FR2887096A1 - Electronic system e.g. radar, architecture, has single synchronization link supplying common synchronization message to sub-modules, and data exchange asynchronous links exchanging messages between two processing nodes - Google Patents

Abstract

The architecture has a single synchronization link supplying a common synchronization message with synchronization data units and a simplified heading to modules. Data exchange asynchronous links allow exchanging messages between two individual processing nodes, where the exchanged messages have processable data units accompanied by a generic heading. Each processing node has an interface to interface with synchronous and asynchronous links. An independent claim is also included for a radar system comprising a set of sub-assemblies having a set of processing nodes.

Description

Generic radar architecture

  FIELD OF THE INVENTION The invention relates to the general field of architectures of complex electronic and / or computer systems and in particular to the field of information exchange structures between the different elements of such architectures. It is particularly interesting for complex systems for which exchanges take the form of irregular data flows generated by time-varying processing functions.

  t o It applies in particular to the construction of electromagnetic or acoustic sensors, such as radars and sonars.

  BACKGROUND OF THE INVENTION - PRIOR ART Current designers of electromagnetic or acoustic sensors are faced with new problems related in particular to the great complexity of the functions performed by the processing subsystems that make up these systems. The complexity of these functions is due as much to the complexity of the operations performed as to the large variation in the quantities of information (information flows) processed from one operating phase of the system to the other. Thus, the same function may have to process both a large data stream and a very short data stream. It may even have to deal with varying alternations of long data streams and short data streams. This implies in particular that for the same function, the processing time can vary to a large extent, this variability consequently generating an irregularity over time of the data flows transmitted from an autonomous functional unit, or processing node. , to another. The functions performed are more and more sophisticated and lead to more and more computing loads and increasingly complex sequencing.

  This present and, above all, future increase in the desired performances does not fit well with the classical architectures around which the current systems are designed. These architectures are generally based on the development of synchronous machines performing tasks that are certainly autonomous, but whose entire sequencing is provided by means of signals delivered by a particular processing node and whose exchanges are closely synchronized. However, the irregularity of the data flows makes the development of such a processing node very difficult.

  Such architectures have the effect of splitting the systems thus conceived o into assemblies of subsets performing processing tasks, each subset being clocked by a set of dedicated synchronization signals provided by a subset responsible for generating the data. different synchronization signals. They hinder a more integrated design approach in which the system is considered in its entirety as a single multifunction machine.

  This fragmentation associated with a too great synchronism of exchanges leads to the system being subjected to a limitation of its processing capacity, in terms of the amount of information processed. This limitation directly attributable to the architecture adopted hampers the necessary increase in the computing capacity of the systems, an increase that the processing nodes making up the subsets would otherwise be able to support separately.

  This limitation makes conventional architectures difficult to adopt for future systems where the level of performance is directly related to computing capacity and adaptability to variations in data flows.

  Another disadvantage of the current architectures is related to the organization of the sequencing of the different processing nodes around a general sequencer responsible for the overall synchronization and which delivers to each of the other processing nodes, the set of the signaling signals. synchronization he needs. This centralized production of synchronization signals is reflected in practice by the presence in developed systems of a very large number of physical links, which makes the manufacture and development of such systems difficult, and hampers their ability to evolve. .

  PRESENTATION OF THE INVENTION One of the aims of the invention is to enable a system comprising a plurality of subassemblies to optimally exploit the computing capacities of each of the processing nodes that constitute them, taking into account the in a given chain of tasks, each processing node is required to perform the task (s) that is incumbent on it (nt) at a rate that can be variable over time, in particular as a function of the variations in the flow of tasks. processed data.

  Another goal is to allow the design of systems that can evolve, a system that can integrate during its life subsets and / or processing nodes whose computing capacity has evolved so that the processing capacity of the system in its integer can increase.

  Another object is to facilitate the production of maintenance and maintenance of increasingly complex systems, in particular limiting the type and number of physical links.

  For this purpose, the subject of the invention is an electronic system architecture, in particular a radar detection system architecture, comprising a plurality of processing nodes in which each node performs a function or part of a function implemented. by the system. This architecture is characterized by: a single synchronization link which delivers to all the modules a common synchronization message which comprises a simplified header, a synchronization top, a base clock and a common reference date, the simplified header for determining to which data stream the synchronization information is applied; a plurality of asynchronous data exchange links, each link allowing the exchange of specific messages between two particular processing nodes; exchanged messages comprising data to be processed, accompanied by a common generic header containing all the information used by the various processing nodes of the system and relating to the operating step of the system to which the data exchanged relates. This header allows each processing node to associate the data contained in a message with a determined phase of the operation of the system and determine how it should process this data.

  According to the invention, the simplified header contained in the synchronization message is part of the generic message header transmitted by the asynchronous links, which corresponds to the same operating step of the system.

  According to the invention, each processing node has own means for interfacing with the synchronization link and for performing its internal sequencing from the general synchronization message. Each processing node also comprises own means for associating the data transmitted by a data exchange link with the corresponding header.

  According to the invention, the synchronization link transmits a timing clock H, a sync top and binary data frames, the sync top and the bit data composing the frames being transmitted by modulating the clock form factor. H, the modulation preserving the time difference separating the rising edges of the clock H. According to one characteristic of the invention, the modulation of the form factor of the clock H comprises 3 states: a state 0 for which the state the top of the clock H has a duration equal to that of the low state, the form factor being equal to / 2, a state 1 for which the high state of the clock H has a duration greater than of the low state, the form factor being greater than '/ 2, - a state -1 for which the high state of the clock H has a duration less than that of the low state, the form factor being less than '/ 2, the values of the bit data of the frames are further coded on the two states 0 and 1, the synchronization top being identified by the code -1.

  According to another characteristic of the invention, the transmission of a data frame on the synchronization link is signaled by the transmission of a synchronization pattern. The transmission of a frame usually corresponds to the start of execution of an operational task by the system. In the context of a radar system, an example of an operational task may consist of a burst. According to the invention, each processing node synchronizes its slow clocks, the frequency of which is a sub-multiple of the frequency of the clock. H, using the synchronization top.

  According to a preferred embodiment of the invention, the synchronization top is sent periodically, accompanied by a data frame containing useful synchronization information or a dummy frame containing insignificant data.

  According to this preferred embodiment, each transmitted frame includes a bit field for indicating whether this frame is a real frame or whether it is a dummy frame.

  According to an alternative embodiment of the invention, the synchronization top is transmitted on the synchronization link in the form of a sequence composed of a succession of bits that can take the values 0, 1, or -1, the structure of the sequence being defined so that recognition by the subsets of the synchronization peak can be achieved by means of a ternary correlator by the detection of a correlation peak.

  According to another variant embodiment, the information contained in the bit frame is encoded by means of a detector and error correction code.

  This decentralized architecture advantageously makes it possible to reinforce the asynchronous operation of the different subsets and processing nodes. Each data exchange between two processing nodes is performed by an asynchronous link, without the intervention of validation or synchronization signals.

  This architecture also allows each processing node to adapt to variations in the flow of data exchanged.

  The system is thus considered in its entirety as a single logical machine whose computing capacity is much greater than that obtained by operating the same processing nodes with a conventional synchronous architecture. The architecture according to the invention also makes it possible to perform a functional cutting of the tasks performed by the system, the same function being able to implement calculation nodes distributed over different physical entities such as printed circuits. The functional cut is then advantageously independent of the physical cut of the system.

  The architecture according to the invention also makes it possible to take advantage of the increase in computing capabilities that a module can benefit from following a new design. It also makes it possible to implement additional functions to the system during its lifetime, without this implementation requiring a significant modification of the existing structure of the system.

  The implementation of a single synchronization link also has the advantage of limiting the number of physical links to be made. It allows each subset or processing node to autonomously synchronize its own clocks from a timing clock H associated with a synchronization top.

DESCRIPTION OF THE FIGURES

  These and other features and advantages will appear more clearly in the course of the description illustrated by the appended figures which represent: FIG. 1: a schematic illustration of the architecture according to the invention, FIG. schematic representation of the interfacing structure of a processing node developed according to the architecture according to the invention; - FIG. 3: the illustration of the schematic structure of the means that comprises a processing node producing an asynchronous process FIG. 4 is an illustration of the schematic structure of the means comprising a processing node performing a synchronous processing; FIG. 5 is an illustration of the principle of the routing of the data carried out in a processing node developed in accordance with FIG. architecture according to the invention.

  FIG. 6: an illustration of an example of data association with the corresponding header, FIG. 7, an illustration of the principle of coding the data bits by modulation of the clock form factor. FIG. 8 illustrates the structure of the signals and digital data transmitted on the synchronization link; FIG. 9 illustrates the principle of synchronization of the slow clocks of the processing nodes by the link synchronization.

DETAILED DESCRIPTION

  We first consider in Figure 1. This figure shows the general characteristics of the architecture of the invention. It represents in the form of a functional synoptic all the functions that can contain a complex electronic system, such as a radar for example. Such a system may comprise a first family of sub-assemblies fulfilling interface functions with an external medium, which medium may for example be the atmosphere, in the case of an electromagnetic sensor such as a radar, but which can also be the marine environment for a sonar-type acoustic sensor, or any specific interface of the physical sensor type. In the specific case of a radar system, the subsets 11 are for example the subsets responsible for transmitting and receiving microwave.

  Such a system may also include a number of sub-assemblies 12 for analog purposes, one of whose functions may for example consist in digitizing the signals exchanged by the subsets 11 with the external medium.

  Such a system may finally comprise digital subsets 13, for example microprocessors or microcontrollers whose functions consist essentially of synchronous or asynchronous processing of data.

  In the case of complex systems, it is easy to see that, although belonging globally to one or another domain, a subset may itself comprise elements belonging to other domains. Thus, for example, it is easy to conceive that a subset 11 whose main function is to interface with the physical world (atmosphere or marine environment for example) has elements 14 - 15, allowing it to exchanging digital data, synchronously or asynchronously, with a processing node such as a microprocessor. Similarly, it is readily conceivable that such a subset has essentially analog elements 16 allowing it to transmit the signals received from the outside environment to a subset 12 responsible for digitizing them.

  The structure of the links between analog machines generally responds to constraints and defined standards which are in particular a function of the nature of the analog signals exchanged. A particular standardization of such links is therefore in principle not useful.

  On the other hand, as far as the standardization of digital data exchanges is concerned, the benefit provided by such standardization can be very important.

  This is why the architecture according to the invention is mainly intended to define the mode of operation of the digital elements that comprise the different subsets making up a complex system, and to standardize the digital data exchanges between these elements. It therefore mainly concerns the synchronous digital domains 17 and asynchronous 18.

  In the remainder of the description, the term "processing node" is used to designate a functional entity that is part of a subset and capable of autonomously and intelligently performing a given synchronous or asynchronous task, according to parameters supplied to it. These processing nodes represent the functional entities to which the architecture according to the invention applies. A given functional unit may, depending on the function it performs, comprise one or more processing nodes. The processing nodes that cooperate to perform a given function can also be located on separate physical entities, printed circuits or racks for example. Regardless of their respective physical locations, the exchange of data between two processing nodes is always ensured by means of an asynchronous link bus. Likewise, each processing node is connected to the synchronization link; As illustrated in FIG. 1, the architecture according to the invention is mainly based on subsets, in particular digital subsets, comprising one or more asynchronous processing nodes exchanging data asynchronously with other processing nodes.

  However, some processing nodes still need to have accurate timing for some of the tasks they perform, they also include synchronous processing nodes.

  The same subset may further include synchronous processing nodes and other asynchronous.

  It is thus possible to characterize the architecture by the nature of the interfaces that comprise a subset adapted to this architecture, allowing the different processing nodes contained in this subset to process and exchange data, and by the nature of the links. allowing the exchange of data between processing nodes.

  A system designed around the architecture according to the invention thus comprises a plurality of asynchronous data exchange links 19 and a synchronization link 110.

  According to the invention, the asynchronous links 19 are essentially "point-to-point" links, a link enabling a processing node belonging to a given subset to exchange information with other processing nodes belonging to the subscriber. same subassembly or a different subset. The synchronization link 110, for its part, is a single link allowing a subset 111 managing the global synchronization of the system to send to all the processing nodes a general synchronization message.

  FIG. 2 is now considered, which presents in a very general manner the structure of the interfaces that equip a subset adapted to the architecture according to the invention. The illustration of FIG. 2 corresponds to a standard subset, realizing functions relating to the various fields of information exchange mentioned above. It is of course conceivable that, because of the tasks it performs, a subset does not have an interface 21 with the analog domain or interface with the external medium. In this case, it will not include an analog interface or a sensor. On the other hand, the generalization of exchanges of information or signals in digital form makes the presence of digital interfaces necessary, whatever the global function fulfilled by the subset.

  An important feature of the invention is to allow a standardization of exchanges, and to ensure each subset an internal operation as independent as possible from that of other subsets. To this end, a subset adapted to the architecture according to the invention comprises one (or more) asynchronous digital interface 23 which enables it to exploit the data transmitted to it by another subset and to condition the data. that it transmits to another subset. It also comprises a synchronous digital interface 24 enabling it to carry out an independent internal sequencing which may refer, if the task to be performed so requires, to an overall chronometry of the system.

  The remainder of the document relates more particularly to these digital links and the corresponding interfaces.

  As has been said previously, the architecture according to the invention implements on the one hand asynchronous links allowing the exchange of data between subsets, on the other hand a synchronization link. The synchronization link simultaneously connects all subsets of the system to which it provides a synchronization message, a base clock H, a synchronization tick, and a date reference. The synchronization link further provides a simplified header enabling a processing node performing synchronous processing to associate the received synchronization message with a particular asynchronous data item itself characterized by the header (label) of the synchronization process. message in which it is included. According to the invention the simplified header transmitted in the synchronization message is a part of the header accompanying the corresponding asynchronous data.

  Depending on the processing node considered, it uses all or part of this information.

  The synchronization link is a multi-address unidirectional link, operating in broadcast mode, on which the synchronization messages are transmitted by the particular subset 111 to the processing nodes of the other subsets. The technical characteristics of the synchronization link according to the invention are described later in the document.

  FIG. 3, which schematically represents the management principle of the data received by a processing node 31 performing asynchronous processing 35, is considered. This processing is clocked by the clock transmitted on the synchronization link 110 by the sub-code. together 111.

  According to the architecture according to the invention, the data received by this processing node are data transmitted by another node through an asynchronous link 19.

  According to the invention, all the asynchronous links 19 have an identical structure. These are bidirectional point-to-point links, such as "full duplex" for example, through which the data 32 is exchanged between the processing nodes. In the case of asynchronous exchanges, the necessary identification of the data transmitted to a system operation phase is ensured by the transmission of a header 33 which accompanies the data transmitted. Thus, the processing node that receives the data is informed, by stripping the header message 33, of how this data is to be processed, without having been informed of the start time of the operating phase. considered.

  If we take as an example a radar system, in which the operating rhythm is given by the instants of pulse burst emission, a signal processing module will be able thanks to the architecture according to the invention of asynchronously process data relating to an emitted burst without having to consider the moment at which this burst was emitted.

  To correctly process a received data item, a processing node performing asynchronous processing includes means 34 for acquiring, analyzing, storing and retrieving this header. These means also make it possible to parameterize the processing performed according to the nature of the data or the data to be processed.

  Thus, in the example illustrated in FIG. 3, the node 31 performs a processing on two data streams originating, for example, from two different processing nodes.

  In order to guarantee that the processing carried out is applied to data belonging to the same operating phase, the means 34 analyzes the respective headers of the two data present on each of the asynchronous links 36 and 37. In the opposite case, the processing is not for example not done. Furthermore, in the case where one of the data is transmitted in advance with respect to the other, the means 34 can proceed, after analysis of the header, to put the data transmitted first on hold. For this purpose the processing node 31 may comprise temporary storage means not shown in the figure.

  The processing node may also include means not shown in the figure for recovering the basic clock H from the general synchronization message and thus create its own clocks from the base clock.

  Once the processing has been applied to the received data, the processing node 31 associates the previously stored header 33 with the processed data 38 and transmits the data preceded by its header to the next node by another asynchronous link 19.

  According to the invention, the header 33 is a generic message, which contains all the information necessary for each processing node of the processing chain to determine how it should process the data associated with this information. head. This message is propagated without change of structure from one processing node to another along the entire processing chain. Thus, at the different stages of its processing, a datum is transmitted from one processing node to another accompanied by a unique and invariant structure header.

  Asynchronous data processing associated with an asynchronous exchange structure is an essential and advantageous characteristic of the architecture according to the invention. This association makes it possible to limit the influence of the variations in the flow of information information processed by the system on the global computing load that the system must support. In particular, it allows the existence of a variable data flow rate between the different processing nodes, the general synchronization constraints being supported at the level of the global system and not at the level of each node.

  FIG. 4, which schematically represents the management principle of the data received by a processing node 41 performing a synchronous processing 45, is considered. This type of processing occurs, for example, in the case where analog signals digitized on the fly by means of of the basic clock of the system, must undergo a first treatment which depends on the operating phase of the system corresponding to these analog data. Such a case occurs for example in a radar system, particularly at the receiver subset. The stream of digitized data corresponding to an emitted pulse burst must sometimes be processed, taking care not to process the first N data received for each pulse of the transmitted burst. Since the number N may vary according to the type of steal issued, it is important for the first processing node to be able to know the precise arrival time of the first data in order to eliminate the first N data. In such a case, the processing node 41 considered has a synchronous processing 42 enabling it to process the data received in accordance with the sequencing described by the synchronization data item 44 transmitted by the synchronous link. The processing node also includes asynchronous processing enabling it to know the parameters relating to the operating phase considered. In the case of a radar, the asynchronous data item 45 contains, for example, information relating to the burst transmitted, as well as information on the moment at which the data will be available for processing.

  Thus, by exploiting the information transmitted by the asynchronous link, the processing node 41 is able to perform a synchronous processing according to the generic synchronization message received by the synchronous link. The concordance between the data transmitted by the asynchronous link and the corresponding synchronization message is also ensured thanks to the simplified header 46 transmitted in the synchronization message. Indeed this header which is for example a part of the header 47 of the corresponding asynchronous message: at the same operating phase of the system can identify and associate without error messages from the two links.

  As can be seen from FIGS. 3 and 4, the synchronization link thanks to the general message that it conveys is advantageously of universal use in the system. It is both usable by a processing node performing synchronous processing and by a processing node performing asynchronous processing, the particular use made of it being determined at the considered processing node which can even host synchronous and asynchronous processing. In this way, the architecture according to the invention makes it possible to design systems whose processing nodes are seen as independent elements performing functions in an entirely autonomous way, each processing node can thus be designed to take care of its own sequencing. In such an architecture, each processing node is considered as a single component of a single machine represented by the system as a whole, the internal exchange structure of this machine not depending on the implanted elements.

  As also illustrated in FIGS. 3 and 4, the synchronization link 110 distributes simultaneously to each processing node a general-purpose base clock H and a synchronization message mainly used by the processing nodes performing synchronous operations. Thus, each processing node can manage its internal sequencing autonomously, asynchronously while remaining linked to the general synchronization of the system by the same basic clock H. However, to prevent the consequences of a possible failure relating to the reception of this clock by one or more processing nodes (parasite, operating alea at the receiving module of the synchronization link of the processing node, etc ...), consequence, which could be felt late, because of the asynchronous operation of the system, the synchronization link advantageously allows each processing node to check if such a failure has not occurred. For this purpose the synchronization link distributes to all the processing nodes additional information representing a reference date permanently updated. This reference date changes at the rate of the base clock H. Each processing node can use this date to compare it to a local date evolving with the rhythm of the basic clock actually received. Thus, the two dates having been previously reconciled, at the start of the processing node for example, the synchronization module can periodically check the match of the clocks and, in case of difference, make a new reconciliation the local date with the reference date. According to one possible variant of implementation, the rephasing of the local date on the reference date can even be performed automatically and systematically each time a new synchronization message is received.

  The synchronization link 110 therefore distributes three distinct pieces of information: a base clock H, a synchronization top and a reference date. This information is also accompanied by a simplified header to associate with particular asynchronous data. This information is accessible to all the processing nodes so that although each processing node usually works asynchronously, synchronization of the entire system remains assured.

  We are now interested in FIG. 5, to explain how the data exchanges on the asynchronous links 19 are ensured. Generally, in a complex system, the same data stream received by the system during a given phase of operation undergoes a certain number of transformations during its passage in the different processing nodes composing the system. Thus, this stream of data can give rise after processing by a processing node 51 (M2 or M3) to a data item 52 (59) which is transmitted to a second processing node 53 (M4) for which it constitutes a data item. Entrance. This data item 52 (59) is processed by M4 and can give rise to one or more output data 54, 55 or 511, respectively transmitted to the processing nodes 57 (M6) and 55 (M5).

  Thus, in a pulse radar type system for example, this data stream is constituted by the signal received in response to the emission of a burst of pulses emitted in a specific mode. This signal is processed after digitization by the different processing nodes that constitute the radar.

  In such an example, the processing node 51 may for example be responsible for performing adaptive filtering on the signal and for delivering a data item 52 corresponding to the signal after filtering. This data item 52 can then be transmitted to the processing node 53 whose role is to normalize the data item 52 with respect to a calculated average value. The processing node 53 can, for its part, communicate the standardized data to a processing node 57 and the calculated average value to another processing node 56.

  In the architecture according to the invention the different processing nodes work asynchronously from each other. On the other hand, the data exchanges are also performed asynchronously by means of the links 19. It is therefore necessary, for a processing node which receives data on an asynchronous link, to be able to process this data correctly. given, that said data is accompanied by an identification message. For this purpose, according to the invention, the data exchanged through asynchronous links are integrated in a message consisting of the actual data to which are associated headers or labels that characterize the data.

  According to the architecture according to the invention, as illustrated in FIG. 5, all the data coming from the same initial data stream, acquired by the system during a given phase of its operation, are associated with one and the same system. generic head 58 (510) which contains all the information relating to this operating phase.

  According to the invention, each module integrated in the system therefore comprises means enabling it to identify, from the associated header, the parameterization of the processing function that it will apply to the input data, received by the processing node. The module also has means for storing this header so as to associate it with the output data resulting from the processing of the received data. Each output data 54, 55, or 511 is associated with the generic header 58, 510 which accompanied the input data 52, 59 so that it could be transmitted to another processing node. Since each output data is associated with a generic header, it may be transmitted to any recipient processing node for subsequent processing.

  In the case of a radar system mentioned above as an example, the header 58 may for example contain the information relating to the burst of transmitted pulses, such as the size of the pulses, their repetition period, as well as the rank of the impulse considered.

  In this example, the data 55 normalized by the processing node 53, resulting from the data 52, is transmitted to the processing node 57 associated with the label 58 while the data item 54 corresponding to the average value calculated by the processing node 53 to from the same data 52, is transmitted to the processing node 56 also associated with the label 58.

  The architecture according to the invention is thus based on a generic data exchange structure which has the advantage of making it possible to insert into an existing data processing chain an interposed function implemented by a processing node. given. This insertion is performed easily by inserting the processing node opening the asynchronous link which conveyed the data item to which complementary processing is to apply and establishing an asynchronous link between the transmitting processing node and the processing node performing the intermediate function. on the one hand, and an asynchronous link between this processing node and the initially receiving processing node. This insertion operation is advantageously performed without it being necessary to modify the initially connected processing nodes.

  We now consider FIG. 6. This figure illustrates the operating principle of the means 34, which comprises a processing node according to the architecture according to the invention, which perform the association of the output data with the header. corresponding. These means mainly provide a storage function, not shown in FIG. 6 but represented in FIG. 3, which makes it possible to keep for each data item received the header that accompanies this datum, and a routing function whose role consists in associated with each output data 63, produced by the processing 61 implemented per considered processing node, the header (label) 64 of the corresponding input data. The routing function is a specific function of the processing performed by the considered processing node. It can for example be implemented by addressing an association matrix symbolized in the figure by the box 65. This matrix that associates one or more output data 63 with one or more labels 64 is developed according to the binding relationship the output data 63 to the input data 62.

  The remainder of the description relates to the nature and structure of the information conveyed by the synchronization link. This part of the description will enable those skilled in the art to appreciate even better the advantages of the architecture according to the invention.

  Referring now to Figures 7 and 8 which schematically show the form of the information transmitted and the sequencing of the exchanges that occur on the synchronization link.

  From a general point of view, the synchronous link according to the invention is a serial type link on which bits of binary data are sent to each subset. A frame comprises a fixed number of binary data types defining fields, each field having a given number of bits, so that the frame itself comprises a determined number of bits. The sequencing as well as the content of the frames transmitted are also constituted by the subassembly 111 responsible for the general synchronization. It should also be noted that, depending on the number of processing nodes that a subset comprises, the signal transmitted by the synchronous link is either directly processed at the subset level or distributed as such to the various processing nodes that constitute it. .

  According to the invention, the synchronous data link permanently transmits a clock clock H of fixed frequency synthesized by subassembly 111, a clock of frequency F = 10 MHz for example. This clock H has the characteristic of being the reference clock of the entire system that carries the architecture according to the invention. It is from this clock H that is distributed to all the subsets by the synchronization link that are manufactured within each processing node, the clocks necessary for the operation of the processing nodes that make up the subset considered. . In some cases these clocks, so-called slow clocks, are clocks of lower frequency than the frequency F of the clock H, clocks that can be obtained by dividing the frequency F, by means of counters for example. In other cases these clocks are clocks of higher frequencies, or even much higher, than the frequency F which are obtained by multiplication of the frequency F, by means of phase locked loop circuit for example. According to the invention, the clock H is a clock having the desired stability.

  In the particular case of a radar signal processing system, this clock must, for example, be of sufficient stability to enable the subsystems responsible for the analog-to-digital conversion of the radar data coming from the receiver subsystem to manufacture their signals. own sampling clocks whose frequency is a multiple of the frequency F. It must also for example be of sufficient precision to allow the synthesis of local oscillators operating at intermediate frequencies several tens of times higher than the frequency F. so that the timing clock H is distributed to all subsets by the synchronization link, the processing nodes advantageously benefit from a precise and stable common frequency reference. Furthermore, the implementation of the frequency multiplier circuits necessary for the realization of these multiple signals of the frequency F is advantageously facilitated by the modulation mode of the form factor of the clock H, which modulation keeps the time difference constant. separating the rising edges of the clock H. The synchronization link also carries, where appropriate, various synchronization information which in particular enables the subsets to associate data originating from subassemblies together. different. This information is coded as serial binary data.

  In order not to interrupt the permanent transmission of the sequencing clock, the synchronization information is transmitted, according to the invention, by performing a particular modulation of the shape factor of the clock H. According to the invention this modulation has, as illustrated in Figure 7, three distinct states. A first state 71 corresponding to a data bit of value 0, a second state 72 corresponding to a data bit of value 1, and a third state 73 corresponding to a data bit of value -1. In this way, in order to transmit on the synchronization link a binary data equal, for example, to 101, it is necessary to modulate the form factor of the clock H over three periods in order to successively transmit a state 1, a state 0 and then another state. Each bit of data thus has a duration equal to T. In the non-limiting example of FIG. 7, the coding of a state 0 is obtained by keeping the form factor at 1/2, the duration of the high state of the clock being equal to that of the low state, while the coding of a state 1 is obtained by modifying the form factor so that the duration of the high state of the clock is equal to 3/4 of the period of H, and that the coding of a state -1 is obtained by modifying the form factor so that the duration of the high state of the clock is equal to period H. The encoding of the clock H is used to transmit binary data: each bit can therefore take only two states (0 or 1). According to the invention, the third modulation state is used as the indicator state, as the synchronization peak. This synchronization top has two main uses.

  The purpose of the synchronization top is first of all to signal the transmission of a data frame on the line. Thus, according to the invention, the transmission of a data frame is always signaled by the transmission of a bit, or top, of synchronization. The timing diagram of Figure 8 describes the different states of the synchronization link.

  In this figure we can distinguish the two possible modes of operation, 83 and 84, of the synchronization line. Mode 83 corresponds to the case where no synchronization information is transmitted. The clock is then transmitted alone in the form of a periodic signal with a form factor equal to / 2. The mode 84 corresponds to the case where synchronization information is transmitted on the synchronization line. In this operating mode, the form factor of the transmitted clock H is modulated so as to transmit on the one hand a synchronization peak signaling the start of transmission of a data frame, and then to transmit the data bits of value 1 or 0, constituting the frame.

  The modulation of the form factor is performed modifying the position of the falling edge 85 of the considered clock slots, while keeping the period of the clock H constant so that the time spacing of the rising edges is kept constant and equal to T. The sequence of operating modes over time is essentially dependent on the application in question. Thus the synchronization link may for example operate in a mode 83 for a given period of time corresponding to an initialization phase, then in a mode 84 for a period of indefinite duration corresponding to the permanent sending of data frames. in an operational operating phase, sending a frame which can be separated from the neighboring items, as illustrated in FIG. 8, by short time intervals during which the link operates in the mode 83.

  We now consider Figure 9.

  According to the invention, the synchronization top also serves to enable the phasing of the set of slow clocks, deduced from the clock H by frequency division (H / 2, H / 4, ...) generated by the different subsets. The mechanism of this phasing is illustrated in Figure 9. It consists in principle, for each subset, to detect the appearance of the start frame bit 86 and synchronize, if necessary, the rising edge of the clock considered with the rising edge of the frame start bit 86. FIG. 9 illustrates the example of two clocks HI and H2 generated by two different subassemblies, by division by 2 of the frequency of H. This example highlights the fact that two clock of the same frequency deduced from H can however be shifted. It also shows that simply by using the synchronization top, constituted by the signaling bit 86 of the passage of a data frame, each subset can, in an entirely autonomous way, proceed to phase up the slow clocks generated. from H. Phasing thus advantageously does not require any particular general procedure. In this way each subset can perform, as soon as its initialization is completed, for example, a phasing of its clocks.

  In a preferred form of implementation of the sync link, the sync top is transmitted periodically, as soon as the subassembly 111 responsible for the overall system synchronization and managing this link is in operation, after a initialization phase for example. In this preferred mode of operation the synchronization top is transmitted periodically, even if no useful synchronization information is transmitted. In the latter case, a dummy frame is transmitted, a frame of which one of the bit fields contains data informing the different subsets that the data contained in the frame are not to be taken into account. In this way the synchronization link is the seat of a permanent transmission of frames which can be significant or not. The synchronization top is thus advantageously transmitted periodically and frequently. The phasing of the clocks of a subset can thus be performed in a very short time. This advantageous characteristic makes it possible in particular for each subset, during the general start-up of the system, to proceed to the synchronization of its clocks as soon as its initialization and that of the subset 111 responsible for the general synchronization are completed, without having to wait for operational synchronization data to flow on the link. The subset 111 only has to transmit dummy frames.

  This mode of synchronization, advantageously simple to implement, therefore has the advantage of not requiring any specific synchronization signal.

  It also advantageously allows each of the processing nodes of the system to perform its own synchronization independently. In this way, when a processing node is forced to carry out its own reset, as a result of an internal operating problem for example, it is able, as soon as it is reset, to resynchronize its clocks with all of the clocks. system clocks without requiring a general resynchronization phase on the part of the system. Likewise, when the system is turned on, no particular clock synchronization procedure is necessary, each processing node synchronizing its own clocks when it has finished initializing.

  With regard to the data frames, it should be noted that these are, as has been said previously, constituted by a fixed number of bits, the bits composing a frame being grouped into fields, each field corresponding to one piece of information. synchronization, coded on a given number of bits.

  The nature and size of the information transmitted varies in practice largely from one application to another. However, certain information whose exchange is an important feature of the structure according to the invention are necessarily present. Thus, a frame according to the invention conveys, as has also been said previously, a date reference common to all the elements of the system. This date makes it possible in particular to reference certain processed information to a more general time reference common to several systems exchanging information for example.

  The frame according to the invention also conveys a simplified header which is a part of the header accompanying the corresponding asynchronous data item and which enables a processing node to associate the received synchronization message with a particular asynchronous data item. - even characterized by the header (label) of the message in which it is included.

  In the preferred embodiment, the data frame according to the invention also comprises a data field enabling the subsets to know the real or artificial nature of the frame in question. In addition to these characteristic data, a frame according to the invention may also comprise a certain number of data charnps whose content is used to define, for example, the date on which the information contained in the frame must be taken into account by the processing nodes. It can thus be specified when it is a frame actually operational in how long, expressed in number of periods T, after reception of the frame, the synchronization information contained will be applicable.

  The table below shows an example, according to the invention, of a data frame structure that can be used in a radar system for exchanges between the synchronization subset 111 and the different processing nodes that constitute the system using an architecture according to the invention.

Field - Data Number Description

  of bits Identifier of 1 bit of synchronization start of frame type of synch 6 code of sync type Delay 8 delay for taking into account the synchro application.

  Value: from 0 to 255 clock cycles H Reservation 13 Available field Identifier of 4 Identification of the frame type data block Counter of 16 synchronization tops counter w blocks of Value: from 0 to 65535. 3 data Update of the 8 Data for updating the date general internal date of the processing nodes according to the general date of the system Test of 8 Cyclic redundancy link integrity test In the example described here, data exchange on the synchronization links are carried by 64-bit frames, the first bit corresponding to the frame start bit which is also used to synchronize the slow clocks.

  The frame taken as an example then comprises a 6-bit field for defining the type of synchronization performed. These six bits can thus be used to define whether it is a frame relating to synchronization at the rate of recurrence, that is to say of the emission of the radar pulses; at the rate of the burst, that is to say at the rate of a pulse group for which the received signals are processed in a coherent manner; or at the rate of the antenna revolution and the passage of the antenna beam to the north.

  The frame taken as an example also comprises, according to the invention, a data field making it possible to know the instant of application of the data contained in the frame. In this example, the "Application delay" field extends over 8 bits, which makes it possible to define a delay ranging from 0 to 255 clock cycles H. This frame also comprises a data field relating to the type of frame transmitted. . This field with a size of 4 bits makes it possible to define whether it is an operational frame, a test frame or a fictitious frame. It also includes an event counter, a dynamic 16-bit, for assigning to the event considered, beginning of recurrence or bursts for example, associated with the information contained in the frame.

  The frame taken as an example still comprises, in accordance with the invention, a data field enabling each processing node to synchronize its internal date to the general date. In the example shown in the table above the general date is set to a 64-bit dynamic whose update is done by nibbles, starting the update by the quartet corresponding to the four low-order bits of the counter who manages the general date. In this way, the synchronization data field on the general date has two separate fields, a 4-bit field for knowing the rank of the transmitted nibble and another 4-bit field giving the value of the corresponding nibble. In the example, the general date is thus globally updated aftertransmission of 16 successive frames.

  The frame taken as an example finally comprises a data field containing information relating to a cyclic redundancy test, the data contained allowing each processing node to determine whether the received frame is free of transmission errors or, if not, of attempt to restore the integrity of the information contained in the frame.

  In a more general manner, it should be noted that the structure of the information exchanged on the synchronization link advantageously lends itself to the implementation of means making it possible to guarantee the integrity of the transmitted information.

  Thus as regards the indication of the presence of a frame, it is possible to transmit, before the frame, instead of the simple top of synchronization, a determined ternary sequence, based on the states 0, 1 and -1, whose recognition makes it possible to identify the synchronization top with the minimum of error. For example, it is possible to materialize the synchronization top by a sequence of 7 bits and to implement a ternary correlator for the recognition of the synchronization top, a correlator whose output provides a value equal to 7 at the exact instant of the pass the sequence corresponding to the synchronization top.

  Similarly, it is possible with respect to the data frame, to size it so that the binary data is coded using a detector code and error corrector. The size of the coded data then depends on the number of errors that one wishes to be able to detect and the number that one wishes to be able to correct automatically. Thus, for example, it is possible to secure the integrity of the data transmitted in a 64-bit frame by providing for coding each group of 16 data bits according to a code, of the hamming code type of distance 4, making it possible to detect three simultaneous errors. and correct two. Such a code has the effect of extending the size of the frame to 88 bits (i.e.4 times 16 + 6 bits).

Claims (9)

  An electronic system architecture comprising a plurality of processing nodes, each processing node performing a function or part of a function implemented by the system, characterized by: a single synchronization link that delivers to all the nodes of processing a same general synchronization message comprising a simplified header, a base clock, a synchronization top and a reference date, - a plurality of asynchronous data exchange links, each link allowing the exchange of specific data between two particular processing nodes, the data exchanged being accompanied by a generic header containing all the information used by the different processing nodes and relating to the operating step of the system to which the data exchanged relates, Each node processing having own means for interfacing with the connection of synchronization and to perform its internal sequencing from the base clock and the synchronization message; each processing node also having own means for processing the headers associated with the input data received by a data exchange link and for associating the output data generated with the corresponding header.   2. The electronic system architecture of claim 1 wherein each processing node has own means for implementing a local date and for periodically comparing this local date to the reference date.   An electronic system architecture according to one of claims 1 or 2, wherein the synchronization link transmits a timing clock H, a sync top and binary data frames, the sync top and the binary data comprising the frames being transmitted by modulation of the shape factor of the clock H, the modulation keeping the time difference separating the rising edges of the clock H. 4. Architecture according to claim 3, wherein the modulation of the form factor of the clock H comprises 3 states: a state 0 for which the high state of the clock H has a duration equal to that of the state bottom, the form factor being equal to '/ 2, - a state 1 for which the high state of the clock H has a duration greater than that of the low state, the form factor being greater than - a state -1 for which the high state of the clock H has a duration less than that of the low state, the form factor being less than '/ 2, the values of the bit data of the frames being coded on the two states 0 and 1, the synchronization top being identified by the code -1.   5. Architecture according to one of claims 3 to 4 wherein the transmission of a frame of data on the synchronization link is signaled by the transmission of a synchronization top.   6. Architecture according to any one of claims 3 to 5, wherein each processing node performs the synchronization of its slow clocks by means of synchronization top.   An architecture according to any one of claims 3 to 6, wherein the synchronization top is issued periodically, accompanied by a data frame containing useful timing information or a dummy frame containing insignificant data.   8. Architecture according to claim 7, wherein each transmitted frame has a bit field to indicate whether this frame is a real frame or if it is a dummy frame.   9. Architecture according to any one of claims 3 to 8, wherein the synchronization top is transmitted on the synchronization link in the form of a ternary sequence whose structure is chosen so that the detection of the top synchronization by the subsets is: performed by means of a ternary correlator by the detection of a correlation peak.   10. An architecture according to any one of claims 3 to 9, wherein the information contained in the bit frame is coded by means of a detector and error correction code.   Radar system comprising a plurality of subassemblies each subassembly comprising a plurality of processing nodes, each processing node performing a function or part of a function implemented by said radar, characterized in that it comprises: a single synchronization link that delivers to all the processing nodes the same general synchronization message including a simplified header, a base clock, a synchronization top and a reference date.   a plurality of data exchange links, each link allowing the exchange of specific data between two particular processing nodes, the exchanged data being accompanied by a generic header containing all the information used by the different processing nodes; and relating to the burst to which the data exchanged relates; each processing node having own means for interfacing with the synchronization link and for performing its internal sequencing from the base clock and the synchronization message; each processing node also having own means for processing the headers associated with the input data received by a data exchange link and for associating the output data generated with the corresponding header.