α + (ω_Β - ω_Α)/Ν .

The divided RF signal obtained is used thus to shift the carrier frequency of signal A to obtain a new signal that has an instantaneous frequency that lies at a fraction of 1/N of the frequency intetva) between A and B from signal A, and (N-1)/N of the interval from signal B. An optical- electrical-optical frequency division has thus been performed.

In the above embodiments, the frequency shift may be realized using an acousto-optic modulator (AO ) for shifts of less than 400 MHz or an electro-optic modulator using amplitude, phase or single-sideband for shifts of up to 50 GHz. The modulator may also be followed by a narrow-bandwidth filtering process such as optical injection locking or Fabry-Perot filtering.

An alternative approach to that described above in order to process the beat or offset frequency is to use an all-optical frequency divider, rather than the combination of photodiode detector and electronic divider.

REFERENCES

I. Kim, K. Croussore, X. Li, and G. Li, All-Optical Carrier Synchronization Using a Phase- Sensitive Oscillator, IEEE Photon. Tech. Lett. 19, pp. 987-999 (2007).

II. M.J. Fice, A. J. Seeds, B.J. Pugh, J.M. Heaton and S. J. Clements, "Homodyne

Coherent Receiver with Phase Locking to Orthogonal-Polarization Pilot Carrier by Optical Injection Phase Lock Loop," Proc. OFC 2009, paper OTuG1 (2009).

III. S.K. Ibrahim et at. , "Novel Carrier Extraction Scheme for Phase Modulated Signals

Using Feed-Forward Based Modulation Stripping," ECOC 2010, We7.A4 (2010). Slavik, R., Parmigiani, F., Kakande, J.K., Lundstrom, C, Sjodin, M., Andrekson, P. , Weerasuriya, W., Sygletos, S., Ellis, A.D., Gruner-Nielsen, L, Jakobsen, D., Herstrom, S. , P elan, R., O'Gorman, Bogris, A., Syvridis, D., Dasgupta, S. , Petropoulos, P., Richardson, D.J., All-optical phase and amplitude regenerator for next-generation telecommunications systems

Nature Photonics 2010 pp.690-695

T. Sakamoto, A. Chiba, A. Kanno, I orohashi, and T. Kawanishi, "Real-Time

Homodyne Reception of 40-Gbs BPSK Signal by Digital Optical Phase-Locked Loop, European Conference on Optical Communications (ECOC 2010), 19-23 Sept., Torino, Italy, 2010.

Claims

CLAIMS What is claimed is:

1 . A frequency modulating device for outputting an optical signal of a desired output frequency based on input optical signals of first and second input frequencies, the device comprising:

first and second inputs for receiving first and second input optical signals of first and second input frequencies;

an optical mixing path arranged to receive and propagate at least respective

components of the first and second input optical signals, or signals phase synchronously derived therefrom, so that the components mix to generate a beating signal component with a beat frequency;

a frequency divider having a bandwidth greater than the beat frequency arranged to receive the beating signal component at the beat frequency and generate as an output a frequency divided signal having a frequency 1/N of the beat frequency, where N is an integer of 2 or more;

an optical modulator operable to shift the frequency of an optical signal input thereto by the frequency of a control signal input thereto, the modulator being arranged to receive a component of the first input optical signal, or of an optical signal phase synchronously derived therefrom, as its input optical signal and to receive the frequency divided signal output from the frequency divider as its control signal, thereby to generate an output optical signal shifted by a fraction 1/N of the beat frequency.

2. The device of claim 1 , wherein the beat frequency is the difference frequency between the first and second input frequencies ω_Α - ω_Β so that the frequency divided signal has a frequency (ω_Α - ω_Β)/Ν and the frequency of the output optical signal is ω_Α + (ω_Α - ω_Β)/Ν.

3. The device of claim 1 , wherein the optical mixing path has a portion in which the respective components of the first and second input optical signals pass through a third-order non-linear medium to induce four wave mixing between the first and second input optical signals.

4. The device of claim 3, wherein the idler used is at an idler frequency oo_idter = 2OOB - ω_Α , where ω_Α and ω_Β are the first and second input frequencies, and wherein the beating signal component is between the idler and the pump so that the beat frequency is u>_idier - ω_Α = 2(ω_Β - ω_Α).

5. The device of claim 4, wherein the device further comprises a pump source operable to generate the first input optical signal as a continuous wave and supply the pump to the first input and wherein, when the second input optical signal is a data signal from a transmission line bearing data which is phase encoded into N states separated equally in phase, the device recreates a carrier which is substantially free of phase modulation as the output optical signal, since the modulator shifts the frequency of the pump to that of the data signal by the operation ω_Α ⁺ 2(u>6 - ω_Α)/2 = ω_Α + ω₈ - ω_Α = ω_Β, and since the idler, and hence the beating signal component, is substantially free of phase encoding as a result of optical phase erasure by the four wave mixing.

6. The device of claim 4 or 5, wherein the first and second optical input signals have orthogonal polarizations and a polarizer is arranged before the frequency divider to filter out the first input optical signal (i.e. the pump) prior to generation of the frequency divided signal.

7. The device of claim 3, further comprising a third input for receiving a third input optical signal of a third input frequency ως, at least a component of which is supplied to the optical mixing path, so that components of the first, second and third input optical signals pass through the third-order non-linear medium to induce degenerate four wave mixing between both the first and second input optical signals and the first and third optical signals, thereby to generate first and second idlers at respective first and second idler frequencies

= 2u>c - ω_Α , where M is an integer and where ω_Α and ω_Β are the first and second input frequencies, and wherein the beating signal component is between the first and second idlers so that the beat frequency is ω^ι - ω,οι_βΓ2 = 2(ω_Β - co_c).

8. The device of claim 7, wherein the device further comprises first and second pump sources operable to generate the first and third input optical signals as continuous waves and supply the first and second pumps to the first and third inputs and wherein, when the second input optical signal is a data signal from a transmission line bearing data which is phase encoded into N states separated equally in phase, the device recreates a carrier which is substantially free of phase modulation as the output optical signal, since the modulator shifts the frequency of the second pump to that of the data signal by the operation u>_c + 2(U>B - ω₀)/2 = >c + U>B - ω₀ = ω_Β, and since the first and second idlers are selected to be of Mth order where M=N-1 so that the first idler is substantially free of phase encoding as a result of optical phase erasure by the four wave mixing, thereby ensuring the beating signal component, is also substantially free of phase encoding.

9. The device of claim 8, wherein the data signal and second pump are directed to propagate along the optical mixing path in opposite directions, and the first pump signal is split into respective components which are directed to propagate along the optical path in opposite directions.

10. The device of any of claims 1 to 9, wherein the mixing components include a first mixing component phase synchronously derived from and shifted in frequency with respect to the first input optical signal, and a second mixing component which is a component of the second input optical signal.

1 1. The device of claim 10, wherein the mixing components further include a third mixing component phase synchronously derived from and shifted in frequency with respect to the first input optical signal in an equal amount and opposite direction to the first mixing component.

12. The device of any of claims 1 to 1 1, wherein the optical modulator is connected to receive as its input optical signal a component of the first input optical signal.

13. The device of any of claims 1 to 1 , wherein the optical modulator is connected to receive as its input optical signal a component of an optical signal phase synchronously derived from and frequency shifted with respect to the first input optical signal.

14. The device of any of claims 1 to 13, wherein the frequency divider comprises an optoelectronic converter having a bandwidth greater than the beat frequency and arranged in the optical mixing path to receive at least a component of the mixed optical signal and generate therefrom an electronic signal at the beat frequency, and includes an electronic frequency division module arranged to receive the electronic signal at the beat frequency and generate the frequency divided signal electronically, and wherein the optical modulator is electrically actuated with the electronic frequency divided signal as its control signal.

15. The device of claim 14, wherein the electrically actuated optical modulator is an electro- or acousto- optical modulator.

16. The device of any of claims 1 to 13, wherein the frequency divider is an all-optical component based on injection locking, and the control signal input to the optical modulator is an optical control signal.

17. A frequency modulation method for outputting an optical signal of a desired output frequency based on input optical signals of first and second input frequencies, the method comprising:

supplying first and second input optical signals of first and second input frequencies; optically mixing at least respective components of the first and second input optical signals, or signals phase synchronously derived therefrom, so that the components mix to generate a beating signal component with a beat frequency;

frequency dividing the beating signal component into a frequency divided signal having a frequency 1 /N of the beat frequency, where N is an integer of 2 or more; and

modulating a component of the first input optical signal, or of an optical signal phase synchronously derived therefrom, with the frequency divided signal to perform a frequency shift of 1/N of the beat frequency.