JPH06508460A - How to generate a common time reference for systems with distributed computing units - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] Generating a common time reference for systems with distributed computing units How to The claimed invention communicates with each other by serial data transfer via a data bus, each have their own clock generation system and are timed to each other via the data bus. exchange information about how computing units operate .

Distributed computing nodes or computing units Data processing configurations connected via data buses are primarily used in industrial and automotive applications. used for local networks. Computing nodes are , consisting of data processing equipment or signal processing equipment for a specific application. . It also includes data transfer between the control device, sensors and regulating devices. That's it Such local networks typically operate under real-time conditions. sand i.e., computer operations and control interventions can be programmed within a specific time window. processes to ensure data is available in a timely manner. do not have. By doing so, the relevant process parameters can be to ensure on-time transfer to the processing node and optimal process operation. It is necessary to For local networks, this is the computer node means that it provides sufficient processing capacity.

In principle, data transfer is performed using the so-called "broadcast communication" system and "point-to-point" Systems are differentiated. In a "broadcast" system, a node transmits data via a bus. and all other nodes can wait to receive. With “point-to-point” connection connects only two nodes to each other in the way defined in each case is possible, and this one node is called the called node.

An example of a data processing system defined above is a system that represents a de facto standard protocol. controller area network (CAN). CAM is a participating node or An open system that can be expanded at any time to add additional computing nodes Open System Interconnection (OSI) standard model recommended as a Based on. In CAN, physical level, transfer level, object level, etc. Which various levels are distinguished.

The physical level essentially defines how signals are transferred. The parameters involved are transmission medium, signal level and timing, transfer rate, and bit representation. . The transfer level includes the kernel of the CAN protocol. The transfer level is It forms the connection link between the virtual and physical levels, and provides bit timing, optical object or message frame definition, transfer acknowledgment, error recognition and and error handling, as well as arbitration, i.e., when multiple computer nodes Responsible for resolving conflicts when accessing services. Finally, at the object level Filtering of objects or messages and top-level applications responsible for processing objects related to the application level. by this, Objects or messages are connected to the bus and the corresponding physical level, transfer level , and object level to multiple application levels and Transmitted in both directions between. The physical level and object level are can vary depending on the particular implementation. However, the transfer level is defined by the protocol.

The basic concept of CAM is common for all nodes in the data processing configuration defined above. The network is based on the concept of virtual data memory, and its structure Through this, we undertake the task of realizing the data in this common virtual memory. protocol data sent from the computing node over the network according to the The data is assembled into a message or object that corresponds to the data frame. form a ct. Within a data frame, for example, the beginning and end of the frame Various fields are defined that indicate the field and define the actual data fields.

Each object defined by a data frame has a frame start bit of Afterwards, define the data name and priority of the object, whether it is dominant or recessive, Contains a header (identifier). Object priority determines bus access. It is determined. Therefore, the value of the header is determined by collision-free arbitration in the network. has a decisive influence on This means that some messages have higher priority than other messages. There is a sage. According to the CAN protocol, the length of the header is 11 bits.

Messages or objects transmitted over the bus are managed within the code. The transmission itself is a broadband transmission. i.e. for a particular receiver node. The code is never addressed. Therefore, the objects transmitted over the bus The object first arrives at all potential recipients. Next, the recipient is the information flow You must decide which messages or objects to receive from stomach. Filtering of important objects from the information flow is done by This is done based on the header.

This is done at the object level and operates according to multiple optimization steps. The control structure for filtering and evaluating priority is known. It is. An example of a so-called complete CAM implementation can be found in the German patent specification Known from No. 3506118. The object level implemented in the patent is , independent of the attached central processing unit (CPU) of the computing node performs management of all data or objects in the network. basic In a CAN implementation, the entire object management is transferred to the electronic control unit. handled by an included central processing unit.

Such a data processing configuration with distributed computing nodes is For example, it is used in real-time systems such as control and regulation systems.

The various elements of the adjustment circuit, such as the target value providing mechanism, actual value providing mechanism, adjustment mechanism, and control members, , forming a distributed component of a data processing configuration or a distributed computing node do. An example of a device that provides actual values is a sensor, and an example of a control member is a controlled system. There is an actuator that acts on the The essential data transferred through the control circuit is the data - Transferred to the data processing arrangement via the bus. Thereby, each computer The receiving node stores the received information assigned to it. For example, control circuit The dynamic behavior of, for example, the sampling rate and the sensor signals and actuators Affected by signal delay time. Delay in sensor or actuator signals appears as wasted time in dynamic behavior, but this is basically ignored in control circuits. Undesirable. As a result of wasted time, for example, the entire control process becomes unstable, Or at least its control performance is impaired. However, such dynamic control behavior Dynamic damage is often determined by controlling the control parameters by taking into account the dead time that occurs. This can be compensated for by adapting the meter.

In a centralized system, the control circuitry is concentrated in one unit and the processing computer The using unit operates according to a single time reference. Therefore, the delay Time and wasted time can be determined and taken into account when processing control algorithms. can. However, in data processing configurations with distributed systems of the type mentioned at the beginning, , each individual component generally has its own clock supply mechanism. Each clock provided A power supply is typically non-uniform with respect to other components or computing nodes. Contains an oscillator that operates periodically. For example, the high Adequate electromagnetic tolerance since using frequency clock signals results in electromagnetic emissions to the surroundings removed to ensure high performance or to reduce cabling costs.

Therefore, in a distributed data processing configuration, each computing node's Because the local time references are different, the receiving node can, for example, There is a problem in that it is not possible to determine the amount of time it will take from the time the data is sent to the time it is processed. This problem is The computing node that generates the signal is responsible for not only the signal value, but also for the signal generation. This occurs even when transmitting intermediate values. Because the receiving node has the time to be transmitted This is because the time reference of the value is not known.

IEEE Transactions of Computers, Vol. C-36, No. 8. Published August 1987, pages 933-940 contain the time Five different causes of error are listed. These causes are as follows .

a) The individual clocks are read at the sending node before the synchronization information is sent. Time elapsed at b) The access time from the node to the bus (arbitration time) is uncertain.C) The sending side node transfer time on the board, transfer time on the bus, and transfer time on the receiving node. d) the time from when the synchronization information is received until the moment of reception is recorded or stored; lapse of time e) Granulation time. In other words, the time information required to resolve the raster time.

The above paper transfers synchronization information from one node to other nodes via a bus. It does not describe the method and how this information is recorded. But the above mentioned error The accuracy that a global time system can achieve when assuming . The result As a result, the local time at each node is immediately corrected based on the correction value to be calculated. It has been confirmed that continuous correction can be performed during the next synchronization period.

In other words, the linear increase curve representing the dependence of count value on time is If you take into account the mask time required, it may rise suddenly or the slope may be twisted. Therefore, the mask time and local time curves coincide later.

One can think of ways to establish a common time reference for all nodes. first A possible method is to read the sending node's clock by making it a synchronous call. It is to simultaneously transmit a message to another node. This method is very complicated It is. After the start pulse, we must first wait and check if transmission is possible. . Even if possible, the obtained time values must be corrected. second possibility The method sets the clock at the sending node to zero and transmits this to other nodes. It is necessary to notify at the same time. This second method uses the sending node's internal clock This is problematic because it is uncertain whether the bus will be free at the time set.

To resolve this issue, you must give timing messages the highest priority. Must be.

Typically, a common time signal carried on a bus separates the There is no way to synchronize the nodes. This is because this method is effective against disturbances on the bus. with the problem that the computing node may consider it to be a synchronization signal. It's your body.

Therefore, in distributed data processing configurations of the type mentioned at the beginning, time synchronization functionality is not offered. Not provided. For example, in known CAM networks, CAN controllers The controller implementation does not transfer time information out of the transport level.

A computing node, essentially a central processing unit (CPU), is , if the message is sent from the node, the transmission frame is set at the object level. give only the command. Object level is the priority given to messages The transmission is performed autonomously according to the following. In this case, the exact moment the message is transmitted is No information is provided to the computing unit at that moment. Sunawa The computing unit determines in advance when the message can be sent. I don't know. The same applies when receiving messages from the CAN bus. Ru.

In this case, the CAN controller uses the receiving program according to the provided filtering. The entire process is executed autonomously and the central computer of the associated computing nodes The peering unit is not given any information about the exact reception time. However, Therefore, the central processing unit of such a computing node stores time information as It can also be received directly from the CAN controller via the output signal or Nor do you receive through the level.

Therefore, it is an object of the present invention to communicate with each other, each growing its own clock generation system and creating a clock generation system for the system. Exchange timing information over a data bus to generate an accurate time reference , a method for operating a computing unit.

This object is achieved by the features of claim 1.

The present invention purports to enable the introduction of a global time reference into distributed data processing configurations. Cultivate advantages. In the present invention, the protocol provided for object exchange is therefore no central clock supply mechanism is required! It is. Support for the global time system This method takes into account the time delay between when the data signal is generated and when it is processed. It becomes possible to Therefore, wasted time can be taken into account, even if For example, the dynamic behavior of the control circuit can be improved. With the help of a global time reference, A computing node that generates a signal may interpret a time value as an information value or a signal value. can be transferred together. This time value changes from the local time reference to the global time reference. is converted to A receiving computing node that processes the information signal forwards Convert the time value to the local time reference of the node, thereby generating the signal. The total delay from when an item is created to when it is processed can be determined. distributed data processing The advantage of being able to scale your configuration to almost any degree can be maintained without significant expense. be done. The method according to the present invention applies to "broadcast" systems and "point-to-point" systems. It is convenient because it can be used for both.

The main advantage of the invention is that the clocks of individual nodes are The point is that it is not synchronized to the clock of a certain node by setting. but, Instead, each node calculates the time at another node by transforming Can be done. In the present invention, it is necessary to correct the local time continuously or discontinuously. is eliminated, and each node can calculate the real time of another node. Therefore, multiple If there are several /- codes whose internal clocks have different stability, the time rate of change is linear or near-linear time variation, with wide range of synchronization even when provides a stable clock. Clock voltage caused by differences in time changes Differences in counter values can be easily resolved by conversion.

Embodiments of the invention are characterized in the dependent claims.

Next, the present invention will be explained with reference to the drawings.

FIG. 1 is a circuit diagram showing the principle of a data processing configuration according to the present invention.

FIG. 2 is a diagram that helps explain timing relationships.

FIG. 3 is a diagram that helps explain the conversion equation.

As shown in FIG. 1, the data processing configuration consists of a CAN unit CER, CEI or Contains CR2. Although each CAN unit represents only an exemplary configuration, It is implemented as a basic CAN unit. In other words, computing Along with the CAN controllers CANR and CNI to CN4, C,”, N1 to CAN4 are provided.The CAM controller accepts Handles filtering and input buffering. Object management and apps Recall/yon is performed by a computing node. individual CA The N units are interconnected by data bus B. Individual CAM Messages or objects exchanged between units are A node sends data through its associated object level and transport level. Arrive at Ta Bus B. Conversely, objects are transferred from the data bus. reach the computing node through the transport level and the object level. wear it. According to the CAN structure, there is also a physical level between the transfer level and data bus B. It has been inserted.

The CAN system considered here operates according to a "broadcast" scheme.

However, the claimed invention described below can also be implemented in a "point-to-point" system. be.

According to the present invention, each node uses its own clock time to clock other nodes' clocks. It can be compared with time. For this purpose, one node sends a synchronous message to the bus. supply the ji. This synchronous message is a bit frame like any other message. It consists of The bit indicates that the message is a synchronous message. including indications.

Synchronous messages sent by a node with or without priority arrive on the bus. All other nodes then receive this synchronization message. All other no When the message is received, the card records the instantaneous counter value. No Synchronous messages sent from a node can include the sending node's counter value. However, it is not necessary to do so. This message is a request to record time. That's fine.

Send a synchronous message, i.e., when a request to read the time is provided. Make sure that each node, including side nodes, reads the time as simultaneously and reliably as possible. Must be.

One possible way to ensure maximum accurate simultaneity is to The first step is to record the counter value at the moment of the starting edge of the page. this moment In between, any receiving node determines whether the received message is a synchronization request or not. I don't know. That can only be determined when the entire message is received.

That is, it is possible to check from the entire message whether it was a synchronization request or not. I can do it. The received message is not related to the synchronization request. If the evaluation indicates that the recorded counter value is deleted again. But sync If the request is indicated as a request, the recorded counter value is retained and the receiver A node can later supply this value to the bus.

After the storage process, the individual nodes transmit the recorded counter values onto the bus and each Nodes become aware of other nodes' internal clocks.

For the most accurate simultaneity, it is not necessary to There is no need to synchronize with Tsuji. Recording the counter value of the node's internal clock Other points in the frame are also possible and can be used. Ru.

For long frames, the CAN protocol requires that after the starting edge of the frame the transmission Repeating the synchronization with a mask, i.e. through triggering by hardware means. I can do it. That is, at the edge of a given bit within a frame, the receiving computer The scanning raster of the scanning node is adapted to the mask of the frame being received.

This is called "hard synchronization." This hard synchronization is required because the sender clock Due to differences in the behavior of the clock and receiver clock, the scan mask may deviates so much from the mask that it is impossible to identify the bits, i.e. to evaluate the frame. This is to avoid becoming Therefore, when the receiving node's local bit The interval counter is newly counted between the starting edge of the received message and the hard synchronization point. start counting. Apart from this, each node's internal clock continues to operate and Do not stop. However, the counter value of the node's internal clock is recorded at the moment of the start edge of the synchronization, or at the moment of hard synchronization.

In each case, the counter value of the node's internal clock at a given moment is recorded. will be recorded. In particular, in the case of CAN, only then the entire message is evaluated and the error error is corrected. Only then can you check if the message is a synchronization request. can be approved.

When the operation of the internal clock is guaranteed to be uniform, the operation of the internal clock is not uniform, but a time comparison at a single point in time is sufficient. individual nodes The counter speeds of the counters representing the internal clock of In some cases, transmission of the counter value at two consecutive points in time is sufficient. counter If the value does not depend linearly on time, more than two time points are required for validation.

According to the present invention, each node can calculate the real time of other nodes.

Therefore, there is no need to perform either continuous or discontinuous correction of local time. this Therefore, if there are multiple nodes with different internal clock stability, Higher degree of synchronization with more stable clocks with linear or near-linear time variations provided. Even when different rates of change exist, the difference in time change, i.e. In other words, differences in clock/counter values caused by nonlinear characteristics can be easily converted. can be resolved.

If you have multiple nodes, make one or at least a few nodes the master node. , the node with the most stable clock.

This way, all nodes post their counter values to the bus in response to synchronization requests. This eliminates the need and avoids overloading the bus with information. and For example, if you select one node as the master node in response to a synchronization command, If so, only this mask node can communicate with other nodes across the bus in response to a synchronization request. All you have to do is notify the counter value of can be associated. Other nodes can interact with each other after converting the mask time appropriately. will be able to communicate with

However, in response to a synchronization request, all nodes first check their counter readings. If the bus should be notified and thereby other nodes, this notification should be properly Which node has the most stable clock using the provided protocol carried out to determine. Each node then determines which node is the master node. We can agree on what to choose. Thereafter, the tag according to the claimed invention Timing information can be relativeized.

According to the present invention, errors a), b), C ) can be avoided. This makes it difficult to measure time and therefore the global time reference. This greatly improves the accuracy when

In this way, in the present invention, immediately after a synchronization request, that is, within the same frame, Nodes need to advertise their own time or stop or adjust their local clock. There's no need to adjust it. According to the claimed invention, only a request to measure time is first issued. , then it is determined whether the request actually requires a time measurement.

If the time measurement is not correct, the measured time is deleted again. measurement The measured time is also stored for future information gaps. is sent to other nodes on the fly.

To generate a common global time reference according to Figure 1, one computing node node and its local time reference to the reference computing node CNR and its local time reference. The school computing CNR via its CAN controller CANR Access the bus. Time synchronization sent from the reference computing node timing information is transmitted to the receiving node if the timing information is included in the transmission protocol. Therefore, it is recognized. via the bus based on whether the transmission protocol can be identified. can recognize whether the transmitted object is valid or not. Receiving side The code first examines the incoming object in the forwarding plane and determines whether the object is Determine whether it is valid. Then, at the object level, this object It is checked whether timing information is included.

Information about the exact moment an object is sent or received is outside the transport level. , to the connected computing nodes. imp Rementation takes place within the CAN controller. Transmission level transmission information or or received information via the object level or in the CAN controller. A computing node is provided via an output signal.

For generating a global time reference, i.e., synchronizing computing nodes. , a computing node can be masked or This reference computing node is selected after you select it as the reference computing node. ing node CNR sends a synchronization object and the synchronization object is received by all other computing nodes approximately simultaneously. . That is, the reception of the object is sufficient at all network nodes. can be considered to be simultaneous with sufficient accuracy. Sending synchronized objects is Determined by the forwarding level of the target computing node and the synchronization object's Reception is determined by the forwarding level of other computing nodes. all rotations The transmission level delivers messages to its associated computing nodes. . When each floating node receives a message from the forwarding level, , remember its local time. i.e. the reference computing node is also Remember that local time. Then, on the reference computing node, It is determined when the initialized object was actually sent. all other compilations The using node records the time the synchronization object arrived. reference The computing node can later share the reference time it remembers with other computers. Notify peering nodes. This reference time is sent by the synchronization object. It's time to believe. Other computing nodes similarly Notify other nodes of the current time. Therefore, each computing A node calculates its own time while referring to a reference time or the time of other nodes. can do.

Timing information from the transfer level can be modified by information from the object level. It should be something that can be displayed. That is, the timing information from the transfer level is transferred to the synchronization a specific object characterized as an object or a time-comparison message can be linked with. For example, if the timing information is is sent directly from the transport level to the computing node via the output signal of , the CAM controller is unqualified with each object that passes on the bus. or modified with information from the object level, For example, by a specific object characterized by a header or an identifier. generates a synchronization signal that is only generated when These features of CAN controller The moment the page or synchronization information is published to the associated computing node. have a predetermined timing relationship with respect to the start bit of the object frame. do. Therefore, the reference computing node object or synchronization The start bit of the converted object is added to the object frame according to the protocol. It takes into account that all included bits are reference bits to be synchronized. It can be done. That is, the bit edge of the start bit is This is the most accurate edge. Therefore, the start bit of the object frame is First, it provides an unqualified time signal that can be modified by the contents of the object. do. Therefore, at the forwarding level, the message is verified to be valid and At the object level, it is confirmed that it is a time object. death Therefore, each time an object arrives, the receiving computing node can determine the local time and remember it.

After the object is identified, the object is If it is determined that the supplied timing information is not included, the memory of the receiving node The current local time will be deleted. But it contains timing information If this is confirmed, the deletion mechanism will not work and the local time will still be remembered. It can be done. By announcing the reference time or the time of other nodes, A local time time relationship is established.

In principle, if the clock is stable, each computing node requires All it does is recognize its own time and a reference time or any other time. be. From its knowledge of these times, each computing node It is possible to extrapolate future local time relative to time. This feature This also applies to de. Therefore, each computing node transmits to the bus object with timing information calculated based on the reference time. can be provided. Therefore, both the sending node and the connected receiving node This timing information can be used for the same event, i.e. reference computing node. can be associated with the reference time of the code. The prerequisite is that all messages or is that the object appears at all network nodes almost simultaneously .

The time reference of a CAN computing node typically oscillates at a constant frequency. It is equipped with an oscillator and a counter. Since the frequencies of the individual oscillators are usually different from each other , the computer time standards are also different. Message or synchronization object Synchronization occurs after a synchronization signal is sent from the transport level to the computing node. Once a point has been established, each computing node can Using knowledge of the frequency of the and the relationship with the local time of other nodes. Other computing If you do not know the frequency of the time reference for each In order to be able to determine the local time reference for the must be executed consecutively. Thereby, in a first approximation, the computer If the counter value according to the time reference of the tying node forms a linear function of time, then is assumed. The nonlinear characteristics of the time base of individual computing nodes are This is taken into account by periodically repeating the periodization process.

In the preferred configuration of computing nodes shown in Figure 1, each node capture registers TR and T1 to T4 and capture registers CRR and Equipped with CRI or CR4. As shown in the symbol, the CAN controller , upon receiving an object from bus B, the associated capture register CRi , a command is given to cause the associated clock generation mechanism Ti to hold the instantaneous time. supply The holding action is triggered by the received object frame start pulse. will be carried out. After evaluating the object frame, this frame contains timing information. , the determined local time of the clock generation mechanism Ti The local time will remain remembered, otherwise the local time will be deleted.

Bus uptime is assumed to play no role and all network nodes Since the objects on the bus exist almost simultaneously in each computing Issuing or transferring a command to record time by the node's CAN controller sending appropriate messages from the transmission level to the computing node, and Subsequent memory of local time on all computing nodes is occur at about the same time.

Figure 2 shows the cost-effectiveness of the computing node CNH, taking into account the global time reference. From computing node CAN1 to computing node CAN2 It explains the time relationship when sending objects. With time reference TR A computing node CANI with a time reference TI with a difference of ΔT1 is , generate a sensor scanning signal at time A. This scanning signal takes priority into account. is applied to the bus at time S and received by CAN units 1-CF2 at time E. .

The CAM unit CE2 has a time reference T2 whose difference from the reference time TR is ΔT2. do. Objects received at time E are processed at time V. therefore, The time difference between the generation of the signal at time A and the processing of the signal information at time V is D. Become. The present invention allows receiving computing nodes to use a common time reference. Based on the reception time of the object and its processing as well as this computer The sending node has informed the receiving node of the point in time A of the transmission relative to a common time reference. In some cases, the point of transmission of the object at the computing node is also It becomes possible to recognize the So, for example, in a periodic process, the total delay time can be taken into account.

Figure 3 shows two linear time dependent clocks on two computing units. An example of a conversion equation in a certain case is shown. Node M and the clock time to the node Take a pause. Clock time can be described by an unmodified equation.

However, Δ is the slope, t is the time, and 0 is the start value or offset.

UM (t) = ΔM - t + OM UK (D = ΔK - t + OK The computing node has an assumed time t, and 2 at 1+-1 After selecting one time reading, the stored clock value UK(t .

) and UK(t+-+) and the transmitted clock values UM(t+) and UM( t, −+) can be used to calculate the following transformation equation.

UM (t+) -UM (t+-+) OK, M = UK (L-+) AK, M (t+-+-t.) corresponding to computing node M Apply the equation. In this equation, the node has a stored point in time and a transmitted Using the time points given, we can calculate the following equation:

UK (t+) UK (t+-+) OM, K = UM (t,,) - Δ M, K (t + - + to) From the above equations and Figure 3, each involved component is It is clear that the computing node needs to remember the time value when reading the time. It is white.

Refer to the qualified timing information for instructions on actually generating the global time reference. I will explain. Reference computing information in transmitted objects The timing information for the code includes the header (wXX Miko) in the object frame. be. A predetermined number is assigned to the header. The header is a time comparison message Notify the receiving node that the Messages can be prioritized It is Noh. This configuration applies commonly to all computing nodes. Ru. Connected by the first object from the reference computing node It may be advantageous to activate the synchronization mechanism of the computed node. each After receiving the first time object, the computing node A synchronization mechanism responds to each received object initiation pulse. Each controller When the peering node receives the second object, the starting pulse is Know that it is time for periodization to occur. This allows you to capture, e.g. The deletion mechanism in the register becomes inoperable and therefore new local time values are not recorded. Not done. by a third object from the reference computing node , the sender reference time of this node is communicated.

With the help of the present invention, timing information about objects can be distributed to all nodes. Introduces a global time reference into distributed data processing configurations for processing in a manner that extends to becomes possible. The synchronization mechanism required for this is provided by the data transfer protocol. It is determined that This type of implementer silane has made it possible to transfer timing information over a serial data bus without the need for a central clocking mechanism. can be obtained directly. Distributed data processing configurations can be scaled up without problems. The benefits are maintained.

SA '185 Continuation of front page (51) Int, C1,5 identification code Office reference number HO4L 12/40 I

Claims (13)

[Claims] 1. They communicate with each other by serial data transfer over a data bus, and each has an internal clock generation system and exchanges timing information via a data bus A method of operating a computing unit, comprising: a) time alignment; a) a start signal is provided to the data bus; b) each computing unit: When the time alignment start signal is identified, it stores its own time value; c) Each computing unit then, at some later point in time, d) transfers its own time value to other computing units; and d) time received by a computing unit from other computing units. remember the value, e) Each computing unit records its own time value that it remembers. It compares the time values of other computing units it has stored and Calculate the current time value of other computing units taking into account the current value A method characterized by: 2. a) Two or more timing start signals are applied to the data bus at consecutive times. Re, b) When each computing unit identifies the time alignment start signal, responds remembers its own time value to c) Each computing unit then, at some later point in time, d) transfers its own time value to other computing units; and d) time received by a computing unit from other computing units. remember the value, e) Each computing unit records its own time value that it remembers. It compares the time values of other computing units it has stored and Calculate current time values of other computing units considering time-in-time values A method according to claim 1, characterized in that. 3. One or more timing start signals in one computing unit Claim 1 or Claim 2, characterized in that the data bus is supplied by The method described in. 4. Characterized by using the characteristic part of the data series as a time alignment start signal The method according to claim 2 or claim 3, wherein 5. Verification data regarding the time adjustment start signal is data including the time adjustment start signal. 5. The method according to claim 4, characterized in that it is provided with a series. 6. The computing unit receives verification data belonging to the timing start signal. If not recognized, each computing unit will have a Delete again its own time value that was stored when the timing start signal appeared. 6. A method according to claim 5, characterized in that: 7. Use the first pulse edge of the data series as the timing start signal. 5. The method according to claim 4, characterized in that: 8. When the data series is long, the data The receiving computing unit uses the synchronization signals included in the series. Synchronize if hardware synchronization of the pulse raster to be generated is provided. The method according to claim 4, characterized in that the signal is used as a time adjustment start signal. Law. 9. communicate with each other via a serial data bus and have their own local time reference. and the objects arriving on the data bus via the object level and transport level. objects arriving from the data bus and vice versa. a data processing configuration with distributed computing units that replace , after receiving the object, from the transport level, the associated computing node A means is provided to forward the message to the A synchronization mechanism is included to synchronize the local time reference with the local time reference. Characteristic data processing configuration. 10. The synchronization mechanism stores local time (TR, T1 to T4). Claim 9, characterized in that it has a storage mechanism (CRR, CR1 to CR4). Data processing configuration described in . 11. If the synchronization configuration includes capture registers (CRR, CR1 through CR4) The data processing arrangement according to claim 9 or 10, characterized in that: 12. Each computing node has a counter as a local time reference. Any one of claims 9 to 11, characterized in that the oscillator comprises an oscillator with Data processing configuration as described. 13. The message transport means can change the transport level directly or at the object level. connected to the associated computing node via a , the data processing configuration according to any one of claims 9 to 12.