JP2017163485A - Optical receiver and optical symbol label identification method in optical receiver - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve problems that, in a complex amplitude label decision using a machine learning technique in the prior arts, examinations including cycle slip are not made, a cycle slip in which phase estimation of a carrier at a reception side is failed occurs due to deterioration of an optical signal caused by phase instability or optical transmission characteristics of a laser, the occurrence is random and difficult to predict, it is difficult to prevent such a phenomenon itself as long as the laser has phase instability, and a more stable optical receiver taking the cycle slip into account is desired.SOLUTION: The present invention discloses a novel receiver system for constructing a boundary of complex amplitude labels by using a machine learning technique while also taking various kinds of phase noise generated in a laser light source used for an optical receiver, in particular, occurrence of cycle slip into account. The receiver system is operated in such a manner that the occurrence of the cycle slip is monitored and when a fixed discrimination condition is satisfied and the occurrence of the cycle slip is detected, a label of a received complex amplitude value is corrected, thereby removing phase offset that occurs in the cycle slip.SELECTED DRAWING: Figure 6

Description

  The present invention relates to an optical receiver (optical receiver) of an optical network. More specifically, the present invention relates to a technique for identifying a symbol label of an optical signal received by an optical signal receiver.

  With the spread of high-definition video distribution services and mobile communications, the amount of traffic flowing through the network has become enormous and continues to increase year by year. Research and development for increasing the frequency utilization efficiency of light is being carried out toward the construction of a large-capacity optical network that can meet such traffic demands.

  In an optical network, information is transmitted by intentionally changing the shape of a light wave as a carrier wave according to information. For example, in quadrature amplitude modulation (QAM), the amplitudes of two orthogonal carriers are changed, and the two carriers are combined and transmitted from the transmission side. In the receiver on the opposite reception side, two amplitude values (a set of amplitude values), that is, an optical complex amplitude value can be obtained by demultiplexing the received light wave into two orthogonal components. Optical complex amplitude values can be plotted as points on a graph called a signal space diagram (also called a signal point diagram). A point of the optical complex amplitude value is also called a “signal point”, and one signal point corresponds to one symbol composed of one or more information data bits. Also, this signal point is associated with a complex amplitude label representing the transmitted information bits. On the receiving side, a complex amplitude label determination boundary for determining a complex amplitude label can be constructed and held from the signal point coordinates on the signal space diagram of the received complex amplitude. By identifying the complex amplitude label based on the constructed boundary, it is possible to identify the information bit and realize transmission of information from the transmission side to the reception side. Further, the correspondence relationship between the complex amplitude label and the complex amplitude value can be increased, that is, in the signal space diagram, the number of complex amplitude coordinates associated with the complex amplitude label can be increased and densely arranged. As a result, the amount of information that can be transmitted by one set of complex amplitudes can be increased, the so-called multilevel can be increased, and the utilization efficiency of the optical frequency can be increased.

  However, by transmitting such an optical signal through an optical fiber, various types of noise are mixed. As a result, the complex amplitude value obtained on the receiving side is obtained by adding dispersion to the complex amplitude value on the ideal complex amplitude coordinate at the time of transmission. That is, the position of the signal point on the signal space diagram obtained on the receiving side varies. Therefore, in the currently general complex amplitude label determination, a two-dimensional signal space diagram is divided so as not to be superimposed on a square region having the number of complex amplitude labels, and each region is associated with one complex amplitude label. . Based on the relationship between each region and the complex amplitude value of the received optical signal, the complex amplitude label is determined regardless of the noise applied to the optical fiber.

  FIG. 1 is a diagram showing determination boundaries in the quadrature amplitude modulation method (16QAM) in which the number of complex amplitude labels is 16, and ideal complex amplitude points of each label. In FIG. 1, two orthogonal thick arrows correspond to two orthogonal carrier waves. One is an in-phase (I) axis, and the other is a quadrature (Q) axis. In 16QAM, a label determination boundary indicated by a dotted line constituting 16 square regions 2 on the IQ plane and an ideal complex amplitude point (signal point) 1 of each of the 16 labels are shown. Each signal point is labeled 0-15. Each signal point (label) is associated with a communication symbol, and one information symbol can represent four information bits. By using the complex amplitude label determination region as shown in FIG. 1, the complex amplitude label can be determined even when a complex amplitude value with added dispersion is received. If the variance added to the transmission complex amplitude value follows a normal distribution, the optimum complex amplitude label determination can be performed only by the above-described determination method. However, the nonlinear effect that actually occurs in the optical fiber is also a major factor that distorts the complex amplitude value, that is, the distribution of signal points.

  FIG. 2 is a diagram for explaining the influence of the nonlinear effect in the optical fiber on the distribution of the reception complex amplitude. FIG. 2A shows the signal point distribution when the influence of the nonlinear effect is small, and FIG. 2B shows the signal point distribution when the influence of the nonlinear effect is large. Referring to (a), when the influence of the non-linear effect is small, each signal point is distributed tightly within, for example, the substantially circular regions 21 and 22 of the dotted line, centering on each ideal signal point shown in FIG. , It is possible to clearly identify which label of signal points 0 to 15 corresponds to. On the other hand, referring to (b), when the influence of the non-linear effect is large, the signal points plotted on the two-dimensional graph vary widely and are distributed. The circular area 21 in (a) corresponds to the elliptical area 23 in (b), the shape of the area in which the signal points are distributed is deformed into an irregular shape and the size thereof is enlarged, and the adjacent label signal point distribution You can see that it overlaps. The circular area 22 in (a) corresponds to the elliptical area 24 in (b), and is in the same situation as the signal points in the circular area 21.

  When distortion occurs as shown in FIG. 2 (b), it is difficult to accurately identify the label when the complex amplitude label is determined based on the area simply divided into squares on the two-dimensional graph. cause. As is generally well-known, the distance between signal points of the ideal complex amplitude coordinate is narrower, that is, the number of different levels distinguished in one carrier is large, and transmitted by one set of complex amplitude values. As the number of types of amplitude levels that can be generated becomes higher, the higher-order multi-level modulation method, the more likely that a complex amplitude label is erroneously determined. Non-Patent Document 1 reports a result of evaluating a transmittable distance when an optical modulation method having a large multilevel is introduced.

  As a technique for coping with the above-mentioned problem of nonlinear distortion, Non-Patent Document 2 discloses a complex amplitude label determination boundary construction technique that can be applied to the nonlinearly distorted complex amplitude distribution described in FIG. . In this technique, a signal with a known complex amplitude label is received before information transmission is started. A discrimination boundary of complex amplitude labels is constructed by a machine learning method from a set of combinations of received signal point coordinates obtained from the received signal and corresponding known complex amplitude labels. This makes it possible to construct a complex amplitude label determination boundary that is highly adaptable to non-linear distortion that occurs in an optical fiber.

  The complex amplitude label determination using the machine learning method shown in the cited document 2 is mainly performed in two phases, a learning phase and a determination phase. In the learning phase, first, a learning signal having a known complex amplitude value is transmitted to the receiving side by a predetermined multilevel transmission method. At this time, the complex amplitude label corresponding to the complex amplitude value constituting the transmitted learning signal is given in advance to the receiving side, and it is known what label is transmitted on the receiving side. Therefore, by receiving the learning signal, a large number of combinations of actually received complex amplitude values (two real values) and complex amplitude labels can be prepared on the receiving side. Hereinafter, the combination of this complex amplitude value and its label is referred to as learning data. Furthermore, a set of learning data is called a learning data set. The learning signal may also be called a teacher signal.

  After receiving the above-described learning signal in the learning phase and accumulating sufficient learning data, a discrimination boundary for complex amplitude labels is constructed by a machine learning algorithm based on the acquired learning data set. In the learning phase, the correspondence between the complex amplitude coordinate and the complex amplitude label on the signal space diagram can be obtained by such an operation. When the learning phase is completed, in the determination phase, the complex amplitude label can be determined from the coordinates of the complex amplitude value received in the actual transmission of the information data, using the obtained complex amplitude label boundary.

Bosco, G .; Curri, V .; Carena, A .; Poggiolini, P .; Forghieri, F., "On the Performance of Nyquist-WDM Terabit Superchannels Based on PM-BPSK, PM-QPSK, PM-8QAM or PM -16QAM Subcarriers, "in Journal of Lightwave Technology, vol. 29, no. 1, pp. 53-61, January 1, 2011 Minliang Li; Song Yu; Jie Yang; Zhixiao Chen; Yi Han; Wanyi Gu, "Nonparameter Nonlinear Phase Noise Mitigation by Using M-ary Support Vector Machine for Coherent Optical Systems," in Photonics Journal, IEEE, vol. 5, no. 6, December 2013

  In complex amplitude label determination using such a machine learning method, a label boundary is constructed using complex amplitudes that are actually distorted obtained by transmitting a learning signal through an optical fiber. Therefore, the complex amplitude label can be determined while reflecting the nonlinear distortion actually received by the signal. In order to construct a practical complex amplitude label determination boundary with high identification accuracy by receiving a learning signal, it is essential to collect learning data without error.

  However, in the conventional complex amplitude label determination, the accuracy of the determination of the complex amplitude label is not sufficient due to the phase instability of the laser light source on the transmission side and the reception side, and further improvement is required. For example, in the 16QAM demodulation process, information is identified by the complex amplitude value of the signal point on the coordinates on the signal space diagram, that is, the amplitude and phase when viewed in polar coordinates. Therefore, in order to accurately demodulate and decode the received signal, the phase of the local oscillation laser light source of the receiver needs to be stable. In order to eliminate the instability of the phase of the laser light source, the optical receiver uses an optical phase locked loop (PLL) that stabilizes the phase of the laser light source or a digital carrier phase estimator that estimates the phase fluctuation from the received complex amplitude value. Use. In a receiver using these, it is inevitable that a phenomenon called a cycle slip that detects a complex amplitude value with an offset added to the phase of the original received complex amplitude with a certain probability is inevitable.

  It is known that the essential cause of cycle slip is the temporal instability of the phase of the laser output optical electric field. For this reason, not only the local oscillation laser on the optical receiver side but also the laser used on the transmitter side can cause cycle slip. A cycle slip occurs due to the phase instability of the laser light source itself, the signal degradation due to the transmission path, the modulation method, and the characteristics of the phase estimation circuit.

  Once this phenomenon occurs, the complex amplitude value is detected in the form in which an offset is similarly added to the subsequent received complex amplitude. As a result, a complex amplitude label determination boundary is constructed based on a learning data set including erroneous learning data. That is, the cycle slip is observed as a sudden shift (offset) of the reference phase, and the amount of the phase shift (offset) varies depending on the modulation method. Specifically, it is an integral multiple of the phase interval between two adjacent points in the rotationally symmetric group of signal points. For example, in the case of 16QAM, the phase offset is generated in units of an integer multiple of 90 degrees, and in the case of 8PSK (Phase shift keying), the phase offset is generated in units of an integer multiple of 45 degrees.

  While the oscillation frequency of radio is several GHz, the oscillation frequency (laser output frequency) of optical fiber communication is about 196 THz (terahertz), and it is not easy to control the phase of such a high frequency. For this reason, the phase of the laser is in a much unstable state as compared with a wireless oscillation circuit or the like. Although it is possible to estimate the occurrence probability from the characteristics of various causes of cycle slip that are currently known, it is impossible in principle to predict the occurrence timing. Therefore, the cycle slip occurs randomly, and can occur at various time intervals from several tens of picoseconds (corresponding to one symbol time length in optical communication) to ∞ seconds.

  FIG. 3 is a diagram illustrating an example of a learning data distribution and a label discrimination boundary when no cycle slip occurs. FIG. 3A is a plot of learning data on a 16QAM signal space diagram with different colors (saturation) for each label, which is generally the same as the signal point distribution of FIG. In addition, the signal points of the learning data are grouped and distributed near each label, and different labels can be recognized. (B) shows the boundary of the label determination constructed based on the learning data of (a), and each label area in the central part is generally quadrangular, and the labels in the peripheral part should not deviate from the quadrangle. However, it is a single area. By using this label determination boundary, for example, when the complex new amplitude value of the received signal is at the signal point 30, it can be determined that the complex new amplitude value has a label belonging to the region 31.

  FIG. 4 is a diagram illustrating an example of a learning data distribution and a label discrimination boundary when a cycle slip occurs. FIG. 4 (a) is a plot of learning data on a 16QAM signal space diagram with different colors (saturation) for each label, similar to FIG. 3 (a). Data signal points are distributed and distributed, and regions with different labels are difficult to distinguish. (B) shows the boundary of label determination constructed based on the learning data of (a), but the shape of each label determination area in the central part is significantly distorted and the size of the area varies. In the peripheral portion, a plurality of regions appear as enclaves divided into two regions, although the same label is represented, such as the two regions 41a and 41b and 42a and 42b. The label determination boundary as shown in FIG. 4B does not include this offset even though the phase of the received complex amplitude value includes an unexpected offset due to the occurrence of cycle slip. This is caused by the construction of boundaries. If an unexpected phase offset is included, it is misunderstood that the intended signal point remains as it is, although it is observed at a position different from the originally assumed signal point position. If the complex amplitude label is determined at the boundary generated by incorrect learning data, the accuracy is not sufficient, and the determination result also includes an error.

  In the complex amplitude label determination using the machine learning method of the prior art, the examination considering the cycle slip has not been made yet. In optical fiber transmission, the received optical signal level is stable because there is no fading as in radio wave propagation, but the phase of a laser light source such as a local oscillation light source is caused by various laser noises. Cycle slip occurs. The occurrence is random and difficult to predict. In addition, cycle slip occurs at a stage before error correction means at the bit level such as differential encoding and gray mapping at the time of bit conversion from a symbol, and includes a phase modulation signal that includes information in the phase of the optical signal Gives an error directly to the "raw" phase of As long as the laser light source has phase instability, it is difficult to prevent the cycle slip phenomenon itself, and a more stable optical receiver considering the occurrence of the cycle slip has been desired.

  The present invention has been made in view of such problems, and its object is to identify a complex amplitude value of an optical signal, which is more stable and accurate than the prior art. It is to construct a complex amplitude label identification boundary and obtain better communication quality.

In order to achieve the above object, the present invention provides a modulation system including at least a part of a plurality of signal points that are rotationally symmetric with respect to an origin on complex amplitude coordinates. Receiving a teacher signal modulated with a data sequence using the modulation method, and sequentially outputting complex amplitudes corresponding to any of the signal points on the complex amplitude coordinates A phase estimation circuit that sequentially receives the complex amplitude output from the phase estimation circuit, calculates a phase of the received complex amplitude, and generates a cycle slip that causes an offset in the phase of the complex amplitude A detection circuit for detecting, and a plurality of sets of corresponding labels indicating the contents of information carried by the sequentially received complex amplitudes and signal points corresponding to the complex amplitudes A circuit for storing a training data set and the stored complex corresponding to the detected complex amplitude based on a phase offset detected by the detection circuit in response to the detection of the occurrence of the cycle slip Based on the circuit for correcting the amplitude value (label) and the stored learning data set including the corrected complex amplitude value (label), different labels of a plurality of signal points corresponding to the complex amplitude are identified. And a circuit configured to construct a boundary for the receiver. As will be described later, the object to be corrected based on the detection of the occurrence of cycle slip may be a complex amplitude value stored in a circuit to be stored or a label.
The invention according to claim 2 is the receiving apparatus according to claim 1, wherein the detection circuit is an average of the plurality of stored complex amplitudes corresponding to each signal point at a rotationally symmetric position in the modulation scheme. An average phase calculation circuit for calculating each phase and a signal point at the rotationally symmetric position that can be superimposed on the complex amplitude coordinates by rotating by a predetermined rotational symmetric angle are grouped into one group. When configuring a group, it corresponds to the phase of the current complex amplitude and the label of the signal point excluding the signal point in the group to which the current complex amplitude belongs and corresponding to the current complex amplitude Based on the phase difference between the phase difference calculation circuit for calculating the phase difference between the average phase and each of the signal points in the group to which the current complex amplitude belongs, a predetermined cycle time is calculated. And a phase difference monitoring circuit that determines whether or not a loop generation determination condition is satisfied, and specifies a label of a signal point that satisfies the cycle slip generation determination condition as an updated label of the current complex amplitude, and To do.

  A third aspect of the present invention is the receiving apparatus according to the second aspect, wherein the modulation scheme is 16QAM, and the four different 90-degree rotational symmetry points with respect to 16 different signal points on the complex amplitude coordinate. Four groups of symbols are configured, and the phase difference calculation circuit corresponds to the phase of the current complex amplitude and the signal point in the group to which the current complex amplitude belongs and corresponds to the current complex amplitude. A phase difference between the average phase corresponding to the labels of the three signal points excluding the signal points to be obtained is obtained, and the phase difference monitoring circuit is configured to include the three signal points as part of the cycle slip occurrence determination condition. Monitoring whether the determined phase difference corresponding to the label is within 45 degrees, and the detection circuit satisfies that the determined phase difference is within 45 degrees to satisfy the current complex amplitude Plus before A count circuit for measuring the number of times that has occurred successively with respect to sequentially received complex amplitudes, and determining the occurrence of the cycle slip when the measured number exceeds a predetermined value. Is further included.

  The invention according to claim 4 is the receiving apparatus according to any one of claims 1 to 3, wherein a learning data signal modulated by a predetermined data sequence is received as the teacher signal, and the learning data signal is received from the learning data signal. A learning mode that operates to store the obtained complex amplitude and a known label corresponding to the data string as the learning data set, and a learning data signal modulated with an unknown payload signal as the teacher signal Then, the complex amplitude obtained from the learning data signal and the label of the complex amplitude are determined, the determined label is bit-converted, and the error-corrected bit is further subjected to bit / label conversion. It is stored as the learning data set, and is operated to perform the boundary determination for the payload signal and the boundary reconstruction for label identification in parallel. Characterized in that it is configured to switch between the modes.

  The invention according to claim 5 is the receiving apparatus according to claim 4, wherein in the tracking mode, the reconstruction of the boundary for label identification is performed at regular intervals.

  As described above, in a receiver that constructs an identification boundary of complex amplitude values of a signal receiver of an optical network, it is possible to construct a complex amplitude identification boundary that is stable and highly accurate, and communication quality is improved.

FIG. 1 is a diagram illustrating determination boundaries in quadrature amplitude modulation (16QAM) and ideal signal points for each label. FIG. 2 is a diagram for explaining the influence of the nonlinear effect in the optical fiber on the distribution of the reception complex amplitude. FIG. 3 is a diagram illustrating an example of a learning data distribution and a label discrimination boundary when no cycle slip occurs. FIG. 4 is a diagram illustrating an example of a learning data distribution and a label discrimination boundary when a cycle slip occurs. FIG. 5 is a diagram showing an outline of the configuration of the receiver of the present invention in the optical node. FIG. 6 is a diagram showing a configuration of a complex amplitude label determination circuit in the optical receiver of the present invention. FIG. 7 is a diagram illustrating the arrangement of 16QAM complex amplitudes used in the optical receiver according to the first embodiment of the present invention. FIG. 8 is a diagram for explaining average phase calculation and grouping in the complex amplitude label determination circuit of the optical receiver of the present invention. FIG. 9 is a diagram for explaining an operation example of group selection in the complex amplitude label determination circuit of the optical receiver of the present invention. FIG. 10 is a diagram for explaining another operation example of group selection in the complex amplitude label determination circuit of the optical receiver of the present invention. FIG. 11 is a diagram for explaining an example of the phase difference calculation operation in the complex amplitude label determination circuit of the optical receiver of the present invention. FIG. 12 is a diagram for explaining the operation of determining the occurrence of cycle slip in the complex amplitude label determination circuit of the optical receiver of the present invention. FIG. 13 is a diagram illustrating a configuration example of a counter used for determining occurrence of a cycle slip in the phase difference monitoring unit. FIG. 14 is a diagram for explaining the communication quality obtained by the complex amplitude label determination circuit of the optical receiver of the present invention in comparison with the prior art.

  The present invention is a novel receiver system that constructs a complex amplitude label boundary using a machine learning method in consideration of various phase noises generated by a laser light source used in an optical receiver, and in particular, generation of cycle slip. To clarify. In the receiver system of the present invention, when the occurrence of the cycle slip is detected by monitoring the occurrence of the cycle slip and satisfying a certain condition, the label of the complex amplitude value received is corrected, and the phase offset caused by the cycle slip is corrected. Works to remove. The receiver system of the present invention can be implemented in various ways, but is suitable for implementation by arithmetic processing using a digital signal processor (DSP), and all processing can be performed in the digital domain. it can. In addition, it can be operated in a learning phase that receives known learning data and constructs a boundary of a complex amplitude label, and a determination phase that performs subsequent normal data reception to determine a label (non-tracking mode) ). Furthermore, it is possible to operate in a tracking mode in which received data bits are re-symbolized by performing bit / symbol (label) reconversion from data obtained from an actual received signal and used as label information for learning data. .

  Hereinafter, various embodiments of the present invention will be described with reference to the drawings, but the present invention is not limited to the following embodiments. In the following description, “complex amplitude” obtained from a received signal is a set of amplitude values of signal points (I-axis amplitude value and Q-axis amplitude value) on a signal space diagram represented by an IQ coordinate plane. Means. The “complex amplitude” can also be expressed in a polar coordinate format, and in that case, it can also be expressed by an amplitude and a phase value.

  The term “complex amplitude label” is a display that is attached to each of a plurality of signal points on the signal space diagram determined by the modulation scheme and represents one or more information bits transmitted at the signal points. In order to avoid redundancy, the label is sometimes simply referred to as “label”. With this label, the contents of the transmission information bits carried by the signal point can be specified. Since any one of the above-described plurality of signal points has a one-to-one correspondence in one modulation symbol time, the label and the symbol also have a one-to-one correspondence at a certain time. Note that in the following description, the terms label and symbol may be used interchangeably. Hereinafter, an outline of a system configuration, an outline of operation of label boundary determination, and a more specific apparatus operation will be sequentially described with reference to the drawings.

In the following, the present invention will be described as an invention of an optical receiver. An optical signal that is used in an optical communication system and that carries information on the phase or amplitude of light and is modulated by a modulation method that can identify this information on the receiving side. As long as it can receive the light, it can be applied to any optical receiver. Naturally, it may be an apparatus including an optical signal transmission function. In addition, in the complex amplitude label determination circuit described below, the present invention can be implemented as an invention of a method for constructing a complex amplitude label boundary by using a machine learning method that also considers the occurrence of cycle slip. Furthermore, the present invention can also be implemented as an invention of a storage medium storing a computer program for executing each step of the method.
1. System Configuration FIG. 5 is a diagram showing a schematic configuration of an optical receiver of the present invention in an optical node. As shown in FIG. 5, the optical receiver 50 of the present invention includes a polarization compensation circuit 51, a phase estimation circuit 52, a complex amplitude label determination circuit 54, and a forward error as a part for realizing the basic receiver function. The correction circuit 55 is used. When the optical receiver 50 receives an optical signal, the polarization compensation circuit 51 first compensates for signal distortion depending on the polarization state. Next, the phase estimation circuit 52 estimates the phase of the carrier wave at the timing of the symbol rate corresponding to the predetermined modulation method and modulation rate, and outputs the complex amplitude of the received signal. The output of the complex amplitude is input to the complex amplitude label determination circuit 54 of the present invention, which will be described later, and the complex amplitude label is determined according to the complex amplitude label determination boundary constructed in advance and the positional relationship between the received complex amplitudes. The Thereafter, the determined complex amplitude label is converted from a symbol to a bit sequence, and the forward error correction circuit 55 performs error detection / correction on the received bit sequence. From the optical receiver 50, an error-corrected bit string 59 is output. The blocks after the complex amplitude label determination circuit 54 are also collectively referred to as a bit string decoding circuit 53.

  As will be described later, the optical receiver of the present invention receives actual information data instead of learning data in addition to the non-tracking mode in which the label determination boundary is constructed at the time of non-normal operation of the receiver using the learning data. However, complex amplitude label decision boundaries can also be constructed. This is an operation mode called the tracking mode (Example 2). At this time, the bit string decoded by the bit symbol conversion unit 56 is converted again into symbols and used as learning data. The switch 57 conceptually shows the switching between the tracking mode and the non-tracking mode. Next, a more detailed configuration of the complex amplitude label determination circuit surrounded by the dotted line in FIG. 8 will be described.

  FIG. 6 is a diagram showing a configuration of a complex amplitude label determination circuit in the optical receiver of the present invention. As shown in FIG. 6, the complex amplitude label determination circuit includes two blocks roughly surrounded by a dotted line, that is, a cycle slip detection / correction unit 78 and a label boundary construction / label determination unit 79. As will be described later, some functions are common to the above two blocks. In the following description of FIG. 6, the mutual connection relationship among the functional blocks, data input / output in each functional block, information, an outline of the operation, and the like will be described. Detailed operations will be described separately together with examples. In the following description, the complex amplitude value 61 received by the optical receiver and output from the phase estimation circuit 52 is input to the switch 80, the non-tracking mode is selected, and the operation of the learning phase is further performed by the switch 81. The state where is selected will be described. At this time, the non-tracking mode is selected by the switch 82, and the complex amplitude label 62 of known learning data provided from any place of the receiver is given to the complex amplitude label determination circuit.

  The cycle slip detection / correction unit 78 in FIG. 6 detects the occurrence of cycle slip from the complex amplitude value 61 received by the optical receiver and output from the phase estimation circuit 52, and corrects the received complex amplitude in response to the detection. It is configured to As sub-blocks, a learning data storage unit 66, a learning data distribution unit 67, an average phase calculation unit 74, an average phase grouping unit 75, a group selection unit 76, an average phase difference calculation unit 63, an average phase difference monitoring unit 64, and a reception complex An amplitude correction unit 65 is included. Arrows between the sub-blocks indicate the flow of information or data, and the name of the arrow indicates the number of data. Each block in FIG. 6 is a block corresponding to each processing content for each of arithmetic processing such as collection of label learning data, detection of cycle slip, correction of received complex amplitude values, construction of a label determination boundary, which will be described later. It is drawn and does not have to be configured as separate hardware blocks. It should be noted that all the functional blocks can be implemented as a group of arithmetic processing using a DSP, for example, and depending on a specific method of arithmetic processing, the functional blocks are not limited to block division as shown in FIG.

  The learning data storage unit 66 stores many combinations of the received complex amplitude and the complex amplitude label 62 that are learning data. The learning data distribution unit 67 outputs the learning data for each learning data to different output ports according to the content of the complex amplitude label. Here, the term “output port” corresponds to the fact that calculation processing of cycle slip detection and label boundary construction described later is performed for each complex amplitude label of the learning data set. In other words, this means that information processing is performed for each complex amplitude label at least in the first stage, and does not necessarily mean that some physical signal is output to a separate output port. Of course, at least a part of the arithmetic processing can be performed by dedicated hardware. In such a case, the physical signal may be output to a separate output port. The description of the output port, the input port, and the like can be considered similarly in the following other blocks.

  In the following description, the arithmetic processing of each block can be performed in synchronization with the complex amplitude value input 61 that is demodulated every moment from the learning data input to the receiver during a predetermined period. However, if a sufficient amount of the complex amplitude value 61 and the corresponding known complex amplitude label 62 can be stored as a learning data set in the storage means or the learning data storage unit 66 not described in FIG. These calculation processes may be performed collectively, or some processes may be delayed. One characteristic of the complex amplitude label determination circuit in the optical receiver of the present invention shown in FIG. 6 is that a more accurate learning data set considering cycle slip is constructed. In the “non-tracking mode”, it is not necessary to process the actual user data, so that the processing of each block does not need to be synchronized with the complex amplitude value input 61 in real time. In order for the calculated value of the average phase to have a certain degree of accuracy, the calculation process of the average phase may be performed in advance, and then the complex amplitude correction process may be performed.

  The average phase calculation unit 74 calculates and outputs the average value of the phases of the received complex amplitude values 61 for each input port, that is, for each complex amplitude label given from the learning data distribution unit 67. Therefore, the average phase calculation unit 74 receives “label number” type data and outputs “label number” type average phase values. The type data of “number of labels” mentioned here corresponds to a known complex amplitude label 62 corresponding to the complex amplitude value 61 at the timing when the learning data is input to the receiver and the complex amplitude value 61 is sequentially input. The signals are sequentially input to one input port, and the average phase is calculated.

  The average phase grouping unit 75 groups a plurality of complex amplitude labels having a rotationally symmetrical positional relationship in an ideal complex amplitude (signal point) arrangement having no dispersion or distortion, and assigns complex amplitude labels belonging to the same group. For the received complex amplitude, the average phase is calculated for each complex amplitude label. The calculated average phase is also grouped. In short, the average phase grouping unit 75 classifies the same number of average phase values as the total number of complex amplitude labels calculated by the average phase calculation unit 74 into a plurality of groups of “number of groups”. Details of grouping of complex amplitude labels will be described later.

  The group selection unit 76 receives the complex amplitude label 62 input to the learning data storage unit 66 and newly stored as a learning data set, and the grouped arithmetic average phase from the average phase grouping unit 75. To do. Based on these inputs, an average phase value corresponding to the complex amplitude label that is in the same group as the input complex amplitude label 62 and that excludes the input complex amplitude label 62 is output. Therefore, the group selection unit 76 receives the average phase value of “number of groups” and outputs the average phase value of “number of groups−1”.

  The phase difference calculation unit 63 calculates the difference between the actual phase of the complex amplitude value 61 newly stored in the learning data set and the plurality of average phases input from the average phase grouping unit 75, and calculates the phase difference. The absolute value is rounded to within 180 degrees and output. The calculated phase difference is output in the same manner as the group selection unit 76. That is, the phase of the complex amplitude value 61 from the learning data input via the optical fiber that is actually subjected to nonlinear distortion is the same group as the input complex amplitude label 62 corresponding to the complex amplitude value 61. The difference from the average phase value corresponding to the complex amplitude label excluding the input complex amplitude label 62 is output.

  The phase difference monitoring unit 64 detects the occurrence of a cycle slip based on the output from the phase difference calculation unit 63 and notifies the correction unit 65 of the received complex amplitude. The absolute values of a plurality of rounded phase differences calculated by the phase difference calculator 63 are input. When the phase difference from the same input port, that is, the phase difference of the same complex amplitude label satisfies a preset condition, occurrence of a cycle slip is detected. For example, the number of consecutive times when the phase difference of the same complex amplitude label is within a preset threshold value is counted, and the occurrence of a cycle slip may be detected when the number is equal to or greater than the preset number of times. After detecting the occurrence of the cycle slip, the phase difference monitoring unit 64 notifies the reception complex amplitude correction unit 65 of the complex amplitude label in which the occurrence of the cycle slip is detected. For example, an input port number where occurrence of cycle slip is detected may be notified to the reception complex amplitude correction unit. The reception complex amplitude correction unit 65 is based on the input input port number, that is, the complex amplitude label in which the occurrence of the cycle slip is detected and the phase difference calculated by the phase difference calculation unit 63, and the phase slip amount due to the cycle slip. And the label of the reception complex amplitude in which the occurrence of the cycle slip is detected and the subsequent reception complex amplitude are corrected.

  The label boundary construction / label judgment unit 79, which is another block of the complex amplitude label judgment circuit in the optical receiver of the present invention shown in FIG. 6, is configured to construct a complex amplitude label judgment boundary and judge a complex amplitude label. Has been. The label boundary construction / label determination unit 79 includes a complex amplitude label determination boundary construction unit 68, a complex amplitude in addition to the learning data storage unit 66 and the learning data distribution unit 67 included in the cycle slip detection / correction unit 78. A label determination boundary storage unit 69 and a complex amplitude label determination unit 70 are included.

  The learning data storage unit 66 stores many combinations of complex amplitudes and corresponding complex amplitude labels that are learning data. The learning data distribution unit 67 distributes each learning data corresponding to each complex amplitude label according to the label for arithmetic processing in the next block. The complex amplitude label determination boundary construction unit 68 constructs a boundary for label determination from the learning data supplied from the learning data distribution unit 67 by a machine learning algorithm. The complex amplitude label determination boundary storage unit 69 stores the label determination boundary on the signal space diagram obtained by the complex amplitude label determination boundary construction unit 68. The complex amplitude label determination unit 70 does not know the complex amplitude label determination boundary obtained by the complex amplitude label determination boundary construction unit 68 in the preceding stage and the complex amplitude label such as the payload signal input to the optical receiver. The corresponding complex amplitude label is determined by comparing with the complex amplitude.

In the complex amplitude label determination circuit in the optical receiver of the present invention shown in FIG. 6, the label / bit conversion unit 71 converts the label determined by the complex amplitude label determination unit 70 into a bit string. The label is a label of a signal point on the signal space diagram, and the signal point corresponds to the symbol on a one-to-one basis, so the label / bit conversion is the same as the symbol / bit conversion. The converted bit string is subjected to error correction processing by the forward error correction unit 72, and then a bit string 77 without error is obtained. That is, the complex amplitude label determination circuit of FIG. 6 performs labeling based on the complex amplitude label boundary constructed from the complex amplitude value 61 inputted from the phase estimator 52 by the learning method considering the cycle slip of the present invention. A determination is made, and an error-free bit string 77 is obtained. Next, the operation of the label determination circuit based on the boundary of the complex amplitude label constructed by the machine learning method considering cycle slip will be described in detail.
2. Operation of complex amplitude label determination circuit <Overview of operation>
Hereinafter, a more detailed operation of the complex amplitude label determination circuit in the optical receiver of the present invention having the configuration shown in FIGS. 5 and 6 will be described.

  The optical receiver of the present invention has two operation modes, a non-tracking mode and a tracking mode, from the viewpoint of the operation of the complex amplitude label determination circuit. In non-tracking mode, the construction of the boundary of complex amplitude label determination using machine learning method considering cycle slip is independent of the reception operation of actual data such as payload signal using known learning data. This mode is implemented. On the other hand, the tracking mode is a mode in which the construction of the complex amplitude label determination boundary using the machine learning method considering cycle slip is performed simultaneously with the reception operation of the actual data whose label such as the payload signal is unknown. In FIG. 6, the non-tracking mode and the tracking mode can be switched by switching the path of the received complex amplitude value 61 or complex amplitude label 62 as learning data by the switch 80 and the switch 82. However, it is not necessary to implement as a hardware switch for switching the signal path as shown in FIG. 6, and it is sufficient that the signal or the information flow is switched equivalently in the arithmetic processing. It suffices that the method of supplying the label to the learning data storage unit and whether to use the received complex amplitude as learning data can be implemented by two different arithmetic processes.

  The operation in the non-tracking mode further includes a learning phase for collecting learning data and building a complex amplitude label determination boundary, and a determination phase for performing complex amplitude label determination of the payload signal. These two phases are also shown by path switching by the switch 81 in FIG.

  In the operation of the complex amplitude label determination circuit of the receiver of the present invention, it is assumed that a complex amplitude value is input as learning data for machine learning at a predetermined timing or with a predetermined time interval. Yes. This data may be known learning signal data or unknown payload signal data. In the learning phase, using the data of a known learning signal, the corresponding complex amplitude label 62 arrives simultaneously with the complex amplitude value 61 from the phase estimation circuit 52 as shown in FIG. It is assumed that the order of the complex amplitude values 61 to be transmitted is determined in advance with the transmission side. Therefore, in the optical receiver of the present invention, a known complex amplitude pattern is stored, and in synchronization with the complex amplitude value 61, the stored complex amplitude label 62 is supplied to the learning data storage unit.

  In the conventional optical receiver, when a cycle slip occurs in the learning phase, a practical complex amplitude label determination boundary cannot be constructed. In the complex amplitude label determination circuit of the receiver of the present invention, the occurrence of cycle slip is first detected by the novel cycle slip detection / correction circuit 78 shown in FIG. When the occurrence of cycle slip is recognized, the complex amplitude value received after the occurrence is corrected according to the detected slip amount. The correction may be applied to the received complex amplitude data or may be applied to the label. Correction by the cycle slip detection / correction circuit 78 eliminates signal point shifts due to cycle slip. Based on a set of combinations of “corrected complex amplitude values” and complex amplitude labels obtained by the cycle slip detection / correction circuit 78, a complex amplitude label determination boundary is constructed by a machine learning algorithm. At this time, since the influence of the cycle slip is eliminated, the complex amplitude label determination boundary does not have the determination region jump as shown in FIG. 4B, and is shown in FIG. It is possible to form a determination region in which the shapes are uniform. More specific effects will be described later.

  In the non-tracking mode, when the “learning phase” for constructing the complex amplitude label judgment boundary in consideration of the cycle slip described above is completed, label judgment is performed on the actual data using the constructed label judgment boundary. Move to “judgment phase”. The decision phase is the normal operation of the optical receiver. The learning phase may be periodically performed at a timing at which reception of the payload signal is unnecessary, or may be performed irregularly. Normally, once an optical fiber line is installed, the nonlinear amount generated in the optical fiber is determined. Therefore, if the learning phase is executed once in the short term, it is only necessary to store the label determination boundary obtained as a result. . If an optical fiber is added or a path is switched, the nonlinear state in which the optical receiver is affected also changes. In addition, it is necessary to cope with a relatively slow change in the environmental temperature of the optical fiber installation site and a change with time. Therefore, what is necessary is just to determine suitably the timing frequency which implements a learning phase.

  The cycle slip detection / correction circuit 78 of FIG. 6 may be used in the determination phase or tracking mode described below. In the determination phase, the complex amplitude label of the payload signal is determined by comparing the reception complex amplitude of the payload signal whose complex amplitude label is unknown with the complex amplitude label determination boundary acquired in the learning phase.

  In the tracking mode, from the forward error correction circuit 55 located at the next stage of the complex amplitude label determination circuit 54 shown in FIG. A label is entered. In general, since there is no error in the bit string 59 after the error correction by the forward error correction circuit, the complex amplitude level reconverted into the label by the bit symbol conversion circuit 56 is accurate. Therefore, the reconverted label can be used as learning data for the received complex amplitude value of the payload signal. The complex amplitude label determination boundary is updated using the received complex amplitude value of the payload signal and the correct complex amplitude label after error correction. If a label decision boundary is established through the learning phase of the non-tracking mode first, based on the boundary data, the payload signal is received by using the tracking mode, and the complex amplitude considering the cycle slip is used. Continue building label judgment boundaries.

  More detailed operation will be described in the embodiment, but the present invention is modulated by a modulation method including at least a part of a plurality of signal points that are rotationally symmetric with respect to the origin on the complex amplitude coordinate. A receiving device (50, 60) that receives a light wave receives a teacher signal (61) modulated with a data string using the modulation method, and outputs a complex signal corresponding to one of the signal points on the complex amplitude coordinate. A phase estimation circuit 52 that sequentially outputs amplitudes, and sequentially receives the complex amplitudes output from the phase estimation circuit, calculates a phase of the received complex amplitudes, and generates an offset in the phase of the complex amplitudes A detection circuit (74 to 76, 63, 64) for detecting the occurrence of a cycle slip, the sequentially received complex amplitudes, and the contents of information carried by signal points corresponding to the complex amplitudes. A circuit (66) for storing a learning data set comprising a plurality of corresponding sets of corresponding labels, and based on a phase offset detected by the detection circuit in response to the detection of the occurrence of the cycle slip, A circuit 65 for correcting the stored label corresponding to the detected complex amplitude, and a plurality of signal points corresponding to the complex amplitude based on the stored learning data set including the corrected label. And a circuit (68) configured to construct a boundary for identifying different labels.

  As described above, the implementation of the complex amplitude label determination boundary considering the cycle slip in the optical receiver of the present invention can be used in either the non-tracking mode or the tracking mode non-tracking mode, or only one of the modes. It may be used in. By the operation of each mode and each phase, the complex amplitude label determination adaptive to the signal distortion caused by the transmission of the optical fiber can be stably performed even in an environment where a cycle slip occurs.

  Hereinafter, specific embodiments for each mode of the optical receiver of the present invention will be described in more detail. First, as Example 1, an operation example in the non-tracking mode will be described. Next, as Example 2, an operation example in the tracking mode will be described.

<Example in non-tracking mode>
FIG. 7 is a diagram illustrating the arrangement of 16QAM complex amplitudes used in the optical receiver according to the first embodiment of the present invention. In the 16QAM system, there are 16 types of signal points on the signal space diagram, and corresponding complex amplitude labels of 0 to 15 exist. In this signal space diagram, when a cycle slip occurs in the optical receiver, an angle that is an integral multiple of the minimum rotation angle of the signal point, that is, a phase offset of 90 degrees, 180 degrees, or 270 degrees is added. obtain. Gray coding is used as the coding method. The condition for determining the occurrence of cycle slip will be described in detail later, but it is determined that a cycle slip has occurred when the input to the phase difference monitoring unit 64 is 45 degrees or less for 5 consecutive times on the same label. It shall be.

  As described above, in the non-tracking mode, there are a “learning phase” and a “determination phase”. First, a learning phase in which a learning data set is acquired and a complex amplitude label identification boundary is constructed based on the learning data set will be described. In the optical receiver of the present invention, the number and length of learning data (teacher signals) used in the learning phase, the order of transmission, and the like are determined in advance with the transmission side. Therefore, the contents of the learning data are known in the optical receiver, and the contents are read out and used in the learning mode by storing them in the memory of the optical receiver.

  Next, the optical signal modulated with the learning data is transmitted from the transmission side to the reception side according to the determined order. As shown in FIG. 5, the optical signal that has arrived at the optical receiver is processed by the polarization compensation circuit 51 and the phase estimation circuit 52, and is input to the complex amplitude label determination circuit 54 as a complex amplitude value for learning. When the learning mode is started, as shown in FIG. 6, the label 62 corresponding to the newly arrived complex amplitude value 61 for learning is used for learning of the complex amplitude label determination circuit in accordance with the input of the complex amplitude value 61 for learning. The data is input to the data storage unit 66. In the following description, the complex amplitude value 61 that has arrived at the optical receiver of interest will be referred to as the “current complex amplitude”. As a result, a combination (learning data) of the complex amplitude value 61 to which dispersion due to nonlinearity or cycle slip of the optical fiber is added and the corresponding complex amplitude label 62 can be acquired on the receiving side. The complex amplitude label corresponding to the input complex amplitude value is accumulated in the learning data storage unit 66. However, the received complex amplitude value 61 is subjected to complex amplitude correction in consideration of a cycle slip, which will be described later, as necessary.

  Next, the learning data distribution unit 67 reads complex amplitude label learning data from the learning data storage unit 66 and outputs the learning data to different output ports according to the complex amplitude label value. As described above, the different output port here means that the subsequent arithmetic processing is performed for each complex amplitude label value, the arithmetic data is processed, and the data processing is performed for each different label. This does not mean that a specific signal or information is output to a specific physical port or output location. Although at least a part of the following arithmetic processing can be realized as hardware, it is not necessary to provide an output terminal as hardware.

  FIG. 8 is a diagram for explaining average phase calculation and grouping in the complex amplitude label determination circuit of the optical receiver of the present invention. The average phase calculation unit 74 calculates and outputs an average phase for each input port (that is, label) with respect to learning data extracted for arithmetic processing. FIG. 8A conceptually shows the operation of the average phase calculator 74, and learning data is input for each of the 16 types of labels. For each label, an average phase value is calculated and output. That is, the average phase calculation unit 74 can be expressed as having 16 input ports 801 and 16 output ports 802. The 16 input ports 801 correspond to individual phase values obtained from the received complex amplitude 61 in the learning data in each of the 16 types (number of labels) of labels. Sixteen output ports 802 correspond to the calculated average phase values for each of the 16 types (number of labels) of labels.

  FIG. 8B conceptually shows the operation of the average phase grouping unit 75, and average phase values calculated for each of the 16 types of labels are input. Then, labels that are rotationally symmetric with respect to the average phase value are grouped, and a set of average phase values corresponding to the number of groups is output. That is, the average phase grouping unit 75 can be expressed as having 16 input ports 802 and having 4 output ports 803 in this embodiment of 16QAM. Sixteen input ports 802 correspond to the calculated average phase value for each of the 16 types (number of labels) of labels. Four output ports 804 correspond to the calculated set of average phase values in each of the four types (number of groups) groups.

  The grouping of average phase values is performed as follows in the case of the 16QAM modulation system of the present embodiment shown in FIG. In the 16QAM modulation system, each signal point on the signal space diagram (IQ coordinate axis plane) is axisymmetric with 16 signal points at equal intervals on each axis, and at the same time with respect to the origin of the IQ coordinate. It is arranged to be a rotation target. In the optical receiver of the present invention, it is necessary to consider cycle slip, and when cycle slip occurs as described above, an offset of any one of 90 degrees, 180 degrees, or 270 degrees is added. The average phase grouping unit 75 defines the four signal points 701, 702, 703, and 704 with the labels 0, 3, 15, and 12 shown in FIG. Here, the signal points of labels 0, 3, 15, and 12 are group 1. Similarly, four signal points 705, 706, 707, and 708 close to the origin are also defined as one group. Circles represented by the same hatching in FIG. 7 constitute a group. That is, the signal points of labels 1, 7, 8, and 14 are group 2, the signal points of labels 2, 4, 11, and 13 are group 3, and the signal points of labels 5, 6, 9, and 10 are group 4. The average phase grouping unit 75 outputs the calculated average phase value for each of the four groups. Eventually, the average phase grouping unit 75 operates to group the labels according to the modulation method, divide the calculated average phase values into groups, and supply them to the next group selection unit 76.

  In this embodiment, an example of a 16QAM modulation system is shown, but this is applicable to all phase modulation systems that are affected by cycle slip caused by phase instability of the optical transmission / reception system. Therefore, for example, the present invention is naturally applicable to a case where eight signal points are arranged at equal intervals in a circular shape on the IQ coordinate plane as in the 8PSK method. In the case of 8PSK, the signal point arrangement has a rotational symmetry of 45 degrees, so when a cycle slip occurs, a phase offset of 45 degrees, 90 degrees,. At this time, since all signal points (labels) are in a rotationally symmetrical relationship with each other, only one group is generated. The grouping depends on the arrangement of signal points of the target modulation system and the phase offset value that can be taken by cycle slip. In other modulation schemes, the concept of average phase calculation and grouping can be applied in the construction of the complex amplitude label decision boundary considering the cycle slip of the optical receiver of the present invention. Next, the operation of the group selection unit 76 will be described.

  FIG. 9 is a diagram for explaining an operation example of group selection in the complex amplitude label determination circuit of the optical receiver of the present invention. The group selection unit 76 is a group to which the label 62 corresponding to the current complex amplitude 61 newly arriving at the optical receiver in the learning mode belongs, and to three labels that are complex amplitude labels excluding the label. Output the corresponding average phase. Referring to the signal point diagram of FIG. 9A, it is assumed that the newly arrived current complex amplitude 61 corresponds to the signal point 901 on the IQ coordinate plane, and the corresponding complex amplitude label 62 is label 0. Consider the case. A set 904 of average phase values of the number of groups from the average phase grouping unit 75 is input to the group selection unit 76.

  At this time, the group selection unit 76 selects the input 903 of the group 1 including the labels 0, 3, 12, and 15 to which the label 0 belongs, as shown in FIG. Three average phase values 905 of labels 3, 15, and 12 excluding the label 0 which is the input label 902 from 1 are output in this order. Here, the signal point of the label 3 is a signal point 90 degrees away from the signal point 901 of the current complex amplitude 61 having the label 0. The signal point of the label 15 is a signal point that is 180 degrees away from the signal point 901 of the current complex amplitude 61 having the label 0. Further, the signal point of label 12 is a signal point that is 270 degrees away from the signal point 901 of the current complex amplitude 61 having label 0. As described above, the output of the group selection unit 76 belongs to the same group as the current complex amplitude in the order of small phase difference (90 degrees, 180 degrees, 270 degrees) from the assumed signal point of the current complex amplitude. The average phase value is output for other labels. Therefore, when the position (order) of the output port 905 is also taken into consideration, the average phase value of the corresponding label is output in ascending order of the phase difference from the signal point 901 of the complex amplitude 61.

  FIG. 10 is a diagram for explaining another operation example of group selection in the complex amplitude label determination circuit of the optical receiver of the present invention. Referring to the signal point diagram of FIG. 10A, the newly arrived current complex amplitude 61 corresponds to the signal point 906 in the IQ coordinate plane, and the corresponding complex amplitude label 62 is assumed to be the label 11. Consider the case. At this time, as shown in FIG. 10B, the group selection unit 76 selects the input 909 of the group 3 including the labels 2, 4, 11, and 13 to which the label 11 belongs. 3, three average phase values 911 of labels 13, 4, 2 excluding the label 11 as the input label 908 are output in this order. Here, the signal point of the label 13 is a signal point 90 degrees away from the signal point 906 of the current complex amplitude having the label 11 by a phase difference. The signal point of label 4 is a signal point that is 180 degrees away from the signal point 906 of the current complex amplitude having label 11 by a phase difference. Further, the signal point of label 2 is a signal point that is 270 degrees away from the signal point 906 of the current complex amplitude having label 11 by a phase difference. The group selection unit 76 outputs an average phase of the number (4-1 = 3) obtained by subtracting 1 from the number of average phases (4) in one group.

  As in the group selection operation example shown in FIGS. 9 and 10, the group selection unit 76 calculates the phase difference performed by the phase difference calculation unit 63 described next on the current complex amplitude 61. In addition, the average phase to be used for the calculation of the phase difference is selected.

  FIG. 11 is a diagram for explaining an example of the calculation of the phase difference in the complex amplitude label determination circuit of the optical receiver of the present invention. The phase difference calculation unit 63 calculates the phase difference between each of the three average phases input from the group selection unit 76 and the current complex amplitude 61. In FIG. 6, it should be noted that the complex amplitude 61 from the phase estimation circuit is provided to both the learning data storage unit via the phase difference calculation unit 63 and the complex amplitude correction unit 65. Referring to the signal point diagram of FIG. 11, the current complex amplitude 61 belongs to group 2 (with labels 1, 7, 8, and 14) and corresponds to the received symbol 1004 with label 7. Is shown. In the learning mode, it is known what information is sent as the learning data signal (teacher signal). Naturally, the complex amplitude estimated by the phase estimation circuit, that is, on the signal symbol diagram of the received symbol The approximate position or phase is known. Thereby, a complex amplitude label identification boundary can be constructed from a learning data set including an actual complex amplitude value 61 and a corresponding complex amplitude label 62 affected by the phase fluctuation of the transmission environment.

The phase difference calculator 63 calculates the phase difference between the phase of the received symbol 1004 and each of the three average phases input from the group selector 76. That is, the absolute value | θr−θ ₁₄ | of the phase difference between the received symbol 1004 and the average phase 1003 of the label 14 is calculated. Similarly, the absolute value of the phase difference | θr−θ ₈ | between the received symbol 1004 and the average phase 1002 of the label 8 and the absolute value of the phase difference | θr between the received symbol 1004 and the average phase 1001 of the label 1 −θ ₁ | is calculated. Here, when the phase difference exceeds 180 degrees, an absolute value obtained by rounding the phase difference to within 180 degrees is output. For example, if the interval between the received symbol 1004 and the average phase 1001 of the label 1 is approximately 280 degrees in the clockwise direction, the phase difference is measured in the opposite direction to −80 degrees, and the absolute value thereof is further obtained. Calculated as 80 degrees. In this way, the phase difference calculation unit 63 estimates that the current complex amplitude has a correct label, and the current complex amplitude and the average phase of the label excluding the estimated label in the same group. The phase difference from the value is obtained. It will be readily appreciated that these calculated phase differences will be in the vicinity of 90 degrees, 180 degrees, 270 degrees (-90 degrees) if cycle slip has not occurred. Conversely, if the calculated phase difference deviates from the above value, it is possible to detect the possibility that a cycle slip has occurred. As described above, from the phase difference calculation unit 63, the number of phase differences (4-1 = 3) obtained by subtracting 1 from the number of average phases (4) in one group having a rotationally symmetric relationship is calculated as follows. Are output to the phase difference monitoring unit 64.

  The phase difference monitoring unit 64 determines whether or not the phase difference obtained by the phase difference calculation unit 63 satisfies a predetermined cycle slip determination condition. Specifically, it is confirmed whether each of the absolute values of the phase difference calculation values input from the phase difference calculation unit 63 is 45 degrees or less, for example. The condition for determining whether or not “45 degrees or less” can be applied when the modulation method is 16QAM of the present embodiment, and is different in the case of other modulation methods having different signal point arrangements. Note that applies. For example, in the 8PSK modulation system in which rotationally symmetric signal points are arranged at intervals of 45 degrees, it can be easily understood that the determination can be made with 22.5 degrees, which is half of that in the 16QAM modulation system, as a threshold value.

  FIG. 12 is a diagram for explaining the operation of determining the occurrence of cycle slip in the complex amplitude label determination circuit of the optical receiver of the present invention. As shown in FIG. 11, a case where the current complex amplitude has a label 7 and the estimation is made for the time being, and a case where a cycle slip has actually occurred immediately before is shown. Referring to (a) of FIG. 12, the current complex amplitude is not the position 1009 in the lower left quadrant when the cycle slip shown in FIG. 11 (a) does not occur, but the position 1005 in the upper left quadrant. It is in. It can be seen that the received symbol 1005 is closer to the average phase position of label 1 rather than the estimated position of label 7. In such a situation, an offset is added to the complex amplitude that has been estimated to have the label 7, and it can be determined that there is a high possibility that a cycle slip has occurred.

Here, in the phase difference calculation unit 63, the phase of the received symbol 1005 estimated as the label 7 and the average phase of the labels 14, 8, and 1 to which the label 7 belongs and the label 7 is excluded. The phase difference has been calculated. That is, the absolute value | θr−θ ₁₄ | of the phase difference between the received symbol 1005 and the average phase 1008 of the label 14 is calculated. Similarly, the absolute value | θr−θ ₈ | of the phase difference between the received symbol 1005 and the average phase 1007 of the label 8 and the absolute value | θr of the phase difference between the received symbol 1005 and the average phase 1006 of the label 1 −θ ₁ | has been calculated. The phase difference monitoring unit 64 is based on the phase difference between the complex amplitude (received symbol) from the phase estimation circuit and each of the three average phases (labels 14, 8, 1) input from the phase difference calculation unit 63. The occurrence of cycle slip is determined. In FIG. 12A, the phase difference between the received symbol 1005 and the average phase 1006 of the label 1 is about 20 degrees. Therefore, the condition that | θr−θ ₁ | is 45 degrees or less is satisfied, and it is recognized that the determination condition is satisfied once, and the possibility of occurrence of a cycle slip at the current complex amplitude is recognized. At this time, a phase offset is added to the received symbol due to the occurrence of a cycle slip.

  (B) in FIG. 12 shows how to determine the occurrence of cycle slip. The state in (a) in FIG. 12 shows that the complex amplitude, which was originally labeled 7, is labeled 1 because of the phase offset due to cycle slip. It can be judged that it is observed near the signal point. As will be described later, when the determination of the occurrence of cycle slip is confirmed, the received symbol is considered to be the complex amplitude of label 1 instead of label 7, and learning is performed on the received signal point (label) determined to be incorrect. A new label is given (updated) as data for use. The occurrence of cycle slip is not determined only by a single determination condition being satisfied, but can also be made conditional on the determination condition being satisfied a plurality of times. This can be implemented by providing a counter in the phase difference monitoring unit 64 as described below.

  FIG. 13 is a diagram illustrating a configuration example of a counter block used for determining occurrence of cycle slip in the phase difference monitoring unit. The counter block 1100 receives three phase differences (90 degrees, 180 degrees, and 270 degrees) from the phase difference calculation unit 63, and determination conditions 1104 to 1106 are determined for each phase difference. Furthermore, dedicated counters 1101 to 1103 are provided for each determination condition. As each determination condition, if the calculated phase difference is 45 degrees or less, the value of the counter corresponding to the input port (label) is increased by one. At this time, if the value of the other counter is other than 0, it is changed to 0. When the increased counter value becomes 5 (determination condition 1107), it is determined that a cycle slip has occurred, and the input port number to the phase difference monitoring unit 64 is notified. That is, among the phase differences calculated by the phase difference calculation unit 63, a label that repeatedly satisfies the condition for determining the occurrence of cycle slip and has a phase average value that is very close to the current complex amplitude is received complex The amplitude correction unit 65 is notified.

  Along with this notification, the values of all the counters 1101 to 1103 in FIG. 13 are reset to 0. Referring again to FIG. 12B, the current complex amplitude was estimated to have label 7, but the actually received complex amplitude is signal point 1010, which is the average phase of label 1. It is very close to the position of 1006. At this time, the reception complex amplitude correction unit 65 is notified that “the average phase of label 1 satisfies the determination condition for occurrence of cycle slip”. In other words, the current complex amplitude was estimated to have label 7, but the reception complex amplitude correction unit 65 indicates that it should be corrected to label 1 as the correct label due to the occurrence of cycle slip. Will be.

  The reception complex amplitude correction unit 65 recognizes how the cycle slip has occurred from the input port number. That is, if the port number input from the phase difference monitoring unit 64 is 1, it can be regarded as slipping 90 degrees, 180 degrees if the port number is 2, and 270 degrees if the port number is 3. it can. Accordingly, the phase difference monitoring unit 64 responds to the detection of the occurrence of the cycle slip and corresponds to the complex amplitude in which the phase offset is detected based on the phase offset detected by the detection circuit (phase difference calculation unit 63). The stored label will be corrected. The reception complex amplitude correction unit 65 corrects the complex amplitude that arrives after detection of the occurrence of the cycle slip so as to cancel the phase offset (phase shift) caused by the cycle slip.

  Various variations of the label correction method described above can be taken. For example, if a cycle slip is detected when the number of consecutive determination conditions is satisfied, (a) 5 labels (symbols) related to cycle slip detection can be excluded from the data used for boundary construction. In addition, (b) there is also an option for performing boundary construction after correcting the cycle slip of the five labels (symbols) retroactively. If the number of consecutive determination conditions for determining the occurrence of cycle slip is too large, a problem arises. For example, when the continuous number is set to a value corresponding to a long time such as 100,000 (100,000 times), cycle slip occurs a plurality of times in the time range, and the once generated phase offset returns to the original state. Situations may occur that behave as if there is no apparent cycle slip. In such a case, the label determination boundary is constructed using data including many errors, and the reception characteristics such as BER of the optical receiver deteriorate.

  Conversely, the highest performance can be achieved by setting the number of consecutive determination conditions to 1 and eliminating all suspicious complex amplitude data. However, if the distribution of the received signal points is dispersed, there is a suspicion of cycle slip and the number of data that cannot be used increases. Therefore, there may be a disadvantage that it takes time for machine learning. Practically, it is considered appropriate to set the number of continuous establishment of the determination condition (determination condition of the counter 1107) to about several.

  In the above description, the label of the learning data is corrected in response to the occurrence of the cycle slip. However, the label remains the same, and the received complex amplitude is changed according to the detected phase offset. May be changed. In short, as the learning data, if there is no error in the correspondence between the complex amplitude and its label, an accurate complex amplitude label determination boundary can be constructed in which the influence of the occurrence of cycle slip is eliminated as a result.

  In addition, the operation and function distribution of each block of the cycle slip detection / correction unit 78 in FIG. 6 described above is described as an example suitable for arithmetic processing by the DSP. It should be noted that the movement and storage of intermediate data between blocks can be performed in various other ways. Therefore, multiple signal points (labels) at rotationally symmetric positions are grouped, the average phase difference is calculated, the phase difference between the phase of the received complex amplitude and the calculated average phase is calculated, and the cycle slip It should be noted that various calculation methods can be applied as long as the occurrence detection and the phase offset estimation by cycle slip can be estimated. The processing of each block is all arithmetic processing using a processor such as a DSP, and part or all of it can be realized by hardware. Although each block is described as a “circuit”, it is not intended to be limited to an electronic circuit or an electric circuit as long as functions and processes equivalent to those described in each block can be performed. In this way, the cycle slip detection / correction circuit 78 detects and corrects the cycle slip, and the combination of the complex amplitude and the value of the complex amplitude label including the corrected one is accumulated in the learning data storage unit 66. Go.

  Therefore, in the optical receiver of the present invention, the detection circuit calculates the average phase of the plurality of stored complex amplitudes corresponding to each signal point at a rotationally symmetric position in the modulation scheme. (74), and when the signal points at the rotationally symmetric positions that can be superimposed on the complex amplitude coordinates by rotating by a predetermined rotational symmetry angle are grouped into one group, Between the phase of the current complex amplitude and the average phase corresponding to the signal point label within the group to which the current complex amplitude belongs and excluding the signal point corresponding to the current complex amplitude Based on the phase difference between the phase difference calculating circuit (63) for calculating the phase difference of the current complex amplitude and each of the signal points in the group to which the current complex amplitude belongs. A phase difference monitoring circuit (64) for determining whether or not the generation determination condition is satisfied, and specifying the label of the signal point that satisfies the cycle slip generation determination condition as the updated label of the current complex amplitude Can be implemented.

  When the number of learning data accumulated in the learning data storage unit 66 reaches 10,000, the process moves to construction of a complex amplitude label determination boundary. First, the complex amplitude label determination boundary construction unit 68 reads all the learning data from the learning data storage unit 67. Then, a complex amplitude label determination boundary is constructed by the machine learning algorithm and stored in the complex amplitude label determination boundary storage unit 69. With the above operation, the learning mode is completed, and preparation for starting transmission / reception of the payload signal is completed.

  Next, the operation of the “determination phase” in which complex amplitude label determination is performed on the payload signal will be described. Similarly to the learning phase described above, the optical signal that has arrived at the optical receiver is processed by the polarization compensation circuit 51 and the phase estimation circuit 52 in FIG. The value is input to the complex amplitude label determination circuit 54.

  However, since the received signal is a payload signal, the content of the payload signal is unknown, unlike the learning phase, and the corresponding complex amplitude label is naturally not input to the learning data storage unit 67. The received complex amplitude value is directly input to the complex amplitude label determination unit 70. The complex amplitude label determination boundary is read from the complex amplitude label determination boundary storage unit 69, and it is confirmed where the received complex amplitude value is mapped in the two-dimensional graph to which the complex amplitude label determination boundary is drawn. That is, the complex amplitude label is determined using the determination boundary of the complex amplitude label constructed in the “learning phase”. Thereafter, the determined complex amplitude label is converted into a bit sequence by the label / bit conversion unit 71 based on the gray coding method, and is input to the forward error correction circuit 72. The error-corrected bit string 77 is output last.

  Since the construction of the complex amplitude label determination boundary considering the cycle slip in the optical receiver of the present invention is implemented, a stable and highly accurate complex amplitude label identification boundary is constructed, and good communication quality can be obtained.

FIG. 14 is a diagram for explaining the communication quality obtained by the complex amplitude label determination circuit of the optical receiver of the present invention in comparison with the prior art. FIG. 14A shows a BER measurement value in the case where the label determination boundary of the 16QAM modulation method is constructed using machine learning without considering cycle slip according to the conventional technique. The horizontal axis of the graph of the BER measurement value indicates the laser line width (MHz) of a laser used as a light source in the optical receiver, and is known as a meter representing instability of the laser. That is, the larger the laser line width, the more the laser noise and the more unstable the phase. In the BER measurement, the BER was measured when a signal having a symbol time of 33 ps (symbol rate of 30 GHz) was received for 10 ⁵ symbols, and this measurement was repeated 10,000 times. The circle plot shows the average value of 10,000 times, and the length of the line shows the maximum value and the minimum value of BER. In FIG. 14A, when the laser line width is large, the BER measurement value variation is very large.

  On the other hand, FIG. 14B shows a BER measurement value in the case of using a 16QAM modulation type label determination boundary constructed using machine learning in consideration of the cycle slip of the present invention shown in FIG. At this time, the determination of the occurrence of the cycle slip is made by correcting the label when the determination condition (phase difference is 45 degrees or less) occurs five times continuously in the learning phase. The borders for label determination are each grouped into one area and are aligned rotationally symmetrically. The BER measurement value has little fluctuation even when the laser line width is large, and the influence of cycle slip is not observed. As described above, in the optical receiver of the present invention, since the label determination boundary constructed in consideration of the cycle slip by the complex amplitude label determination circuit can be used, high communication quality not affected by the cycle slip can be obtained.

<Example of operation in tracking mode>
In this embodiment, an operation example in the tracking mode in the optical receiver of the present invention will be described. The tracking mode is realized in a state where the switch is turned on in the tracking mode side in FIG. The complex amplitude label decision boundary has been established by the above-described operation in the non-tracking mode, or the information on the complex amplitude label decision boundary has been given to the optical receiver by the network administrator and stored, and transmission of the payload signal is started. It shall be.

  Also in the tracking mode, as in the non-tracking mode, the optical signal that has arrived at the optical receiver is processed by the polarization compensation circuit 51 and the phase estimation circuit 52, and the complex amplitude value is input to the complex amplitude label determination circuit 54. Since it is a payload signal, the complex amplitude label is unknown. The reception complex amplitude is input to both the complex amplitude label determination unit 70 and the learning data storage unit 66 shown in FIG.

  The complex amplitude label determination unit 70 determines the complex amplitude label as in the non-tracking mode. The determined label is converted into a bit sequence by the label / bit conversion unit 71 and input to the forward error correction unit 72.

  The forward error correction unit 72 corrects bit errors using an error correction algorithm. The obtained bit sequence is converted again into a complex amplitude label value by the bit / label conversion circuit 73. By inputting the reconverted complex amplitude label value to the learning data storage unit 67, a combination of the complex amplitude 61 of the payload signal and the correct complex amplitude label can be newly set as learning data. An error is not included in the bit string after the forward error correction through the differential encoding, and the error in the actual operation is immediately corrected even if a cycle slip occurs.

  After the payload signal is accumulated as learning data, these data are input to the complex amplitude label determination boundary construction unit 68 and used to construct a new complex amplitude label determination boundary. The new complex amplitude label determination boundary is overwritten in the complex amplitude label determination boundary storage unit 69, and the subsequent payload signals are determined as complex amplitude labels based on the new complex amplitude label determination boundary whose accuracy has been improved.

  If a complex amplitude label decision boundary with some degree of accuracy has already been established through the “learning phase”, an appropriate prime amplitude label decision boundary can be obtained by using payload signal reception, which is a normal operation of an optical receiver. Can be maintained. The interval at which the complex amplitude label determination boundary is updated may be larger than the interval at which the complex amplitude arrives.

  Therefore, the optical receiver of the present invention receives a learning data signal modulated with a predetermined data sequence as a teacher signal, and corresponds to the complex amplitude obtained from the learning data signal and the data sequence. A learning mode that operates to store a known label as the learning data set; a learning data signal modulated with an unknown payload signal as the teacher signal; and the complex amplitude obtained from the learning data signal; Determining a label of the complex amplitude, bit-converting the determined label, storing a label obtained by further bit / label-converting an error-corrected bit as the learning data set, and determining a label of the payload signal And is configured to switch between tracking modes that work to reconstruct the boundary for label identification in parallel It can be implemented as objects.

  As described above in detail, the present invention makes it possible to construct a complex amplitude identification boundary that is stable and highly accurate in a receiver that constructs a complex amplitude value identification boundary of a signal receiver of an optical network. Communication quality can be improved.

  The present invention is generally applicable to an optical communication system. In particular, the present invention can be used for an optical receiver that receives an optical signal using phase modulation.

1, 30, 701 to 708 Signal point 2, 31 Determination region 50 Optical receiver 51 Polarization compensation circuit 52 Phase estimation circuit 53 Bit stream decoding circuit 54, 60 Complex amplitude label determination circuit 55, 72 Forward error correction unit 56, 73 bits / Symbol converter 59, 77 Bit string 61 Complex amplitude value (input)
62 Complex amplitude label (input)
71 Label / bit conversion section 78 Cycle slip detection / correction section 79 Label boundary construction / label determination section 1100 Counter circuit

Claims (5)

In a receiving apparatus that receives a light wave modulated by a modulation method including at least a part of a plurality of signal points that are rotationally symmetric with respect to the origin on a complex amplitude coordinate,
A phase estimation circuit that receives a teacher signal modulated with a data string using the modulation method and sequentially outputs a complex amplitude corresponding to any of the signal points on the complex amplitude coordinate;
A detection circuit that sequentially receives the complex amplitude output from the phase estimation circuit, calculates a phase of the received complex amplitude, and detects occurrence of a cycle slip that causes an offset in the phase of the complex amplitude;
A circuit for storing a learning data set comprising a plurality of sets of corresponding labels displaying the content of information carried by the sequentially received complex amplitudes and signal points corresponding to the complex amplitudes;
A circuit for correcting the stored label corresponding to the detected complex amplitude based on the phase offset detected by the detection circuit in response to the detection of the occurrence of the cycle slip;
A circuit configured to construct a boundary for identifying different labels of a plurality of signal points corresponding to the complex amplitude based on the stored learning data set including the corrected label. A receiving device. The detection circuit includes:
An average phase calculation circuit for calculating an average phase of the plurality of stored complex amplitudes corresponding to each signal point at a rotationally symmetric position in the modulation scheme;
When the signal points at the rotationally symmetric positions that can be superimposed on the complex amplitude coordinates by rotating by a predetermined rotational symmetry angle are grouped into one group, when one or more groups are formed, Calculate the phase difference between the phase and the average phase corresponding to the signal point label of the signal point in the group to which the current complex amplitude belongs and excluding the signal point corresponding to the current complex amplitude A phase difference calculation circuit that
Based on the calculated phase difference with each signal point in the group to which the current complex amplitude belongs, it is determined whether or not a predetermined cycle slip occurrence determination condition is satisfied, and the cycle slip occurrence determination condition is satisfied The receiving apparatus according to claim 1, further comprising: a phase difference monitoring circuit that identifies a signal point label as an updated label of the current complex amplitude. The modulation scheme is 16QAM, and four groups of four symbols that are 90-degree rotationally symmetric with respect to 16 different signal points on the complex amplitude coordinates are configured.
The phase difference calculation circuit includes three signal points excluding a signal point corresponding to the current complex amplitude and a phase of the current complex amplitude and a signal point in the group to which the current complex amplitude belongs. A phase difference between the average phase corresponding to the label of
The phase difference monitoring circuit monitors whether the obtained phase difference corresponding to the labels of the three signal points is within 45 degrees as part of the cycle slip occurrence determination condition,
The detection circuit includes:
Satisfying that the obtained phase difference is within 45 degrees, in addition to the current complex amplitude, a count circuit that measures the number of times that has occurred successively with respect to the sequentially received complex amplitude, The receiving apparatus according to claim 2, further comprising: a counting circuit that determines the occurrence of the cycle slip when the measured number of times exceeds a predetermined value. A learning data signal modulated with a predetermined data sequence is received as the teacher signal, and the complex amplitude obtained from the learning data signal and a known label corresponding to the data sequence are used as the learning data set. A learning mode that works to memorize,
A learning data signal modulated with an unknown payload signal is received as the teacher signal, and the complex amplitude obtained from the learning data signal and a label of the complex amplitude are determined, and the determined label is bit-converted. Then, the error-corrected bits are further subjected to bit / label conversion labels, stored as the learning data set, and the boundary determination for label determination and label identification of the payload signal is performed in parallel. The receiving apparatus according to claim 1, wherein the receiving apparatus is configured to switch between tracking modes.   The receiving apparatus according to claim 4, wherein in the tracking mode, the reconstruction of the boundary for label identification is performed at regular intervals.