Station-to-station full duplex communication in a communications network

Background of the Invention

1. Field of the invention

The present invention relates in general to the field of communication networks, in particular networks which have both half-duplex communication channels and full-duplex point-to-point communication channels.

Second Discussion of the state of the art

A dramatic increase in the use of small computers and workstations has led to an increased need for local area networks (LANs) that are able to serve dozens or hundreds of stations. A typical LAN contains a number of stations that are interconnected by a series of communication links. (The terms "link" and "channel" are used interchangeably below.) A station may consist of a personal computer or PC, a workstation, a bridge, or a number of other information processing or storage devices. Some of the stations may be "half-duplex stations", which means that they are able to either transmit or receive information at a given time, but not both at the same time. Other stations may be "full duplex stations" that are capable of transmitting, receiving, or doing both at a time. Similarly, some of the communications links may be "half-duplex links" capable of supporting only half-duplex communications, while other links may be "full-duplex links" capable of supporting both half-duplex and full-duplex communications. Any two stations can be effectively connected via a communication "path" that is (1) half-duplex along its entire length (made up entirely of half-duplex links), (2) partially half-duplex and partially full-duplex, or (3) full-duplex along its entire length ,

The main function of a LAN is to create a communication channel or a communication link, via which a certain station can possibly communicate with one or more other stations from a remote location. A specific connection within a LAN is characterized by a protocol that defines when and how stations can send and receive information over this connection. Each station using this connection must operate in accordance with the protocol to communicate effectively with other stations and to avoid interference with other stations. A typical protocol thus establishes rules or conditions that relate to the time sequence or timing, data formats and the like.

An example of a protocol that is widely used in business environments is defined in ISO 8802-3: 1989 (E) and is sometimes referred to as "Ethernet". The protocol specified in ISO 8802-3: 1989 (E) is an example of a type of protocol known as "multiple access by carrier detection with collision detection" or "CSMA / CD" (carrier sense multiple access with collision detection). As explained below, CSMA / CD is a "competition" protocol in which a plurality of stations compete for and share a single communication channel through which information is transmitted using a "packet switching" method.

According to the CSMA / CD protocol specified in ISO 8802-3: 1989 (E), each station must first check the channel before sending information through a communication channel to determine whether any other station is transmitting at the same time. If no other station is currently transmitting, the station that checked the communication channel (taking further requirements into account) may send its information. However, if another station is already transmitting, the first station must wait for a period of time and check the channel again. Thus, a mandatory requirement of the CSMA / CD protocol is that only one station may legitimately send information at any given time.

The requirement that each station must check for activity on the communication channel before sending is an aspect of operational control or "scheduling" created by the CSMA / CD protocol. It is this operational control that enables numerous stations to communicate effectively by sharing a single channel.

However, the scheduling created by the CSMA / CD protocol represents a drain on resources or performance within the LAN. While scheduling allows a number of stations to share a single communication channel over a period of time, this becomes costly reduced bandwidth since only one station is allowed to send information at any given time. More specifically, according to the CSMA / CD protocol, only one station can legitimately send information at a given time (even if the sending and receiving stations and the connection that connects them can all perform full duplex communication), compared to full duplex communication In which a station can send and receive information at the same time, only half of the amount of information is transmitted per unit of time.

A major disadvantage of the CSMA / CD protocol is thus the mandatory requirement of half-duplex communication, which effectively limits the bandwidth of a communication link to half the maximum possible bandwidth between two full-duplex stations that are connected by a full-duplex link.

Even if a station has checked the channel correctly before transmission and has determined that no other station is currently transmitting on the channel, a "collision" of two transmissions can nevertheless occur in certain situations. As an example, assume that a station has just started broadcasting, but the broadcast information has not yet reached its intended destination due to the delay in propagation. At the same time, a second station, ready to transmit, checks the channel unaware that the previous (or other) broadcast is still in progress and determines that no other station is transmitting and is beginning to transmit its own information , The information sent by the second station can then "collide" with the information sent by the first station. As a result of the collision, it may happen that both messages sent do not reach their desired destinations.

To eliminate the consequences of collisions, ISO 8802-3: 1989 (E) provides a "collision detection signal" that accompanies each shipment. Each station monitors the collision detection signal, and if a collision occurs, the affected stations try the broadcast again until they are either successful or a maximum number of allowed unsuccessful attempts is reached.

For the collision detection method to work correctly, all sending stations must always have enough time to detect the occurrence of a collision. As a result, the physical size of the LAN must be limited so that the propagation delay between any two stations does not exceed a defined maximum propagation delay. Compliance with the maximum propagation delay time guarantees that a given broadcast will either end successfully or a collision will be detected within a limited period of time before the broadcast ends.

In order to comply with the maximum propagation delay required by the CSMA / CD protocol, the maximum physical distance between any two stations must be limited. A connection using the CSMA / CD protocol is usually limited to a maximum distance of 2.8 to 4.5 km. Such a short maximum distance between two stations often prevents the use of such connections in applications in which remote stations are separated from one another by relatively large distances.

Thus, another major disadvantage of the CSMA / CD protocol is the requirement for maximum propagation delay, which effectively limits the physical distance between stations and the distance covered by the LAN. US-4 288 868 teaches a satellite communication system for voice signals and data signals, in which half-duplex data traffic is carried out simultaneously with, but separate from, full-duplex voice traffic between two ground stations. A second receiver enables all ground-based stations to follow both data traffic and voice traffic.

Summary of the invention

In brief, the present invention provides for the establishment and maintenance of full duplex communication between two full duplex stations connected by a point-to-point full duplex link. The stations can be parts of a LAN or an extended LAN that has both full-duplex and half-duplex stations as well as full-duplex and half-duplex connections. The stations may also be interconnected by other types of communications, such as a wide area network or wide area network (WAN), satellite link, and the like.

In general, full duplex communication can be established between any two stations, provided that the communication link (s) connecting the stations is / are able to support full duplex communication. In cases where some stations or one of the links connecting them is unable to support full duplex communication, such stations can operate in a conventional manner using half duplex communication.

The invention in its general form relates, according to claims 1 and 7, to a method and a device for establishing full duplex communication and for determining whether a communication link between first and second stations is available for full duplex communication.

The present invention works to automatically establish and maintain full duplex communication in a manner transparent to the user. Once the full duplex communication of the invention is established, it effectively doubles the bandwidth of the connection connecting the two stations.

In addition, since two stations that are in full-duplex communication through a full-duplex link cannot experience "collision" of broadcasts by definition, the present invention also eliminates the need for collision detection and for normally considering one of a CSMA / CD protocol required maximum propagation delay. As a result, the physical distance between the participating stations can advantageously be increased.

The present invention can be carried out in individual stations connected by a communication link. Initially, a station employing the invention can communicate in a conventional manner (ie, according to the CSMA / CD protocol) over the connection using half-duplex communication. To begin the process of initiating full duplex communication, which can be in response to instructions stored in a station or upon receipt of a specified instruction from an external source, the station sends a request message over the connection using half duplex communication. The request message serves as an indicator that the station is ready for full duplex communication. The request message also serves as a test to determine whether the point-to-point connection between the station sending the request message and a second station receiving it is able to support full duplex communication between the stations.

When a second station receives a valid request message indicating that the connection is capable of supporting full duplex communication and the second station is ready for full duplex communication, the second station sends a response message over the connection to the first station using half-duplex communication to indicate its readiness. Like the request message, the answer message also serves as a test to determine whether the point-to-point communication link between the first and second stations is capable of supporting full duplex communication. After the first station has received a valid reply message, full duplex communication can begin between the two stations.

Once full duplex communication begins, the two participating stations constantly monitor the point-to-point communication link to ensure that both stations and the link continue to function properly to maintain full duplex communication. In the event of failure of one of the two stations or another interruption of the connection, the station returns to half-duplex communication, which allows communication to continue through the connections in a conventional manner.

Brief description of the drawings

This invention is set out in detail in the accompanying claims. The above and other advantages of this invention will be apparent from the following description taken in conjunction with the drawings.

1 is a diagram of an extended local area network;

Figure 2 is a diagram of two bridge stations shown in Figure 1;

3A-3C are flow diagrams of a method for establishing and maintaining point-to-point full duplex communication between two stations in accordance with the present invention; and

Fig. 4 is an overview indicating the contents of the messages of the method shown in Figs. 3A-3C.

Detailed description of an exemplary embodiment overview

The logical topology of an Ethernet LAN that works according to the protocol specified in ISO 8802-3: 1989 (E) is a bus (a shared medium) that connects two or more stations. The stations communicate by exchanging (sending and receiving) a basic unit of information called a frame. The scheduling of broadcasting on the shared medium is controlled by the CSMA / CD access method to distributed media.

The basic requirement for performing full duplex communication in an Ethernet LAN is that the topology be point-to-point. This implies that there is a point-to-point connection that includes a pair of separate, bidirectional duplex communication paths. Existing Ethernet interfaces are used during operation in full-duplex Ethernet mode, and all services at the data connection customer interface are provided transparently. The standard Ethernet frame and packet formats are used without modification. (A packet consists of a preamble, a frame start separator (SFD), a frame (which contains data) and a broadcast end separator (ETD).) While operating in full duplex Ethernet mode, the CSMA / CD media access protocol is like this Simplifies that it ignores carrier detection and collision when sending and receiving.

It is possible to misconfigure a network in such a way that a full duplex link interface is connected to a half duplex ethernet link interface. Although this connection is physically possible due to connector and media compatibility, it cannot be allowed to start full duplex operation because the protocols for each connection interface are significantly different. The main difference lies in the fact that the half-duplex Ethernet protocol requires the incoming carrier to be monitored before transmission and the collision signal to be monitored during transmission for correct network operation. The full duplex link interface does not require carrier monitoring since the physical channel, unlike a shared bus in the case of the CSMA / CD half duplex link, is a full duplex point-to-point link.

For full duplex connections, a deterministic and self-configurable verification procedure is necessary to ensure that the physical channel is full duplex point-to-point and that both ends of a point-to-point connection are capable of full duplex operation. The incorrectly configured case of a Volduplexinterface connected to a half duplex interface would lead to an interruption of the network connected to the Halfduplexinterface. The interrupt occurs because the full duplex transmitter can transmit whenever a packet is to be sent, regardless of the state of the incoming carrier. When this packet is sent to the half-duplex link, a carrier can be generated at any time. This violates the CSMA / CD access protocol and results in a later collision, packet loss, problems with inter-packet gap shrinkage, etc.

The verification method created by the present invention initializes itself into a half-duplex Ethernet mode until it determines that both ends of the point-to-point connection are full-duplex interfaces. Performing the verification process in half-duplex Ethernet mode prevents network interruption if the connections are configured incorrectly. As soon as the verification process is successfully completed, both connection interfaces switch to full duplex operation. The link is periodically tested during full duplex operation to ensure that the link is still working. This creates stability for the verification process. An error during the verification process prevents a connection interface from operating in full duplex mode, thereby preventing a possible network interruption.

1 shows an expanded local area network 1, which comprises three local area networks (LANs) 2a, 2b and 2c. Each LAN 2 can contain one or more media segments 4, network terminations 6, stations 5, repeaters R and bridge stations B. Each media segment 4 represents a physical communication link, and all segments 4 combine the stations 5, the repeaters R and the bridge stations B with each other. A network termination 6 is connected at each point where a media segment 4 physically ends.

For example, each station S may represent a personal computer, a workstation, or other information processing or storage device.

Each repeater R is a device that connects one media segment 4 to another and buffers all information received from one segment and sends it to another element (“repeated”). As required by the CSMA / CD protocol, each repeater R operates continuously using half-duplex communication and is the only type of device that can connect two media segments. The CSMA / CD protocol also requires that when repeater R receives a "fragment" of information that contains less than a specified minimum number of bits, it must automatically append additional bits to the fragment to make it the required minimum Increase size before sending (repeated) the information. More specifically, according to the protocol specified in ISO 8802-3: 1989 (E), a repeater must expand all fragments whose length is less than 96 bits to a minimum length of 96 bits.

Each bridge station B represents, for example, a special station type that functions like a station S and analogously to a repeater R. In contrast to two media segments, a bridge station B can connect two or more LANs within a single LAN, thereby forming an extended LAN. Thus, for example, the three LANs 2a, 2b and 2c in FIG. 1 are connected to one another by two bridge stations B in order to form the extended LAN 1. In general, a bridge station B receives and stores information from a LAN, checks the destination of the information and, if the destination is reached directly or indirectly, routes (sends) the information through one of the other LANs connected to the bridge station B towards the intended one Aim to the appropriate LAN. However, if the destination is part of the same LAN (or possibly a LAN connected to it) from which the information is received, the bridge station B does not forward such information. The information from the bridge station B is thus either stored or forwarded.

The media segments 4 can be constructed from a suitable medium, such as a coaxial cable, an optical fiber or a combination thereof, which creates a communication channel through which the stations S can send information to one another, to the repeaters R and the bridge stations S and receive them from there , Each station S is physically connected to a media segment 4 by, for example, a media connection unit (MAU) 8 or 9. Each MAU 8, 9 can be a separate physical unit or an integral part of the station S to which it is assigned. The MAUs 8 represent 10Basis5 MAUs as specified in ISO 8802-3: 1989 (E), while the MAUs 9 represent 10Basis2 MAUs.

The information flow via the LANs 2 is regulated according to a protocol. Each station 5 and bridge station B typically contains a LAN interface that sends and receives information from the LAN 2 of which it is a part. The LAN interface often contains a microprocessor that can be programmed to process information according to the protocol. The LAN interface can also have further circuits in order to convert electrical signals used internally by the station into optical signals and vice versa. Thus, each station S that contains a compatible interface can communicate with any other station S with a compatible interface via the LAN.

Two examples of the operation of the LANs 2 are briefly described below. With further reference to FIG. 1, it is assumed that the LANs 2 operate according to a CSMA / CD protocol such as ISO 8802-3: 1989 (E). ISO 8802-3: 1989 (E) requires that each station (including each bridge station B) is assigned an address by which that station can be uniquely identified, either on a global basis or within a local management level. Whenever a station sends information, the information thus contains a "source address" that uniquely identifies the origin of the information. Similarly, all information sent contains a "destination address" that uniquely identifies the desired destination of the information.

It is also assumed that a station 10 desires to communicate with a station 12. It is also believed that stations 10 and 12 can perform full duplex communication and that full duplex communication is preferred due to the increased speed at which information can be transmitted. The segment of the LAN 2a that connects the stations 10 and 12 contains a repeater 14 and is a half-duplex segment. Due to the configuration of the LAN 2a, all information exchanged between the stations 10 and 12 must pass through the repeater 14. Repeater 14, however, is not suitable for full duplex communication and must operate continuously using half duplex communication. Thus, the presence of repeater 14 (or possibly other devices) and the half-duplex segment in the point-to-point path between stations 10 and 12 constitute a barrier to establishing full-duplex communication between stations 10 and 12.

In contrast to the previous example, it is now assumed that two bridge stations 16 and 18, both of which can perform full duplex communication, wish to exchange information using full duplex communication. It is further assumed that a media segment 17 connecting the bridge stations 16 and 18 is a full duplex connection. It should be noted that a "full duplex link" (such as media segment 17) may include, for example, a satellite link, an underwater cable, one or more links that are part of a large area network, or a number of other communication links that are capable of to support full duplex communication.

It should also be noted that there are no repeaters R in the point-to-point path (ie, media segment 17) between the bridge stations 16, 18. Under these circumstances, it may be possible to establish full duplex communications between the bridge stations 16, 18 according to the method described above.

FIG. 2 shows the bridge station 16, 18 shown in FIG. 1 in more detail. The bridge station 16 has two assigned MAUs 20a, 20b. MAU 20a is connected to a half-duplex segment 22a, while MAU 20b is connected to the full-duplex segment 17. Similarly, the bridge station 18 has two associated MAUs 20c, 20d, MAU 20c being connected to the full duplex segment 17 and MAU 20d being connected to a half duplex segment 22b. The full duplex segment 17 comprises two separate one-way communication paths 24a, 24b, one of which is used to send information from the bridge station 16 to the bridge station 18 (24a) and the other (24b) is used to transmit information from the To send bridge station 18 to bridge station 16 (24b).

Each bridge station 16, 18 comprises a plurality of main functional components including a pair of MAU interfaces 26. Each bridge station 16, 18 also has a memory area 28, a central processing unit (CPU) 30, a programmable logic array (PAL) 32 and a read-only memory (ROM) 33 on. It should be noted that other components may be included in addition to, or instead of, the components shown. The in Fig. 2 Hardware components shown are generally conventional, commercially available electronic devices that can be obtained from a number of sources.

The conventional functions of the bridge stations 16, 18 are briefly described below. The bridge station 16 can receive information through half-duplex communication over the half-duplex segment 22a. The received information is temporarily stored in the memory 28a. The information received from the bridge station 16 can consist, for example, of a series of frames with data, each containing an address indicating the desired destination of the associated frame. The CPU 30a checks the destination address of a given frame and determines whether that frame should be forwarded to the bridge station 18. The CPU 30a can make this determination since it already has "knowledge" of which destination addresses are reached via the bridge station 18. When the destination address of a given frame is reached by bridge station 18, bridge station 16 forwards (sends) that frame to bridge station 18 (via path 24a) where it is temporarily stored in memory 28b. The bridge station 18 then forwards the frame to the appropriate destination via the segment 22b.

The ROMS 33a and 33b can be used, for example, to store program instructions for the CPUs 30a and 30b, respectively, including instructions for performing the selective method of "storing and forwarding" just described. The ROMS 33a and 33b can also be used to store all or part of the program instructions for establishing full duplex communication between the bridge stations 16 and 18, as described below.

3A, 3B and 3C illustrate a method for establishing full duplex communications between two stations connected by full duplex communications, which operates according to a CSMA / CD protocol. The steps of the method can be carried out, for example, by circuits which are assigned to a single station. The circuits of the bridge stations 16 and 18 (Fig. 2) can be used, for example, to carry out the method and thereby establish full-duplex communication between these two bridge stations. However, it should be understood that similar circuitry for performing the method may be included in or associated with any suitable station served by the LAN including, for example, personal computers, workstations, and the like.

Initially, a station suitable for both full-duplex and half-duplex communication operates in a conventional manner using half-duplex communication according to the CSMA / CD protocol. In other words, the station initially "assumes" as a standard condition that full duplex communication is not possible and only half duplex communication can be used. The station then starts (START) either in response to a specified command received from the station or simply as an automatic process for establishing full duplex communication.

The station enters a request state or REQUEST STATE by sending (feeding) a request or REQUEST message through a connection of the LAN and by starting a request state or REQUEST STATE send interval timer, as shown at Schntt 34. In the preferred embodiment, a setpoint for the REQUEST STATE send interval timer is approximately 10 seconds. However, it should be understood that an appropriate duration for the REQUEST STATE transmit interval timer, as well as for the timers described below, depends on the specific requirements or limitations of a particular application. In the preferred embodiment, all steps performed within the REQUEST STATE use half duplex communication in accordance with the requirements of the CSMA / CD protocol, since the station sending the REQUEST message does not yet know whether it is possible to communicate with any other station on a full duplex basis communicate.

The REQUEST message comprises two packets, as shown in FIG. 4. The first packet of the REQUEST message is called a "test packet" and contains less than the minimum number of bits required for a fragment due to the CSMA // CD protocol. In the preferred embodiment, the test packet is 80 bits in length, while the CSMA / CD protocol requires a minimum fragment size of 96 bits (including preamble and SFD). The first 64 bits of the test package contain a preamble and a frame start separator (SFD). The last 16 bits of the test packet include a test packet identifier, which is a fixed bit sequence that uniquely identifies the test packet and distinguishes it from conventional or standard packets. In the preferred embodiment, a test packet identifier is used that uses the "1111 1111 1000 0000" bits, with the leftmost bit being sent first in time.

Since the length of the test packet is less than the minimum required by the CSMA / CD protocol, the test packet is automatically expanded to the required minimum length by each repeater that it traverses while traversing the LAN. In other words, if the test packet encounters a repeater while traversing the LAN, the presence of which represents a barrier to establishing full duplex communication, the test packet is expanded to the minimum length by the repeater. Otherwise, the test package remains below the minimum at its original length. Thus, a station that receives a REQUEST message that contains a test packet extended to the required minimum length "knows" that the test packet hit a repeater and that full duplex communication cannot be supported by the communication path that the test packet followed.

The test package is followed by a "standard package" with a length of N bits, where N is greater than or equal to the required minimum. In the preferred embodiment, which operates in accordance with the standard set forth in 150 8802-3: 1989 (E), the minimum required after the preamble and SFD is 512 bits. The standard package includes a preamble followed by a frame delimiter, a destination address, a source address, data and a frame check sequence. In the standard packet of the REQUEST message, the destination address is actually a standard address, which is a group address. A group address is used since a REQUEST message is intended to be an "invitation" to other available, but not yet identified, stations with which the station sending the REQUEST message can start full-duplex communication. Thus, by using a default group address as the destination address for a REQUEST message, this REQUEST is "multicast" to a defined group of stations, which are potential candidates for full-duplex communication. In the preferred embodiment, the default group address contains the bits "1001 0000 0000 0000 1101 0100 0100 0000 1000 0000 1100 0000", with the leftmost bit being sent first. The section of the source address of the standard packet uniquely identifies the station that sends the REQUEST message.

When the REQUEST message is sent, the station that sent the message checks whether a REQUEST message or START message is coming in from the connection, as shown at step 36. At this point, it is allowed to send either a REQUEST message (from another station or possibly the original REQUEST message that has "looped back") or a START message from another station received as described in detail below. If no message is received, the station checks at step 38 to determine if the REQUEST STATE transmit interval timer has expired. If so, the station returns to step 34 and, subject to a special timing requirement, retransmits the REQUEST message as before.

A special timing requirement is necessary because the present invention, as described above, allows the construction of an extended LAN in which two given stations can have a physical distance that is greater than that permitted by the CSMA / CD protocol. In other words, if the maximum distance between stations required by the CSMA / CD protocol is not maintained, there is no guarantee that the CSMA / CD protocol will work correctly to prevent two or more stations from operating synchronously and constantly try to resend a message at the same time, causing repeated collisions. Since the present invention can be connected in connection with an extended LAN, successive retransmissions of the REQUEST message are thus separated from one another by variable time periods, in order to ensure that two or more stations do not remain in synchronization with one another and repeatedly attempt to execute a REQUEST Send message. The provision of these variable time periods between retransmissions of the REQUEST message can be referred to as a "jitter time component". In the preferred embodiment, the “jitter-time component” can have a desired value in the range of approximately +/- 250 ms.

If, at step 38, the REQUEST STATE send interval timer has not yet expired, the station returns to step 36 and continues to check whether a message is coming in from the connection.

If either a REQUEST message or a START message was received according to steps 36 and 40, the receiving station first determines whether the received message contains both a valid test packet and a valid standard packet. It should be understood that a test packet and a standard packet can be received at different times, even if they together represent a single message. This principle is applicable to other types of messages described below.

A valid test packet is one that has retained its original minimum length, at least 16 bits of preamble, a frame delimiter, the correct test packet identifier, and no more than seven additional bits following the test packet identifier. Receiving a test packet that has been extended to the minimum length represents a failure of the method and a return to step 34 to begin again. A standard package is validated by examining the frame check sequence that is used to detect errors in the standard package.

If valid test and standard packets are received, the source address of the standard packet is checked in step 42. If the source address is the same as that of the station that received the packet, which means that the receiving station was the origin of the packet, this is considered an error in the process, which then returns to step 34. However, if the source address of the received packet is different from that of the receiving station (ie the packet was sent from another station), the process continues to step 44.

Starting at step 44, the station enters a START STATE and continues to operate using half duplex communication in the preferred embodiment. At step 44, the station stores the source address of the previously received message (packet) and starts a START STATE timeout timer. In the preferred embodiment, a setpoint for the START STATE timeout timer is approximately 4 seconds. The station then sends a START message and starts a START STATE broadcast interval timer. In the preferred embodiment, a target value for the START STATE send interval timer is approximately 250 ms.

The START message comprises two packets, as shown in FIG. 4. The first packet of the START message is a test packet that is similar to the test packet of the REQUEST message. The second packet of the START message is a standard packet that is similar to the standard packet of the REQUEST message except that the destination address is no longer a default group address but the address that was saved in step 44. That is, the START message is addressed to the station that sent the previously received REQUEST or START message.

After sending a START message, the station that sent the message checks whether a message is coming in from the connection, as shown in step 48. At this point it is allowed to receive either a START message (from another station or possibly the original START message that "came back through the loop") or an ACKNOWLEDGE message from another station, as described in detail below. Any REQUEST message received at step 45 has no effect on the process. Thus, if a REQUEST message is received at this step, the process will "loop" through steps 48, 50 and 52 until either the START STATE timeout timer or the send interval timer expires.

If neither a START message nor an ACKNOWLEDGE message is received at step 48, the station determines at step 50 whether the START STATE timeout timer has expired. If so, which means that a maximum allowed time to receive a message has elapsed, the station returns to step 34 to begin the process again. If not, the station proceeds to step 52 and checks whether the START STATE transmit interval timer has expired. If the START STATE transmit interval timer has not expired, the station simply returns to step 48 and continues to check if a message is coming in. If the START STATE transmit interval timer has expired, the station returns to step 46 and sends the START message again.

Again, if either a START message or an ACKNOWLEDGE message is received at step 48, the station proceeds to steps 54 and 56, which are analogous to steps 40 and 42 discussed above. At step 58, the source address of the received message is compared to the address that was saved during step 44. If the addresses are not the same, meaning that the two received messages are from different stations, this is considered an error in the process, which then returns to step 34 to start again. If the addresses are the same, which means that both received messages came from the same station, the station enters a FULL DUPLEX state at step 60.

At step 60, a FULL DUPLEX timeout timer is started and in the preferred embodiment the station switches to full duplex communication. In the preferred embodiment, a setpoint of the FULL DUPLEX timeout timer is approximately 100 s. The station can now simultaneously send information to the station with which it exchanged the REQUEST and START messages and receive information from it. At step 62, the station sends an ACKNOWLEDGE message and starts a FULL DUPLEX send interval timer. In the preferred embodiment, a setpoint for the FULL DUPLEX send interval timer is approximately 10 5. The ACKNOWLEDGE message comprises two packets, as shown in Figure 4, the first of which is a test packet that is the test packets of the REQUEST and START messages is similar.

Once an ACKNOWLEDGE message is sent, the station checks at step 64 whether an ACKNOWLEDGE message is coming in from the connection. At step 64, receipt of a START message has no effect on the process, which would simply "loop" through steps 64, 66, and 72 until either the FULL DUPLEX timeout timer or the send interval timer expires. If a REQUEST message is received at step 64, it means an error in the process, which then returns to step 34 to begin again.

If an ACKNOWLEDGE message is not received at step 64, the station determines at step 66 whether the FULL DUPLEX timeout timer has expired, which means that a maximum allowed time to receive an ACKNOWLEDGE message has expired. If so, then at step 68 the station can test the data link to determine if it is working properly. If the connection works correctly at step 70, the station returns to 34 to begin the process again.

Again, if, according to step 66, the FULL DUPLEX timeout counter has not yet expired, the station determines at step 72 whether the FULL DUPLEX transmit interval timer has expired. If not, the station returns to step 64 and continues to check for an ACKNOWLEDGE message. If so, the station returns to step 62 to send an ACKNOWLEDGE message again.

If an ACKNOWLEDGE message is received at step 64, the station proceeds to steps 74, 76 and 78, which are analogous to steps 54, 56 and 58, respectively, discussed in detail above.

Since stations operating in the FULL DUPLEX state can receive or send information at any time or can do both at the same time, the bandwidth of the connection connecting the stations is effectively doubled compared to conventional half-duplex operation according to the CSMA / CD protocol. When sending information in the FULL DUPLEX state), each station should generate a minimal delay between information packets to ensure sufficient separation for correct reception by the receiving station. In the preferred embodiment, a minimum gap between packets of 9.6 µs is used.

Because the FULL DUPLEX state allows full duplex communication only between stations that are point-to-point connected through a full duplex communication link, no station in the FULL DUPLEX state is required to comply with the carrier detection or collision detection aspects of the CSMA / CD protocol. More specifically, for a station in FULL DUPLEX state, it is unnecessary to check the communication link for activity before sending information. It is also not necessary for any station in the FULL DUPLEX state to monitor the connection for possible collisions. This is permitted because, by definition, there is never a collision of information sent between the two point-to-point connected stations working in the FULL-DUPLEX state and there can be no time when one station is not allowed, information to the other to send.

In addition, since the collision detection aspect of the CSMA / CD protocol can actually be ignored by a station operating in the FULL DUPLEX state, there is no corresponding maximum propagation delay that should be observed. As a result, the physical distance between the two participating stations can advantageously be increased.

In an alternative embodiment of the present invention, a station may send a REQUEST message as described above, but may use full duplex communication instead of half duplex communication. This alternative embodiment can be used, for example, when a full duplex station "knows" in advance that it is connected to a communications link that supports full duplex communication. In such a case, the station sending the REQUEST message need not test the suitability of the connection before starting full-duplex communication, but must determine whether a second station is ready for full-duplex communication. Thus, in this alternative embodiment, a second station can send a START message using full-duplex communication instead of half-duplex communication. The two stations can then exchange ACKNOWLEDGE messages as described above.

The foregoing description has been limited to a specific embodiment of this invention. However, it is understood that variations and modifications can be made in the invention while retaining some or all of the advantages of the invention. It is therefore an object of the appended claims to cover all such variations and modifications.