WO2000060803A1

WO2000060803A1 - Reconstruction of messages distributed on frequency and time basis in a data collecting network

Info

Publication number: WO2000060803A1
Application number: PCT/FR2000/000694
Authority: WO
Inventors: René KOPP; Christian Visse
Original assignee: Telediffusion De France
Priority date: 1999-04-06
Filing date: 2000-03-17
Publication date: 2000-10-12
Also published as: FR2792138B1; FR2792138A1

Abstract

Each message (ME) transmitted from terminals to a base station (SB) is segmented into bursts (B1-B10) successively distributed on consecutive carriers and over successive time intervals. Each burst comprises a 3-bit or 4-bit code word indicating its position and the rank relationship between its carrier and the carrier of an adjacent burst. To reconstruct the message correctly in the base station, a flag representing the order relationship in the code word of a first burst is produced first of all. The data of each other burst are concatenated with those of the preceding bursts when the order relationship in the burst code word with the preceding burst carrier corresponds to the order relationship represented by the flag of the preceding burst. The flag then becomes representative of the rank relationship between the carriers of the burst and of the following burst.

Description

Reconstitution of messages distributed friguentially and temporally in a data collection network

The present invention relates generally to a digital broadcasting network between a broadcasting base station and user receiving terminals. The network is particularly intended to digitally broadcast television programs according to the European standard DVB-T (Digital Video Broadcasting- Terrestrial) from the base station, and in the context of interactivity between user terminals and the broadcasting station. base, to collect in the base station data transmitted by user terminals, such as digital television sets, through a return channel with narrow frequency band, also called uplink.

The return channel is of the frequency distribution type of a message and / or time distribution type of a message.

Frequency distribution consists in distributing a message sent by a terminal over M consecutive carriers among a set of N carriers during a time interval of predetermined duration, M and N being integers such that M <N. The aforementioned message is segmented into M elementary transmission units constituting data packets, called "bursts" in the following description, with a predetermined constant rate on the M carriers.

The temporal distribution consists in distributing a message transmitted at the predetermined rate on a single carrier among the N carriers, that is to say in segmenting the message into Q bursts occupying Q successive time intervals on the single carrier.

More generally, the successive bursts in a message can be distributed in time over at least two successive time intervals, starting with the lowest frequency carrier available among the M carriers in the first time interval, then starting with the most low frequency of the M carriers in the following time intervals.

For example, the set of N carriers are ordered according to the increasing order of the frequencies of the carriers.

In order to reconstitute in the station the messages segmented into bursts, transmitted by the terminals, each burst includes a code word indicating the position of the burst in the message and the order relation between the frequencies of the carrier on which said burst and the 'one of the bursts adjacent to this one are emitted.

The invention aims to reconstruct a burst message without error by considering only the minimum of information included in the code words of the bursts.

To this end, a method for reconstructing messages comprising at least one burst transmitted from terminals to a base station, the bursts of a message being successively distributed frequently over at most M consecutive respective carriers among N predetermined ordered carriers, with 1 <M <N, during each of several successive respective time intervals, each burst comprising data and a code word indicating the position of the burst relative to the start and the end of the message and the order relation between the frequencies of the carriers on which said burst and one of the bursts adjacent to said burst in the message are transmitted, is characterized in the station basic by the following steps to reconstitute a message: produce one of three types of flag respectively representative of one of three order relations contained in the code word of a first burst of the message, no flag being produced in response to a message with a code word indicating a single burst, concatenate the data of each burst of the message other than the first burst with the data of the preceding bursts of the message when the order relation in the code word of the burst between the frequencies of the carriers on which said burst and the burst preceding said burst in the message are transmitted, corresponds to the order relation represented by the flag linked to said preceding burst ent, and

- replace the flag with a flag which is representative of the order relation between the frequencies of the carriers of the burst with the concatenated data and of the following burst in the message, deduced from the code word in the burst with the concatenated data, when that -this is not the last burst of the message. The respective time intervals in which the bursts of a message are distributed may be consecutive, or else spaced by a constant number of time intervals known by the terminals and the base station. When the flag associated with a burst is representative of an order relation in a direction opposite to the ordering of the carrier frequencies, the next burst having a code word containing an order relation for the preceding burst carrier according to the same direction as the ordering of the carrier frequencies, is sought on at most the M carriers in the time interval following that containing said burst, before performing the step of concatenating the data of the following burst.

In practice, the number M of carriers allocated to a message is always less than or equal to a predetermined number less than N.

In this case, the code words of the bursts which follow a burst of the message, the code word of which first contains a carrier order relation with the next burst of the message in the opposite direction to the ordering of the carrier frequencies and which are emitted in a second time interval of the message are analyzed successively until a burst having a code word containing an order relation for the preceding burst carrier in the same direction as the scheduling of the carrier frequencies is found. The burst carrier index found constitutes the smallest carrier index of the message and is memorized. According to a preferred embodiment, a first pointer initially placed in the rank of the carrier frequency of the first burst of the message is decremented up to the rank of the carrier frequency on which the next burst to be sought is transmitted, when said next burst to be searched the first burst in the message with this carrier frequency. The largest message carrier index can also be stored, by searching for it among the bursts of the message sent in the first time interval of the message. Preferably, a second pointer initially placed in the rank of the carrier frequency of the first burst of the message is incremented each time that the flag associated with the burst preceding the burst having data to be concatenated is representative of an order relation according to the same meaning that the scheduling of the carrier frequencies, and the preceding burst is in the first time interval of the message.

The concatenation of the bursts of the message is stopped and the message is ignored in particular as soon as the difference of the second and first pointers is equal to the predetermined number of consecutive carriers on which a message is likely to be distributed. As will be seen in the detailed description of the process, the concatenation can be stopped on other criteria based on bursts inconsistent with the logic of reconstitution of a message.

The step of concatenating each burst can be carried out as long as the size is less than a maximum message size.

According to other aspects of the method of the invention, the burst position can be one of the following three positions: first burst at the start of the message, intermediate burst between the start and the end of the message, and last burst at the end of the message. The burst order relation can be one of the following three: less than, equal to and greater than, for a first message burst relative to the frequencies of the carriers on which said first burst and the burst following said first burst in the message are sent, and for a final burst of message relative to the carrier frequencies on which said last burst and the burst preceding said last burst in the message are sent.

The code word contained in an intermediate burst preferably signals two order relations between the frequency of the carrier on which said intermediate burst is transmitted and the frequencies of the carriers on which the bursts adjacent to said intermediate bursts in the message are transmitted. The code word contained in a first burst of a message only concerns an order relation with the only burst adjacent to it, that is the next burst, and the word contained in a last burst of a message concerns only a order relation with the burst preceding this last burst.

When a message contains only one burst, this unique burst is signaled by a code word different from the other code words. According to a first embodiment, the code words are of eleven types and include four bits.

According to a second embodiment, the code words are less numerous and are of eight types with three bits. In this case, at least one of the code words is common to a first burst and to an intermediate burst or to a last burst of the message, preferably having common carrier frequency order relationships.

Other characteristics and advantages of the present invention will appear more clearly on reading the following description of several preferred embodiments of the invention with reference to the corresponding appended drawings in which: - Figure 1 is a schematic block diagram of a broadcasting and collecting network with frequency and time multiplex between a base station and several user terminals; FIG. 2 is a diagram showing the distribution of the bursts of two messages in a frequency and time multiplex of the access network;

- Figures 3A and 3B respectively show the format of a frame and the format of a burst; - Figure 4 is a table showing eleven types of bursts according to a first embodiment;

- Figures 5 and 6 are respectively frequency and time multiplex diagrams showing the frequency and time location of each burst of several messages with the meaning of their code words; Figure 7 is a frame reconstruction algorithm for the central station, according to the invention; - Figures 8, 9 and 10 are algorithms included in that of Figure 7, detailing steps respectively related to three types of flags

; and

- Figure 11 is a table of eight code words according to a second embodiment.

Referring to Figure 1, a radio broadcast and collection network includes a base station

SB and several user terminals TEl to TEX distributed in any manner in a service area covered by the base station.

The broadcasting and collection network is an asymmetrical bidirectional transmission network.

The base station is a master station for all of the terminals, and in particular provides them a pilot frequency and a time reference. The base station also manages the allocation of frequencies or time intervals during a cycle identified by the time reference, to each of the terminals wishing to transmit data to the base station. The downlink from the base station to the terminals is for example included in the form of teletext packets inserted in the blanking intervals of a conventional analog television signal, or else implemented by frequency-multiplex modulation orthogonal coded COFDM (Coded Orthogonal Frequency Division Multiplex).

According to the direction from the terminals to the base station, a return channel supports low-speed data transmitted by the terminals to the base station, based on a synchronous frequency division multiple access (SFDMA) multiple access transmission. ). Multiple access means that the terminals can simultaneously communicate with the base station, without the communications interfering with each other, using the same return channel. Depending on the carrier and time slot assignment to a given TEx terminal by the base station, with 1 <x <X, the TEx terminal transmits mx data messages on M carriers among N carriers, or on a carrier for one or more consecutive predetermined time intervals in a cycle determined by the time reference retrieved by the terminal.

For example, the terminal is included in a television set which allows the user of the set to transmit data messages to the base station in response to selected instructions on the keyboard of the station's remote control, in order to participate interactively for example in televised games, teleshopping, prepayment of television programs, etc. More generally, this application is part of the broadcasting of digital terrestrial DVB-T television programs, although the invention can be applied to a broadcasting system by satellite or by cable. Such a radio broadcasting and collection network is for example described in French patent application FR 2 709 388 filed on August 24, 1993.

The invention is directed more particularly to a radio network for collecting data and thus to the return channel with access to SFDMA transmitting data messages transmitted by distributed user terminals TE1 to TEX to the base station SB. SFDMA access is based on both a frequency multiplex and a time multiplex.

With reference to FIG. 2, the time multiplex is divided into cycles each comprising j 2 -1 consecutive time intervals, which are numbered by a burst index IB varying

J cyclically from 0 to 2 -1, as shown on the abscissa in FIG. 2. The frequency multiplex comprises N carrier frequencies spaced by a predetermined step in a frequency band of the order of megahertz and numbered by a carrier index IP varying from 0 to Nl according to the increasing order of frequencies. For example, the integer J is equal to 16 and the number N of carriers is between 50 and 2000; the modulation of the carriers by the data is a differential phase modulation of the D-PSK type, for example with 4 or 8 states. According to the prior art, each terminal transmits on one or more modulated carriers with a low speed during a predetermined time interval allocated by the base station. For example, a message ml sent by a terminal TE1 is distributed over three respective carriers of index IP = 2, 3 and 4 during the first time interval identified by IB = 0 of the time cycle, as shown in Figure 2. The carriers produced by the TEl terminal are slaved to a pilot frequency, for example the vision carrier of a television signal transmitted by the base station to the terminals, and have their power adjusted beforehand under the control of the base station so that the corresponding signal received by the base station has a predetermined level. Thus the modulated carriers are synchronized on a common time reference so as to constitute a frequency multiplex which is completely demodulated in a single receiver of the base station by discrete Fourier transform TFD.

The temporal distribution results implicitly from the segmentation of data messages into elementary transmission units, hereinafter called "bursts", analogous to bursts or bursts, the transmissions of which in the time intervals are synchronized with the temporal reference transmitted by the base station. According to the prior art, a m2 message sent by a TE2 terminal can be distributed over several consecutive respective time intervals, for example three in number, on the same respective carrier here of carrier index IP = 2, as shown in the figure 2.

As shown in Figure 3A, a data frame produced by a terminal includes a header EN, a data field proper D, and a cyclic redundancy check sequence CRC. The header EN contains in particular the address of the terminal and the address of the recipient of the frame, that is to say that of the base station SB of the network, the number of the frame and the length of the field of data. According to the invention, as shown in FIG. 2, such a frame can be segmented over M consecutive carriers and over Q consecutive time intervals. A frame has a maximum length of NBO = 4096 bytes. For example, the frame is transmitted in the form of a message ME having ten bursts B1 to B10 indicated in FIG. 2.

FIG. 3B shows a message to be sent after the frame of FIG. 3A has been segmented into three bursts B1, B2 and B3. Each burst includes a long DB data field and a short segmentation field. The data field DB has a length expressed in bytes, determined as a function of the data rate in the terminal, and is filled with a few tens of bytes of the frame TR, typically 53 bytes maximum. The segmentation field consists of a code word MC of a few bits which is used to reassemble the data fields DB of the bursts of the message in the frequency and time multiplex of FIG. 2, that is to say to concatenate the DB data fields of the message bursts of FIG. 3B in order to reconstitute the frame TR in the base station. In FIG. 3B, the last burst B3 comprises the end of the data field D of the frame TR and the control field CRC as well as possibly stuffing bits BB before the corresponding code word MC.

The set of carriers is ordered in the increasing direction 0 to Nl of the IP carrier index, from preferably in correspondence to the increasing direction of the carrier frequencies.

The bursts composing a message, such as for example the ME message with ten bursts shown in FIG. 2, are first distributed over a first burst index time interval IB = 1 according to this example, from the first available frequency among the M = 3 respective carrier frequencies of carrier index 3, 4 and 5 for the message ME. According to the example illustrated in FIG. 2, the first available frequency on which the first burst B 1 of the message ME is transmitted in the index time interval IB = 1 is the index carrier IP = 4. The preceding burst the message ME and identified with IB = 1 and IP = Pm = 3 can be the last burst or the only burst of a message sent by another terminal, or be empty of data. Then during each following time interval, that is to say in each of the three following columns in FIG. 2, the bursts of the message ME are distributed over the M carriers according to the increasing order of the frequencies. For example after the carrier of maximum index PM = 5 on which the second burst B2 is transmitted, the third burst B3 in the following time interval of index IB = 2 is transmitted on the carrier of minimum frequency Pm = 3 in all M = 3 carriers assigned to the message ME. The burst following the message ME and identified with IB = 4 and IP = PM = 5 can be the first burst or the only burst of a message sent by another terminal, or be empty of data.

For example, for a frequency band of the multiplex of 1 MHz, 62, 125, 250, 500, 1000 or 2000 carriers can be distributed equally. The duration of a burst B1 to B10 is for example 500, 250, 125, 62, 31 or 15.6 ms. The number of useful bits per burst is between 23 and 53 bytes.

As a variant, the respective time intervals in which the bursts B1 to B10 of a message ME are distributed, are not consecutive and are spaced by a number of predetermined time intervals NIT previously stored in the terminals TE1 to TEX and the base station SB; for example, the burst sets [B1, B2], [B3, B4, B5], [B6, B7, B8] and [B9, B10] of the ME message are distributed respectively in the burst index intervals 1+ NIT, 1 + 2NIT, 1 + 3NIT and 1 + 4NIT.

In a frequency and time multiplex, all bursts are sent with the same minimum rate. However, the distribution of bursts of the same message not only on consecutive carriers but also over consecutive time intervals contributes to increasing the bit rates of the ME messages according to the invention compared to the bit rates of the ml and m2 messages already known. For example, for each time reference cycle at 2 -1 time intervals, the message ME is distributed over Q = 4 respective time intervals and M = 3 respective carriers, which gives it a flow rate substantially equal to 3.3 times the flow of the message ml only distributed over a single time interval and over three carriers, or substantially equal to 3.3 times the flow of the message m2 distributed over three time intervals and only over a carrier. In general, relative to a time interval, a message distributed over M carriers has a flow rate equal to the product of M by the minimum flow rate of a burst during a time interval, and a message distributed over Q intervals of burst time at a rate equal to the product of Q by the maximum rate.

The code word MC in a current burst BC is used to identify the burst BC with respect to the burst BP preceding the burst BC and with respect to the burst BS following the burst BC in the message received ME to be processed in the base station SB. In the 3-burst message shown in FIG. 3B, when the current burst BC is the second burst B2, the preceding burst BP is the first burst B1 and the next burst BS is the third and last burst B3. The coding considers that certain bursts in a message may be on a carrier of the carrier index lower than the carrier index associated with the first burst (FIRST BURST) at the start of the message. According to this first embodiment, eleven types of burst are necessary to reconstruct any message and thus reassemble a frame TR, passing from FIG. 3B to FIG. 3A. The eleven burst types require a 4-bit MC code word.

The table in Figure 4 indicates the eleven code words OB to LBU defining the types of burst. In the designation of code words, the letter B indicates a current burst BC, the letter 0 (One) a burst constituting in itself a message; the letter F

(First) the first burst of a message; the letter I

(Intermediary) an intermediate burst between a first burst and a last burst in a message; and the letter L (Last) a last burst in a message. The letter D (Down) or U (Up) in a code word indicates a carrier with an index lower or immediately higher than the IP index of the current burst BC in a message, which carrier is associated with burst BS following the current burst BC when this is a first burst, or at the BP burst preceding the current burst BC when this is a last burst, or else at the preceding burst or following the current burst when this is a intermediate burst, the letter D or U designating the index for the previous burst being indicated before the letter D or U designating the index for the next burst. The letter E (Equal) in a code word designates the carrier common to the current burst BC and to the burst BS following the latter when the current burst is a first burst or an intermediate burst, or common to the current burst BC and to the BP burst preceding this one when the current burst is a last burst or an intermediate burst.

Two examples of messages are illustrated in the matrices in FIGS. 5 and 6 by assuming that the carriers are number N of three and that the first messages ME1 and ME3 start with a first burst on the first carrier of index 0 and at the reference time of the base station. For each burst of each message ME1 to ME5, the associated code word MC is indicated. The rule of transmitting the bursts of a message according to the increasing direction of the frequencies is typically applied in the messages ME1, ME2 and ME5. In these examples, it has been assumed that a message can only be transported on at most three parallel consecutive index carriers, although in practice the maximum number PC of consecutive carriers for a message is equal to 16, with PC <N = 32. The message ME3 in FIG. 6 only includes three bursts on the carrier of index 0. The message ME4 in FIG. 6 only consists of a burst with MC = OB. As a variant, boxes of the matrices in FIGS. 5 and 6 are empty of data: for example, no data burst is transmitted in place of the third and fourth bursts of the message ME1 and the second and third bursts of the message ME3.

The message reconstruction and frame reassembly method according to the invention is illustrated by the algorithm in FIG. 7 in association with the algorithms in FIGS. 8 to 10. These algorithms are installed in the form of software in the station PROM memory basic SB.

The message station reconstruction method consists in scanning the N carriers in frequency in each column of FIG. 2 in ascending order of the carrier frequencies, in order to successively write the N bursts of a column received during a burst time interval in a column memory of the base station, then to write the bursts of the received column respectively in buffer memories in which are reconstructed in parallel the messages column received by column received according to the burst indexes croissants. The messages are thus reconstituted "on the fly", that is to say as and when the bursts occur in the column matrix, by concatenating the data fields of the bursts of each message according to the reconstruction method of the invention described below in order to reassemble them in a frame. Thus, several algorithms according to the reconstruction method according to the invention are developed in parallel according to the messages being reconstructed.

Each message being reconstituted is written burst by burst by associating a flag (flag) f (BC) with the code word MC of each burst BC. The pennant of the current burst BC indicates the sign of the difference IP (BS) - IP (BC) of the carrier indexes of the burst following BS and of the current burst BC, when the message comprises at least two bursts, that is to say when the BC burst is not a single OB burst or the last LBD, LBE, LBU burst of the message.

With reference to FIG. 7, at the start of a current message at an initial EO step, no flag can be associated with a previous burst, ie f (BC) = 0, and the first burst of the current message is read in order to start the corresponding frame (step E01). A frame size variable TA expressed in bytes is set to zero and must be less than or equal to the maximum length NBO of the frames.

The code word MC of the first burst is read in order to produce the first flag in the following steps El to E4 when the message comprises at least two bursts. If in step E1, MC = OB, which corresponds to a message with a single burst, the frame is processed in the base station SB (step E1i). Then the next burst is analyzed to find the start of another message in step E5, returning to step E0 and holding f (BC) = 0.

Steps E2, E3 and E4 consider the type of the first burst of the message so as to associate with it a respective flag indicating the magnitude of the IP index (BS) of the carrier of the next burst in the message compared to the IP index (BC) of the first burst, i.e. the sign of the IP difference (BS)

- IP (BC). If the first burst has a word MC = FBD in step E2, the flag f (BC) is set equal to fd (down)

(step E21) to indicate that the IP index (BS) of the next burst BS is lower than the IP index (BC) of the first burst BC, that is to say IP (BS) - IP (BC) <0, and it will be necessary to decrement the burst index IB during the analysis of the next burst. If the first burst and the following burst are on the same carrier, the word MC is equal to FBE in step E3 and the flag f (BC) is set to fe (equal) (step E31) to indicate that there is n there is no change of carrier, i.e. IP (BS) = IP (BC). If the first burst has a word MC = FBU in step E4, the flag f (BC) is set equal to fu (up) (step E41) to indicate that the IP carrier index (BS) of the following burst BS immediately succeeds the IP index (BC) of the first burst, with IP (BS) = IP (BC) +1. Then after steps E21, E31 and E41, the second burst of the message is sought (step E6) as a function of the flag of the first burst, as will be seen below in steps SI, S4 and S7. If no code word MC is detected, that is to say if the burst contains no bytes of data, the following message is sought in step E5 which maintains or resets the flag f (BC) as well as a CIB burst index change variable.

In addition, at steps E22 and E42 respectively after steps E2 and E4, two pointers PM and Pm are set equal to the IP carrier index (BC) of the first burst when the carrier index must vary from the first burst to next burst in the message. The PM and Pm pointers indicate the maximum and minimum limits of the IP carrier index of the message being processed during a cycle and are respectively incremented and decremented as the corresponding frame is reconstructed burst by burst in order to check that the consecutive carriers of the message are less than or equal to PC = 16. Also, the message size variable TA is increased by the number of data bytes OC (BC) in the data field DB of the burst BC in steps E22 and E42 and in a step E32 after step E31.

Then after the processing of the first burst of the current message between one of the steps E2, E3 and E4 and the step E6, each subsequent burst of the message is processed in one of the steps SI, S4 and S7 as a function of the flag f (BC) of the previous burst. Thus, step SI is selected for f (BC) = fd, that is to say when the burst BP preceding the current burst has a code word MC equal to FBD or IBDD; step S4 is selected for f (BC) = fe, that is to say when the burst BP preceding the current burst has a code word MC equal to FBE or IBE; and step S7 is selected for f (BC) = fu, that is to say when the burst BP preceding the current burst has a code word MC equal to FBU, IBDU or IBUU.

When f (BC) = fd in step SI, the algorithm shown in Figure 8 is run. This means that the current burst BC to be analyzed is located in the next burst interval and on a carrier whose index is equal to or less than all the carrier indices of all the preceding bursts in the message. Thus, in the following steps SU and S12, the burst index IB is incremented by one and the variable for changing the burst index CIB is set to 1, and the current carrier index IP is set to the smallest Pm index of message bursts already received.

Step S12 is linked in a loop to steps S2 and S3 to find the first carrier of lower index which carries a current burst of type IBUU or LBU, that is to say of which the preceding burst is on a upper carrier (Up), and in a step S13 to decrement the variable Pm by one unit. This loop is only carried out when the first burst having the minimum carrier index in the message is sought, ie the burst B3 of index IP = 3 in the message ME according to the example of FIG. 2. The code words MC bursts are thus analyzed in the decreasing direction of the carrier index from the initial index Pm. For example, with reference to the message ME in FIG. 2, the pointer Pm is equal to 4 after the analysis of the second burst B2, the burst B4 in the following interval IB = 2 having a code word MC = IBDD different from IBUU and LBU is read since Pm is equal to 4 at the start of the message ME, and finally by decrementing Pm by one unit, the burst B3 containing a code word MC = IBUU (step S2) is found.

In the loop, at steps S14 and S15 between steps S13 and S12, the difference PM-Pm between the limits PM and Pm of the carrier index for the current message is compared to PC = 16 and the variable Pm is compared to zero in order to refuse to analyze a message which is distributed over more consecutive carrier PCs.

A step S16 is also included in the loop, between steps S12 and S2, to check the consistency between successive bursts in a message and particularly that the current burst before the one sought containing a code word IBUU or LBU is not already there. - even linked to a flag of type fd, which would mean that the current burst in the loop belongs to another message and that the sought burst is lost. Thus, when the condition in step S14, S15 or S16 is not satisfied, the current message is ignored and the method proceeds to the next message in step E5. When the current burst BC has a code word MC = IBUU in step S2, the frame size variable TA is incremented by the number of bytes OC in the data field DB of the current burst and compared with the maximum size of NBO message = 4096 bytes, in the following steps S20 and S21. If the current size TA is less than or equal to the normal size NBO, the data of the current burst BC are read to complete the frame with the intermediate current burst of type IBUU in the base station SB in step S22. Since the current burst of IBUU type must be followed by a burst with a larger IP carrier index, the flag f (BC) is set to fu in the next step S23, and the method proceeds to step E6 to analyze the next burst by step S7 (figure 7).

When the current burst BC has a code word MC = LBU in step S3, the size parameter is incremented by the number of bytes OC and compared with the maximum size NBO in steps S30 and S31 identical to steps S20 and S21 , and the data of the current burst BC are read in a step S32 when TA <NBO. The frame terminated with the last burst of type LBU is then processed in the base station SB in step S32. The flag f (BC) is set to zero in the next step E5 to seek the next message and in particular the first burst of the following message in one of the steps E1 to E4.

For a current burst received with which the flag f (BC) of the preceding burst is equal to fe, step S4 is followed by the steps shown in FIG. 9.

The current burst BC having the same carrier as the previous burst, the next step S40 locates the current burst in the time interval indicated by the IP burst index incremented by one. Then the code word MC of the current burst is analyzed to determine the IBE or LBE type thereof in steps S5 and S6.

As after steps S2 and S3 (FIG. 8), step S5, S6 is followed by a step S50, S60 to increase the size variable TA with the number of bytes OC in the data field DB of the current burst BC, and a comparison step S51, S61 to compare the parameter TA with the maximum message size NBO.

When the current burst BC has a code word MC equal to IBE in step S5 and the variable TA is less than NBO in step S51, the data of the current burst BC are read to complete the frame in a step S52 with the IBE type intermediate current burst. The state of the flag f (BC) is kept in the state fe in a step S53. The process continues towards step E6 to analyze the next burst by step S4.

When the current burst BC has a code word MC equal to LBE in step S6 and the variable TA is less than or equal to NBO in step S61, the data of the current burst BC are read to end the frame by the last current burst of type LBE and processing the completed frame in the base station SB in step S62. The next step E5 sets the flag f (BC) to zero and searches for the next message.

Returning to FIG. 7, when the current burst to be processed is associated with a preceding flag f (BC) of type fu in step S7, the algorithm shown in FIG. 10 is carried out. If the flag f is not filled (state 0) or is different from the types fd, fe and fu, the code word of the current burst is erroneous and the algorithm goes to step E5. Referring to step S7 in Figure 10, when f (BC) = fu, the maximum limit pointer PM of the carrier index for the current message is incremented by one in step S71 only when the CIB burst index change variable for the message is still zero in step S70, that is to say when the current burst is sought in the first time interval occupied by the message, the interval IB = 1 for the message ME in FIG. 2. Then the IP carrier index is incremented by one in step S72. As in step S14

(Figure 8), in the next step S73, the difference PM

- Pm of the carrier index limits for the current message is compared to PC = 16 and the current message being processed is refused when this difference is greater than or equal to PC. Otherwise, one of the following steps S8, S9 and S10 is performed depending on the type MC of the code word of the current burst. Steps S8, S9 and S10 relate respectively to a current burst of the IBDU, IBDD and LBD type and are followed by respective steps S80-S81, S90-S91 and S100-S101, similar to steps S20-S21 or S30-S31 (figure 8), according to which the size variable TA is incremented by the number of bytes OC in the data field of the current burst BC and compared with the maximum message size NBO.

When the current burst BC is of the IBDU type in step S8 and the size variable TA is less than or equal to NBO, the frame is completed by the current burst in the next step S82 and the flag f (BC) is set. equal to fu in step S83, since the burst following the current burst must have a larger IP carrier index. In an analogous manner, a step S92 reads the current burst BC to complete the current frame of reconstruction in a step S92 when in step S9 the current burst is of the IBDD type and in step S91 the variable of size TA is less than or equal to NBO. The next step S93 sets the flag f (BC) to fd since the next burst must have an IP carrier index lower than the current burst of the IBDD type. After step S83 or S93, the method returns to step E6 to find the next burst of the current message by step S7 or SI. Step S10 is followed by the two steps S100 and S101 when the code word MC of the current burst is of the LBD type, that is to say constitutes the last burst of the message following a burst of lower carrier index. When the size variable TA is less than or equal to NBO in step S101, the current burst BC is read in step S102 to process the completed frame in the base station SB, and the flag f (BC) is set to zero in the next step E5 which searches for the next message.

According to a second embodiment of the invention, the field of the code word MC is reduced to three bits, without thereby generating errors on the frame to be reconstructed compared to the first embodiment described above for a four-bit code word .

In the second embodiment, the coding table has only eight lines, as shown in FIG. 11, instead of eleven lines, by assigning to three pairs of burst types three respective code words, which makes it possible to delete three words of code, i.e. 11 - 3 = 8 code words. According to the example illustrated in FIG. 11, the three pairs of code words combined are the following: The first pair of OB / LBE code words concerns an OB type burst which alone constitutes a frame and a last LBE type burst having a previous burst on the same carrier. The equalities MC = OB in step E1 in FIG. 7 and MC = LBE in step S6 in FIG. 9 are replaced by the equality MC = OB / LBE. If the code word MC = OB / LBE is detected in step E1 without a flag f (BC) being assigned to it in step EO, this burst is necessarily an OB type burst constituting a frame. With reference to FIG. 9, if in step S6 the code word MC = OB / LBE is detected, the current burst is necessarily a last frame burst of LBE type if the flag f (BC) is of the fe to the previous step S4.

- The second pair of FBE / IBE code words concerns a first burst and an intermediate burst on the same carrier. In FIGS. 7 and 9, the equality MC = FBE in step E3 and the equality MC = IBE in step S5 are replaced by the equality MC = FBE / IBE. When in step E3 (FIG. 7), the code word MC = FBE / IBE is detected in a burst for which the flag f (BC) is equal to zero, the flag f (BC) necessarily becomes of the type fe in the next step E31. In step S5 (FIG. 9), when the flag f (BC) = fe, the code word = FBE / IBE detected is necessarily associated with an intermediate current burst of IBE type; otherwise the fe flag would not be detected, which would impose a first FBE type burst.

- The third pair of code words FBU / IBDU relates to a first burst of type FBU and an intermediate burst of type IBDU whose subsequent bursts are on carriers of higher indexes. The equality MC = FBU in step E4 in figure 7 and the equality MC = IBDU in step S8 in FIG. 10 are replaced by the equality MC = FBU / IBDU. At the start of a message, when no flag is assigned, ie f (BC) = 0 in step E0, the code word MC = FBU / IBDU detected in step E4 must correspond to a first burst FBU type. The next step E41 puts the flag f (BC) in the state fu. When the flag of the previous burst is such that f (BC) = fu in step S7

(FIG. 10), the code word MC of type FBU / IBDU detected in step S8 is necessarily associated with an intermediate burst of type IBDU.

Many other torque triplets can be considered to reduce the field of the code word from 4 to 3 bits. For example, the three pairs are the following common code words: MC = FBE / LBE, MC

= FBD / IBUU, MC = OB / LBD.

Claims

1 - Method for reconstituting messages (ml, m2, ME) comprising at least one burst transmitted from terminals (TEl, TEX) to a base station (SB), the bursts of a message being successively distributed frequency over at most M consecutive respective carriers among N predetermined ordered carriers, with 1 < M < N, during each of several successive respective time intervals, each burst comprising data (DB) and a code word (MC) indicating the position of the burst relative at the beginning and at the end of the message and the order relationship between the frequencies of the carriers on which said burst and one of the bursts adjacent to said burst in the message are transmitted, characterized in the base station by the following steps for reconstituting a message :

- produce (E2-E21, E3-E31, E4-E41) one of three types of flag (f(BC)) respectively representative of one of three order relationships contained in the code word (FBD, FBE, FBU) of a first burst of the message, no flag being produced in response to a message with a code word indicating a single burst (OB),

- concatenate (S22, S32; S52, S62; S82, S92, S102) the data of each burst of the message other than the first burst with the data of the previous bursts of the message when the order relationship (S2, S3; S5, S6; S8, S9, S10) in the code word (MC) of the burst between the frequencies of the carriers on which said burst and the burst preceding said burst in the message are transmitted, corresponds to the order relationship (SI; S4 ; S7) represented by the flag (f(BC)) linked to said previous burst, and - replace (S23; S53; S83, S93) the flag (fd, fe, fu) with a flag (fu; fe; fu, fd) which is representative of the order relationship between the frequencies of the burst carriers with the concatenated data and the next burst in the message, deduced from the codeword in the burst with the concatenated data, when this is not the last burst (LBD, LBE, LBU) of the message.

2 - Method according to claim 1, comprising, when the flag (f (BC)) associated with a burst is representative of an order relationship in a direction opposite to the ordering of the carrier frequencies, a search (S12, S2, S3, S13, S14) of the next burst having a codeword containing an order relation for the carrier of the previous burst according to the same direction as the ordering of the carrier frequencies, on at most the M carriers in the interval time following that containing said burst, before carrying out the step (S22, S32) of concatenating the data of the next burst.

3 - Method according to claim 2, comprising a step (S12, S2, S3, S13) of successively analyzing the code words (MC) of the bursts (B4, B3) which follow a burst (B2) of the message (ME ) whose codeword first contains a carrier order relationship with the next burst of the message in the opposite direction to the ordering of the carrier frequencies and which are transmitted in a second time interval of the message, until finding a burst (B3) having a codeword containing an order relationship for the previous burst carrier

(B2) in the same direction as the ordering of the carrier frequencies in order to memorize the index of carrier of the burst found constituting the smallest carrier index of the message.

4 - Method according to claim 2, comprising a step (S13) of decrementing a first pointer (Pm) initially set (E22, E42) to the rank of the carrier frequency of the first burst (FBD, FBU) of the message, until at the rank of the carrier frequency on which the next burst to be searched for is transmitted, when said next burst to be searched for is the first burst in the message having this carrier frequency.

5 - Method according to any one of claims 1 to 4, comprising a step (S71) of searching for the largest carrier index among the bursts (Bl, B2) of the message (ME) transmitted in the first time interval of the message in order to memorize the largest carrier index.

6 - Method according to any one of claims 1 to 4 comprising a step (S71) of incrementing a second pointer (PM) initially set (E22, E42) to the rank of the carrier frequency of the first burst (FBD, FBU ) of the message, each time that the flag (f(BC)) associated with the burst preceding the burst having data to concatenate is representative of an order relationship in the same sense as the ordering of the carrier frequencies, and the burst previous is in the first time slot of the message (S70).

7 - Method according to claims 4 and 6, according to which the concatenation of the bursts of the message is stopped and the message is ignored (S14, S73) as soon as that the difference of the second and first pointers (PM - Pm) is equal to a maximum number (PC) of consecutive carriers on which a message is likely to be distributed.

8 - Method according to any one of claims 1 to 7, according to which the step of concatenating the data of each burst is carried out as long as the size (TA) of the message is less than a maximum message size (NBO).

9 - Method according to any one of claims 1 to 8, according to which the burst position is one of the following three positions: first burst (OB, FBD, FBE, FBU) at the start of the message, intermediate burst (IBE , IBDU, IBDD, IBUU) between the start and end of the message, and last burst

(LBD, LBE, LBU) at the end of the message, and the burst order relationship is one of the following three: less than, equal to and greater than, for a first burst (FBD, FBE, FBU) of message relative to the frequencies of the carriers on which said first burst and the burst following said first burst in the message are transmitted, and for a last message burst (LBD, LBU, LBU) relative to the frequencies of the carriers on which said last burst and the burst preceding said last burst in the message are transmitted.

10 - Method according to any one of claims 1 to 9, according to which the code word (MC) contained in an intermediate burst (IBE, IBDU, IBDD, IBUU) signals two order relationships between the frequency of the carrier on which said intermediate burst is transmitted and the frequencies of the carriers on which the bursts adjacent to said intermediate bursts in the message are transmitted.

11 - Method according to any one of claims 1 to 10, according to which when a message contains only one burst, this unique burst is signaled by a code word (MC = OB) different from the other code words.

12 - Method according to any one of claims 1 to 11, according to which the code words (MC) are of eleven types and each comprise four bits.

13 - Method according to any one of claims 1 to 11, according to which at least one of the code words (OB/LBE, FBE/IBE, FBU/IBDU) is common to a first burst (OB, FBU, FBE) and to an intermediate burst (IBDU, IBE) or to a last burst (LBE) of the message, preferably having common carrier frequency order relationships.

14 - Method according to any one of claims 1 to 13, according to which the code words are of eight types and comprise three bits.