Multilevel Coherent Optical System

1. Field of the Invention

The present invention relates to the communication systems using optical signals propagating through single-mode optical fibres and, in particular, a method of and an apparatus for generating, transmitting and receiving a multilevel optical signal.

2. Description of the Related Art

Reliable and economically competitive, coherent optical transmission systems which can be made available at short and medium terms allow novel network architectures to be provided regarding long-distance and high-performance connections and multi-user LAN (Local Area Network) and MAN (Metropolitan Area Network) connections as well. In particular, the very large bandwidth of the single-mode optical fibres (thousands of GHz) can be suitably exploited by providing optical FDM-systems (Frequency Division Multiplexing) in which the selection of the desired channel can be obtained by shifting the frequency of the local oscillator. This allows passive optical networks with very high traffic capacity (thousands of gB/s) to be carried out. However, two important aspects restrict on one hand the bandwidth of the single channel and limit on the other hand the maximum number of channels which can be tuned by the user. In the first instance, in fact, the main restriction is due to the bandwidth of the photodiodes and the electronic circuits, while regarding the second instance it should be considered that the frequency range which can be tuned by the user depends on the tunability characteristics of the laser used as local oscillator.

In order to increase the information rate of any channel, systems have been provided in which the information to be transmitted is coded with more than two levels instead of being coded using only the two binary levels as it is customary for providing a high signal reception sensitivity. By transmitting multilevel signals an improvement of spectrum efficiency expressed in terms of information rate per unit of occupied band is obtained at the cost of a reduction of the sensitivity. The known systems with two or more levels resort to the digital amplitude and phase keying (APK) or to the digital phase shift keying (PSK) or polarization shift keying (SPSK) of the electrical component of the electromagnetic field associated to a coherent optical wave generated by a laser source.

In particular, according to the previous state of art, EP-A-0 277 427 discloses methods of an devices for processing an optical signal by altering the polarization state thereof under control of a signal at a predetermined scrambling frequency.

EP-A-0 280 075 discloses an optical low-noise superheterodyne receiver for modulated optical signals in which a received light signal is coupled to a coherent light signal having the same polarization. Then such signals are combined so as to provide two pairs of optical signals, the signal of each pair having the same polarization perpendicular to that of the other pair, and fed to photoelements which provide electrical signals. Such electrical signals are then summed to each other after demodulation and after at least a phase shifting of one of such signals.

In "Electronics Letters" Vol. 26, No. 4 of 15 Feb. 1990 there is disclosed the performance of coherent optical transmission systems using multilevel polarization modulation based upon equipower signal constellations at the vertices of regular polyhedra inscribed in to the Poincare's sphere.

The present invention seeks to provide a method of generating a multilevel signal with a better performance than the known systems with regard to the signal reception sensitivity on the same number of employed levels. Within such general aim the invention seeks to provide in particular a transmitting and a receiving apparatus carrying out the above mentioned method.

Such aims are achieved by the invention defined and characterized in general in the claims attached to the following description in which the present invention is disclosed by way of a non-limitative example with reference to the accompanying drawing, in which:

FIG. 1 is a block diagram of a transmitting apparatus for a multilevel optical signal according to the present invention;

FIG. 2 is a block diagram of the detecting stage and the intermediate frequency stage of a receiving apparatus according to the invention;

FIG. 3 is a block diagram of a multilevel signal processing stage based on the determination of the coefficients of the inverted Jones matrix in a receiving apparatus according to the invention;

FIG. 4 is a block diagram of a multilevel signal processing stage based upon an algorithm for providing and uptodating the values of the components of the reference vectors in the receiving apparatus of the invention;

FIG. 5 is a block diagram of the circuit of the stage of FIG. 4 for uptodating the values of the components of the reference vectors;

FIG. 6 is a diagram of the logarithm of the error probability P.sub.e versus the number of the received photons per bit F for different values of the level number N;

FIG. 7 is a graph for the comparison of the sensitivity of the receiving apparatus (N-4Q) according to the invention, expressed in terms of the logarithm of the number of received photons per bit F versus the level number N, with the sensitivity of a N-PSK apparatus (N-level Phase Shift Keying), a N-APK apparatus (N-level Amplitude and Phase Keying), and a N-SPSK apparatus (N-level Polarization Shift Keying with detection by Stokes parameters); and

FIG. 8 is a graph for the comparison of the sensitivity of the receiving apparatus according to the invention, expressed in terms of the logarithm of the number of received photons per bit F versus the level number N, with the limit performance of the transmitting apparatus defined by the Shannon expression of the transmitting channel capacity.

The electrical field E(t) of an electromagnetic wave having angular frequency .omega..sub.o and propagating through a single-mode optical fibre can be written as follows:

where the phase terms x.sub.1 and x.sub.3 and the phase quadrature terms x.sub.2 and x.sub.4 are the components on the reference axes x and y of the polarization state, i.e. the vector representing the electrical field according to a given polarization. Vector X=(x.sub.1, x.sub.2, x.sub.3, x.sub.4) can be associated to any state of such electromagnetic field, the components of which being such that:

where P is the transmitted optical power;

The schematic block diagram of a transmitter according to the invention is shown in FIG. 1: a laser source 1 generates a linearly polarized optical carrier having a frequency, for example, of 10.sup.14 Hz, so as to form an angle of 45.degree. with respect to the reference axes x and y. The phase of such optical field is modulated by a phase modulator 2 with a message, for example a voltage having a time variable amplitude .alpha.(t), which is generated by a coder 10 from a binary sequence m(t) representing an information to be transmitted. After the phase modulation the components of the polarization state on axes x and y are split by a polarization selection beam splitter 3. It should be noted that the reference axes x and y are defined by the orientation of splitter 3. In the upper branch the polarization of the signal is rotated by 90.degree. by a polarization rotator 4 so as to align it with that of the signal in the lower branch. The phase of the latter signal is modulated by a modulator 5 with a message .beta.(t) also generated by coder 10. The two signals having the same polarization are mixed by a directional coupler 6, the outputs of which will be as follows:

where A.sup.2 is proportional to the transmitted optical power. The polarization state of signal s.sub.1 is then rotated by 90.degree. by a polarization rotator 7' so as to make it orthogonal to that of signal s.sub.2, the phase of which is modulated by a modulator 8 with a message .gamma.(t) generated by coder 10. The resulting signals are then coupled by a polarization selection directional coupler 9 to provide the optical signal to be transmitted through the fibre, the x and y polarization components of which have the following phase terms and phase quadrature terms: ##EQU1## where the function .alpha.(t), .beta.(t) and .gamma.(t) can have values between 0 and 2.pi. according to the selected codification method.

In particular, such functions are generated by coder 10 according to the following criteria. A succession of bits representing the information to be transmitted are fed into coder 10. Such succession is divided in groups of bits, each group of bits representing a symbol of the alphabet used by the coder. Thus the succession of bits is transformed in a succession of symbols. In case a N-level signal is transmitted and, for the sake of semplicity, under the assumption that N is a power of 2, each symbol is formed by m bits where m=2 log N. Each symbol can be univocally associated to a point of the sphere in the four-dimensional space in which the electromagnetic field is represented, such point being determined by the vector X=(x.sub.1, x.sub.2, x.sub.3, x.sub.4) or by a tern of generalized spherical coordinates .alpha., .beta. and .gamma. and by the radius of the sphere, i.e. the square root of the transmitted optical power. Therefore, the transmission of a symbol corresponds to the transmission of a well defined state of the electrical field. As the succession of bits m(t) are fed into the coder, an association between symbols and points at the coordinates .alpha., .beta. and .gamma. is effected; the latter are then entered into a digital-to-analog converter and transformed to the voltages .alpha.(t), .beta.(t) and .gamma.(t) which are the control signals of the demodulators 2, 5 and 8. It should be noted that the states of the electrical field are completely determined by the three angular coordinates as the transmitted optical power in the apparatus of FIG. 1 remains constant.

The block diagram of the stage detecting the optical signal and of the intermediate frequency stage of a receiving apparatus according to the invention is shown in FIG. 2.

The optical signal modulated in phase and polarization and generated by a transmitter of the type shown in FIG. 1 and transmitted through a single-mode fibre 11 is entered into a "90.degree. optical hybrid" 13 along with a coherent optical signal generated by a laser source operating as local oscillator 12. Such signal of the local oscillator having a frequency which differs from that of the transmitted signal carrier by a predetermined amount between 10.sup.8 and 10.sup.9 Hz is linearly polarized at 45.degree. with respect to the reference axes x and y. The 90.degree. optical hybrid 13 is a known device having two inputs and two outputs and providing at one output the sum of the input signals and at the other output the sum of one input signal and the other input signal the phase of which is shifted by 90.degree.. In such a case, therefore, the output signals are the phase component and the phase quadrature component of the beat signal.

The x and y components of the polarization state of the output signals of the optical hybrid 13 are then split by polarization selection beam splitters 14 and 15 defining with their orientations the reference axes x and y, and separately detected by four photodiodes 16, 17, 18 and 19. The four electrical intermediate frequency signals are then filtered by bandpass filters 20, 21, 22 and 23 centered about the intermediate frequency and having a double as high bandwidth as the figure rate R.sub.s, i.e. the inverse of the transmission time of a symbol. A phase locked loop (PLL) 28 and four multipliers 24, 25, 26 and 27 allow the four intermediate frequency signals y.sub.1, y.sub.2, y.sub.3 and y.sub.4 at the outputs of the filters 20-23 to be translated to base band. Such signals are then fed to four lowpass filters 29, 30, 31 and 32 having the same bandwidth as the figure rate R.sub.s so as to provide four base band signals z.sub.1, z.sub.2, z.sub.3 and z.sub.4 proportional to the estimated values of the components of vector X which are mainly impaired by the detection noise.

Two preferred embodiment of the processing apparatus have been proposed for providing and updating the estimated values of the components of vector X from the base band signals z.sub.1, z.sub.2, z.sub.3 and z.sub.4. Such apparatus allow the fluctuations of the polarization state of the optical signal due to the propagation through a single-mode fibre to be compensated by merely electronic techniques.

The operation of the first apparatus, the block diagram of which is shown in FIG. 3, is based on the determination of the inverse Jones matrix. As it is known, the effects due to the propagation through a single-mode optical fibre can be taken into account by the Jones unit operator providing the input-output relation between the polarization states of the optical field. Since such relation is linear, the application of the inverse Jones operator to the received signal allows the polarization state of the transmitted optical signal to be determined. Vector Z having the base band signals z.sub.1, z.sub.2, z.sub.3 and z.sub.4 as components is multiplied in Unit 33 by the inverse Jones matrix U.sup.-1 so as to provide the estimated values of the components of vector X. The coefficients of the matrix are determined by an algorithm based upon the consideration that the fluctuations of the polarization state (0, 1-1 Hz) due to the fibre birefringence are much slower than the figure rate (10-1000 Hz). The algorithm is implemented on the base of the calculation of the time averages of the signals z.sub.1, z.sub.2, z.sub.3 and z.sub.4 at coefficient units 34, 35, 36 and 37 in time intervals much longer than the symbol period. i.e. the transmission time of a symbol, and much shorter than the characteristic period of the polarization fluctuations. The elements of the Jones matrix depend linearly on the averages of the signals z.sub.1, z.sub.2, z.sub.3 and z.sub.4, as the coefficients of such linear relation are the averages of the four coordinates of the reference points evaluated in the set of the N feasible transmitted symbols and stored in calculation unit 38. Therefore, if the averages of the signals z.sub.1, z.sub.2, z.sub.3, z.sub.4 are known, a linear system of four equations with four unknown values can be implemented, the solution of which calculated in calculation unit 38 provides the real and imaginary parts of the coefficients of the Jones matrix, the inverse of which is then calculated in unit 33. This algorithm causes the coefficients of the Jones matrix to be uptodated at the end of any time period at which the time averages of the signals z.sub.1, z.sub.2, z.sub.3 and z.sub.4 are evaluated, thus allowing the apparatus to follow the fluctuations of the polarization state due to the single-mode fibre birefringence. The decision, i.e. the recognition of the state of the multilevel signal received at a given time, is effected in comparison unit 39 by comparing the estimated vector .epsilon. of components .epsilon..sub.1, .epsilon..sub.2, .epsilon..sub.3 and .epsilon..sub.4 with the reference vectors corresponding to the feasible transmitted symbols, the components of which have been stored in unit 39 when adjusting the apparatus. In particular, such comparison is effected by calculating the distances between the point on the surface of the sphere in the four-dimensional space corresponding to the estimated vector .epsilon. and the points determined by the reference vectors. Among the feasible transmitted symbols it is selected the symbol corresponding to the point determined by the reference vector having the shortest distance from the point of coordinates .epsilon..sub.1, .epsilon..sub.2, .epsilon..sub.3 and .epsilon..sub.4. The output signal of unit 30 is fed to an User apparatus 50.

The operation of the second apparatus processing the multilevel signal is on the contrary based upon an algorithm allowing the values of the coordinates of the reference points to be initially determined and uptodated. i.e. the components of the reference vectors on the surface of the sphere in the four-dimensional Euclidean space. The schematic block diagram of such processing apparatus is shown in FIG. 4. The apparatus determines initially the reference vectors by means of a suitable initialization sequence and subsequently effects the continuous uptodating of the components of such vectors, the values of which are fed to decision circuit 45 in which a decision is taken by the above described procedure based upon the calculation of the distance between the point corresponding to the received symbol and the reference points. The decision circuit 45 in case of a N-level signal has 4N memory cells in which the components of the N reference vectors are stored. In the time interval between two successive uptodatings the decision circuit 45 estimates the received symbol and associates it to any of the N symbols which can be transmitted. The uptodating of the components of any reference vector is carried out by calculating the mean value of the vector components which are estimated by the decision circuit during the uptodating interval as corresponding to that reference vector. At the end of any uptodating interval, which is chosen also in this case much shorter than the characteristic periods of the polarization fluctuations and much longer than the symbol period, the reference vectors are replaced by those corresponding to the novel components, the mean values of which calculated by the above described method have been stored in the 4N memory cells.

In the diagram of FIG. 4 the uptodating operation is effected by updating 40 formed of four circuits 41, 42, 43 and 44, each of them comprises a switch 46 and N mean value circuits 47 for the calculation of the mean value of the signal selected by the switch. After having estimated the received symbol, the decision circuit 45 supplies the control signal formed of the components of the reference vector corresponding thereto to the four blocks 41, 42, 43 and 44. Such control signal causes any base band signal z.sub.1, z.sub.2, z.sub.3 and z.sub.4 to be entered through switch 46 into circuit 47 for the calculation of the mean value corresponding to the reference symbol selected by the decision circuit 45 among N feasible symbols which can be transmitted. Therefore, during the uptodating interval the outputs of the circuits 41, 42, 43 and 44 supply the signals which are to be used at the uptodating time to calculate the mean values of the components of the novel reference vectors which are then stored in the 4N memory cells of the decision circuit 45. The resulting processing signal of block 45 is supplied to an user apparatus 50.

The performance of the apparatus has been valued in view of the statistics of the detection noise. In order to optimize the performance, the reference states of the transmitted optical field have been selected such as to reduce to a minimum the optical power necessary to achieve a predetermined error probability. In case of a N-level signal such choise consists in determining the position of N reference points on the sphere of the four-dimensional Euclidean space. From an analytical point of view the optimization of the performance can be achieved by an algorithm which minimizes the multi-variable function establishing the relationship between the error probability P.sub.e and the coordinates of the N reference points. The problem cannot be analytically solved in closed form so that a numeric algorithm has been used to minimize the above mentioned multi-dimensional function for 3.ltoreq.N.ltoreq.32.

Some results regarding feasible configurations of N reference points obtained by the minimization algorithm of multi-variable functions and using the downhill simplex method are shown in the following tables I, II, III, IV.

In particular Table I shows the values of the angular coordinates .phi., .PSI. and .theta. corresponding to the points of the sphere of the four-dimensional Euclidean space having standardized unit radius which are associated to the reference states of the electromagnetic field in case of an optimized five-level configuration. The angular coordinates are bound to the components x.sub.1, x.sub.2, x.sub.3 and x.sub.4 defining the state of the electromagnetic field by the following relations:

Table II shows the values of the distances between the reference points on the sphere of standardized unit radius in case of a five-level configuration; in this case the distance of any couple of points is the same, and when that result is obtained, that is the best for simmetry reasons.

Table III shows the values of the angular coordinates .phi., .PSI. and .theta. corresponding to the points on the sphere of the four-dimensional Euclidean space having standardized unit radius which are associated to the states of the electromagnetic field in case of an eight-level configuration.

Table IV shows the values of the distances between the reference points on the sphere having standarized unit radius in case of an eight-level configuration. In such case it was not possible to arrange the eight reference points on the four-dimensional sphere in such a way that they are at the same distance from one another. Nevertheless the optimum configuration has a high simmetry as any point has six first near points at a distance equal to the radius of the sphere multiplied by .sqroot.2 and only one second near point at a double as high distance as the radius of the sphere.

In FIG. 6 the performance of the apparatus is shown by the logarithm of the error probability P.sub.e versus the photon number per bit F for a number N of levels equal to 4.8 and 16, respectively.

In FIG. 7 the sensitivity of the apparatus is shown by the logarithm of the photon number per bit versus the number N of levels at an error probability of 10.sup.-9. In such figure the performance of the apparatus according to the invention designated by N-4Q is compared with that of a N-level heterodyne PSK apparatus (N-PSK, N-Phase-Shift-Keying), a N-level heterodyne APK apparatus (N-APK, N-Amplitude-Phase-Keying), and a N-level polarization modulation apparatus with detection by Stokes parameters (N-SPSK, N-Stokes-Parameter-Shift-Keying), the former two being described in K. Feher "Digital MODEM Techniques", Advanced Digital Communications, Prentice-Hall Inc., Eaglewood Cliffs, N.J., 1987, the third one being described in an article of S. Betti, F. Curti, G. De Marchis, E. Iannone, "Multilevel Coherent Optical System Based On Stokes Parameters Modulation" which is being published on the Journal of Lightwave Technology.

In FIG. 8 the limit performance of the transmitting apparatus conditioned by the Shannon equation regarding the channel capacity is shown. The apparatus according to the invention suffers from a penalty with respect to the Shannon limit of 8.5 dB for N=16, 7.4 dB for N=32 and 7.8 dB for N=64, respectively. The performance of the apparatus according to the invention with respect to the compared apparatus tends to improve as the number of levels increases as illustrated in the following Table V showing the improvement in dB of the performance of the apparatus according to the invention with respect to that of N-SPSK and N-PSK apparatus.

While only one embodiment of the invention has been illustrated and described, it should be appreciated that several changes and modifications can be made without parting from the scope of the invention.