Clock recovery for optical transmission systems

Abstract

A receiver for an optical communications system which corrects distortion of a received signal. A clock recovery system utilizing a feedback and feedforward system are provided. The feedback loop comprises a phase detector and a clock source, while the feedforward loop comprises the phase detector and a delay element for delaying the output of distortion correction system. The feedback loop has a significantly lower bandwidth than the feedforward path. There are also provided methods of optimizing tap weights and of acquiring initial tap weights.

Claims

1. A method of initially acquiring tap weights for a finite impulse response (FIR) filter used to correct distortion in an optical communication receiver, the method comprising: acquiring and storing a series of samples of a received optical signal, the series of samples being acquired in an order in time; applying a blind optimization algorithm to the series of samples to obtain first estimated tap weights for the FIR filter in order to equalize the received optical signal, wherein the blind optimization algorithm is applied to the series of samples alternately in the order that the series of samples is acquired and in a reverse order that the series of samples is acquired in a digital processing system; activating a clock recovery system in the optical communication receiver based on the first estimated tap weights to remove a clock offset for subsequent samples; after the clock recovery system is activated, acquiring and storing a further series of samples of the received optical signal, wherein the clock offset is removed from the further series of samples; and applying the blind optimization algorithm to the further series of samples to obtain second estimated tap weights having an accuracy higher than the first estimated tap weights.

2. The method according to claim 1, further comprising transferring the second estimated tap weights to the FIR filter.

3. The method according to claim 2, further comprising correcting distortion in the received optical signal with the FIR filter.

4. The method according to claim 1, further comprising applying the second estimated tap weights to the FIR filter and correcting distortion in the received optical signal based on the second estimated tap weights.

5. The method according to claim 1, further comprising activating a decision system in the optical communication receiver to decode values represented by the received optical signal.

6. The method according to claim 5, further comprising entering a tracking mode, in which a tap update algorithm is activated and configured to update the second estimated tap weights.

7. An apparatus comprising: an optical receiver configured to acquire and store a series of samples of a received optical signal, the series of samples being acquired in an order in time; a clock recovery system; a digital signal processor configured to: apply a blind optimization algorithm to the series of samples to obtain first estimated tap weights, wherein the digital signal processor applies the blind optimization algorithm to the series of samples alternately in the order that the series of samples is acquired and in a reverse order that the series of samples is acquired; wherein the optical receiver is further configured to activate the clock recovery system based on the first estimated tap weights to remove a clock offset for subsequent samples; wherein, after the clock recovery system is activated, the optical receiver is configured to acquire and store a further series of samples of the received optical signal and the clock offset is removed from the further series of samples; wherein the digital signal processor is further configured to apply the blind optimization algorithm to the further series of samples to obtain second estimated tap weights having an accuracy higher than the first estimated tap weights; and a finite impulse response (FIR) filter configured to equalize the received optical signal with the second estimated tap weights.

8. The apparatus of claim 7, wherein the digital signal processor is further configured to transfer the second estimated tap weights to the FIR filter.

9. The apparatus of claim 7, wherein the FIR filter is further configured to apply the second estimated tap weights and correct distortion in the received optical signal based on the second estimated tap weights.

10. The apparatus of claim 7, wherein the optical receiver is further configured to activate a decision system to decode values represented by the received optical signal.

11. One or more non-transitory computer readable storage media encoded with computer readable instructions which, when executed by a processor, cause the processor to: acquire and store a series of samples of a received optical signal, the series of samples being acquired in an order in time; apply a blind optimization algorithm to the series of samples to obtain first estimated tap weights for a finite impulse response (FIR) filter in order to equalize the received optical signal, wherein the blind optimization algorithm is applied to the series of samples alternately in the order that the series of samples is acquired and in a reverse order that the series of samples is acquired in a digital processing system; activate a clock recovery system in an optical communication receiver based on the first estimated tap weights to remove a clock offset for subsequent samples; after the clock recovery system is activated, acquire and store a further series of samples of the received optical signal, wherein the clock offset is removed from the further series of samples; and apply the blind optimization algorithm to the further series of samples to obtain second estimated tap weights having an accuracy higher than the first estimated tap weights.

12. The one or more non-transitory computer readable storage media of claim 11, wherein the instructions further cause the processor to correct distortion in the received optical signal with the FIR filter.

13. The method of claim 1, further comprising: shifting the second estimated tap weights toward a tap weight center.

14. The method according to claim 13, wherein the shifting is performed using a digital interpolation function.

15. The apparatus of claim 7, wherein the digital signal processor is further configured to shift the second estimated tap weights toward a tap weight center.

16. The apparatus of claim 15, wherein the digital signal processor is configured to shift the second estimated tap weights using a digital interpolation function.

17. The apparatus of claim 10, wherein the optical receiver is configured to enter a tracking mode in which a tap update algorithm is activated and configured to update the second estimated tap weights.

18. The one or more non-transitory computer readable storage media of claim 11, wherein the instructions further cause the processor to shift the second estimated tap weights toward a tap weight center.

19. The one or more non-transitory computer readable storage media of claim 18, wherein the second estimated tap weights are shifted using a digital interpolation function.

20. The one or more non-transitory computer readable storage media of claim 11, wherein the instructions further cause the processor to enter a tracking mode and activate a tap update algorithm configured to update the second estimated tap weights.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) Embodiments of the invention will be described, by way of example, with reference to the following drawings, in which:

(2) FIG. 1 shows power and phase eyes for a QPSK signal;

(3) FIG. 2 shows an example QPSK transmitter;

(4) FIG. 3 shows part of a coherent receiver;

(5) FIG. 4 shows a block diagram of a digital receiver for an optical communications system;

(6) FIG. 5 shows an example of an FIR filter;

(7) FIG. 6 shows an example of a receiver for a dual polarisation system;

(8) FIG. 7 shows a block diagram of a receiver using an analogue phase lock loop;

(9) FIG. 8 shows a block diagram of a receiver using a combined digital and analogue phase lock loop;

(10) FIG. 9 shows a Jitter mask demonstrating the effect of delay on the feedback loop of FIG. 8 with a 1 MHz bandwidth;

(11) FIG. 10 shows an example of an optical receiver using combined feedback and feed forward signals for clock recovery;

(12) FIG. 11 shows a jitter mask;

(13) FIG. 12 shows a block diagram of an optical communications receiver;

(14) FIG. 13 shows a phase detecting system;

(15) FIG. 14 shows a graph of the required phase detector output;

(16) FIG. 15 shows a clock lock detecting system;

(17) FIGS. 16A and 16B show examples of tap weights in an FIR filter;

(18) FIG. 17 shows a method of clock acquisition;

(19) FIG. 18 shows an example of clock and tap weight acquisition; and

(20) FIG. 19 shows a different view of the method of FIG. 18.

DETAILED DESCRIPTION

(21) Embodiments of the present invention are described below by way of example only. These examples represent the best ways of putting the invention into practice that are currently known to the Applicant although they are not the only ways in which this could be achieved.

(22) The description sets forth the functions of the example and the sequence of steps for constructing and operating the example. However, the same or equivalent functions and sequences may be accomplished by different examples.

(23) A 40 Gbit/s DP-QPSK transmission system is utilised as the basis for the following description, but the techniques described are applicable to a range of transmission formats and rates without unduly burdensome modification.

(24) FIG. 10 shows a block diagram of a clock recovery system, using a feedback loop 90 and a feedforward path 91. Phase Detector 92 detects the phase of the output of equaliser 74. The phase detector 92 output is fed back to Voltage Controlled Oscillator (VCO) 72 which provides a clock for the Analogue to Digital Converter (ADC) 36 sampling. The VCO 72, ADCs 36, equaliser 74 and phase detector 92 form a Phase Locked Loop (PLL). The phase detector 92 output is also fed forward to digital delay element 94 which acts to delay the samples output from the equaliser 74. The digital delay element 94 provides a time adjustment function acting on the samples passed to that element. The digital delay element 94 can be implemented using an interpolator, for example as described in H. Meyr, M. Moeneclaey, S. A. Fechtel, Digital Communication Receivers, Wiley & Sons, ISBN 0-471-50275-8. Chapter 9, which is incorporated herein by reference.

(25) The PLL has a relatively low bandwidth of the order of 50-100 kHz and removes the frequency offset and slowly varying phase offset from the clock. The low bandwidth of the PLL means that the processing delay of equaliser 74 does not cause significant gain peaking in the feedback loop, as seen when the bandwidth of the PLL is sufficient to also remove the fast varying phase offset.

(26) The delay element 94 removes higher frequency phase jitter remaining on the signals by acting as a digital interpolator to delay the position of the samples output from equaliser 74. The Low pass filter 95 prior to the delay element has a bandwidth of the order of 1-4 MHz.

(27) The Low pass filter may be combined with the phase detector, provided the filter bandwidth is at least about 10 times that of the PLL, such that it effectively only affects the feedforward path. Regardless of whether the low pass filter is incorporated into the phase detector, or is separate, the bandwidth of the external loop should be lower than the bandwidth of the phase detector and delay element 94.

(28) The combination of a lower frequency feedback loop and a delay element controlled by a feedforward signal mitigates at least some of the problems described with prior systems and enables a clock recovery system to meet the required standards.

(29) FIG. 11 shows the G.8251 Jitter mask and a series of plots for the system of FIG. 10. Dashed line 100 marks the minimum jitter values which must be tolerated. The simulation assumed a processing delay of 372 ns (125 clock cycles), and a 100 kHz analogue closed-loop bandwidth. Even with this relatively large processing delay, the standard can be comfortably met with a Low Pass Filter with a 5.4 MHz 3 dB point. FIG. 12 shows a block diagram of a receiver system for correcting distortion, performing clock recovery, and decoding received data. The system shown in FIG. 12 relates to a single polarisation and is duplicated (apart from the photodiodes and ADCs which are shared) for the second polarisation.

(30) As explained previously, four photodiodes 35 receive four outputs from 90 hybrids mixing the optical signal with the optical Local Oscillator. The outputs of those photodiodes are digitised using ADCs 36, the sampling clock of which is provided by VCO 72 configured in conjunction with phase detector 92 as described in relation to FIG. 8.

(31) The upper pair of photodiodes (solid signal lines) are for a first output of the polarisation splitter described above and the lower pair (dashed signal lines) for the second output of that splitter. The signals from each pair are combined to give a complex valued signal (composing of real and imaginary parts) 110, 111 and passed to the equaliser 112. As explained previously, the butterfly equaliser structure corrects for the unknown polarisation of the incoming optical signal.

(32) Delay element 113 acts on the output of the equaliser 112 to remove any high frequency phase jitter not tracked by the PLL. The output of the delay element 113 is passed to a carrier recovery system 114 to remove the Local Oscillator offset. Example carrier recovery systems are disclosed in Viterbi, A. (1983), Nonlinear estimation of PSK-modulated carrier phase with application to burst digital transmission, Information Theory, IEEE Transactions on 29(4), 543-551, and H. Meyr, M. Moeneclaey, S. A. Fechtel, Digital Communication Receivers, Wiley & Sons, ISBN 0-471-50275-8. pp. 311-322, which are incorporated herein by reference.

(33) Decision circuit 115 decides values of the symbols (in this example DP-QPSK system the output from the equaliser represents a QPSK signal and therefore each symbol carries two bits) and outputs the decided symbols at 119.

(34) A decision-directed tap update system 116 is provided to control and maintain the tap weights of the equaliser 112 such that the system continuously monitors and tracks the incoming signals. The tap update system 116 operates a tap update algorithm which takes inputs of the input signal 117, the equalised signal 118 and the decided symbols 119. To ensure correlation between the input signal and the equalised signal 118 a second delay element 1100 is utilised on the input signal feed to the tap update system to mimic the delay applied by the first delay element. The input signal passed to the update algorithm is therefore actually a delayed input signal 1101 such that the algorithm does not see the delay introduced by the first delay element 113. The decided symbols signal 119 also differs from the input signal 1101 and equalised data 118 as the carrier offset has been removed. The decided symbols must therefore be re-spun such that they correlate with the other signals. This is achieved by applying 1102 the output of the carrier phase estimator 1103 to the decided symbols.

(35) The tap update algorithm may be any suitable algorithm for providing the required functionality, and various options are known in the art. By way of example, the following description is given in relation to a Least Mean Squares (LMS) algorithm.

(36) Firstly, an error vector of the difference between the equaliser output and the decided symbol is calculated as below:
.sub.k=y.sub.kp.sub.k

(37) We note the use of a line over variables to indicate a complex value, and bold type to indicate a matrix. Where y.sub.k is the decided symbol output:
y=d(k)e.sup.+j{circumflex over ()}.sup.k

(38) And p.sub.k is the equalizer output:
p=w.sup.Hv
where H is the Hermitian transpose (or conjugate transpose). The tap coefficients are then calculated:
w.sub.k+1=(1)w.sub.k+e.sub.k*v.sub.k

(39) Where w=tap weight matrix, =leakage factor, is the update rate and v is the unequalised input signal matrix. The delayed input signal v.sub.k is multiplied by the error value, .sub.k, and by , the tap update rate. The first terms in the equation apply the leakage factor to the previous tap weights, which causes them to decay over time.

(40) The receiver system of FIG. 12 thus provides a system for equalising a received signal, performing clock recovery, dynamically updating the equaliser to track changing conditions, and determining the value of the received symbols. As noted previously, a comparable system may be utilised to process the second polarisation of a DP-QPSK system.

(41) FIG. 13 shows a block diagram of a phase detector suitable for a DP-DQPSK system. The operation of such a phase detector is described in Oerder, H. (May 1988), Digital filter and square timing recovery, Communications, IEEE Transactions on 36(5), 605-612 and Zhu, M. S. M. (November 2005), Feedforward symbol timing recovery technique using two samples per symbol, Circuits and Systems I: Regular Papers, IEEE Transactions on [Circuits and Systems I: Fundamental Theory and Applications, IEEE Transactions on] 52(11), 2490-2500, incorporated herein by reference.

(42) For clarity, the following description is given in relation to the first signal and blocks shown in FIG. 13, but applies to each of the blocks duplicated in parallel in FIG. 13.

(43) As described in the above references the incoming signal 120 is first multiplied by a half-rate sine or cosine clock. In the implementation shown in FIG. 13 which operates using two samples per symbol, this is implemented by passing alternate samples to separate data paths 121 and 122. Path 121 receives the even (A) samples and path 122 receives the odd (B) samples. Even and odd are used as labels only to distinguish one sub set of samples from the other. The samples are then passed through low pass filter 123, for example a Finite Impulse Response (FIR) filter to remove high frequency components from the signal which would distort the phase detector output when working with 2 samples per symbol. The complex value (u+vj) is then squared at block 124 to give a clock phase vector 125.

(44) The four clock phase vector signals are added in adders 120, 121. Adders 120, 121 are selectable to allow the selection of one or more of the clock phase vector signals. When in steady state it is likely to be most preferable for the phase detector to operate on all signals (i.e. both polarisations) to give the best accuracy. However, during startup, or at other times, it is possible that the output of the equaliser for one polarisation is poor and thus the phase detector may utilise only the good polarisation to detect the phase. The phase detector may therefore be switched dynamically based on the system's performance to utilise one or both of the polarisation signals.

(45) The clock phase vectors are then averaged at blocks 122, 123 and converted to a phase value using a tan()tan.sup.1 (N/D)) block 124. Averaging the phase vectors rather than the phase value makes the system more robust against cycle slips. The parameters of the Average blocks 122, 123 define the bandwidth of the phase detector.

(46) Unwrap block 125 removes discontinuities at +/ radian intervals and allows the phase detector to track phase changes over multiple unit intervals as shown in FIG. 14 (a plot of Phase Detector Output (Unit Intervals (UI)) against Input Phase (UI)) which shows an idealized output. As shown in FIG. 14 the output should be linear since the signal is utilized as a feedforward signal to control Delay Element 113. Delay Element 113 cannot provide an infinite delay and so saturation block 126 saturates the signal at a predetermined value (2 unit intervals in FIG. 14).

(47) The phase value output is split and may be adjusted 127, 128 according to tap weight phase detector 1202 as described below. The outputs 129, 1200 are utilised as inputs to the VCO 72 and Delay Element 113, respectively.

(48) FIG. 15 shows a block diagram of a clock frequency offset monitor 1201 which may be provided to analyse and indicate whether the clock recovery system has locked to the received data signal. Successive phase values are compared at block 140 to give the phase change between those values. Block 141 performs an unwrap function to prevent false indications when the phase moves over an edge of the tan function. Infinite Impulse Response (11R) filter 142 filters the signal resulting in a signal 143 indicating the offset between the received data clock and the local clock. When this value is stable, lock has been obtained. The indication of clock lock may be utilised by the processing system to control decoding of data or for general system control.

(49) FIR filters, such as those used in the equaliser, can act as variable delay elements by interpolating between samples. This occurs by the tap weights shifting to the left or the right. FIGS. 16a and 16b show plots of tap weights of an FIR filter, each having the same impulse response but the tap weights in FIG. 16b are shifted off-centre to introduce a delay compared to the weights in FIG. 16a. The centred weights shown in FIG. 16a are better able to correct for increased distortion than the off-centre ones shown in FIG. 16b, and therefore it is preferable for the tap weights to be centralised for optimum equaliser performance.

(50) The equaliser 112, being an FIR, attempts to track and correct any slow clock phase drift which causes the tap weights to move from their central position. The off-centre tap weights are less able to equalise increased distortion and therefore system performance may be degraded. The slow clock phase drift should be corrected by the PLL, not the equaliser.

(51) A tap weight phase detector 1202 is provided to monitor the centralisation of the tap weights and provide a correction signal indicative of the tap weight centre offset.

(52) The correction signal output by tap weight phase detector 1202 is split for use by the VCO 72 and delay element 113. The signal for the VCO 72 passes to multiplier 1203 where it is multiplied by a coefficient K.sub.tapPLL. The signal for the delay element passes to multiplier 1204 where it is multiplied by a coefficient K.sub.tapDE. K.sub.tapPLL and K.sub.tapDE define the magnitude of the correction signal that is added to the phase value for the VCO 72 and Delay Element 113 respectively. Adders 127 and 128 add the correction signal to the phase signal for the VCO 72 and delay element 133 respectively.

(53) In a first example, K.sub.tapDE is set to zero such that the correction signal is only applied to the VCO 72. In a second example, K.sub.tapDE may be non-zero such that the correction is applied to both the VCO 72 and Delay Element 133.

(54) The correction signal applied to the VCO 72 (and Delay Element 113 if K.sub.tapDE0) causes the tap weights to re-centralise under the action of the tap update system.

(55) In an alternative implementation, the correction signal could be utilised to directly adjust the tap weights. The signal should be applied slowly to allow the VCO 72 and Delay Elements 133 to track the change thereby avoiding degradation of the equaliser performance.

(56) The tap weight centre for the x-polarisation can be calculated using the following equation:

(57) ${\hat{n}}_x=\frac{\underset{n}{.Math.}n\left(|{\stackrel{\_}{h}}_{{\mathrm{xx}}_n}{|}^2+|{\stackrel{\_}{h}}_{{\mathrm{yx}}_n}{|}^2\right)}{\underset{n}{.Math.}\left(|{\stackrel{\_}{h}}_{{\mathrm{xx}}_n}{|}^2+|{\stackrel{\_}{h}}_{{\mathrm{yx}}_n}{|}^2\right)}$

(58) Where h.sub.xx.sub.n and h.sub.yx.sub.n are the nth elements of the complex-valued sub matrices h.sub.xx and h.sub.yx for the X-X and Y-X filters respectively which equalise the X-polarisation:

(59) $\stackrel{\_}{w}=\left(\begin{array}{c}{\stackrel{\_}{h}}_{\mathrm{xx}}\\ {\stackrel{\_}{h}}_{\mathrm{yx}}\end{array}\right)$

(60) A comparable equation is used for the Y-polarisation. The tap weight phase (the difference between the tap weight centre and the central tap) is then given by:
.sub.tapX=n.sub.centre{circumflex over (n)}.sub.X

(61) Where n.sub.centre is the number of the central tap.

(62) The tap weight phase is combined with the output of the phase detector, as shown in FIG. 13, to correct the phase of the VCO 72 and Delay Element 113, 1100, which in turn results in the tap weights moving to the centralised position. The tap weight phase signal output may be a combination of both the X and Y polarisations, or may be based on one or the other polarisation. It may be desirable to select from one or both polarisations depending on the relatively quality of the signals, or if one of the polarisations has not correctly acquired stable tap weights.

(63) On startup of a transmission system the receiver has no knowledge of the correct clock phase or tap weights for the equaliser. For the phase detector to acquire the phase of the signal (and hence for the clock recovery system to operated) the equaliser must, at least to a certain degree, equalise the received signal. However, determination of the tap weights for the equaliser relies on the clock recovery system.

(64) FIG. 17 shows a flow chart of a system for acquiring the clock phase and initial tap weights to allow startup of the system. FIG. 19 provides an alternative view of the methods of FIGS. 17 and 18 to exhibit the functions performed by each part of the system and the transfer of data between the ASIC and the DSP during the acquisition 187 and tracking 188 phases. Reference numerals on FIG. 19 relate to the blocks of FIGS. 17 and 18.

(65) At block 160 a series of samples 180 of the data 183 is acquired by the ASIC 181. For example, 4000 samples may be acquired, which at a typical sample rate of two samples per symbol, gives 2000 symbols. At block 161 the series of samples is transferred to a Digital Signal Processor (DSP) 182 for processing according to the programming of that DSP. At block 162 the DSP applies a blind-optimisation algorithm to the samples 180 in order to identify an initial set of tap weights which may be used to equalise the incoming signal.

(66) The use of a DSP 182 associated with the ASIC 181 for performing the algorithm is convenient as it may be programmed to perform the specific algorithms required. Since the algorithm need not be applied in real time there is a reduced requirement on the performance of the DSP 182 compared to the processing system which processes the received data in real time. Providing this function in the ASIC 181 is likely to be substantially more expensive and complex than utilising a DSP 182 and therefore the use of two devices may be more cost effective. However, the method can equally be applied within a single device if appropriate. In addition, ASIC implementations will typically work in a highly parallel manner, and the update rate may be limited by feedback delays. A non-real time implementation allows substantially higher update coefficients to be used, and the same data may be processed several times until the tap weights have converged. The use of a short block of data means that clock offsets (which may be 100-200 ppm) are not significant to prevent acquisition of an initial set of tap weights.

(67) At block 163 the initial tap weights are transferred to the ASIC 181 and applied to the equaliser. The equaliser output is now a partially corrected signal from which the phase detector can operate. At block 164 the PLL is activated. Provided the channel is stable, the output of the equaliser will be valid and the PLL will acquire the clock at block 165 and will remove the clock offset.

(68) Once the clock has been acquired, at block 166 the digital delay elements and decision-directed feedback systems are activated and the system enters tracking mode. Provided the initial tap estimates are sufficiently accurate the system will optimise the tap weights, PLL and delay elements. In order for this to succeed the tap weights must be sufficiently accurate for the decision directed update to function, which requires approximately <10.sup.2BER.

(69) Algorithms for the blind-compensation of DP-QPSK are known, for example as described in Raheli, R. & Picchi, G. (1991), Synchronous and fractionally-spaced blind equalization in dually-polarized digital radio links, in Communications, 1991. ICC 91, Conference Record. IEEE International Conference on, pp. 156-161 vol. 1, incorporated herein by reference. A further known blind acquisition method which may be applicable in relation to the method of FIG. 17 is a constant modulus algorithm as described in Godard, D. (1980), Self-Recovering Equalization and Carrier Tracking in Two-Dimensional Data Communication Systems, Communications, IEEE Transactions on [legacy, pre-1988] 28(11), 1867-1875, incorporated herein by reference.

(70) The optimisation algorithm utilised in block 162 may be applied a number of times to the series of samples. A particularly efficient method is to apply the algorithm to the samples alternately forwards and backwards such that any clock offset is not relevant. The update coefficients used in the algorithm may be considerable higher than are utilised in a continual optimisation system as it is only desired to acquire an initial set of tap weights, not to provide continual performance. High update coefficients tend to lead to sporadic changes in tap weights and hence degradation in performance, but provide a more rapid convergence to the initial tap weights.

(71) Prior to the tap weights being transferred to the ASIC they may be re-centred 184 by shifting the taps in response to a calculation of the tap weight centre, or by applying a shift using a digital interpolation function.

(72) The method shown in FIG. 17 assumes that the acquisition phase, polarisation or channel model does not change significantly during the period of the acquisition. It is reasonable to assume that the polarisation and channel model does not change during this phase (approximately 20-500 ms), and a maximum clock offset of 250 ppm is reasonable. With that offset a drift of 0.5 unit intervals will occur in the 2000 symbols, which equates to a shift of 1 equaliser tap. This is not a significant shift and so the required assumptions are reasonable. If the assumptions are not met, then a new series of samples may be acquired and the process restarted.

(73) FIG. 18 shows a flow chart of an extension of the method of FIG. 17 to improve the equalisation before tracking mode is activated. After the clock has been acquired at block 165 a further series of samples 185 are acquired at block 170. At block 171 those samples 185 are transferred to the DSP 182 and again processed using the blind optimisation algorithm at block 172. Since the clock offset has been removed from this second series of samples a more accurate set of tap weights can be derived. At block 173 those tap weights (after optional recentering 186) are transferred to the equaliser and applied to the received signal.

(74) At block 174 the decision-directed tracking and digital delay elements are activated and the receiver enters tracking mode. Frame detection may then be obtained and the system goes on to continuously track the channel.

(75) All parameters and results given in the above description relate to a 40 Gbit/s DP-QPSK transmission system and are given to describe that system. All parameters may be modified in conventional manners as required based on the particular system being utilised without departing from the invention.

(76) Where functions or algorithms have been described as being performed by a particular device or type of device, this is for example only and is not intended to be limiting in any way. As will be appreciated any suitable mode of implementation may be utilised as appropriate.

(77) The blocks and demarcation between functions described above is given by way of example only, and as will be appreciated the functions may be demarked and distributed in any suitable way. Where a function has been described as being performed by a particular type of device (for example an ASIC) it will be appreciated that the principles described herein are also applicable to other methods of implementing those functions (for example in software running on a DSP).

(78) The term Voltage Controlled Oscillator is used to describe a component which produces an output signal for clocking the ADCs in dependence on an input signal. As will be appreciated, it is not intended to restrict the component to one which is controlled directly by a varying voltage, but rather to describe the function of the component. For example, the component may equally be controlled by a digital signal indicating values.

(79) Although not shown or described explicitly, it will be appreciated that, as is well known in the art for PLLs, a loop filter will be incorporated at a convenient location in the various PLLs described herein. That filter may be conveniently located in close proximity to the clock source, and may form part of the same device as that source. Alternatively, the filter may be located at any convenient location and provided by any convenient means.

(80) The manner of implementation of the techniques described herein is dependent on the particular system and the implementation is within the capabilities of the skilled reader once they have been made aware of the functions required by this document.

(81) The applicant hereby discloses in isolation each individual feature described herein and any combination of two or more such features, to the extent that such features or combinations are capable of being carried out based on the present specification as a whole in the light of the common general knowledge of a person skilled in the art, irrespective of whether such features or combinations of features solve any problems disclosed herein, and without limitation to the scope of the claims. The applicant indicates that aspects of the present invention may consist of any such individual feature or combination of features. In view of the foregoing description it will be evident to a person skilled in the art that various modifications may be made within the scope of the invention.

(82) Any range or device value given herein may be extended or altered without losing the effect sought, as will be apparent to the skilled person.

(83) It will be understood that the benefits and advantages described above may relate to one embodiment or may relate to several embodiments. The embodiments are not limited to those that solve any or all of the stated problems or those that have any or all of the stated benefits and advantages.

(84) Any reference to an item refers to one or more of those items. The term comprising is used herein to mean including the method blocks or elements identified, but that such blocks or elements do not comprise an exclusive list and a method or apparatus may contain additional blocks or elements.

(85) The steps of the methods described herein may be carried out in any suitable order, or simultaneously where appropriate. Additionally, individual blocks may be deleted from any of the methods without departing from the spirit and scope of the subject matter described herein. Aspects of any of the examples described above may be combined with aspects of any of the other examples described to form further examples without losing the effect sought.

(86) It will be understood that the above description of a preferred embodiment is given by way of example only and that various modifications may be made by those skilled in the art. Although various embodiments have been described above with a certain degree of particularity, or with reference to one or more individual embodiments, those skilled in the art could make numerous alterations to the disclosed embodiments without departing from the spirit or scope of this invention.

(87) The term computer is used herein to refer to any device with processing capability such that it can execute instructions. Those skilled in the art will realize that such processing capabilities are incorporated into many different devices and therefore the term computer includes PCs, servers, mobile telephones, personal digital assistants and many other devices. Similarly DSP or ASIC is not intended to restrict the invention to any particular type of processing device, but those terms are simply used to refer to one possible implementation.

(88) The methods described herein may be performed by software in machine readable form on a tangible storage medium. The software can be suitable for execution on a parallel processor or a serial processor such that the method steps may be carried out in any suitable order, or substantially simultaneously.

(89) This acknowledges that software can be a valuable, separately tradable commodity. It is intended to encompass software, which runs on or controls dumb or standard hardware, to carry out the desired functions. It is also intended to encompass software which describes or defines the configuration of hardware, such as HDL (hardware description language) software, as is used for designing silicon chips, or for configuring universal programmable chips, to carry out desired functions.

(90) Those skilled in the art will realize that storage devices utilized to store program instructions can be distributed across a network. For example, a remote computer may store an example of the process described as software. A local or terminal computer may access the remote computer and download a part or all of the software to run the program. Alternatively, the local computer may download pieces of the software as needed, or execute some software instructions at the local terminal and some at the remote computer (or computer network). Those skilled in the art will also realize that by utilizing conventional techniques known to those skilled in the art that all, or a portion of the software instructions may be carried out by a dedicated circuit, such as a DSP, programmable logic array, or the like.