CN1939020A - Communication system - Google Patents

Abstract

The invention relates to a communication system having a central station and a plurality of substations. The central station is configured to execute a compensation process for compensating for degradation of data from said plurality of outstations, the compensation process having a plurality of adjustable characteristics controlled by a set of parameters. The central station stores a corresponding parameter set for each substation; and for each substation, the central station applies a compensation algorithm to data from the substation using the parameter set associated with that substation. During the initialization phase, the central station tests multiple sets of start-up parameters against data from a given substation and selects the set that provides the best compensation. The selected group is stored for subsequent compensation of data from that substation.

Description

Communication Systems

technical field

The present invention relates to a central station and, in particular, to a central station configured to compensate for degradation of data from a plurality of substations.

Background technique

Compensation routines for correcting signal distortions are well known. However, this known procedure is not always suitable when the source of the data arriving at the central station varies on short timescales, or when the arriving data is severely distorted.

Contents of the invention

According to an aspect of the present invention there is provided a central station for receiving data from a plurality of outstations, said central station being configured to, in use, perform degradation for data from said plurality of outstations performing a compensation process having at least one adjustable characteristic controlled by a parameter set, wherein the compensation process comprises the steps of: (i) compensating data from substations using different start-up parameter sets; (ii) determining the quality of data compensated using said different sets of activation parameters; and (iii) based on the determined quality, selecting an activation parameter set to compensate subsequent arrivals of data from said outstation.

In this way, multiple start-up parameter sets can be tested so that an appropriate initial parameter set can be selected to compensate for the incoming flow data. This will allow a station to quickly reach a state where it can use a satisfactory set of parameters for data from a given substation.

Preferably, in order to be able to perform finer adjustments to the compensation process, the compensation process has a plurality of adjustable characteristics, and the parameter groups respectively include a plurality of parameters for controlling the plurality of adjustable characteristics.

Preferably, the central station is configured to store a set of parameters for each substation. This would reduce the need for the central station to calculate or select new parameter values for a substation each time it receives data from that substation, making it easier for the central station to Distortion is compensated for, especially if data from any given outstation arrives at the central station in short bursts. The parameters may be stored locally at the central station, or may be stored by the central station at a remote location.

Preferably, the parameter sets stored for each substation are selected according to steps (i) to (iii) above. Preferably, said selection for each outstation is performed in an initialization phase when data from a newly connected outstation is received at said central station. Selected groups selected for data from a given outstation may be stored for subsequent use of data from that outstation, possibly after data from a different outstation has been received and/or compensated for. However, a stored parameter set for a given substation may be refined, for example using an adaptive compensation process, and the refined values stored for later use with respect to data from that substation.

A start-up parameter set may be selected from other start-up parameter sets by comparing the quality achieved using a different set with a predetermined quality threshold. Preferably, however, the selected set is selected by comparing the quality (ie a measure of quality) of the compensated data achieved using different starting sets, the selected set being the set achieving the highest quality.

Preferably, the data used for comparing said different sets of parameters (at least for a given outstation) is the same data, which data can be duplicated at said central station for different sets of activation parameters.

Preferably the set of start-up parameters is evaluated using test data from the outstation, preferably a copy of the test data is stored at the central station prior to its arrival from the outstation. This allows the stored test data to be compared with the compensated test data from the substation so that the quality of the compensated data can be assessed. The different sets of launch parameters may be applied in parallel to the test data, or alternatively each parameter set may be applied sequentially to the data.

The compensation process may further comprise the steps of: sampling the data stream from a substation at a plurality of time positions within the data stream; and executing a corresponding function for each sample, preferably each corresponding function is determined by a corresponding parameter or parameter group to control. The corresponding functions can each correspond to a characteristic of the compensation process. Each function can be a simple weighting function, and each parameter is a weighting coefficient. However, these parameters control other functions that apply to all sample points. For example, in case the compensation process includes a Fourier transformation step, these parameters can be used to weight the Fourier coefficients respectively.

Preferably, the plurality of substations are set to send data, so that data from different substations arrive at the central station continuously: that is, when wavelength division multiplexing is adopted, data from different substations Data does not overlap, at least data sent on the same frequency channel does not overlap. The substations may transmit bursts of data (preferably digital data) one by one in a cyclic or other sequential manner. Preferably, scheduling and other timing commands are broadcast by said central station to instruct said plurality of outstations when to transmit data.

Since the scheduling order stored at the central station provides an advance notice of when the substations will send data, it is convenient to select the application of the incoming data at the central station according to the stored scheduling instructions. The parameter set used when describing the compensation process (ie, the scheduling command may at least in part determine which parameter set or which parameter is selected).

The scheduling instruction may include an instruction for a specified substation to allow the substation to send traffic data for a specified period (either a continuous period or a segmented period). Thus, preferably, the scheduling instructions will allow substations to send a limited amount of data. For example, upstream traffic can be set up as a flow of multiple frames, each frame containing multiple cells or other subsections, and the scheduling information can indicate or allow substations in one or more specified cells of the frame Send data in. Once a substation has responded to a scheduling command, the substation preferably waits for further scheduling commands before sending further traffic data (although timing data or other management data may be retransmitted without a scheduling command). Thus, once data from a substation has been sent as a result of a first dispatch instruction, the substation preferably waits for further dispatch instructions before sending further traffic data.

In one embodiment, said central station will receive a data stream having a plurality of consecutive stream parts, each stream part having a different label associated therewith, at least some stream parts having different distortion characteristics than other stream parts, Wherein said central station is operable to access scheduling information from which corresponding labels associated with incoming flow portions can be deduced. The central station may then: infer a label for the data stream portion from the scheduling information, and select a parameter set for the stream portion associated with the inferred label based on the inferred label.

Preferably, said compensation process includes an adaptive algorithm for improving the value of said parameter upon receipt of data from a given substation. The plurality of substations can be configured to transmit data known at the central station in order to train the adaptation algorithm. When the adaptive algorithm is only partially trained on data from a given outstation, some or all of the parameters that are only partially improved may be stored and saved for later reception from the same outstation used for further data. Thus, preferably, the central station is structured such that when the source of arriving data changes from a first outstation to a second outstation, the central station: (i) stores a modified version of the data from the first outstation values for later retrieval; and (ii) retrieving parameters previously stored for the second outstation.

In order to aid in the training of the adaptive algorithm (except for selecting a set of start-up parameters), the central station may store predetermined data, and the or each substation may store corresponding predetermined data, and the or each The substations are designed to transmit their reservation data to the central station, so that the stored reservation data can be compared with the received reservation data at the central station. Preferably, in order to make it easier for the adaptive algorithm to assess how well the distorted data is corrected, at least some of the data may be the same (between the substation and the central station, different substations may have different training data). The adaptive algorithm may then compare the corrected data to the original data before distortion (ie, the data at the time of the initial transmission). Such "known" data may be included at the time of manufacture in the persistent memory chips of the central station and the plurality of outstations.

Preferably, an optical network, such as a fiber optic network, is provided to connect the central station to the substations. However, it is also possible to use wireless communication between the central station and the plurality of outstations. If an optical network is used, the network preferably has at least one branch point (also known as a "splitter") so that signals from multiple substations can be passively multiplexed onto a common carrier, These signals thus arrive at the central station as a time-division multiplexed data stream. Thus, preferably, the communications network is configured to operate in time division multiple access (TDMA) for traffic traveling from said outstations to said central station (i.e. "upstream" traffic), whereby each branch The station transmits during the transmission period, the timing of the transmission period is determined in response to one or more command instructions from the central station, and the transmission of each substation is selected to reduce the intersection of data from different substations risk of overlapping or arriving at the central station simultaneously.

Description of drawings

Other aspects of the invention are specified in the appended independent claims. The invention will now be described in more detail, by way of example only, with reference to the accompanying drawings, in which:

Figure 1 shows a communication system with a central station and a plurality of substations according to the present invention;

Figures 2a and 2b respectively illustrate the broadcast and interleaved multiplex transmission at the branch junction in the communication system of Figure 1;

Figures 3a and 3b show the downstream and upstream frame formats, respectively;

Fig. 4 schematically shows a functional representation of a central station in the communication system of Fig. 1;

Figure 5 schematically shows a functional representation of one of the substations shown in Figure 1;

Figure 6(i) to (iv) show data bursts arriving at a central station from a plurality of substations;

Figure 7 is a table showing the steps in the compensation algorithm;

Figures 8a and 8b together show different columns of a table representing various steps in an adaptive compensation algorithm;

Figures 9a and 9b together show different columns of a table representing steps in another adaptive algorithm for processing data from different outstations;

Figure 10 shows a flowchart with the main steps of Figure 9;

Figure 11 shows an alternative representation of the compensation algorithm; and

Figures 12a and 12b show other examples of data arriving from multiple outstations;

The flowchart of Figure 13 illustrates an algorithm in which different activation coefficient sets are compared in parallel;

The flowchart of Figure 14 illustrates an algorithm in which different activation coefficient sets are compared in a serial fashion; and

Figures 15(i), (ii) show compensation for data bursts when different sets of activation coefficients are tested on the data.

Detailed ways

1 shows an optical network 10, also known as a passive optical network (PON), in which a central office 12 (also known as an optical line terminal or OLT) is connected to a plurality of substations 14 ( Each substation is also called an optical network unit or ONU). The fiber optic network 16 includes a trunk fiber optic section 18 to which a plurality of branch fiber optic sections 20 are connected at junctions 21 formed by couplers or splitters. Each of the branch fiber optic sections 20 may have a corresponding substation 14 connected thereto. Additionally, some or all of these fiber sections may have corresponding other couplers 21 connected thereto for connection to a plurality of other branch sections.

The junction 21 is arranged such that the intensity of light propagating from the central station 12 to the substation 14 (referred to as the "downstream" direction) is distributed at the junction 21 among the branch or sub-branch fibers in a preferably uniform manner. In the opposite or "upstream" direction, ie towards the central office, the light from the branch fibers is passively combined at junction 21 . Typically, a passive optical network comprises a number of junctions and sub-junctions, each junction n typically having 32 "branches", ie 32 branches joined to a common backbone.

To communicate with substations, the central station sends broadcast messages in the downstream direction, usually received by all substations, with tags or identifiers indicating which of the substation(s) is the intended recipient .

Figure 2a illustrates transport in the downstream direction. In the upstream direction, successive substations transmit data in respective time slots 201, which are set so that when the light from different branch fibers is merged at junction 21, the data from different substations should not overlap. In this way, data from the substations is passively interleaved or equivalently time multiplexed at the junction into a frame structure 202, as illustrated in Figure 2b. The central station 12 accesses data from each station by reading time slots from that station (ie, using a Time Division Multiple Access (TDMA) protocol). In PON, ATM (Asynchronous Transfer Mode) cells are usually used for communication between the substations and the central station. An example of a frame structure with multiple ATM cells is shown in Figure 3a (for downstream direction) and Figure 3b (for upstream direction). ATM cells are 424 bits long and are separated by guard bands in the upstream direction to account for timing errors in transmissions from substations. In addition to the data payload, each ATM cell contains a header that includes addressing fields and other defined types of information. In the downstream direction, the frames have signaling cells spaced apart, and the ATM cells carry the traffic data that exists between these signaling cells. In the example of Fig. 3b, the signaling cells are PLOAM cells used by OLT cells to allocate (or equivalently, grant) bandwidth, ie "Physical Layer Operations, Administration and Maintenance" cells. PLOAM cells also provide synchronization and are used for ranging, error control, security, and other "maintenance" functions. The rate at which PLOAM cells are sent can be limited by the central station (OLT) and can vary according to the requirements of the OLT and outstations (ONU). Signaling in the upstream direction can be achieved using the upstream field in an ATM cell.

Central station 12 is shown in more detail in FIG. 4 . The central station 12 includes: a network input stage 40 for receiving data from substations via a common optical fiber (trunk optical fiber); a network output stage 41 for sending data to a substation via a common optical fiber or another optical fiber for transmission in the downstream direction ; The central controller 42 connected to the input stage 40 and the output stage 41 is used to control the moment when each substation sends data; the compensation module 44 is used to process the data received from the substation, so that the data may suffer Compensate or equivalently "equalize" any distortion of the substation; a data output stage 46 for outputting the compensated data from the substation to a receiver (not shown); and a data input stage for converting the data to be transmitted Input to the substation.

Central controller 42 and compensation module 44 are implemented in at least one processor facility 50 having at least one processor for processing data. Additionally, both the central controller 42 and the compensation module 44 have access to a memory facility 52 for storing data. The memory facility 52 may be distributed between the central controller and the compensation modules, which may have access to local RAM memory 521 and/or fast access memory 522 . The memory facility 52 may comprise a predetermined coefficient memory 523 . The central controller may also include a clock 43 for timing and synchronization, and a buffer 45 for short-term storage of data to be sent to the network.

The network input stage 40 has a photodetector 55, which is used to convert the received optical signal into an electrical signal, so that the compensation module can process the received optical signal in the electrical domain, preferably, the compensation module Implemented on one or more electrical chip devices. A laser 57 (or other light source) is provided at the output stage 41 for generating an optical signal to be sent to the substation, modulating the optical signal, or generating the optical signal in response to an electrical signal from the controller 42 .

Substation 14 is shown in more detail in FIG. 5 . The substation has: an input stage 60 for receiving an optical signal from the central station 12; an optical compensation stage 62 for compensating possible distortions of the signal from the central station along the fiber path; an interface stage for (i) receiving traffic from at least one user terminal for sending onto the optical network and (ii) sending the traffic carried by the optical network to the user terminal; an output stage 66 for sending the traffic onto the optical network; and a timing stage 68 for controlling the timing of traffic from local users via interface level 64. The input stage 60 comprises a photodetector 61 for converting the optical signal into a corresponding electrical signal so that the compensation stage 62 of the substation can process the signal in the electrical domain.

Timing stage 68 is connected to interface stage 64 to receive traffic intended for transmission onto network 16 . The properly timed flow is delivered as an electrical signal to an output stage having a laser device 67 (or other source) for converting the flow and other electrical signals into optical signals. The timing stage 68, which is preferably connected to the input stage via a compensation stage 62, can thus time the transmission of the signal according to instructions received from the central station. Specifically, the timing stage 68 includes a clock 69 and a buffer 70 for short-term storage of data prior to transmission. For the operation of the substation, a processor 75 having at least one processor and memory means 77 implemented on one or more chip means can be provided.

In one mode of operation, the central controller 42 of the central station is configured to broadcast a frame structure (eg, as part of a PLOAM cell) with a signaling cell containing: a frame synchronization signal; and A scheduling instruction to instruct a selected substation to transmit in one or more corresponding return cells, or equivalently, to instruct a selected substation to transmit data with a corresponding delay relative to the receipt of the frame synchronization signal instruction. Central controller 42 stores the scheduling instructions in memory 52 (preferably, a rapidly accessible portion of this memory, such as its RAM portion). This enables other components of the system, such as the compensation module, to use these scheduling instructions to infer the identity of the substation from which the data was received.

Preferably, the clock 69 at each substation is configured to be synchronized with data in the broadcast frame, such as a cell sync signal or other signal (possibly scrambled data). The timing stage 68 of the substation 14 can then use the signal from the clock to control the delay stage so that the data is delayed by a specified time relative to the receipt of the frame synchronization signal before the data is transmitted. As a result, the central controller 42 can "schedule" data transmissions from the outstations. This scheduling can be affected according to a specified period (which may be variable), especially if dynamic bandwidth allocation is employed, as will often be the case during normal operation.

As an alternative to the case of assigning a window to each outstation during the initialization phase, the controller may instruct one outstation to transmit data and instruct the remaining outstations to refrain from sending data for an unspecified amount of time until a send instruction is sent . Thus, the central station can exchange a series of messages with a given outstation without the other outstations participating.

To account for different transit times from different substations, the central station will implement a "ranging" process in which the central station instructs selected substations to send back signals for a specified amount of time, or to echo the signal immediately after receipt of the instruction (alternatively, the central station may instruct selected substations to echo the signal for a specified amount of time after receipt of the cell sync signal, which is typically equal to time interval is sent from the substation). Based on the time elapsed between the sending of the command and the arrival of the echo signal, the central controller can calculate the differential delay offset and send the offset value to the substation. The substation's timing stage 68 then stores the offset value so that it is included in any future timing instructions from the central controller. In this way, data sent by the substation will be sent with an additional delay corresponding to the offset value.

In order to maintain synchronization, the timing stage 68 of each substation can be configured to send a synchronization check signal at time intervals. Preferably, the check signals are sent from the substations at corresponding specified delays after receiving the frame synchronization signal, or on the contrary, the check signals from the substations arrive at the central station at corresponding specified positions in the return frame. The central controller 42 stores a map that associates the specified location of each inspection signal with the identifier or address of the corresponding substation. Thus, if the check signal is not received at the expected time position of the return frame, the controller may issue a command to the corresponding substation, which command includes an indication of the time difference between the moment of receipt of the check signal and the expected moment, The substation's timing stage responds to this command so that delay stage 70 changes the delay by an amount corresponding to the indicated time difference. In this way, out-off-time substations can be brought back into sync.

The compensation module 44 is configured to run a compensation algorithm for processing or equalizing the data to correct the data for any inter-symbol interference (ISI) or other distortions. The compensation algorithm has a number of tunable properties, each of which represents an aspect of the way the compensation algorithm processes data. These adjustable properties are controlled by a set of coefficients (parameters) such that each coefficient is associated with a corresponding property. The choice of values for these coefficients will depend on the degree of distortion and the nature (type) of distortion. The nature of the distortion may depend on factors such as the distance the signal has traveled along the fiber, the material properties of the fiber, the bit rate, and possibly the environmental conditions in which the fiber is exposed.

Typically, with existing PONs, the fiber distance separating the substations from the central station is about 20 km, but fiber lengths of at least 50 km, 100 km, 150 km or even 200 km are desirable. The distance between the central station and the substations usually varies according to the substation, with a difference of 20km, 50km or even 100km. In addition to facilitating the compensation of abrupt distortions (due to different substation distances) and increased distortions due to longer absolute distances, the invention also facilitates the use (at the corresponding substation) of Broader bandwidth light source (broadband sources such as Fabry-Perot lasers are cheaper than narrow bandwidth sources such as DFB lasers, but result in higher distortion).

In particular, it should be understood that, in the context of the present invention, the nature and degree of distortion may be different for data from different outstations. Furthermore, while the distortion of incoming data at the central station may vary from cell to cell (i.e., as fast as about every 424 bits), for data from a given outstation the distortion, if any, may be only Change slowly.

The compensation module stores the corresponding set of coefficients for each substation in the form of a table which contains a set of identifiers in one row, one for each substation, and the corresponding set of coefficients in another row, This table contains mapping information that maps each substation identifier to a set of coefficients. From the stored scheduling information, the compensation module deduces the identity of the substation from which each arriving data cell originates. Each time it is concluded that data will arrive from a different substation, the compensation module retrieves the set of coefficients stored for that substation and runs the algorithm on the data from the current substation based on the newly retrieved coefficients . Thus, as a result of the scheduling process, the compensation module knows in advance from which substation the next cell to be received will originate. Therefore, the compensation module can quickly switch between appropriate coefficients to compensate for the distortion of the next arriving cell. As a result, adjacent cells will have very different amounts of distortion, because the compensation algorithm does not have to optimize the process from scratch for each cell, but can quickly call back the appropriate set of coefficients, thus maintaining the appropriate values in the compensation algorithm.

Preferably, the compensation algorithm is adaptive, ie preferably, the compensation algorithm has an optimization procedure that improves the values of a set of components starting from a set of initial values. In this case, when the source of arrival data changes from a first station to a second station, the compensation module is configured to: (i) store the improvement calculated for the data from the first substation for later retrieval ; and (ii) retrieving or recalling previously calculated coefficients for the second outstation. Preferably, the trigger to perform steps (i) and (ii) is provided by dispatch instructions from the controller which will indicate in advance the identity of the substation from which future arriving data will originate. Thus, the central controller can provide a signal whenever the source of received data is to be changed from one substation to another. Alternatively, the compensation module may only store (in local random access memory 521 ) or otherwise access scheduling information (generated by the central controller) representing the intended source of received data.

The operation of the compensation module can be elucidated with reference to Figures 6(i) to (iv), each of which shows how the signals from substations A, B, C and D are distorted Varies over time. Here, time increases from left to right, and larger width segments represent larger distortions. Referring to FIG. 6(i) at t=0, the initial coefficients stored for substation A are retrieved, and the data received from substation A is processed. Initially, the data had a high distortion level, but over time, the optimization procedure of the algorithm improved the coefficients, making the compensation algorithm more effective in reducing the distortion and making the signal from substation A less distorted. Small. At t=2, the data from A is terminated, and the improvement coefficient for substation A is stored by mapping to that substation. The initial coefficients stored for substation B are then recalled and the compensation algorithm is run on the data from substation B to reduce distortion. Substation B is closer to the central station (either with a narrower bandwidth light source or equivalently a "higher standard" light source) than the other substations, so that the data from B has a lower distortion level. At t=3, the data from substation B is terminated, the improved coefficient calculated for this substation is stored, and the new coefficient for substation C is called back. Similar steps are performed when data from substation C is terminated at t=5 and data from substation D is received instead. At t=7, the previous improvement coefficient stored at t=2 is retrieved for outstation A. Since the previously improved coefficients are used, the processed data can be processed more efficiently, resulting in a lower distortion level than would have been obtained if the original coefficients had been used. In Figure 6(i), this is represented by the narrow line width of the data from substation A at t=7. In FIG. 6(i), each ONU is continuously trained before performing the standard operation of the normal scheduling process. This training or optimization process of the compensation algorithm can be implemented as part of the PON start-up phase together with the ranging process, or as an independent compensation process or "equalization" start-up phase after PON completion. Once the compensation algorithm has "locked in" and determined the correlation coefficients to be within specified tolerances for an outstation, these coefficients are stored in memory 52 (or local RAM 521 ) for use in standard operation. For simplicity, a general sequential schedule is illustrated. However, the PON scheduling process may be more complicated in response to requests from substations.

As can be seen from Fig. 6, the adaptive compensation can be distributed (distribute), so that the coefficients of a substation are improved for the first time, and when data from the substation is received next time, the improved coefficients are retained and A second, further improvement is made wherein data from another outstation is received in the intermediate period between the first and second improvements.

Figure 6(iv) illustrates an alternative "cold" initialization process, where the initial optimization process for each of the multiple outstations is split among multiple cells, and again local RAM is used to store the relevant coefficient. In a similar manner to the process shown in Fig. 6(i), the process of Fig. 6(iv) includes the following steps: (a) retrieve the coefficients of substation A; (b) receive data from substation A; (c) use The data received by substation A improves the coefficients of substation A, and stores the improved coefficients; (d) repeats steps (a) to (c) for substations B, C, and D; and (e) retrieves Modified coefficient of substation A. However, in FIG. 6(iv), since the compensation algorithm has stored the improved coefficients for substation A, these coefficients can be retrieved such that when additional data from substation A is received at t=4, the The compensation algorithm (specifically, its optimizer) can continue training and further improve the values of the coefficients. In this way, the compensation algorithm can continue from the point where it previously stopped at t=1.

After the training or "initialization" period illustrated in Figures 6(i) and 6(iv), the compensation module 44 (specifically, the compensation algorithm) recalls the set of correlation coefficients for the ONU whose signal is being processed at the time. When a new ONU is connected to the PON, the new ONU can be initialized in the manner of a single burst as shown in FIG. 6(ii). However, in order to minimize the impact on other users, it is preferable to perform this initialization on a plurality of cells as exemplified in FIG. 6(iii). Among them, the scheduling algorithm controls TDMA and distributes compensation optimization on multiple cells. However, storage of the weighting coefficients enables the compensation algorithm to restart using its previously saved coefficients, thus allowing it to continue converging to the optimal response instead of starting from scratch again (sending multiple cells together in the same ranging window is It is possible and more efficient, but the same basic principles apply, see Figure 12a and Figure 12b, Figure 12a shows the initialization phase of sending multiple cells in the ranging window, Figure 12b shows the Example of a case where a new substation is added).

It can be understood that ranging and scheduling can be achieved with a bit error rate (BER) as low as 10-4, so that scheduling can be performed before compensation is completed, so that scheduling can be applied during the operation of the compensation algorithm.

In order that the optimization program can more easily refine the values of the coefficients (and optionally activate coefficient sets, see below), preferably the substations send known data to the central station. For example, the known data (such as a pseudo-random bit sequence) is introduced into the memory of the corresponding substation in advance during the manufacturing stage, so that the optimization program can perform feedback or recursive processing, thereby comparing the processed data with the known data, A convergence value is generated based on the difference between the processed data and the known data. Once the convergence value reaches a threshold value, the coefficients are considered suitable, and the central station retains the values of the coefficients calculated when the convergence value reached the threshold value. When this occurs, the central station is ready to receive traffic data (ie, receive unknown data from the substations).

ISI distortion on a link may vary over time. (Examples that could cause this change include: transmitter degradation in the ONU, pattern noise, PMD fluctuations, adding new fibers, and using a different transmitter). Therefore, it is expected that the PON can periodically (or continuously) re-optimize the weighting coefficients of each ONU. This can be achieved by periodically sending special training (non-traffic forwarding) cells in a manner similar to that of the Physical Layer Operations, Administration, and Maintenance (PLOAM) cells used therein in ATMPON (BPON). Alternatively, these coefficients can be fine-tuned using the traffic forwarding cells themselves.

In order to provide starting values for the coefficients, the central station stores a number of possible starting (equalizer) coefficient sets S1, S2, . . . , sj. They can be pre-computed so that each set is suitable for different amounts or different types of signal degradation (ISI), eg due to different fiber lengths. Thus, rather than "training" the algorithm's coefficient sets according to their default values, the central station applies each of these different sets (sequentially or in parallel) to incoming data during the non-traffic-carrying initiation phase. This incoming data sequence will be known (for example this data could be stored at substations and central stations during the manufacturing phase). The central station then simply selects the coefficient set that provides the best performance or the smallest error (ie, the highest quality signal) by comparing the known data with the incoming data. In this way, the start-up coefficient sets are individually tested and the most suitable (preferable) coefficient set is selected. The central station can use this selected set of coefficients as the set of coefficients for fixed (non-adaptive) equalization, or as the initial set of coefficients to be improved when using an adaptive algorithm.

The flowchart of Figure 13 illustrates the start-up phase. Wherein, at step 0, the known data is received from the new substation. This data is copied, and each copy of this known data is compensated by a compensation procedure (such as a horizontal tap described below). Each copy of the known data is processed in parallel using a different coefficient set (or vector). At step 2, the signal quality of the data processed using different sets of coefficients is evaluated. For example, this evaluation may be done by determining the correlation between high and low values of received and previously "known" data (if the data is in numerical form). At step 3, the set of coefficients providing the greatest signal quality is selected and stored, which is applied immediately to subsequent arriving data if the equalization (compensation) process is adaptive. Otherwise, if the equalization is adaptive, the set of coefficients is refined or "tuned" at steps 4 and 5 before being stored for later use at step 7 .

The flowchart of Figure 14 illustrates an algorithm in which different sets of coefficients are compared in a serial fashion. Wherein, at step 1, each activation coefficient group (vector) is applied to each of multiple copies of the known arrival data. The signal quality of the thus compensated data is assessed and once all sets of coefficients have been tested, the coefficients giving the highest signal quality are retrieved in a manner similar to Figure 13 (which shows selection of coefficients by parallel processing) group for later use.

Figures 15(i) and (ii) illustrate the evolution of the amount of distortion over time in the case where an activation coefficient set is selected from one of a number of possible sets. (In stage (a), the height of the bar represents the amount of residual distortion after compensation using different activation groups, while in stage (b) the height of the bar represents the situation where the improvement was made using the adaptive algorithm The amount of residual distortion after compensation for the selected coefficient set below). For simplicity, a serial implementation example is shown in Fig. 15(i), showing how each of the different sets of coefficients is used in turn, and then the best set is selected, which is then further optimized using adaptive techniques.

Figure 15(ii) illustrates the case of a PON in which multiple ONUs share bandwidth, showing during an initial "cold" start-up in which ONUs are trained serially before performing scheduled standard operations. (This training process can be implemented as part of the PON start-up phase together with ranging processing etc., or as a separate "equalization" start-up phase after the PON start-up phase is completed). Once the equalization algorithm "locks" and determines the ONU's correlation coefficients, these coefficients are stored in local RAM for use in standard operation. (For simplicity, a normal sequential scheduling process is illustrated. The PON scheduling algorithm can be smarter in response to requests from ONUs). The diagram given in Fig. 15 shows serial attempts by the equalizer using different sets of coefficients. If multithreading is available, multiple sets of coefficients can be tested simultaneously, reducing the time required to determine the optimal set.

This "improved initialization procedure" using multiple priming sets is a powerful enhancement because it overcomes concerns about the length of the training sequences—long sequences are no longer needed because the device only needs to select the best set of coefficients instead of doing Fully optimized. With a reasonable number of defined coefficient sets, potential engineering constraints that may be required if the network must deliver a certain minimum level of performance are also overcome.

Another advantage of testing multiple sets of activation coefficients is that the testing can be performed even when the distortion is too great for ranging and/or scheduling. In particular, the starting group can be tested without knowing the identity of the originating outstation. This can occur if the initial level of distortion (eg bit error rate) is too high for the central station to correctly read the arriving data. Once a suitable set of coefficients has been found, this set can be used to read subsequent arrival data so that the central station can obtain the identity if it contains the identity.

One example of a compensation algorithm is known as transversal filter processing. In general, the operation of the process involves sampling at multiple points (ie, temporal positions or "taps") along the incoming data stream. Taps may be spaced apart in the incoming data stream at intervals smaller than one bit (eg, each half or quarter of a bit period). Furthermore, the sampling interval need not be regular, but could be irregular, eg to take into account the complexity of the arriving data. Preferably, taps are located at consecutive or "adjacent" bit positions. Thus, for each target bit to be compensated, weighted samples of the adjacent bits of the target bit can be mixed with the target bit. Specifically, the weighted adjacent bits are usually added to the target bit, or subtracted from the target bit. (Due to distortion or overlap between bits, the value of a bit is no longer 0 or 1, but can be a continuous range of values).

Preferably, adjacent bits will include at least one immediate neighbor of the target bit, e.g. for a 3-tap filter process, the target bit and its two (trailing in this example) neighbors in the data stream sampling. For a 5-bit tap, the sampled neighbors will include the next 4 nearest trailing neighbors of the target bit, and so on. To weight the sampled bits, a weighting function is applied to each corresponding bit, the weighting function for each bit being controlled by a respective one of the coefficients in the group corresponding to the substation whose data is to be corrected. Preferably, the weighting function is a simple factorization function (eg multiplication).

In order for the filter process to work, the compensation module stores a given number of successively arriving bits in corresponding memory locations in the memory 52 , for example in a shift register 522 within the memory 52 . In essence, in operation, (i) weight each successive data bit by the coefficient associated with the memory location in which the data bit is located; (ii) save the weighted value; and (iii) then perform a Shifting such that data in one memory location is replaced by data in the immediately following bit slot (ie, the slot following the slot that has arrived). Repeating steps (i), (ii) and (iii) in sequence results in a value being produced for each cycle which is the combination of the sampled data at each memory location. Specifically, for each target data bit, the combination includes the target data bit and weighted bits that follow the target bit in specified neighbors.

More specifically, the data received at the central station can be represented by R(x) and take the values R1, R2, R3, ..., Rx-1, Rx, Rx+1, etc. at the various bit positions , Rx is the target bit to be compensated. Wherein, the mark "x" indicates the position of the bit in the stream of bit slots. The "compensated" data/"output" E(x) is given by the sum of the sampled bits weighted by a coefficient, i.e., by:

$\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; R</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}$

Thus, for a "3-tap" filter (where n=3), the compensated data is given by:

E(x)=c ₀ R _x +c ₁ R _x-1 +c ₂ R _x-2 It is also useful to use the vector notation E=CR, where C contains the coefficients (c ₀ , c ₁ , c ₂ ,  … , c _n-1 ), and R is the sampled data $\left(\begin{array}{c}{R}_x\\ R_{x-1}\\ R_{x-2}\\ ...\\ R_{x-\mathrm{no}}\end{array}\right).$

Figure 7 shows a table summarizing the main steps of the compensation algorithm. Coefficients c0, c1 and c2 (which are static in this example) correspond to memory locations M1, M2 and M3, respectively. The rows labeled M1, M2 and M3 show for each step which data bits R1-Rx are loaded into the corresponding memory location M1-M3. The last (right-hand side) column shows the compensation values obtained for the equalized data: eg the equalized data for the R3 bit slot position is given at stage 3c.

The following gives a description of a possible compensation algorithm that is adaptive such that the algorithm can perform an optimization or training procedure to adjust the values of the coefficients used to compensate for the distortion of the signal from a given substation Make improvements. The "compensated" data/"output" E(x) is still described by the formula given above, although the coefficients are now allowed to vary and adjusted to provide an improved filter to "filter out" or reduce distortion. Modifying the formula to reflect the time-dependent nature of the coefficients gives:

$\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; R</span>}}_{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>}$

Thus, for a "3-tap" filter (where n=3), the compensated data is given by:

E(x)＝c ₀ (x).R _x +c ₁ (x).R _x-1 +c ₂ (x).R _x-2

Initially, all memory locations are set to zero. Set the coefficients to their initial values (this is usually their "expected"/average value to help reduce convergence time). Adjusting these coefficients will involve an algorithm that responds to some indication of signal quality. If some data is known, the algorithm may be an error function that measures only the difference between the "compensated" data and the known data. In this case, the optimization seeks to reduce the error. Here, a known training sequence is used so that the error can be measured directly. (Optimization for "blind" can however be attempted by using another signal quality metric such as "eye opening").

Suppose the known, sent data is given by K(x). It can be written as K1, K2, K3, ..., Kx-1, Kx, Kx+1...etc. The received data is still given by R(x) and takes values R1, R2, R3, . . . , Rx-1, Rx, Rx+1 . . . and so on. The compensated data E(x) takes values E1, E2, E3, . . . , Ex-1, Ex, Ex+1. The error e(x) between the compensated data and the known data is simply K(x)-E(x). The adaptive algorithm will adjust the coefficients in such a way as to minimize this error.

Various optimization techniques can be used: one example is to reduce the (mean squared) error by making small changes to the coefficients and choosing increments that reduce the (mean squared) error (if the form of the function is known, one can use The derivatives of this function make more informed estimates to revise these coefficients). Using vector notation, if E=C.R, then the revised estimate can be written as E'=(C+δC).R such that e'(=K-E')<e(=K-E). The new coefficients are given by C'=C+δC.

As a summary of the described processing (using known data), the following steps are performed (steps VI and VII can be omitted if the data are incomplete in the first few cycles):

I. Initially, all data memory locations are set to zero, and the coefficients take initial values

II. Shift the data

III. Sample the signal at location x, save it to the first data location

IV. Computing Compensated Data E

V. Call back the known data and calculate the error e

VI. Calculate the "gradient", that is, determine the change in the coefficients to reduce or minimize the error

VII. Adjust coefficients to new values as determined in VI

VIII. Go to Step II

The initial value can be calculated as a function of the offset delay value already calculated by the central controller for each substation, or another delay-related parameter representing the transit time from the substation to the central station can be used, since the transit time Usually associated with the fiber path length, which in turn represents the possible degree of distortion. Figures 8a and 8b show tables representing the steps involved in the adaptive compensation algorithm. Figures 8a and 8b show the same rows but different columns of the table.

The following is a description of an algorithm that explicitly allows data from different substations to require different filter coefficients c1, c2, c3, etc. The "compensated" data/"output" E(x) is described by the formula for the adaptation example given above, although now there is a separate set of coefficients for each ONU. Filters function in the same way, but use knowledge of the order of transmission from the scheduling algorithm to jump between data groups of different ONUs. Written as a step-by-step process:

I. Initially, all data memory locations are set to zero, and the coefficients take their initial values for normal operation

II. New Cell One uses knowledge from the scheduling algorithm to determine which ONU is transmitting

III. Retrieve the coefficients of the ONU

Equalize the data

IV. Shift data

V. Sample the signal, save it to the first data location

VI. Calculation of Compensated Data E

If the data is known, optimize the coefficients (otherwise go to step X)

VII. Call back known data and calculate error e

VIII. Computing "gradients", i.e. determining changes in coefficients to minimize error

IX. Adjust the coefficients to new values as determined in VIII

end of loop

X. If end of cell is reached, store coefficients and go to step II, otherwise go to step IV

Figures 9a, 9b and 9c illustrate the table when the data are known and the coefficients are adjusted (9a, 9b and 9c show the same rows but different columns of the table). If the data are not known, skip steps VII to IX and keep the coefficients fixed.

The modeling process has shown the feasibility of PON initialization at the discussed BER of 10 ^-3 to 10 ^-4 . Table 1 below shows a good probability of receiving data for priming (booting, ranging, etc.) at this BER, although the performance is not acceptable for customer traffic.

Log(BER) The probability of getting n consecutive error-free bytes n=1 n=2 n=3 n=4 n=5 n=10 n=20 n=30 n=40 n=43 1 0.4305 0.1853 0.0798 0.0343 0.0148 0.0002 5E-08 1E-11 2E-15 2E-16 2 0.9227 0.8515 0.7857 0.725 0.669 0.4475 0.2003 0.0896 0.0401 0.0315 3 0.992 0.9841 0.9763 0.9685 0.9608 0.9231 0.8521 0.7865 0.726 0.7088 4 0.9992 0.9984 0.9976 0.9968 0.996 0.992 0.9841 0.9763 0.9685 0.9662 5 0.9999 0.9998 0.9998 0.9997 0.9996 0.9992 0.9984 0.9976 0.9968 0.9966 6 1 1 1 1 1 0.9999 0.9998 0.9998 0.9997 0.9997

Table 1

The above table shows that there is a good probability of receiving data from the limited number of bytes required to communicate during the initial activation phase before equalization is performed. If a failure occurs, the ONU can simply resend. The following table (Table 2) shows the increasing probability of success after 5 attempts.

Log(BER) Probability of getting n consecutive error-free bytes after 5 attempts n=1 n=2 n=3 n=4 n=5 n=10 n=20 n=30 n=40 n=43 1 0.9401 0.6411 0.3401 0.1603 0.0718 0.0011 2E-07 5E-11 1E-14 0 2 1 0.9999 0.9995 0.9984 0.996 0.9485 0.6729 0.3747 0.1851 0.1479 3 1 1 1 1 1 1 0.9999 0.9996 0.9985 0.9979 4 1 1 1 1 1 1 1 1 1 1

Table 2

Figure 11 illustrates the main steps in the adaptive algorithm of Figures 9a, 9b and 9c. Figure 11 shows an alternative representation of the adaptive compensation algorithm. Therein, initially, the data from the input stage 40 of the central station is "sampled" to extract bit slot values and weighted using the coefficient c0. A function with a one bit delay is applied by the transform stage 901 (here the transform Z ⁻¹ ) before the next bit slot is sampled and weighted by the coefficient c1 at the weighting stage 902 . Another transform stage with a further delay of one bit delay is applied before sampling the data, which is then weighted by another coefficient, and so on for a further one bit delay. The weighted samples are then summed at summation unit 903 . Here, the transformation stage 901 , the weighting stage 902 and the summation unit 903 are software units implemented in a processor and a memory (eg located in the processor means 50 and the memory means 52 of the central station). The set of coefficients can then be changed from one ONU to the next.

Other families of equalization (compensation) algorithms can be used, such as: linear equalization; decision feedback equalization (and feed-forward); nonlinear equalization; maximum likelihood sequence detection (MLSD), also known as maximum likelihood detection ( MLD); a possible "blind" technique; and the constant modulus algorithm (CMA).

Embodiments of the present invention can be better understood by way of example with reference to the following steps (where the central station is referred to as an OLT and the outstations are referred to as ONUs).

1. The OLT broadcasts to all ONUs. After the ONU is powered on, the ONU listens to the downstream traffic from the OLT, and the ONU can obtain bit and frame synchronization according to this operation (it can obtain bit and frame synchronization according to the framed data)

2. The OLT broadcasts to all ONUs. The new ONU(s) listen and receive all broadcast cells from the OLT. The ONU extracts network parameters (initial transmit power level and ranging related delays, format of frames, preambles to be used upstream, etc., details of when to send registration and ranging requests, etc.). The OLT periodically broadcasts these parameters downstream in defined bytes, so that all the ONU has to do is listen long enough to extract the relevant information. For example, the OLT may send bytes in the defined bytes in the header to indicate that a ranging window occurs between bytes XXX to YYY within the frame. (Alternatively, these bytes may represent a TTT period after the end of frame ZZZ, or a TTT period after the OLT has sent a given sequence of bytes).

3. It will be appreciated that at any single ONU, for its TDMA time slot, the signal from the OLT may have experienced the same distortion (ISI). Each frame or cell has a known sequence of bytes that can be used as input data to the compensation algorithm. Alternatively, cells dedicated to known data for compensation may be sent at regular intervals (eg, every 128 frames). Either way, the ONU(s) can use this incoming data to train its/their compensation algorithms and optimize their coefficients. The head end includes a higher standard laser, so it does not have to use ISI compensation in the downstream direction. However, the option to use ISI compensation can also be retained if desired. This step also facilitates the execution of step 7, which helps to speed up the convergence of the compensation algorithm for the upstream direction.

4. The ONU adjusts its settings to those received from the OLT, and [after synchronizing itself with the OLT in (1)], determines an available window (some Ranging and registration are not required by the PON protocol, but are more general requirements).

5. In the registration window, the ONU will send a request to the OLT to register using one or more of the predetermined set of coefficients. Different approaches are possible: (A) The ONU applies the sets of coefficients to the data simultaneously and in parallel. Some results will be completely corrupted by ISI, but some results will give good data. The ONU selects the group that gives the best performance. The selected coefficients can then be accepted and fixed, or used as initial input parameters for further optimization by an adaptive equalization algorithm; (B) the ONU can apply each coefficient group sequentially in the same start-up/ranging cycle on data. Also, some results may be corrupted, but the ONU only selects the group that gives the best performance. These coefficients can then be accepted and fixed, or used as initial input parameters for further optimization by an adaptive equalization algorithm; (C) if the ranging window is too short, the ONU can alternatively use Send multiple requests on . The OLT will apply different sets of coefficients in each registration window until it finds a set that gives sufficient performance to initiate communication between the OLT and the ONU. The OLT can either wait until it has used each set of coefficients to select the best set of coefficients, or after communication has been established, can apply a more "smart" approach to determine the best set to use.

As an optional improvement, if the ONU determines a measure of ISI on the link or coefficients required for an equalization algorithm when listening to downstream data from the OLT in steps 1 to 3, this information can be uploaded to The tour is sent to the OLT. Once one of the above methods has found a sufficient set of coefficients to extract the data, this information can be used to select the best set of coefficients based on the coefficients predicted by the ONU.

After selecting the appropriate coefficients for the equalization algorithm, the OLT is able to extract information from the incoming data. This data contains requests from ONUs. The OLT processes the request (using appropriate security measures if necessary) and broadcasts an appropriate response. The ONU(s) listen for the response in the appropriate downstream bytes. If the registration fails, the ONU can repeat the attempt at a later registration window.

6. Ranging. The ranging window is established such that other ONUs do not transmit and therefore do not interfere with each other and are not interrupted by ranging/activation. In some protocols such as GPON, a working ONU is stopped during the ranging process performed on an ONU-by-ONU basis. For alternative protocols, it may not be necessary to stop a working ONU. During the "ranging" window, the ONU sends messages at known times relative to the start of the ranging window. The OLT receives this message, and determines its time of arrival, from which the round-trip delay, and thus the transmission Delay. This transmission delay is then used to determine how much additional delay needs to be added at the ONU to ensure its upstream data is in sync. This information is then sent downstream to the ONU. (For other downstream data, it is broadcast along with an appropriate identifier enabling the ONU to determine its own data).

7. Optimization of the equalization coefficient. Assume that in step 5 a good starting estimate for the compensation (equalization) coefficients was determined. In simple implementations this may be sufficient so that no further adjustments to these coefficients are required, in which case the next step can be performed. However, further optimization can also be performed on the equalization coefficients if desired. There are also many options.

One method is to broadcast a message to stop all other ONUs and enable new ONUs to send a sufficiently long training sequence of known bytes to the OLT. Once the OLT optimizes the compensation coefficients for this new ONU, or the maximum time period is reached, these compensation coefficients are stored for this ONU. The OLT may then broadcast another downstream message indicating the end of backoff, enabling registration, ranging or backoff of another ONU, or enabling a return to normal operation.

An alternative is to define a fixed-length training sequence that can be sent as part of the ranging message or in a separate "backoff" window.

Another approach is to use extended messages in the ranging window, as this avoids making the protocol more complex due to additional windows. This is possible because, relatively speaking, there is a lot of time available in the ranging window. The window must be correctly sized to handle these longer messages, but the overhead required should be minimal. If desired, collision detection can be used to help handle situations where different ONUs use the same ranging window: i.e., collision detection multiple access (CDMA) can be used to ensure that only one ONU at a time tries to equalize. In the event of a collision, the ONU backs off a random amount and monitors the ranging window so that only one ONU attempts ranging and equalization at a time. Since this ONU is registered and ranging at this point, it can be controlled by the PON scheduling algorithm, so it can now be scheduled, so that compensation constants can be saved for this ONU, and the compensation can be optimized if required Scale to multiple odometry/optimization windows.

After a set number of optimization sequences, the initial optimization is considered complete once the error falls below a certain magnitude, or once the increment of the coefficients or the change in the error falls below a certain magnitude. Further optimization can be achieved during normal operation, see step 9 below.

8. Once the OLT has optimized the compensation of the ONU and completed the registration and ranging, it can broadcast an update to other ONUs indicating that it can register/rang and equalize another ONU. (This step is optional, especially when the training sequence has a fixed length and is located at a defined position relative to the downstream frame). If there are no "collisions" (where more than one ONU is transmitting at the same time), more than one optimization can be processed at a time.

9. Normal operation. Ranging the network such that all upstream messages are synchronized together so that they each arrive at their correct position within the frame. The ONU is also compensated. For subsequent operations, various options exist. After selecting the best set of predetermined coefficients, and (as above) also optionally further optimizing them at the initiation phase, there are a number of options for "normal operation":

A) The coefficients can be kept fixed for the ONU so that the OLT uses the same set of coefficients whenever the ONU transmits.

B) The OLT may occasionally attempt to evaluate the predetermined coefficients it uses by occasionally comparing signal quality from multiple predetermined groups (all of them, or only those predetermined groups that are closest to the current predetermined group in terms of ISI compensation) The group is re-optimized.

C) The OLT may occasionally try to re-optimize the coefficients themselves as in previous implementations.

For B) and C), the OLT can use known bytes in the header of standard data cells, PLOAM cells, or other dedicated known "training" cells for this optimization. The re-optimization in B) and C) can be performed periodically, eg every x minutes, daily, etc., or it can be a process that is prompted when the measure of signal quality drops below a certain threshold. (The threshold can be an absolute value, or a value relative to the initial performance after optimization).

The current generation of PONs typically operates at 622Mbit/s, typically has 32 branches (ie 32 ONUs per OLT) and a maximum range of 20km. However, future generations of PONs may need to operate at higher bit rates and over longer distances to bypass outer core transmission equipment. This may involve (but is not limited to) a bit rate of 10 Gbit/s over a fiber distance of ~100 km. Even in current systems, distortion or other intersymbol interference (ISI) (which may be due to factors such as chromatic dispersion (CD) or polarization mode dispersion (PMD) along the fiber) is considered undesirable, and This distortion or ISI may be less desirable at higher bit rates or for systems where substations have wider linewidth light sources (ie "lower gauge" light sources).

It can be seen that at least some of the above embodiments make use of the knowledge of the PON multiple access protocol for the sequential transmission of OLTs of TDMA time slots. As part of the scheduling process, the OLT knows in advance from which ONU the next cell to be received originates. The equalizer (compensation process) can use this knowledge to quickly switch between suitable weighting coefficients to the weighting coefficients of the next ONU to compensate for the ISI of the next cell. This means that adjacent cells may have very different amounts of ISI, since the algorithm does not have to optimize the process from scratch for each cell, but can quickly recall the last set of coefficients and thus maintain a suitable value in the equalization algorithm value. Thus, the OLT will have multiple (e.g. 10) fixed sets of equalization coefficients representing different amounts of ISI, rather than having to fully "train" the adaptive equalization algorithm to optimize its coefficients during the initiation phase . (This can eg correspond to different fiber lengths). During the initiation phase, the device simply applies each of these different parameter sets (sequentially or in parallel) to the incoming data, rather than "training" the coefficients of the algorithm starting from their default values. Since the data sequence is known, then the device need only select the set of coefficients that give the best performance/minimum error. It can then use this fixed set of coefficients as the set of coefficients for fixed (non-adaptive) equalization, or as the initial set of coefficients to be improved when using an adaptive algorithm.

A useful consequence of employing electronic equalization is that the parameters or coefficients used by the equalization algorithm exist in memory and can be easily copied to the output where they can provide input to signals from a variety of ONUs Quality monitoring function. This would be a useful tool for network operators in helping to provide monitoring of the PON. This data can be fed into a secondary application that understands the PON topology (which can correlate data from different ONUs) to help locate faults and degradations in the PON.

One or more of the above-described embodiments would be convenient for existing PONs (e.g., PONs operating (at least prior to the present invention) according to standards such as the ITU G983 standard, and/or PONs operating at high bit rates) to operate. This is achieved by allowing the compensation procedure to be applied to data where the distortion characteristics vary over short timescales (eg due to the path of the data changing over such short timescales). Current generation PONs tend to utilize high quality optical components such as external modulation and distributed feedback ("DFB") lasers to help overcome the effects of signal degradation due to ISI. As mentioned above, it is desirable to use cheaper components such as Fabry-Perot lasers, but their relatively poor performance often hinders their application. The above-described embodiments enable the use of lower specification components, resulting in considerable cost savings for operators seeking to deploy access networks. The potential savings are significant and could transform the economics of PON deployment, making it a viable option for access networks. This advantage may have larger potential applications than the high bit rate, long distance applications discussed.

Further considerations include:

• Preferably, the receiver and equalization algorithm will be able to handle burst mode traffic.

· Other equalization (compensation) algorithm families can be used, for example: linear equalization;

Decision feedback equalization (and feedforward); nonlinear equalization; maximum likelihood sequence detection (MLSD), also known as maximum likelihood detection (MLD); possible "blind" techniques; and constant modulus algorithm (CMA ).

• The above embodiments can be used for other ISI sources: poor transmitter extinction ratio; (time-varying) PMD; mode partition fluctuation combined with dispersion.

• A possible alternative to using RAM would be to use multithreading in parallel processors.

• In another embodiment, coefficients (parameters) may be monitored to monitor the performance of one or more transmission paths.

Claims (23)

1, a kind of central station that is used for receiving data from a plurality of substations, described central station is configured to carry out in use and is used for compensation process that the deterioration from the data of described a plurality of substations is compensated, described compensation process has at least one tunable characteristic by parameter group control, wherein, described compensation process may further comprise the steps: (i) use different start-up parameter groups that the data from substation are compensated; (ii) the quality of using the described data after described different start-up parameter group compensates is measured; And, select a start-up parameter group that the data from the follow-up arrival of described substation are compensated (iii) according to measured quality.

2, central station according to claim 1, wherein, described central station is configured to store a parameter group at each substation.

3, central station according to claim 2, wherein, arriving according to step (i) (iii) is that each substation is selected institute's stored parameters group.

4, according to any described central station in the aforementioned claim, wherein, execution in step (i) is to (iii) in response to the reception of the data of the substation that connects making a fresh start.

5,, wherein,, come from other start-up parameter groups, to select a parameter group by the quality of using described different start-up parameter group to obtain is compared according to any described central station in the aforementioned claim.

6, according to any described central station in the aforementioned claim, wherein, use and the start-up parameter group is assessed, preferably from the test data of substation, before the described test data from substation arrives, the copy of described test data is stored in described central station.

7, according to any described central station in the aforementioned claim, wherein, described parameter group includes a plurality of parameters.

8, according to any described central station in the aforementioned claim, wherein, the substation that described central station is configured for from the overall optical network receives data.

9, according to any described central station in the aforementioned claim, wherein, described compensation process is carried out in electric territory.

10, according to any described central station in the aforementioned claim, wherein, described compensation process may further comprise the steps: from a plurality of time locations place in the data flow of substation to the described line sampling that flows to; And to each sample execution function corresponding.

11, according to any described central station in the aforementioned claim, wherein, described central station is configured to send dispatch command to substation, and to receive data from this substation, described dispatch command comprises this substation of permission and sends the order that data reach the fixed time.

12, central station according to claim 11, wherein, described central station is configured to send further dispatch command, so that can send this further data under the further data conditions of needs.

13,, wherein, be to come selected according to institute's stored scheduling instruction to going into the parameter group of using when office data is used described compensation process according to claim 11 or the described central station of claim 12.

14, according to any described central station in the claim 11 to 13, wherein, described dispatch command comprises and has identified the identifier that allows to send from it substation of data, and wherein uses the described identifier to retrieve the parameter group that is associated with the substation that is identified.

15, according to any described central station in the aforementioned claim, wherein, described compensation process comprises adaptive algorithm, this adaptive algorithm is constructed such that when the data that receive from substation, use the described data that receive, the value of the parameter that is used for this substation is improved with respect to one group of initial value.

16, central station according to claim 15, wherein, described central station is constructed such that in the source that arrives data when first substation changes over second substation, described central station: (i) storage is at the improvement value from the parameter of the data of described first substation, to be used for the retrieval of back; (ii) retrieval is before at the described second substation stored parameters; (iii) in response to the further data from the expection or the arrival of described first substation, retrieval is at the previous improvement value of the parameter of described first substation; (iv) and preferably the parameter to described first substation is further improved.

17, the method for a kind of operation communication system, this communication system comprises central station and is connected to a plurality of substations of described central station, described central station can be used for carrying out and is used for the compensation process that carries out from the deterioration of the data of described a plurality of substations, described compensation process has at least one tunable characteristic by parameter group control, said method comprising the steps of: (i) use different start-up parameter groups that the data from substation are compensated; (ii) the quality of using the data after described different start-up parameter group compensates is measured; And, select a start-up parameter group that the data from the follow-up arrival of described substation are compensated (iii) according to measured quality.

18, method according to claim 17, wherein, described communication system comprises the optical-fiber network that is used for described central station is connected to described a plurality of substations.

19, according to claim 17 or the described method of claim 18, wherein, described optical-fiber network comprises and being used for from the signal multiplexing of at least two substations branch's contact to the common light carrier wave, described a plurality of substation is formed at a plurality of time intervals and sends data, makes in use data from described a plurality of substations by passive time division multiplexing.

20, according to any described method in the claim 17 to 19, wherein, the storing predetermined data of described central station, and described or each substation is stored corresponding tentation data, and wherein said or each substation sends to described central station with its tentation data, makes the described tentation data that is stored in described central station to be compared with the tentation data that is received.

21, method according to claim 20 wherein, compares tentation data of being stored and the tentation data that is received, so that the quality of using the data after described different start-up parameter group compensates is measured.

22, according to claim 20 or the described method of claim 21, wherein, the described tentation data that is stored in substation comprises at least some total data with the described tentation data that is stored in described central station.

23, a kind of communication system that comprises central station and a plurality of substations, described central station is constructed to carry out and is used for compensation process that the deterioration from the data of described a plurality of substations is compensated, described compensation process has at least one tunable characteristic by parameter group control, wherein, described central station stores at least one parameter group at each substation, and for each substation, described central station uses the parameter group that is associated with this substation that the data from this substation are used backoff algorithm.

Images