WO2013072702A1

WO2013072702A1 - Optical signal transmission systems

Info

Publication number: WO2013072702A1
Application number: PCT/GB2012/052853
Authority: WO
Inventors: Sim Heung LEE; David George Cunningham; Richard Vincent Penty; Ian Hugh White
Original assignee: Cambridge Enterprise Limited
Priority date: 2011-11-18
Filing date: 2012-11-16
Publication date: 2013-05-23
Also published as: GB201119922D0

Abstract

We describe a method of communicating data using an optical WDM (wavelength division multiplexed) system, the method comprising : encoding said data into symbols on a first plurality of optical carriers of said WDM system; receiving said optical carriers at a second plurality, M, of optical receivers, wherein each said optical receiver receives more than one said optical carrier such that said optical carriers mix and a said optical receiver receives symbols carried by more than one said optical carrier; and processing electrical outputs from said optical receivers representing said symbols on said mixed optical carriers to extract said encoded data.

Description

Optical Signal Transmission Systems

FIELD OF THE INVENTION This invention relates to systems and methods for optical WDM (wavelength division multiplexing) signal transmission systems.

BACKGROUND TO THE INVENTION Dense wavelength division multiplexing (DWDM) is a very powerful technique for increasing the capacity of communication systems. However, current systems require precise wavelength alignment and registration, these typically requiring components such as laser diodes and multiplexers to be designed to operate at precise wavelengths. For a number of components, precise temperature control is also required, and hence the requirement of accurate wavelength registration for DWDM systems can lead to significantly increased capital and operating costs. Various methods have been used to avoid temperature control, apart from that of simply increasing the separation between spectral channels such as in Coarse Wavelength Division Multiplexing. For example, studies have been carried out on using a thermal wavelength tunable lasers which can realize a stable wavelength WDM link without the need for thermo-electric cooling. Here the lasers are automatically tuned so that their wavelength remains constant irrespective of temperature. However this approach requires lasers with separate modulators for 10 Gb/s operation and beyond, along with advanced electronic control to avoid mode-hopping.

Background prior art can be found in US 7,970,280, US 7,412,170, and US6,525,853. Further background can be found in our online abstract for Paper 8284-5, for SPIE Photonics West Conference 8284 in January 2012, entitled "Uncooled MIMO WDM system using advanced receiver signal processing techniques", by S. H. Lee, Richard V. Penty, Ian H. White, and David G. Cunningham: "Current DWDM systems require components such as laser diodes and multiplexers to operate at precise wavelengths with precise temperature control".

We will describe a new form of narrow channel spacing WDM system which uses uncooled lasers, DE/MUXes and receivers. It removes control of laser wavelengths and thus energy consumption of their temperature control systems at the expense of a slightly more complicated receiver design. The system allows the wavelength of the lasers to drift but uses a cyclic demultiplexer at the receiver so that the separate WDM signals can be decoded with carefully designed signal processing algorithms.

SUMMARY OF THE INVENTION

According to the present the invention there is therefore provided an optical WDM (wavelength division multiplexed) communications system, the system comprising: an optical WDM transmission system having an electrical signal input to receive input data for transmission and a WDM optical output to output a WDM optical signal having a first plurality, N, of carrier wavelengths encoding said input data; and an optical WDM receiving system having a WDM optical signal input and an electrical signal output to output decoded data; and wherein said receiving system comprises a second plurality, M, of optical receivers, wherein each optical receiver of said plurality of optical receivers is configured to be able to simultaneously receive more than one said carrier wavelength.

The inventors have recognised that, paradoxically, there can be an advantage in allowing the optical signals to mix and then having relatively complex signal processing in the electrical domain to descramble these mixed signals. The descrambling process is akin to radio frequency MIMO (multiple-input-multiple-output) transmission systems in which a single receiver receives signals from all the transmitters, which then results in the need to descramble the mixed signals. In an RF transmission system this is unavoidable because the transmissions from a plurality of antennas inherently mix as they propagate through the MIMO channel. In an optical WDM transmission system the different optical carriers enable the signals to be separated and it is thus counterintuitive to deliberately allow the signals to mix. This is because the processing to extract the signals from one another also tends to enhance the noise, increasing the required power budget. Nonetheless when viewed from the perspective of the system as a whole, this can be traded against other benefits, in particular tolerance of the transmission system to carrier drift.

In a DWDM system the lasers and AWG (arrayed waveguide gratings) are prone to thermal drift and in a practical system must be temperature controlled. This consumes significant electrical power as well as introducing reliability issues associated with the cooling systems. Embodiments of the present invention can therefore offer advantages over existing systems and, in particular, may be uncooled. In such an embodiment of the system, although each optical receiver is configured to simultaneously receive more than one carrier wavelength, during operation, because the laser wavelengths drift, some receivers may receive multiple carrier wavelengths and others none, as the spacings between the carriers change.

In some preferred embodiments of the system, the optical receiving system comprises electrical signal processing circuitry coupled to the M optical receivers to process signals from the optical receivers to provide a set of n data outputs where n is no more than M. These data outputs may then be combined to provide the decoded output data from the system. The (real-time) signal processing circuitry may be circuitry in the analogue and/or digital domain, optionally including software or firmware.

The above described approach provides optical redundancy such that, in embodiments, an optical path carrying a data bit of the input data includes at least two of the optical receivers. Thus in embodiments the number of optical receivers M is greater than the number of optical carriers N, for example the centre wavelengths of the optical receivers being regularly spaced and the optical carriers being irregularly spaced. The same number of optical receivers as optical carriers may be employed for even values of N. Alternatively, the incoming data for transmission may divided into n streams and then encoded across M carriers (the same as number of receivers), a fewer number of data streams than carriers then being extracted, to provide transmitted signal redundancy. Optionally a data bit, and/or a data stream comprising the input data, may be encoded across multiple optical carriers as well as employing more optical receivers than optical carriers.

In embodiments the optical receiving system comprises a filter system between the optical WDM input and the optical receivers. This filter system may but need not necessarily comprise an AWG (for example in principle a set of Fabry Perot filters may be employed). As mentioned previously, each receiver receives more than one carrier wavelength, and where a cyclic AWG is employed the passbands may wrap around so that the carrier wavelength at one extreme of the filter system may overlap with the other extreme of the filter system. The number of carriers received by a particular receiver depends upon the width of a passband of the filter system, but is at least two and may be more. (Even where, by design, the system is arranged such that a single receiver receives only two carrier wavelengths, in use the carriers may drift and bunch up so that one receiver may see several carrier wavelengths and another none).

In embodiments the filter system has a set of pass bands each centred at a respective centre wavelength. The passband of a filter is arranged such that it overlaps a significant portion of the adjacent passband, and is not more than 3dB down at the crossing-point wavelength of an adjacent passband, in embodiments not more than 2dB down or 1 dB down: It is important to maintain the transmission level at the crossing point between two passbands sufficiently high that when a laser wavelength drifts from one passband to another, the receiving system still receives a signal with good signal to noise ratio. (It is this, rather than the passband attenuation at the centre wavelength of adjacent passband, which is relevant - for example, in a MATLAB simulation described later we use a AWG with 0.5dB loss at crossing points, but 9dB loss at the adjacent passband centre wavelength). In this way the overall response of the filter system is approximately flat, for example to within 3dB, 2dB or 1 dB rather than exhibiting deep nulls. Conveniently this is achieved by using an AWG, more particularly a cyclic AWG, as previously described.

As previously mentioned, in embodiments the number of carriers is no more than the number of receivers, more particularly the number of centre wavelengths of the filter system. This may be achieved by providing the centre wavelengths of the filter system at regular intervals and omitting carriers from a set of otherwise regularly spaced carriers at the centre wavelengths, for example omitting one carrier in every three to provide, in effect, 50% optical receiver redundancy. Thus in embodiments the centre wavelengths of said optical receivers are substantially regularly spaced and, in a quiescent state, the carrier wavelengths match these centre wavelengths except that every ith receiver a carrier wavelength is omitted (where in embodiments i = 3). Thus, broadly speaking, after every (i-1 ) carrier wavelengths the spacing to the next carrier is increased.

In one embodiment of the receiving system the output of each respective optical receiver is coupled to a signal decoder comprising a set of taps, each tap being coupled to the output of one of the optical receivers. In embodiments optionally, but not necessarily, one tap is provided for each optical receiver, but an approximation may be achieved with fewer taps. Each tap is arranged to apply a complex weight (amplitude and phase) to the signal from the receiver to which it is connected to provide a weighted signal, and each decoder comprises a signal combiner to combine these weighted signals to provide, from the set of decoders, a set of electrical signal outputs. In embodiments one such electrical signal output is provided for each carrier wavelength, although this is not necessarily the case where, for example, the number of carrier wavelengths is the same as the number of receivers and transmit signal redundancy as described above is employed. Nonetheless in embodiments the number of decoders is no more than the number of optical receivers. In embodiments the receiving system also includes a channel estimation system coupled to the optical receivers to determine the complex weights. In one approach one channel estimation system is provided for each decoder. However but this is not essential and there are many ways in which a channel estimation system may be implemented and, for example, the channel estimation system may be shared between multiple decoders. In one approach a channel estimation system comprises a minimum mean square error (MMSE) estimator, but many other approaches may be employed, for example, other linear estimators such as zero-forcing techniques, or more complex non-linear estimators. Again these may be implemented in either the analogue or the digital domain, or a combination of both. Where the receiver signal processing operates in the digital domain, preferably the signals from the receivers, which may comprise photodiodes, are over-sampled such that there are multiple samples, for example two or four samples, per bit interval. This facilitates the bit alignments decoding process. As previously mentioned, in embodiments the transmission system includes an encoder to encode the input data such that an incoming data bit is encoded as one or more symbols for simultaneous transmission on each of a plurality of the carrier wavelengths. There are many such encoding techniques which may be employed, some of which provide error correction, for example a convolutional coder or (borrowing from the RF domain) a turbo encoder (which includes an interleaver).

In an optical WDM system typically the symbols are encoded as RZ (return to zero) or NRZ (non-return to zero) symbols, but optionally other pulse amplitude modulation schemes such as Manchester encoding or a multi-level code such as PAM-4 (Pulse Amplitude Modulation-4) may be employed. In a related aspect the invention provides an optical WDM (wavelength division multiplexed) receiving system, the system having a WDM optical signal input and an electrical signal output to output decoded data, wherein said receiving system comprises an optical plurality of optical receivers, and wherein the system further comprises an optical filter system between said optical signal input and said optical receivers, wherein said optical filter system has a set of passbands each centred at a respective centre wavelength, and wherein one said passband overlaps at least one adjacent said passband such that, for the or each said passband, transmission at a crossing-point wavelength of the two adjacent passbands is not more than 3dB down on transmission at a centre wavelength of said the or each passband, such that each optical receiver of said plurality of optical receivers is configured to simultaneously receive more than one said carrier wavelength of a WDM optical signal at WDM optical signal input.

In a further related aspect the invention provides an optical WDM transmission system, the system having an electrical signal input to receive input data for transmission and a WDM optical output to output a WDM optical signal having a first plurality, N, of carrier wavelengths encoding said input data; and wherein said transmission system includes an encoder to encode said input data such that an incoming said data bit is encoded as one or more symbols for simultaneous transmission on each of said plurality of carrier wavelengths.

In a corresponding method the invention provides a method of communicating data using an optical WDM system, the method comprising: encoding said data into symbols on a first plurality of optical carriers of said WDM system; receiving said optical carriers at a second plurality, M, of optical receivers, wherein each said optical receiver receives more than one said optical carrier such that said optical carriers mix and a said optical receiver receives symbols carried by more than one said optical carrier; and processing electrical outputs from said optical receivers representing said symbols on said mixed optical carriers to extract said encoded data.

Preferably the processing comprises combining the electrical signals from the M optical receivers into n data streams where n is no more than M; these may then be combined to provide an output data stream. In embodiments the encoding comprises encoding a data bit into symbols carried by more than one optical carrier. Additionally or alternatively in some preferred embodiments of the method the number of optical carriers is less than the number of optical receivers. Further preferably, the optical receivers have substantially regularly spaced centre wavelengths whilst the optical carriers are more sparsely or irregularly spaced.

In a further related aspect the invention provides an optical WDM (wavelength division multiplexed) communications system for communicating data, the system comprising: an optical WDM transmission system to encode said data into symbols on a first plurality, N, of optical carriers of said WDM system; an optical WDM receiving system comprising a second plurality, M, of optical receivers, to receive said optical carriers; and wherein N is no more than M.

The skilled person will appreciate that features of the above described aspects and embodiments of the invention, and those of the dependent claims of the earlier described aspects of the invention, may also be employed in embodiments of this further aspect of the invention. Thus in embodiments each optical receiver may receive more than one optical carrier such that the optical carriers mix and an optical receiver receives symbols carried by more than one optical carrier. The WDM system may then include a signal processing system to process electrical outputs from the optical receivers representing the symbols on the mixed optical carriers to extract the encoded data.

BRIEF DESCRIPTION OF THE DRAWINGS

These and other aspects of the invention will now be further described, by way of example only, with reference to the accompanying figures in which:

Figures 1a to 1d show, respectively, an optical DWDM communications system according to an embodiment of the invention, AWG passbands for the system of Figure 1a, examples of AWG with different passbands, and an example of unevenly distributed carrier wavelengths. Figures 2a to 2d show, respectively, a system block diagram for an optical communications system of the type shown in Figure 1 a, a system block diagram for an alternative optical communications system according to an embodiment of the invention, an example decoder for the systems of Figures 2a and 2b, and an example analogue channel estimation system;

Figure 3 shows a variant of the system of Figure 1 a, incorporating feedback; Figures 4a and 4b show graphs of signal-to-noise ratio (SNR) degradation at, respectively, different AWG 3dB passbands, and at different input channel spacings (Δλ) for the system of Figure 1 d; Figures 4c and 4d show shows the accumulated eye diagrams of 48 AWG outputs and recovered signals from 32 decoders; Figure 5 illustrates power budget estimation for a 25km single mode fibre (SMF) link;

Figures 6a to 6c show decoded eye diagrams in an embodiment of the 4 x 10 Gb/s system, illustrating SNR and BER performance at different carrier wavelength spacings, respectively for Figures 6a to 6c 0.58nm, 0.78nm, and 0.98nm;

Figures 7a to 7c show, respectively, SNR performance, BER performance, and corresponding eye diagrams, all for a carrier wavelength spacing of 0.78nm (after 25 km of SMF link); and Figure 8a and 8b show decoded eye diagrams in an embodiment of an 8 x 12.5 Gb/s system after 5 km and 25 km of SMF link, respectively; and Figure 8c to 8e show, respectively, the SNR degradation at different carrier wavelength spacings, BER performance, and power penalties for a 12.5 Gb/s channel after 25km of SMF link. DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS

We will describe a new form of uncooled WDM system which uses standard lasers, DE/MUXes and receivers. The system allows the wavelength of the lasers to drift but uses a cyclic demultiplexer at the receiver with more detection channels than at the transmitter so that the separate WDM signals can be decoded with the assistance of signal processing. By careful design of the overall system, the signal processing for cancelling spectral crosstalk itself is low cost and has low power consumption, being similar to that currently used in the 10GbE Ethernet standard. The system is designed with narrow channel separations, so that in extreme cases, if too many signal channels fall within a single receiver channel and hence too much crosstalk limits the electronic signal decoding performance, modest D.C. current tuning of the laser wavelength can be used to increase the separation. Although the system allows wavelengths to drift, by tracking the individual signals and cancelling crosstalk, channel restoration and hence low error rate uncooled WDM transmission is achieved.

In embodiments the signal processing used to decode the spectral channels may employ a similar approach to that of linear radio frequency multiple input multiple output (MIMO) systems, assuming there is no fibre transmission and chirp. Each AWG output y = (Outi ...Out_M) can be approximated by a linear time-invariant system of inputs x = (Ιπτ ... ln_N). The unknown wavelengths are identified during system training, where each laser is assigned to send a known sequence and received by the photodiode array. The received average power level is then used to determine the initial system (channel) response H = (CH ...C_MN), due to the physical constraints of the system, its entries, C_MN, must all be non-negative real numbers and the sum of each row and column must be greater than zero. The received signals, quantized by an 8-bit ADC, are weighted and summed according to the system response matrix:

And y = Hx + n

If H were the N ^χ N identity matrix, where M = N, then no processing of y would be required to recover x. However, this is invariably not the case so signal processing is used to recover the original signals, x. in an analogous approach to MIMO wireless communications. One implementation of this recovery process is zero-forcing equalisation. This attempts to estimate a matrix, W, such that WH = I_N, where I_N is the N x N identity matrix. In the zero-forcing case W is given by the Moore-Penrose pseudoinverse of H:

W = (H^HH)^_1 H^H

The transmitted signals can then be estimated as:

x = Wy

In the ideal case, this solution forces the cross-channel interference to reduce to zero (hence the name). Zero-forcing can be implemented with a filter, the coefficients of which are determined by transmitting a training signal over the channel.

Drawbacks of this simple MIMO DWDM system are the complicated matrix inverse computation and the singularity problem that arises if H is rank deficient (i.e. singular) meaning that it is not possible to compute (H^HH)^"1 and hence W. Consider a 6 χ 6 uncooled WDM MIMO system. If each of the wavelengths were located between two AWG output peaks, the ch nnel matrix would be of the form:

Here, the channel matrix, H, can be singular and not invertible for some coefficients, for example if a_k = b_k = 0.5 for k = 1 ...6. This corresponds to the case where each transmitted wavelength falls exactly halfway between the two nearest AWG response peaks. Generalising from this, it can be see that for an N x N matrix of the form:

The rows of the matrix are always linearly dependent (i.e. the matrix is singular) for even values of N. Stated another way, this matrix is singular when there is an even number of transmit/receive channel pairs in the WDM system. This can be seen by the fact that for even N, it is always possible to express the N^th row of the matrix, r_N, in terms of the other rows as follows:

JV-1

k=l

where r_k \s the k^th row of the H and a_k are positive real coefficients.

There do exist realistic scenarios, such as when the transmit wavelengths fall halfway between the peak wavelengths of the detectors, in which the singularity problem arises. This singularity issue can be addressed by introducing additional receivers such that the receivers outnumber the transmitters. This ensures that the rows of the channel matrix, H, form a set of linearly independent equations. In an N x N wireless MIMO system, introducing additional receivers creates additional diversity gain, on top of the spatial multiplexing gain achieved by having N independent channels. This additional diversity gain increases the robustness of the link and acts to support and enhance the spatial multiplexing process. Similarly, an improved WDM MIMO system design is achieved by adding more receivers than transmitters (an N χ M system where M>N). Provided the wavelengths are reasonably separated, the crosstalk can be suppressed in the electronic signal processing unit at the receive end despite the temperature variation. The original data streams can be recovered by weighting and summing all received signals. As environmental conditions to change slowly, a new set of weights is calculated using minimize mean square error (MMSE) algorithm only every 25ms this requiring low power and processing overhead. The MMSE algorithm tries to find a set of weights W that minimises E{(W · Out - In)²}, where E{} denotes the expectation operator. The weights are then used in an electrical signal processor based crosstalk cancellation unit, for example similar to that used in 1 0 GbE Ethernet.

Referring to Figure 1 , this shows a schematic diagram of an un-cooled DWDM optical communication system 100 according to an embodiment of the invention. The transmission system comprises a set of DFB (distributed feedback) lasers 1 02a-102N, which may be integrated on a common substrate for a compact implementation. The laser wavelengths define the carrier wavelengths of the DWDM communications system and, by design, may be substantially regularly spaced with gaps in the spacing as described later. However in embodiments these lasers are allowed to operate at unknown wavelengths. If integrated on a common substrate the wavelength separation of the lasers will remain roughly constant with temperature variation but this is not assumed.

The laser outputs are combined by an optical combiner 104, for example, comprising an Arrayed Waveguide Grating (AWG), coupled to a single mode fibre 106.

At the receive end of the system a second AWG 108 is provided, preferably a cyclic AWG so that a carrier wavelength shifting beyond one end of the passband range wrapped around to the opposite end of the passband range. The cyclic AWG 108 provides a set of M optical outputs which are detected by an array of photodiodes 1 10a-1 10M. In the example system of Figure 1 a there are N carrier wavelengths and a larger number M of photodiodes. For example in one embodiment there are 50% more photodiodes than carrier wavelengths. The output of each photodiode 1 10a is provided to a decoder 1 12a-N operating in the electrical signal domain, these decoders providing a prospective set of data outputs 1 14a-1 14N, which may afterwards be combined into a single, common data output stream, the combiner is not shown in Figure 1 a, for clarity. Each decoder comprises a plurality of taps 1 16a-1 16M, one coupled to receive a signal from each photodiode. A tap multiplies the photodiode signal by a complex weight to adjust the amplitude and phase of the signal, as described further later. These adjusted signals are then combined, in this example in a summer 1 18 to provide a data output 1 14. The decoder also includes a clock recovery circuit (CR), a minimum mean square area (MMSE) channel estimation circuit, and a limiter, as described further later. The particular configuration of these decoder modules may be varied and is illustrated only schematically in Figure 1 a. For example the output 1 14 may comprise soft data, for example representing the probability of a bit or symbol, or may be hard data, that is data on which a bit decision has been made, for example by the limiter. Soft data may be employed, for example, by a subsequent error correction stage (not shown). Referring next to Figure 1 b, this shows two sets of AWG passbands, the upper set representing a conventional AWG, the lower set illustrating an example of an AWG employed in embodiments of the invention. As can be seen, by comparison with the conventional approach, an AWG used in embodiments of the invention has broad filter passbands which deliberately introduce crosstalk. Thus the overall response of the comb filter defines a relatively flat-top passband, and in the illustrated example an individual passband is only 1 dB down at the crossing-point between two adjacent passbands. In this way a single photodiode receiver receives optical signals from a plurality of carrier wavelengths, providing a MIMO channel between the transmitter and receiver, as illustrated by the dashed lines in Figure 1 a, because a single receiver receives signals from multiple transmitter lasers. Broadly speaking it has been found that the flatter the top of the AWG filter profile, the better is the performance. (Although in embodiments the comb filter response shown in the lower portion of Figure 1 b is provided by an AWG, this is not essential, in particular if there is only a small number of carrier wavelengths, for example 8 or less wavelengths, a combination of passive optical splitter and wavelength selected filters could be used).

The optical communication system described with reference to Figures 1 a and 1 b introduces many decibels of inter-symbol interference (ISI). Thus in embodiments of the system the crosstalk is greater than 1 dB, 3dB, 5dB, 7dB or 10dB. In embodiments the eye diagram of the signal from any particular receiver may even be closed. The subsequent processing in the electrical domain recovers the signals undergoing crosstalk in the MIMO channel. The advantage of this approach, however, is that there is no need for precise alignment of the transmit laser wavelengths with the centre frequency of the AWG comb at the receiver end, and thus the system is tolerant of wavelength shifts and other imperfections.

Figure 1 c illustrates alternative example filter passband configurations in which the AWG has a set of passbands each centred at a respective centre wavelength with 100GHz spacing and in which the passband of a filter is arranged such that it overlaps a significant portion of the adjacent passbands. The passband attenuation at the centre wavelength of adjacent passband is not particularly relevant: It is instead important to maintain a significant high transmission level at the crossing point between two passbands that, when the laser wavelength drifts from one passband to another, the. receiving system.still receives a signal with good signal to noise ratio.

Figure 1d illustrates one example arrangement of the carrier wavelengths in which the nominal carrier wavelengths are regularly spaced, but in which every third carrier is omitted. In embodiments the regular spacing may be approximately 00GHz (0.78 nm). Thus, in effect, extra 'spacing' is provided for each pair of carrier wavelengths. Whereas only a small number of carrier wavelengths a gap may be inserted, for example, between the two central channels. At the receive end, one receiver (photodiode) is provided for each regular wavelength interval and thus, in this example, there are 50% more photodiodes than carrier wavelengths. By using a laser channel spacing which is not constant, more particularly by introducing gaps in the regular wavelength spacing, the system helps to maintain a good signal to crosstalk ratio amongst all the channels at the receiver.

To illustrate another example arrangement of the carrier wavelengths: In embodiments the carrier wavelengths are evenly distributed and the regular spacing may be approximately 150GHz (1.17 nm). This arrangement provides good system performance because all channels are received with very small crosstalk.

Figure 2a shows a block diagram of a WDM optical communications system 200 incorporating the communications system of Figure 1a. thus in. system 200 a data input 202 is provided to a channel encoder 204 which may provide forward error correction if required, and then to a modulator 206 to modulate the DFB lasers 102, for-example using NRZ modulation. As in Figure 1a, there are N lasers/carrier wavelengths, and N^■ receivers, photodiodes) providing signals to N decoders 112 which provide respective data stream outputs 114 to an output stage 208 which provides a recovered data output 210. The decoders 1 2 also demodulate the received signal; the output stage 208 combines the output streams into a single data output 210 and also applies error correction, for example using a vector Viterbi decoder, and de-interleaving (where the channel encoder also interleaves the data).

Referring to Figure 2a, this shows a variant optical communication system 250 similar to that of Figure 2a, and in which like elements are indicated by like reference numerals. In Figure 2b, however, there are M dfb lasers and M carrier wavelengths, although each receiver 1 10 still receives signals from a plurality of carriers, and there are fewer decoders 1 12 than receivers 1 10. In the arrangement of Figure 2b, the channel encoder is arranged to encode the data over multiple carriers simultaneously, so that there is transmitted signal redundancy. A corresponding de coder in output stage 208 recovers the encode signal transmitted over the M carrier wavelengths. Although in the arrangement of Figure 2b the decoders 1 12 and output stage 208 are shown as separate blocks, in practice this may be combined into a single module.

Thus referring now to Figure 2c, this shows one example of a decoder 1 12. The decoder has a set on inputs 206, one from each photodiode, optionally coupled to a set of analogue-to-digital converters 264, where the decoder operates in the digital domain. In this latter case preferably the input signals are over sampled, to say four samples per bit to adequately capture the phase, and then later decimated. The input signals are then applied to the set of taps 1 16 which receive respective complex weights (amplitude and phase) from a channel estimator, in this example MMSE channel estimator 266. The weighted outputs of the photodiode signals are combined in summer 1 18 to provide the data output 1 14. This signal is also provided to a clock recovery circuit 268 which is used to determine sample times for bit error estimation. Thus in one approach a hard bit decision 270 is made on the output data and the hard bit value is compared with the actual signal level from the eye diagram (see also Figure 2d) in the channel estimator 206 to determine an estimate for the bit error which the channel estimator 206 can then minimise by adjusting the weights of the taps. Where hard rather than soft bit data is desired the output of the bit decision module 210 can provide output 1 14 of the decoder. The skilled person will appreciate that there are many ways of implementing decoder 1 12.

Digital signal processing requires a high bit rate and it can be preferable to employ analogue signal processing. Figure 2d illustrates, how an approximation to an MMSE channel estimate may be determined in the analogue domain. Thus an incoming signal 220 is hard-limited 222 at a timing determined by an edge of the recovered clock, to make a bit decision, and this is compared 224 with the actual signal, and the result is integrated 226 to determine an approximation to an MMSE value. One such arrangement may be provided for each photodiode output. Optionally a clockless version of the arrangement of Figure 2d may alternatively be employed. In embodiments of the above described approach, a least mean squared (LMS) equaliser stores an initial channel estimate (or a reference version of a training signal) and then iteratively adjusted filter tap weights to minimise the mean squared error between the equalised signal and the stored reference. One simple technique for determining an initial channel estimate is to transmit on only one WDM channel at a time to determine one row/column of the channel estimate matrix at a time, thus avoiding crosstalk (a signal on one wavelength is received by all the photodiodes; then just the next wavelength is used and so forth). Alternatively more sophisticated, statistical techniques may be used to determine a channel estimate whilst transmitting multiple wavelengths simultaneously. The channel estimation process can be relatively slow because the weights are quasi static, varying only very slowly as temperature drifts. The use of a cone filter of the type shown in the lower portion of Figure 1 b as compared with the upper portion of Figure 1 b avoids deep nulls which would otherwise result in a channel estimate matrix which is ill-conditioned and could not easily be inverted to determine the original, input signal to the channel.

Figure 3 shows an optical WDM communications system 300 similar to that shown in Figure 1 (and in which like elements are indicated by like reference numerals) and incorporating a feedback control mechanism. Thus in the arrangement of Figure 3 a feedback communications path 302, which may be separate from or combined with fibre 106, is provided from one or more of decoders 1 12 back to the laser transmitters 102. The feedback path 302 controls a dc current tuning module 304 which controls the lasers 102 to adjust their wavelengths. In embodiments the system is designed with narrow channel separations and in extreme cases, if too many signal channels fall within a single receiver channel, excessive crosstalk can limit the electronic signal decoding performance. This can be detected by a decoder by detecting a level of ISI (for example by observing the errors in the bit decisions) or using, for example, a more general performance metric such as a bit error level in the data output of one or more decoders. In response to this a modest tuning of the laser wavelength(s) may be used to increase the carrier wavelength separation. Where the (DFB) lasers are monolithically integrated in general they will drift together but crosstalk can be a problem where the lasers are not integrated and drift at different rates, in which case the lasers can be tuned to separate them out, thus decreasing the overall overlap between channels. Thus, broadly speaking in embodiments a feedback signal is sent back to a controller at the transmitter side by a convenient path, based on the performance of the received signals such as a bit error ratio, the feedback signal being used to change the biasing conditions of the transmitters. The effect of improving the BER can be to reduce or minimise the power consumption of the overall communications system.

In other approaches the feedback path 302 may be employed for additional or alternative control strategies, for example to add a channel (carrier wavelength) where a 'collision' between carrier wavelengths is detected and/or to provide increased (electrical) channel encoding at the transmitter end, for example by increasing a level of forward error correction and/or by re-transmitting data over one or more channels.

Extending this concept the skilled person will recognise that rather than employing fewer carrier wavelengths than receivers, in related systems the same number of carrier wavelengths and receivers may be used and, instead, the data encoded such that it is transmitted with optical carrier redundancy. With this approach a block of data bits for transmission may be encoded so that it is transmitted over multiple carrier wavelengths simultaneously so that the optical redundancy is in the carrier wavelengths rather than the number of receivers.

Thus broadly speaking we have described uncooled MIMO WDM systems using advanced receiver signal processing techniques. In embodiments there is no need for thermo-electric cooling (of the lasers/optical combiner/AWG): although the system allows the carrier wavelengths to drift, by in effect tracking the individual signals and cancelling crosstalk, channel restoration and hence low error rate uncooled WDM transmission may be achieved.

In some preferred embodiments the multi-wavelength source at the transmitter comprises an array of lasers such as an array of EML's (electroabsorptive modulated lasers) or an array of laser sources, the WDM carrier wavelengths may alternatively be generated by other techniques including, but not limited to, a frequency cone generator from single wavelength or multiple wavelength sources. In embodiments the sources are integrated on the same substrate or mounted on the same heat block and are not temperature controlled, so that the comb of wavelengths comprising the optical outputs of the laser array is permitted to drift within a temperature operating range. In embodiments the wavelength spacing between adjacent lasers in the array comprises a set of constant wavelength spacings, for example at 100GHz, with additional gaps, for example also at 100GHz, introduced every I channels where I is, for example, two. Thus in embodiments there are 50% more photodiodes than transmitters, which helps to maintain a desired signal-to-crosstalk ratio amongst the channels at the receiver. The cyclic AWG preferably has a relatively flat-top passband and low insertion loss for all the input wavelengths. This is deliberately arranged to provide substantial channel crosstalk, to facilitate effective decoding of the MIMO signals. The signal processing in the electrical domain may be based upon any of a range of algorithms including, but not limited to BER detection, zero-forcing equalisation, a minimum mean square error (MMSE) determination, and the like, to recover the original signals. Each decoding unit may comprise a set of analogue amplifiers (taps), an electrical combiner, a limiter, an (output) ADC (analogue-to-digital converter), a clock recovery unit, and an analogue or digital processor to calculate the MMSE and to control the weight of each tap. The signal processing for cancelling the spectral crosstalk may thus be relatively low cost and have a low power consumption.

Embodiments of the system work well with a large number of channels because a good signal-to-crosstalk ratio amongst the channels is assisted by leaving gaps every few, for example, every two, channels. Optionally additional coding, error correction and signal processing techniques may also be employed in the optical communications system including, for example, for error correction, feedback/noise cancellation techniques and the like.

The feasibility of the scheme was evaluated using a MATLAB simulation of a 32 ^χ 10Gb/s system with a 150 GHz basic channel separation (equivalent to a 100 GHz system but omitting one carrier in every three to provide). A Gaussian laser response NRZ signal with 47.1 ps rise time (20% - 80%) is assumed for all 32 channels. 32 zero line width lasers with an initial centre wavelength from 1528.77nm to 1564.65nm, are assumed to drift 0.8nm slowly to longer wavelengths at the same rate. The combined NRZ signal is then filtered by a 48 channel 100GHz cyclic Gaussian shape AWG and then detected by a 7.5GHz 4th order Bessel-Thomson photo-receiver. MMSE algorithm is used to recover the original signals and measure the noise enhancement factor. Figure 4a shows the worst-case received signal-to-noise ratio (SNR) degradation with different AWG 3dB pass-bands. The channel to channel loss is significantly smaller in a filter with wider pass-band and results in a smaller noise enhancement factor. Figure 4b shows the tolerance of input channel spacing (Δλ). If the input channel spacing is allowed to drift from 0.55nm to 1 .75nm, there results a 5dB additional SNRe degradation in worst-case states. Figure 4c shows the accumulated eye diagrams of 48 AWG outputs over 1 second (equivalent to a wavelength drift of 0.8nm). In this 32 ^χ 48 MIMO DWDM system, we intentionally drop one channel after every two channels at the transmitter to maintain a high signal to crosstalk ratio. In the initial state, the transmitters' wavelengths are detected at 32 AWG outputs as shown in columns 1 , 3, 4 and 6. These wavelengths then slowly drift to the next AWG outputs when the temperature increases, as shown in column 2, 3, 4 and 1 . The photodiodes after these AWG outputs contain partial or mixed signals from multiple transmitters and therefore require electronic signal processing for data recovery. The WDM MIMO decoder successfully recovers the transmitted signals from the 48 AWG outputs after 50 iterations of weight updating, equivalent to an interval of 20 ms. This is shown in Figure 4d.

Figure 5 shows the estimated power budget for a four channel system. A post-amplifier is used to compensate the insertion losses from the optical combiner (6dBo) and AWG (3.7dBo) in the equipment-limited demonstration, and the attenuation of 25 km SMF (5dBo). Using a very basic restoration algorithm, a 2.5dBo system penalty from noise enhancement factor is expected in the worst case states. This penalty can be improved by an additional feedback or cancellation mechanism. A further 2dBo penalty is expected from MMSE calculation errors.

A proof-of-principle 4 channel system has shown the feasibility of the uncooled DWDM system we describe. Averaged signal traces were captured using oscilloscope and analyzed by MATLAB based offline signal processing to recover the original signals. The transmitter contains four C-band DFB lasers at 0.78nm spacing. All lasers are controlled to drift to longer wavelengths with a step of 0.01 nm simultaneously for 80 iterations (equivalent to a 0.8nm drift). A Mach-Zehnder intensity modulator is used to encode a 10Gb/s pseudorandom bit sequence (PRBS) data onto each wavelength. All NRZ signals are then combined using couplers and launched into a tunable filter with 0.9nm 3dB passband, which is used to replace a 100GHz cyclic AWG for more flexibility. The optical filter is set to 10 different centre wavelengths with 100GHz spacing, but then combined into 6 sets of outputs during offline signal processing to simulate a 6 channel cyclic AWG. The output NRZ data is then directly detected using an oscilloscope. 200 bits of averaged waveform are captured in each iteration. Output noise power is calculated separately based on the receiver noise, ADC noise and noise enhancement factor from the MMSE algorithm.

Figure 6 shows the decoded eye diagrams, calculated SNR and BER performance at different input spacing. Slight differences between the output powers of each laser results in different SNR start values. The estimated BER performance is worse than the SNR equivalent BER because of imperfections of the restoration algorithm. At 0.98nm spacing, channels 1 and 4 suffer more power penalty due to the decrease in signal to crosstalk ratio from the output of the cyclic AWG. At 0.58nm spacing, the insufficient channel spacing leads to a higher crosstalk power in channel 2 and 3. A significant SNR degradation is observed for both channels. The worst-case state SNR degradation is about 4dB for all four channels at 0.78nm spacing. This penalty occurs when the signal wavelength is located halfway between two AWG output peaks. This worst-case penalty could be further decreased by using a more sophisticated electronic decoder - decision-feedback equalization and/or noise cancellation would improve the output signal quality. Also, a hybrid transverse FIR equalizer will improve the dispersion penalty. A modest D.C. current tuning of the lasers can be used to detune the operating wavelength away from this worst case scenario when it is being detected at the receiver. The same experiment was repeated with the presence of 25km SMF and EDFA (Erbium doped fibre amplifier) to validate the system performance with chromatic dispersion (~17ps/nm-km). The input channel spacing is set to 0.78nm. SNR and BER performance are shown in Figures 7a and 7b, which are similar to the results without SMF. Figure 7(c) shows open eyes are obtained after signal processing.

A proof-of-principle 100 Gb/s system with 8 channels is also demonstrated to show the scalability of the WDM MIMO system. The transmitter contains 8 C-band DFB lasers. Each laser is externally modulated to encode a 12.5 Gb/s 2¹¹ - 1 NRZ PRBS, with an average spacing of 1 .17 nm. All lasers are controlled to simultaneously drift to longer wavelengths in 0.01 nm steps for a total of 80 steps (equivalent to 0.8 nm drift). The combined signal is amplified by an EDFA to compensate for the insertion loss of the couplers and then launched into SMF. The received signal is filtered by a tuneable filter with a 3 dBo pass-band of 0.9 nm, which replaces the 100 GHz cyclic AWG for more flexibility. The filter is set to 15 different centre wavelengths with 100 GHz spacing. These 15 outputs are then combined into 12 sets of outputs during offline processing to simulate a 12-channel cyclic AWG. The output data is directly detected using an oscilloscope. 200 bits of averaged waveform are captured at each step. The noise power is calculated based on the receiver noise and noise enhancement factor from the minimum MSE algorithm.

Figure 8a and 8b show example decoded output eye diagrams after 5 km and 25 km of SMF for a selected channel separation (1 .17 ± 0.0 nm). Figure 8c shows the tolerance of the system to input channel spacing. Because bias tuning of the laser diodes can accommodate for all but ±0.3 nm of relative drift between lasers and the absolute wavelength drift is fully accommodated for by the signal processing, we have focused our attention on the system degradation for this tolerance level. Channels are separated with an average spacing of 1 .17 nm, but are allowed to be offset by up to ±0.3 nm, with a minimum wavelength separation of 0.78 nm. Figure 8d shows example off-line processed BER curves and Figure 8e shows the power penalties for the selected channel separation. The worst-case power penalty is ~4.5dBo for all eight channels, which occurs when a wavelength is located halfway between two AWG output peaks. The penalties are in good agreement with the simulation study.

We have thus described and demonstrated a MIMO WDM concept for low-cost communication applications, allowing the use of uncooled standard components. Simulation results show that the original signals are restored successfully using the scheme we describe, even when wavelength drift causes a channel to drift from its corresponding receiver and be detected by an adjacent receiver. We have neglected the effect of fibre nonlinearity that arises from narrowed channel spacing as the number of channels increases, but in any case, an equalisation algorithm can be used to overcome these effects in a DWDM system H.S. Carrer, IEEE Globecom 2004, Vol 2, pp. 1005 - 1010. The laser line-width, chirp and signal processing latency induced additional penalties are neglected in this work but their effects are expected to be small. This system is expected to work with a larger numbers of channels - for example, a high signal-to-crosstalk ratio among all channels can be provided with additional 100 GHz gaps being introduced between every two adjacent channels. Thus we have also worked, for example, on a system with 8 lasers, a 22-channel AWG, and 8 outputs. The 40 Gb/s and 100 Gb/s experimental systems exhibit a degradation of less than 5 dBo as the temperature varies. Experimental results with MATLAB based offline signal processing have demonstrated the feasibility of this uncooled MIMO WDM system, and we have also worked on systems with real-time signal processing.

No doubt many other effective alternatives will occur to the skilled person. It will be understood that the invention is not limited to the described embodiments and encompasses modifications apparent to those skilled in the art lying within the spirit and scope of the claims appended hereto.

Claims

CLAIMS:

1 . An optical WDM (wavelength division multiplexed) communications system, the system comprising:

an optical WDM transmission system having an electrical signal input to receive input data for transmission and a WDM optical output to output a WDM optical signal having a first plurality, N, of carrier wavelengths encoding said input data; and an optical WDM receiving system having a WDM optical signal input and an electrical signal output to output decoded data; and

wherein said receiving system comprises a second plurality, M, of optical receivers, wherein each optical receiver of said plurality of optical receivers is configured to be able to simultaneously receive more than one said carrier wavelength.

2. An optical WDM (wavelength division multiplexed) as claimed in claim 1 wherein one or both of said transmission system and said receiving system

incorporates optical redundancy such that an optical path carrying a data bit of said input data includes at least two said optical receivers.

3. An optical WDM communications system as claimed in claim 2 wherein said transmission system is configured to encode said data bit onto at least two said carrier wavelengths.

4. An optical WDM communication system as claimed in claim 1 , 2 or 3 wherein said optical receiving system comprises electrical signal processing circuitry coupled to said M optical receivers, to process signals from said optical receivers to provide a third plurality, n, of data outputs, wherein n is less than M.

5. An optical WDM communications system as claimed in any one of claims 1 to 4 wherein M is greater than N such that said WDM communications system has more of said optical receivers than of said carrier wavelengths.

6. An optical WDM communications system as claimed in claim 5 wherein centre wavelengths of said optical receivers are substantially equally spaced; wherein, in a quiescent state, said carrier wavelengths match said centre wavelengths of said optical receivers; and wherein every a said carrier wavelength is omitted every ith receiver, where i is an integer.

7. An optical WDM communications system as claimed in any preceding claim wherein said optical WDM receiving system comprises an optical filter system between said optical input and said optical receivers, wherein said optical filter system has a set of passbands each centred at a respective centre wavelength, and wherein one said passband overlaps at least one adjacent said passband such that, for the or each said passband, transmission at a crossing-point wavelength of the two adjacent passbands is not more than 3dB down on transmission at a centre wavelength of said the or each passband.

8. An optical WDM communications system as claimed in any preceding claim wherein said receiving system comprises a cyclic AWG (arrayed wavelength grating) between said optical input and said optical receivers, wherein said AWG provides a set of filtered optical outputs for said optical receivers, each filter having a respective centre wavelength, and wherein said AWG does not provide greater than 3dB attenuation from one said centre wavelength of said AWG to an adjacent crossing-point wavelength of two adjacent passbands of said AWG.

9. An optical WDM communications system as claimed in claim 7 or 8 when dependent on claim 5, wherein said centre wavelengths are regularly spaced, and wherein said WDM optical signal is sparser than said centre wavelengths such that a number of said carrier wavelengths is less than a number of said centre wavelengths.

10. An optical WDM communications system as claimed in any one of claims 5 to 9 wherein said receiving system comprises a set of decoders each having: a set of taps, wherein each tap is coupled to the output of an optical receiver and configured to apply a complex weight to the output of an optical receiver to provide a respective weighted signal, and a signal combiner to combine said weighted signals; wherein said set of decoders provides a corresponding set of electrical signal outputs, one for each said carrier wavelength, for outputting said decoded data; and wherein said receiving system further comprises at least one channel estimation system, coupled to said optical receivers, to determine said complex weights.

1 1 . An optical WDM communications system as claimed in claim 10 wherein said channel estimation system comprises an MMSE channel estimation for each of said electrical signal outputs.

12. An optical WDM communications system in claim 10 or 1 1 wherein said taps and signal combiners operate in an analogue electrical signal domain.

13. An optical WDM communications system as claimed in claim 10 or 1 1 wherein said taps and signal combiners operate in a digital signal domain, and wherein said receiving system includes a set of M oversampling analogue-to-digital converters, one for each of said optical receivers, to provide digitised signals for said taps.

14. An optical WDM communications system as claimed in any preceding claim when dependent on claim 4 wherein said transmission system includes an encoder to encode said input data such that an incoming said data bit is encoded as one or more symbols for simultaneous transmission on each of said plurality of carrier wavelengths.

15. An optical WDM communications system as claimed in any preceding claim further comprising a feedback communications path between said receiving system and said transmitting system, wherein said receiving system includes a system to detect a performance of the receiving system and in response to send a feedback signal over said feedback communications path, and wherein said transmitting system includes a controller to control a spacing of said carrier wavelengths responsive to said feedback signal.

16. An optical WDM (wavelength division multiplexed) receiving system, the system having a WDM optical signal input and an electrical signal output to output decoded data, wherein said receiving system comprises an optical plurality of optical receivers, and wherein the system further comprises an optical filter system between said optical signal input and said optical receivers, wherein said optical filter system has a set of passbands each centred at a respective centre wavelength, and wherein one said passband overlaps at least one adjacent said passband such that, for the or each said passband, transmission at a crossing-point wavelength of the two adjacent passbands is not more than 3dB down on transmission at a centre wavelength of said the or each passband, such that each optical receiver of said plurality of optical receivers is configured to simultaneously receive more than one said carrier wavelength of a WDM optical signal at WDM optical signal input.

17. An optical WDM (wavelength division multiplexed) transmission system, the system having an electrical signal input to receive input data for transmission and a

WDM optical output to output a WDM optical signal having a first plurality, N, of carrier wavelengths encoding said input data; and wherein said transmission system includes an encoder to encode said input data such that an incoming said data bit is encoded as one or more symbols for simultaneous transmission on each of said plurality of carrier wavelengths.

18. A method of communicating data using an optical WDM (wavelength division multiplexed) system, the method comprising:

encoding said data into symbols on a first plurality of optical carriers of said WDM system;

receiving said optical carriers at a second plurality, M, of optical receivers, wherein each said optical receiver receives more than one said optical carrier such that said optical carriers mix and a said optical receiver receives symbols carried by more than one said optical carrier; and

processing electrical outputs from said optical receivers representing said symbols on said mixed optical carriers to extract said encoded data.

19. A method as claimed in claim 18 wherein said processing comprises combining said electrical signals from said M optical receivers onto a third plurality, n, data streams, wherein n is less than M, to extract said encoded data.

20. A method as claimed in claim 18 or 19 wherein said encoding comprises encoding a said data bit into symbols carried by more than one said optical carrier.

21 . A method as claimed in claim 18, 19 or 20 wherein a number N of said optical carriers is less than a number M of said optical receivers.

22. A method as claimed in claim 21 wherein said M optical receivers have a substantially regularly spaced centre wavelengths, and wherein wavelengths of said optical carriers are irregularly spaced.

23. A method as claimed in any one of claims 18 to 22 further comprising allowing a temperature of said optical WDM system to vary, to vary said wavelengths of said optical carriers such that during said communicating one or more of said optical receivers receives symbols carried by two or no said optical carriers.

24. An optical WDM (wavelength division multiplexed) communications system for communicating data, the system comprising:

an optical WDM transmission system to encode said data into symbols on a first plurality, N, of optical carriers of said WDM system;

an optical WDM receiving system comprising a second plurality, M, of optical receivers, to receive said optical carriers; and

wherein N is less than M.