WO2017216597A1

WO2017216597A1 - Deterministic and smoothed re-multiplexing of transport streams within single frequency networks

Info

Publication number: WO2017216597A1
Application number: PCT/IB2016/001129
Authority: WO
Inventors: Ludovic Poulain; Eric Deniau
Original assignee: Enensys Technologies
Priority date: 2016-06-17
Filing date: 2016-06-17
Publication date: 2017-12-21
Also published as: EP3472951A1

Abstract

The present invention proposes a method and a device for deterministic re- multiplexing of transport streams, such as digital television signals primarily intended to be broadcast over Single Frequency Networks. DTT frame delimiters are inserted in streams by head-ends. The resulting streams are received by the transmitters that operate DTT frame after DTT frame. TS packets for a current DTT frame are filtered from the input streams and stored. Once all the TS packets for the current DTT frame have been retrieved, they can be inserted in the output remuxed stream, in the stream portion corresponding to that DTT frame, according to timestamping. The applied deterministic insertion scheme is based on a request management process allowing to smooth packet insertion within the DTT frame, in order to insert a maximum number of retrieved TS packets in the output stream while fulfilling their timing constraints.

Description

DETERMINISTIC AND SMOOTHED RE-MULTIPLEXING OF TRANSPORT STREAMS WITHIN SINGLE FREQUENCY NETWORKS

FIELD OF THE INVENTION

The present invention proposes a method and a device for deterministic re- multiplexing of transport streams, such as digital television signals primarily intended to be broadcast over Single Frequency Networks, SFN, made of transmitter sites synchronously transmitting on the same frequency. In a more general way, the method or device can be applied to all applications where the generation of an output digital signal from one or several incoming streams must be deterministic.

An embodiment of the invention is dedicated to the broadcasting of digital terrestrial television, DTT. Standards for DTT include DVB-T, DVB-T2, ISDB-T, DMB-T, DTMB, ATSC-1 or ATSC-3. Other applications could involve the use of satellite or cable transmission standards, such as the ones defined by DVB, ATSC, DAB, DTMB or ARIB.

BACKGROUND OF THE INVENTION

A Single Frequency Network (SFN) is a broadcasting network where several transmitters simultaneously send the same signal at the same time over the same frequency channel in the same geographical or SFN area. An SFN area is a geographical zone covered by a set of transmitters, the number of which is greater than or equal to one.

The advantage of SFN broadcasting is particularly significant for terrestrial networks. This is because this technique allows bandwidth of the radio spectrum, which is a scarce and hence expensive resource, to be saved. In order to achieve SFN broadcasting, these transmitters need to be finely synchronized and transmit exactly the same output signal on the same frequency at the same time.

The bitstream to be broadcast by the SFN transmitters is generally divided into a continuous sequence of temporal frames, named for instance megaframes in DVB-T and T2 frames in DVB-T2. To correctly drive the transmitters in simultaneously broadcasting the same signal, the transmission time for the temporal frames is defined in the input stream received by the transmitters. For instance, the transmission time of a temporal frame is defined in a corresponding MIP packet for DVB-T and in a T2MI timestamp for DVB-T2. Often, transmitters on terrestrial networks are fed by satellite signals, and local adaptation of content received in input streams is required in an SFN region.

A specific transmitter feeding approach referred to as "single illumination technology" seeks to use an existing DTH stream to also feed the SFN transmitters for DTT broadcasting, thereby avoiding additional bandwidth to be used for transmitting the same television programs to the SFN transmitters.

A DTH, or Direct-to-Home, stream refers to a transmission directly intended for home viewers. The latter only use DTH receivers to process the DTH stream and display television programs. A well-known DTH implementation is conventional DVB-S or DVB-S2 satellite broadcasting, intended for direct reception by end-user devices such as set-top boxes.

A challenge with this single illumination scenario is that, in order for all the transmitter sites to operate in a SFN compliant way in the same SFN area, they need to perform exactly the same adaptation process to produce exactly the same output bitstream from the received input DTH stream or streams. In addition, the input DTH streams should remain directly usable by the home viewers.

Publications WO 2012/1 16743 and WO 2015/91603 are known that disclose methods for deterministic re-multiplexing of transport streams in Single Frequency Networks (SFN) made of transmitter sites synchronously transmitting on the same frequency. The methods comprise the following steps performed at one of the transmitter sites:

receiving one or more input streams;

retrieving, from a received input stream, timing information defining temporal frames for an output SFN stream to be broadcast by the transmitter sites on the same frequency;

generating a remux stream made of remux stream portions, from the one or more input streams, each remux stream portion corresponding in time to a temporal frame defined by the retrieved timing information; and

using the generated remux stream to broadcast an output SFN stream. In WO 2012/1 16743, TMP packets are provided in the DTH input stream, which also includes TS packets. A TMP packet defines a time window, or temporal frame, relative to 1 PPS signal. A number of packets is defined in this temporal frame, given the modulation scheme used, and associated timestamps Τ_Τχ(Ν) are defined. The rules for re-multiplexing at each packet position Τ_Τχ(Ν) of the output SFN stream is as follows: the input TS packet having the lowest time stamp T_ARRIVAL between the start of the temporal frame and T_TX(N) is inserted; if no TS packet satisfies the criteria, a NULL packet is inserted; and all TS packets not yet inserted after a predefined margin TBACKLOG are dropped.

In WO 2015/91603, a reference service with MIP packets and NULL packets is added in the DTH streams at the head-end before broadcasting. This reference service provides the temporal stream structure to be used by the transmitter sites for their output SFN streams. Each TS packet of the received input stream is successively considered to perform the re-multiplexing, provided it has to be forwarded in the SFN. Given an associated timestamp of the received TS packet, it is inserted, on the fly, in replacement of a NULL packet of the output SFN stream that, which NULL packet has a later timestamp T_out(i), obtained from the timestamps provided by the MIP packets.

These approaches are not satisfactory, in particular when the input streams experience bursts in the incoming TS packet rate. Indeed, in that case, a backlog of input TS packets delays their insertion processing.

On one hand, the processing of delayed input TS packets may occur after the next temporal frame starts, resulting in a switching of some TS packets to another temporal frame.

This moving to the next temporal frame is detrimental to the required synchronization between all the transmitter sites.

For instance, transmitters that are restarted during the bursts or just after cannot know which TS packets are switched to another temporal frame. Thus, as long as the backlog has not been cleared, the transmitter sites no longer have the same output SFN stream.

On the other hand, the bursts may lead to unnecessarily deleting TS packets.

For instance, as the backlog of TS packets increases, the number of TS packets not yet inserted after the predefined margin T_BACKLOG also increases. So does the number of TS packets that are finally dropped, thereby resulting in a loss of quality of the output SFN stream. Thus, situations of data overflow or bursts are not efficiently handled.

This inefficiency is exacerbated by the fact that "holes" of TS packets may precede such bursts of packets within the same temporal frame. A consequence is that a large number of NULL packets may be inserted during the "hole" of packets (for instance in the first part of the temporal frame), while numerous relevant TS packets may be dropped at the end of the temporal frame due to a burst occurring during the temporal frame.

Also, the resulting output SFN stream must include Service Information data, composed of tables, which need to be extracted from the input streams and possibly adapted or computed. These operations have to be taken into account during the deterministic re-multiplexing in order to comply with both imposed cycling values and available space in the outcoming stream such as described in WO 2015/91603.

SUMMARY OF THE INVENTION

The present invention seeks to overcome the foregoing concerns.

In this context, embodiments of the invention provide a method for deterministic re-multiplexing as introduced above, wherein generating a remux stream portion corresponding in time to a temporal frame includes:

retrieving, from the received input stream or streams, all the TS packets to be forwarded for broadcast in the SFN that have an associated timestamp belonging to the time range of the temporal frame; and

after all the TS packets to be forwarded have been retrieved, applying an insertion scheme to the retrieved TS packets to insert them as packets in the remux stream portion only, regardless of a transmission time of the packets forming the remux stream portion.

Indeed, the remux stream is to be output for generating the output SFN stream, and thus each packet forming the remux stream is output at a respective transmission time.

According to these embodiments, the TS packets belonging to a temporal frame are inserted only in the corresponding remux stream portion, i.e. the stream portion matching their associated timestamps. A switch from one temporal frame to another for the TS packets is thus avoided, increasing the determinism of re- multiplexing and synchronization between the SFN transmitter sites.

In addition, since the packet insertion process is performed after all the TS packets have been retrieved, better insertion within the remux stream portion can be obtained. In particular, the inventors have noticed that the constraints on the transmission time Τ_Τχ(Ν) or T_ou,(i) of a packet position in the remux stream to allow a TS packet to be inserted at that position in the prior art is useless and prejudicial to efficiency of the re-multiplexing. In this context, these embodiments of the present invention porpose to not take into account these specific transmission times of the packets forming the remux stream portion, for the packet insertion. For instance, there is no comparison between a timestamp associated with a retrieved TS packet to insert and a timestamp (i.e. transmission time) associated with a packet position in the remux stream portion, to allow or not the retrieved TS packet to be inserted at this packet position.

As a result, all the packet positions forming the remux stream portion are available for packet insertion. Thus, a higher number of TS packets can be inserted at the end, even in case of packet bursts. Better packet insertion smoothing is thus achieved, resulting in fewer packet drops.

One may note that since the re-multiplexing method is fully deterministic, all the transmitter sites can perform the same processing as defined above, thus achieving to produce the same remux stream from the same input streams, and then to produce the same output SFN stream. The SFN system requirements are fulfilled.

Correspondingly, the embodiments of the invention also provide a deterministic re-multiplexer or deterministic DTT stream adapter for re-multiplexing transport streams at a transmitter site in a Single Frequency Network, SFN, made of transmitter sites synchronously transmitting on the same frequency, the deterministic re-multiplexer comprising:

a receiver for receiving one or more input streams;

a metadata extractor for retrieving, from a received input stream, timing information defining temporal frames for an output SFN stream to be broadcast by the transmitter sites on the same frequency;

a re-multiplexing unit for generating a remux stream made of remux stream portions, from the one or more input streams, each remux stream portion corresponding in time to a temporal frame defined by the retrieved timing information; and

an SFN output module for using the generated remux stream to broadcast an output SFN stream,

wherein the re-multiplexing unit generating a remux stream portion corresponding in time to a temporal frame includes:

a packet parser for retrieving, from the received input stream or streams, all the TS packets to be forwarded for broadcast in the SFN that have an associated timestamp belonging to the time range of the temporal frame; and

a packet insertion module for applying, after all the TS packets to be forwarded have been retrieved, an insertion scheme to the retrieved TS packets to insert them as packets in the remux stream portion only, regardless of a transmission time of the packets forming the remux stream portion.

The deterministic re-multiplexer has the same advantages as the method defined above

Optional features of embodiments of the invention are defined in the appended claims. Some of these features are explained here below with reference to a method, while they can be transposed into system features dedicated to any deterministic re-multiplexer according to embodiments of the invention.

In embodiments, the timing information defining the temporal frames is retrieved from a private section of TS packets composing a reference service within the received input stream. The private section in the TS packets may be as defined in ISO/IEC 13818-1 standard.

This configuration ensures the input streams to be kept compatible with DTH transmission. Indeed, as being part of private data, the timing information for implementation of the present invention does not impact conventional DTH receivers.

In specific embodiments, a TS packet of the reference service includes a conventional PCR field and, within its private section, a frame PCR sampling the time at which a temporal frame starts in the input streams. In particular, the time range of the temporal frame may be defined by two consecutive frame PCRs provided in the reference service.

According to a specific feature, the private section further includes a synchronization timestamp defining a transmission time of a temporal frame for broadcasting in the SFN. In other words, the TS packet considered in the reference service includes three separate items of information: conventional PCR timestamping the TS packet, frame PCR according to the invention which defines a temporal frame, and synchronization timestamp that indicates the transmission time of same temporal frame in the Single Frequency Network.

The frame PCR and the synchronization timestamp are based on different clocks, namely a 27MHz GPS-based clock for the PCR counter while it is a T2 clock for the synchronization T2MI timestamp or a 10MHz clock for the synchronization MIP timestamp (STS). As a consequence, providing the above three items guarantees efficient determinism in the generation of an output SFN stream at the SFN transmitter sites.

Thus, another invention may be directed to a method for deterministic re- multiplexing as introduced above with reference to the prior art, which is supplemented with the provision of the above three items of timing information through a reference service. PCR_fr and synchronization timestamp may be preferably, but not necessarily, included in a private section of the reference service TS packets.

In other specific embodiments, a private section of TS packets composing the reference service includes modulation parameters to be used by the transmitter site for broadcasting the output SFN stream. More generally, the private sections of the TS packets forming the reference service may be used to define information to control the transmitter sites to broadcast in the single Frequency Network.

The private section of a single TS packet may be used to convey such information for each temporal frame. However, due to the size of such information, the latter may be split into several TS packets: for instance a first packet including the timing and frame structure information, and one or more subsequent packet including the other information such as the modulation parameters.

In some embodiments, at least two input streams are received by the transmitter site that both include a separate reference service embedding timing information defining temporal frames, and the method further comprises selecting one of the input streams as a master input stream from which the timing information is retrieved. The same rule to select the master input stream is used by all the transmitter sites forming the same SFN region.

In some embodiments, retrieving all the TS packets to be forwarded for broadcast includes storing TS packets to be forwarded that belong to a first input stream into a first memory module and storing TS packets to be forwarded that belong to a second and separate input stream into a second and separate memory module. In other words, the TS packets are stored in various memories depending on the input stream to which they belong. The present invention does not limit the number of input streams to one or two. There are as many memory modules for storing audiovisual TS packets as the number of input streams.

Using separate memory modules advantageously avoids having simultaneous access to the same memory to be handled.

Note that the TS packets to be forwarded may be filtered from the input streams using conventional filtering technics.

In specific embodiments, a first memory bank in a memory module is used to store TS packets being retrieved for the temporal frame, while a second and separate memory bank in the same memory module stores all the TS packets already retrieved for a previous temporal frame and is simultaneously used to generate the remux stream portion corresponding to the previous temporal frame. Thus a switch between the two memory banks can be made when the processing switch from one temporal frame to the other.

Simultaneous processing can thus be efficiently obtained, thanks to the frame basis approach of the present invention.

In other specific embodiments, the method further comprises generating signaling tables (e.g. DVB/SI tables) for the temporal frame from the received input stream or streams and storing each table, as one or more TS packets, in a dedicated memory module. Additional memory modules are thus used.

The signaling tables can be built from scratch from the input streams, or can be formed by a mere filtering of signaling tables existing in the input streams.

In other specific embodiments, the method further comprises generating a Mega-frame Initialization Packet, MIP, for the temporal frame from modulation and timing information contained in a private section of TS packets composing a reference service within the received input stream, and storing the MIP as a TS packet in a dedicated memory module. An additional memory module is thus used.

This provision applies for instance for DVB-T or DVB-H standards.

In some embodiments relating to the insertion of the TS packets, applying an insertion scheme includes the following step by each of the one or more memory modules storing one or more TS packets retrieved from the input stream or streams for the temporal frame:

successively sending, to a packet insertion module during the time length of the temporal frame, one or more requests for successively inserting the stored TS packet or packets in the remux stream portion corresponding to the temporal frame.

The packet insertion module is thus the module dedicated to perform, over time, the insertion of the TS packets from the memory modules into the remux stream portions. Preferably, the memories are of a FIFO type so that the oldest TS packets are inserted first and the same order of TS packets as in the input stream or streams is kept.

One may note that when a single TS packet is stored in the memory module, a single request is sent during the time length of the temporal frame, at an appropriate time instant. This is for instance the case for the MIP packet which should be inserted as the last packet in the remux stream portion corresponding to the current temporal frame. According to a specific feature, successively sending one or more requests includes periodically sending a request for inserting a respective stored TS packet in the remux stream portion, wherein the sending period is computed based on a ratio between a number of TS packets to be inserted and stored in the memory module, and a number of packets composing the temporal frame.

The number of packets composing the temporal frame can be computed from the rate of the remux stream, which itself derives from the modulation parameters preferably embedded in the input stream as defined above.

This provision smooths the insertion of the TS packets over the whole remux stream portion corresponding to the current temporal frame.

In other embodiments relating to the insertion of PSI/SI table, applying an insertion scheme includes the following step by one memory module storing a signaling table as one or more TS packets for the temporal frame:

successively sending, to a packet insertion module during the time length of the temporal frame, requests for successively inserting a multiple number (preferably 16) of the stored table in the remux stream portion corresponding to the temporal frame.

By choosing an appropriate number of copies of the table inserted in the remux stream portion, the management of a continuity counter of the tables inserted in the output SFN stream is simplified. For instance, inserting the same number of tables as the number of possible values for the continuity counter ensures the continuity counter to start with the same value at each new temporal frame.

For instance in the case of DVB-T ou DVB-T2, sixteen identical tables are preferably inserted in the remux stream portion. This is because the continuity counter is a 4-bit value.

In other embodiments relating to the insertion of the TS packets, applying an insertion scheme includes the following step by one memory module storing a signaling table as TS packets for the temporal frame:

successively sending, to a packet insertion module during a table period corresponding to a multiple (preferably 16) of the time length of the temporal frame, requests for successively inserting the stored TS packets in the remux stream portion corresponding to the temporal frame.

In particular, if the number of stored TS packets is not aligned with the number of possible values of a continuity counter for the signaling table, a stuffing table with a discontinuity indicator can be added in the memory module at the end of the stored TS packets. This simplifies the management of a continuity counter in the tables inserted in the output SFN stream.

This provision may apply for instance to SDT and EIT tables that are quite large. It thus limits the bitrate of the signaling tables, in particular due to large signaling tables.

In some embodiments, the packet insertion module gives priority, for TS packet insertion at a current packet position of the remux stream portion, to a memory module for which a higher number of requests has not yet been served compared to a memory module for which a lower number of requests has not yet been served.

This approach contributes to efficiently smooth the insertion of the TS packets over the whole temporal frame.

In specific embodiments, at same number of requests not yet been served, the packet insertion module gives priority to a memory module storing TS packets of an input stream compared to a memory module storing signaling tables.

According to a specific feature, the packet insertion module gives the highest priority to insertion of a Mega-frame Initialization Packet, MIP. In other words, as soon as a request to insert a MIP is received, the packet insertion module performs the insertion of this MIP in the remux stream portion. Of course, this applies for DVB-T or DVB-H standard. As it derives from these standards, such request and thus insertion are preferably done for the last packet position in the remux stream portion (corresponding to the last position in the megaframe). This is because the MIP usually announces the start time of the next megaframe.

In some embodiments, the packet insertion module and the memory modules are cadenced at a packet rate of the remux stream. In other words, the requests and the insertions are performed at the same rhythm, at each new packet position in the remux stream portion.

In some embodiments, each time a TS packet retrieved from the input stream or streams (i.e. usually audiovisual TS, and not signaling table packets) is inserted in the remux stream portion, it is deleted from the corresponding memory module, and the TS packets remaining in the memory module or modules at the end of the remux stream portion are discarded. Although the present invention makes it possible to smooth the insertion of bursts of TS packets, a limit is that no more TS packets can be inserted than the number of packets available in the remux stream given its bitrate (or modulation parameters). As a consequence, the full determinism of the present invention is made at the cost of having some TS packets discarded or dropped in case of heavy bursts.

In some embodiments specific to the DVB-T2 standard, the method further comprises, by the transmitter site, generating a T2-MI stream from the remux stream, wherein T2 frames of the T2-MI stream are aligned with the remux stream portions (and thus are aligned with the temporal frames defined by the timing information embedded in the input stream). The alignment is to be understood with respect to the TS packets: all and only the TS packets forming a remux stream portion (or current temporal frame) are used to feed the BaseBand frames (or T2-MI packets) of a corresponding T2 frame.

This illustrates an application of the present invention to the generation of DTT signals.

In other embodiments, two input streams are received from two separate head-ends. Of course, a higher number of input streams may be received from a variable number of head-ends or national/regional transmission sites.

In some embodiments, the method further comprises inferring timing information for a next temporal frame from the timing information retrieved for a previous temporal frame. This is to offer robustness of the deterministic re-multiplexing in case of reception errors.

At least parts of the method according to the invention may be computer implemented. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, micro-code, etc.) or an embodiment combining software and hardware aspects that may all generally be referred to herein as a "circuit", "module" or "system". Furthermore, the present invention may take the form of a computer program product embodied in any tangible medium of expression having computer usable program code embodied in the medium.

Since the present invention can be implemented in software, the present invention can be embodied as computer readable code for provision to a programmable apparatus on any suitable carrier medium. A tangible carrier medium may comprise a storage medium such as a hard disk drive, a magnetic tape device or a solid state memory device and the like. A transient carrier medium may include a signal such as an electrical signal, an electronic signal, an optical signal, an acoustic signal, a magnetic signal or an electromagnetic signal, e.g. a microwave or RF signal. BRIEF DESCRIPTION OF THE DRAWINGS

Further advantages of the present invention will become apparent to those skilled in the art upon examination of the drawings and detailed description. Embodiments of the invention will now be described, by way of example only, and with reference to the following drawings.

Figure 1 shows an example of an overview of a broadcasting system 1 in which the present invention may be implemented;

Figure 2 schematically illustrates a deterministic DTT stream generator in a head-end of Figure 1 ;

Figure 3 illustrates an embodiment of reference service insertion in an input DTH-compatible MPTS, for implementation of the present invention;

Figures 4 and 5 illustrate exemplary implementations for declaring DTT markers in private sections of TS packets in a DTH stream, according to embodiments of the invention;

Figure 6 schematically illustrates a deterministic DTT stream adapter according to embodiments of the invention;

Figure 7 illustrates an exemplary implementation of the deterministic DTT stream adapter of Figure 6 with more details;

Figure 8 illustrates a timestamping of input TS packets based on a reference service;

Figure 9 illustrates the assignment of input TS packets to temporal frames, according to embodiments of the invention; and

Figure 10 illustrates an exemplary management of a memory module for storing EIT data.

DETAILED DESCRIPTION

The invention will now be described by means of specific non-limiting exemplary embodiments and by reference to the figures.

Figure 1 shows an example of an overview of a broadcasting system 1 implementing single illumination technology. Embodiments of the present invention may be implemented in such system to perform a deterministic SFN re-multiplexing at SFN transmitter sites.

The figure shows that one or main DTH streams 20_!, 20₂ are generated that can be used by DTH consumers 30, using conventional DTH receivers. The DTH streams may be generated by one or more DTH head-ends 10_1t 10₂, which may be national and/or regional head-ends.

As depicted in the Figure, the DTH streams 20i, 20₂ are conveyed over satellite links from the head-ends 10i, 10₂, meaning that the transport streams TS they comprise are encapsulated over DVB-S or DVB-S2 physical layer. This is however not a requirement, and other types of distribution systems and standards may be used.

As conventionally known, each DTH stream is merely a transport stream made of a sequence of multiplexed TS packets according to the ISO/I EC standard 13818-1. The DTH stream includes services, also known as programs, which are made of elementary streams (identified by PIDs) within the DTH stream. The services or programs are signaled through signaling tables.

The principle of single illumination technology is to generate a same DTH stream at the head-end that can be on one side received by DTH consumers 30 through any DTH receiver, and on the other side processed by adapters provided at SFN transmitter sites to generate DTT multiplexes that are SFN-compatible.

Figure 1 shows two separate SFN areas 4i, 4₂ (more generally reference 4) that use two separate single frequencies to broadcast a DTT multiplex.

At each transmitter site 40^ 40₂, 40₃, 40₄ (more generally reference 40) in an SFN region, a deterministic DTT stream adapter or re-multiplexer 41 1 , 41 2, 41₂- 1 , 41₂-2 (more generally reference 41 ) receives the different DTH streams as inputs, generates an output DTT stream 42 1 , 42 2, 42₂-1 , 42₂-2 (more generally reference 42) from the input DTH streams, and supplies it to a demodulator 43i-1 , 43i-2, 43₂-1 , 43₂-2 (more generally reference 43) for SFN broadcasting over the single frequency of the SFN area 4.

The output DTT stream 42 may be generated according to any DTT standard, for instance according to the DVB-T standard (in which case it includes MIP packets) or the DVB-T2 standard (in which case it is a T2-MI stream). Of course, other standards such as ATSC or DMB-T may be contemplated.

An issue with the single illumination technology is that all SFN transmitter sites in the same SFN area need to produce exactly the same output SFN stream from the received input DTH stream or streams, to fulfil the SFN system requirements. In the same time, the input DTH streams may not be substantially modified, so that they remain directly usable by the DTH consumers 30.

To allow deterministic SFN re-multiplexing at the SFN transmitter sites, the head-ends should produce a stream that is compatible with both DTH reception and DTT stream generation. To do so, they add additional information in the DTH stream. This is made by deterministic DTT stream generators, DDSGs, 12-,, 12₂ (more generally reference 12) depicted in the Figure, as described below.

One type of head-end is show in reference 10i which comprises a conventional program and SI generator 11 i generating a multi-program transport stream, MPTS, including a plurality of services and corresponding signaling tables. The MPTS complies with MPEG and DVB standards and is intended for DTH transmission.

The MPTS is fed to deterministic DTT stream generator 12i which adds DTT additional information to allow deterministic SFN re-multiplexing at the SFN transmitter sites.

As described below, deterministic DTT stream generator 12i may insert a reference service in the incoming TS stream, which reference service carries the additional information. This is however not a requirement, and a single elementary stream (PID) without defining a service in the signaling tables may be used, provided all the SFN transmitter sites 40 know it.

The DTH TS stream adapted with the inserted reference service is then fed to a modulator, preferably a DVB-S or DVB-S2 modulator, for broadcast (20i) to the DTH consumers 30 and SFN transmitter sites 40.

Another type of head-end is show in reference 10₂. A conventional program and SI generator 1 1₂-2 generated a multi-program transport stream, MPTS, including a plurality of services and corresponding signaling tables. The MPTS complies with MPEG and DVB standards and is intended for DTT transmission.

This DTT-intended MPTS is fed to deterministic DTT stream generator 12₂ which operates as described above to insert a reference service with additional information to drive the SFN transmitter sites to perform deterministic SFN re- multiplexing.

The resulting adapted DTH TS stream is fed to an MPEG re-multiplexer 14 which also received another MPTS from conventional program and SI generator 1 1₂-1. This other MPTS complies with MPEG and DVB standards and is intended for DTH transmission.

The two MPTS are multiplexed and the resulting multiplex feeds modulator 13₂ for broadcast (20₂) over satellite links.

The multiplex can be received by any DTH receiver 30, and can also be processed by the SFN transmitter sites 40 according to embodiments of the invention to regenerate the DTT multiplex. Although generator 11₂-2 and deterministic DTT stream generator 12₂ are shown located in the same head-end as generator 11₂-1 , re-multiplexer 14 and modulator 13₂, they can be located into another head-end which feeds DTH head-end 10₂.

The operations at the head-ends to insert the additional information for deterministic SFN re-multiplexing at the SFN transmitter sites are further described below with reference to Figures 2 to 5. The additional information include timing information defining temporal frames, from which a split of the input streams can be made in a fully deterministic fashion by the SFN transmitter sites 40.

As proposed below, each deterministic DTT stream generator 12 inserts two sets of data (forming metadata) in the input MPTS, more particularly through one or more TS packets of a reference service:

- First, a conventional PCR value (as described in ISO/I EC standard 13818-1 ) corresponding to the local PCR counter sampled when the packet of the reference service is inserted in the MPTS;

- Second, DTT markers containing required information to perform the DTT transmission. DTT markers include for instance either information related to megaframe and T configuration for a DVB-T transmission, or information related to T2 frame and T2 configuration for a DVB-T2 transmission. In particular, it includes (MIP or T2MI) timestamp, as well as modulation parameters. Other information may also be included.

At each SFN transmitter site 40, a deterministic DTT stream adapter 41 1 , 41 2, 412-1 , 412-2 (more generally reference 41 ) is provided according to embodiments of the invention that generates an output DTT stream from the received DTH streams 20, for SFN broadcast.

To do so, a deterministic (MPEG) re-multiplexing of transport streams in Single Frequency Networks (SFN) made of transmitter sites synchronously transmitting on the same frequency is conducted. It aims is to generate one remux TS stream containing all services that will be present (i.e. need to be forwarded for the SFN area) in the output DTT stream 42, from the incoming DTH streams 20, and to update PSI/SI tables where appropriate.

The remux TS stream can then be encapsulated using any requirement of a DTT standard (e.g. DVB-T or DVB-T2) to create an output DTT stream feeding modulator 43 for SFN broadcast. A method for deterministic re-multiplexing TS streams comprises several steps performed at one of the transmitter sites, and preferably at each transmitter site 40 of one or all SFN areas 4:

receiving one or more input streams 20;

retrieving, from a received input stream 20, timing information defining temporal frames for an output SFN stream to be broadcast by the transmitter sites 40 on the same frequency;

generating a remux stream made of remux stream portions from the one or more input streams, each remux stream portion corresponding in time to a temporal frame defined by the retrieved timing information; and

using the generated remux stream to broadcast an output SFN stream.

In an inventive deterministic re-multiplexing method according to embodiments of the invention, generating a remux stream portion corresponding in time to a temporal frame includes:

retrieving, from the received input stream or streams, all the TS packets to be forwarded for broadcast in the SFN that have an associated timestamp belonging to the time range of the temporal frame;

Thus each temporal frame as defined in the input streams is processed one after the other, thereby allowing a deterministic split of the input streams to be obtained at each transmitter site. The use of the same insertion scheme by the transmitter sites ensures the re-multiplexing to be fully deterministic.

Furthermore, working (for insertion) on the whole temporal frame without constraint on the transmission time of each packet forming the corresponding remux stream portion advantageously allow each packet of the remux stream portion to receive any TS packet retrieved from the input streams. Efficient packet insertion smoothing is thus achieved, resulting in fewer packet drops compared to known technics.

Operations at the transmitter sites 40, and more particularly at the deterministic DTT stream adapters 41 are further described below with reference to Figures 6 to 10. Figure 2 schematically illustrates a deterministic DTT stream generator 12 in a head-end 10 which receives an input DTH-compatible MPTS from a program and SI generator 11. The deterministic DTT stream generator 12 inserts additional information in the DTH-compatible MPTS for controlling transmitter sites 40 in their process of generating and broadcasting SFN streams.

The insertion of the additional information is done so that the DTH- compatible MPTSs remain compatible with DTH transmission, i.e. it remains directly usable by DTH receivers.

To achieve that, the deterministic DTT stream generator 12 inserts, in it, a reference service (in the meaning of DVBs service or MPEG programs) dedicated to convey the additional information. For instance, the service type for the reference service is chosen as a user-defined service type, e.g. service type equal to 6.

The reference service is made of TS packets that each includes a conventional PCR field (ISO/IEC 13818-1 ) in the Adaptation field of the TS packet. TS packets of the reference service are used to convey the additional information or "markers", preferably in a section defined as being private in the above-mentioned standard.

In the case of a DVB-T transmission by the transmitter sites 40, MIP marker information is provided in private sections of the TS packets. In the case of a DVB-T2 transmission, the L1 current is also contained in such private sections.

Deterministic DTT stream generator 12 receives a GPS signal to drive a PCR counter 120. For instance, the PCR counter is generated on a 27MHz clock synchronous to the GPS clock.

In case of a plurality of head-ends 10, the various deterministic DTT stream generators 12 shall have the same PCR counter, i.e. synchronous counters based on the GPS clock.

As an absolute timestamp is required to generate the DTT timestamp (e.g. T2 timestamp or MIP timestamp), the absolute timestamp may be generated in a deterministic fashion by counting PCR clock pulses after a reference date, for instance January 1^st, 2016. Indeed, the reference date initializes the PCR counter to zero, and after each pulse clk_27, the counter PCR is incremented.

So when a deterministic DTT stream generator 12 starts, it first waits for the PPS and the GPS time from the GPS receiver, and then computes the number of seconds since the reference date. The current value of the PCR counter may thus be inferred given the number of clk_27 pulses per second. Next, at each new pulse clk_27, the PCR counter is incremented.

Also at each PPS pulse, the current value of the PCR counter is compared to the theoretical value of the PCR counter (given the reference date). If the difference between the two values is superior to 1 ms, the PCR counter is adjusted to take the theoretical value. This allows PCR drift to be regularly corrected between the various head-ends 10.

The PCR counter is thus available to DTT marker generator 121 and reference service inserter 122, in particular to set the PCR value of each reference service TS packet inserted in the DTH-compatible MPTS.

The deterministic DTT stream generator 12 is configured to a specific DTT standard. For instance, if it is configured for DVB-T, its DTT marker generator 121 works locally as a MIP generator to produce the additional information including MIP information (MIP timestamp, MIP continuity counter). On the other hand, if it is configured for DVB-T2, its DTT marker generator 121 works locally as a T2 Gateway to produce additional information including T2MI information (T2MI timestamp, T2MI- Continuity Counter, T2MI-L1 ).

DTT modulation settings 123 feed DTT marker generator 121 to generate DTT markers.

It includes generating pulses at the time when a DTT temporal frame starts, for instance a megaframe for DVB-T or a T2 frame for DVB-T2. The PCR counter 120 is sampled at each start time, producing a frame PCR, noted PCR_fr, Periodicity of the pulses is defined by the modulation parameters used for SFN broadcast.

It also includes generating conventional DTT markers for each successive DTT temporal frame as mentioned above, in particular MIP timestamp, MIP payload (modulation parameters, ...) and MIP continuity counter for DVB-T, or T2MI timestamp, T2MI-Continuity Counter (T2MI superframe index, T2MI frame index, T2MI packet count), T2MI-L1 for DVB-T2.

The DTT markers are fed to reference service inserter 122 which build TS packets based on these DTT markers. The DTT markers are inserted in private sections of the TS packets. The TS packets, timestamped with conventional PCR, are inserted in the DTH-compatible stream in replacement of NULL packets, hence requiring no bitrate adaptation.

Therefore, a TS packet of the reference service includes a conventional PCR field and, within its private section, a frame PCR, PCR_fr, sampling the time at which a temporal frame starts in the streams. This information will help the transmitter sites 40 to retrieve TS packets belonging to each temporal frame for SFN broadcast.

As successive PCR_fr are inserted in the DTH stream, the time range of the temporal frame is defined by two consecutive frame PCRs provided in the reference service.

To provide the other DTT markers, the private section of the TS packets further includes a synchronization timestamp (MIP timestamp or T2MI timestamp for instance) defining a transmission time of the temporal frame for broadcasting in the SFN. This is to efficiently drive the transmitter sites 40 in SFN broadcasting an output SFN stream.

Also, a private section of TS packets composing the reference service includes modulation parameters to be used by the transmitter site for broadcasting the output SFN stream.

Figure 3 illustrates an embodiment of reference service insertion in an input DTH-compatible MPTS.

The input DTH-compatible stream received by deterministic DTT stream generator 12 is shown on top of the Figure. It shows a plurality of TS packets, some of which are NULL packets. The DTT frame pulses are also indicated to show the timing relationship between them and the TS packets.

In the lower part of the Figure, the adapted stream is shown, wherein TS packets of the reference service (shown as "ref in the Figure) have been inserted.

In this example, the first NULL packet after the DTT frame pulse (i.e. first NULL packet with a PCR following PCR_fr) is replaced by a service reference packet which contains the conventional PCR value (sampled at the transmission time of that TS packet), as well as some DTT marker information in a private section. Preferably, the DTT marker information of this first service reference packet in the DTT frame includes the frame PCR, PCR_fr, defining the start time of the DTT frame, and the synchronization DTT timestamp (MIP timestamp or T2MI timestamp for instance) defining a transmission time of the temporal frame for broadcasting in the SFN.

This first service reference packet in the DTT temporal frame thus includes three items of timing information: conventional PCR, frame PCR and DTT timestamp.

It is not mandatory that the timing information is included in the first inserted packet (i.e. in replacement of the first NULL packet following the DTT frame pulse). Any other NULL packet within the temporal frame can be used. However, forcing to use the first NULL packet advantageously guarantees the timing information to be transmitted as soon as possible.

Figure 4 illustrates an exemplary implementation for declaring these DTT markers in the private section of a first TS packet as defined in ISO/IEC 13818-1 , for both DVB-T standard (Figure 4a) and DVB-T2 standard (Figure 4b).

The private section length is expressed in bytes. In Figure 4a, MIP CC corresponds to the continuity counter for the current MIP packet, while STS is the synchronization MIP timestamp.

In Figure 4b, T2MI superframejdx is the superframe index for the current temporal frame (corresponding to a T2MI/T2 frame). It is incremented at each new superframe as defined in DVB-T2 standard. T2MI framejdx is the frame index of the current temporal frame within the current superframe. It is incremented at each new temporal frame. T2MI packet_count is the T2MI counter of the first packet of the temporal frame. BW is the bandwidth intended for SFN broadcast (because it is not specific in T2MI L1 packet). T2 timestamp is the synchronization T2MI timestamp for the current temporal frame. seconds_since_2000 is an absolute timestamp. Maximum network delay defines an allowed maximum delay (in ms) for the SFN. Next, the BUFS parameters for each PLP are defined.

Back to Figure 3, the second NULL packet encountered in the input DTH- compatible stream is replaced by a second service reference packet which contains a conventional PCR and additional DTT markers in a private section, in particular DTT markers relating to modulation parameters (MIP payload for DVB-T and T2-L1 packet for DVB-T2). This it contains the full set of DVB-T or DVB-T2 parameters (and more generally DTT parameters) to be used by the SFN transmitter sites 40 to generate the DTT frame pulses.

This additional DTT markers related to modulation parameters may be larger than the possible payload of a single TS packet, for instance in the case of DVB- T2 where a set of modulation parameters is defined for each physical layer pipe (PLP). In that case, they can be split into two or more TS packets that are successively inserted in NULL packets of the input DTH-compatible stream.

Figure 5 illustrates an exemplary implementation for declaring these DTT markers in the private section of a second TS packet as defined in ISO/IEC 13818-1 , for both DVB-T standard (Figure 5a) and DVB-T2 standard (Figure 5b). One may note that all the DTT parameters required by the transmitter sites 40 to generate the structure of the output SFN signal are provided through these private sections.

As shown in Figure 3, deterministic DTT stream generator 12 may further replace a NULL packet of the input DTH-compatible stream at regular intervals (typically 40 ms complying with conventional requirements as defined for instance in ETR 290) by a service reference packet which contains only a conventional PCR value.

Due to NULL packet replacement, no bitrate adaptation is required. The bitrate of the reference service to convey the above metadata as described in Figures 3 to 5 is evaluated to be about 40 Kbit/s.

The reference service bitrate may be reduced if needed, for instance, if the input DTH-compatible stream offers very low bitrate of NULL packets. This is to offer more data bitrate to the television or the like programs.

A first step for bitrate reduction may be obtained by inserting less often the PCR-only packets, or even by disregarding them.

A second and additional step for bitrate reduction may be obtained by inserting the additional DTT markers related to modulation parameters with a lower rate. That is less often.

In all case, the first reference service packet with includes PCRfr and the synchronization DTT timestamp is kept inserted for each temporal frame. This is because each temporal frame needs to be declared as soon as possible for the transmitter sites 40 to be able to deterministically form the corresponding remux stream portions.

To illustrate this, in case of DVB-T with a megaframe period equals to 0.6 s, inserting modulation DTT markers every 5 s (without PCR-only packets being inserted) makes it possible to reduce the bitrate of the metadata to about 3 Kbit s.

In case of DVB-T2 with a T2 frame period equals to 0.25 s, inserting modulation DTT parameters every 5 s (without PCR-only packets being inserted) makes it possible to reduce the bitrate of the metadata to about 7 Kbit/s.

It can be noted that in case of remultiplexing of the resulting stream (for instance through multiplexer 14 of DTH head-end 10₂ of Figure 1 , the conventional PCR contained in Adaptation field of each reference service packet is updated using conventional mechanisms, while, on the other hand, the value PCR_fr remains unchanged. This is because it is stored in a private part of the reference service packets. Thus, despite modification of the multi-program transport stream, the value PCR_fr always refers to the same instant of DTT temporal frame generation.

One may also note that compared to above-mentioned known publications, no MIP packets are directly inserted as TS packets in the DTH streams transmitted in the satellite broadcasting network.

Deterministic DTT stream generator 12 does not perform any processing on the signaling tables (and more generally any MPEG PSI/DVB SI data) provided in the input DTH-compatible stream for DTH broadcast. At most, the reference service can be added to these data and thus declared in the PAT table. However, this is not mandatory, and ghost PIDs known by each transmitter site 40 may be used.

If other and specific PSI/SI data are required for DTT transmission (i.e. each transmitter site 40 must know to generate an appropriate output SFN stream), they can be added in ghost PIDs (not corresponding to mandatory PIDs defined in DVB standard) not to disturb the DTH transmission/reception. Examples of such other and specific PSI/SI data include NIT table for the DTT multiplexes, SDT tables for other DTT multiplexes, EIT tables for other DTT multiplexes.

The DTH stream or streams 20 thus generated by the DTH head-ends 10 are broadcast to the transmitter sites 40 of one or more SFN areas 4. In each SFN area, these streams are processed by the transmitter sites to generate the same output SFN signal for SFN broadcasting in the area. This processing involves a deterministic re-multiplexing according to embodiments of the invention.

Figure 6 schematically illustrates a deterministic DTT stream adapter 41 according to embodiments of the invention. The deterministic DTT stream adapter 41 receiving one, two or more input DTH streams to generate a SFN compatible DTT stream to feed DTT modulator 43.

To do that in a similar way as the other transmitter sites 40 of the same SFN are, a splitting of the input streams into temporal frames is made in a deterministic fashion, based on timing information retrieved from one received input stream. Next, all the TS packets to be forwarded for broadcast in the SFN that have an associated timestamp belonging to the time range of a current temporal frame are retrieved from the received input stream or streams, and the same insertion scheme is applied to them after all these TS packets to be forwarded have been retrieved. The insertion scheme inserts them as packets in an output remux stream, in particular only in a portion thereof that correspond to the current temporal frame, regardless of a transmission time of the packets forming the remux stream portion. Due to this deterministic re-multiplexing, the streams generated by the one deterministic DTT stream adapters 41 of the same SFN area 4 allow exciters of modulators 43 to generate RF signals strictly identical one to each other, based on DTT configuration (in particular modulation parameters) embedded in the received input streams.

The deterministic DTT stream adapter 41 also updates PSI/SI tables accordingly

In some cases, a terrestrial standard encapsulation in order to provide the remux stream in an appropriate format for the standardized modulator 43.

The Figure illustrates both implementations to DVB-T and DVB-T2 standards.

For instance, in the particular case of DVB-T2, a T2MI encapsulation process is performed in order to generate a T2MI stream 42 to be fed to the DVB-T2 modulator 43.

In the case of DVB-T, the deterministic re-multiplexing may directly include

MIP packets in the remux stream, so that the latter is already formatted to be fed to the DVB-T modulator 43 for SFN broadcast.

Figure 7 illustrates an exemplary implementation of the deterministic DTT stream adapter 41 with more details.

It comprises pre-processing units 410 for timestamping the TS packets of the n received input streams 20 (n being 1 or more); a metadata unit 420 for retrieving relevant metadata from the input streams 20; a re-multiplexing unit 430 for generating a deterministic remux stream from the input streams, given the retrieved metadata; and a DTT encapsulator 450 to format the remux stream into a DTT stream to be fed to modulator 43.

It is recalled here that each input DTH stream 20 received by the deterministic DTT stream adapter 41 includes a reference service made of TS packets with conventional PCRs.

Each pre-processing unit 410 operates on one single input stream 20. It generates a timestamp for each TS packet of the input stream based on the timestamp (conventional PCR) of the packets forming the reference service. This makes it possible to have a timestamping of each TS packet that is independent of the arrival time of the packets.

Preferably, as illustrated through Figure 8, the reference service PCR is used to timestamp each TS packet located between two reference service packets. Let be Ref(i) and Ref(j) the PCRs of two consecutive packets Ϊ and 'j' in the reference service of the input stream 20, each TS packet 'p' between them is timestamped T_in(p) according to a linear approximation.

In particular, the following formula is used:

, PCR(Ref(j)) - PCR (Ref(Q)

T_in p) = PCR(Ref(i)) + ^{, KJ))} _N L-L y^. * _Ni where N is the number of TS packets between the two consecutive reference service packets and Nj is the number of TS packets between packet 'p' and the first reference service packet Ϊ.

The operations of the pre-processing unit 410 are continuously performed as the input streams are received.

In parallel, the metadata unit 420 operates on the input streams to obtain the relevant metadata for DTT stream generation.

In particular, it is in charge of retrieving, from a received input stream, timing information defining DTT temporal frames for an output SFN stream to be broadcast by the transmitter sites on the same frequency, that is PCR_fr values inserted in the private sections of the reference service packets.

It includes a master selector 421 to select a reference stream from amongst the received input streams 20 (if there is a plurality of them). This is to get a time reference for the re-multiplexing operation, that can be the same selected one for all the transmitter sites of the SFN area.

By default, input stream 1 is selected as the master input. Input stream 1 may be the first physical entry in the deterministic DTT stream adapter 41 , being noted that the same input configuration is used by all the transmitter sites.

Once the master input is known, the metadata from this input stream are retrieved by the metadata parser or extractor 422 and are stored in dedicated memory 423. This includes the retrieval of the frame PCRs, PCR_fr, defining the time limits of the DTT temporal frames for the DTT stream to be generated, as well as the retrieval of other DTT parameters such as the synchronization timestamp (MIP or T2MI timestamp) and the modulation parameters.

More precisely, the metadata from the private parts as shown in Figures 4 and 5 are retrieved by the metadata parser 422.

In sequence, the metadata extractor 422 parses the master input stream and extracts the private section shown in Figure 4 from the reference service to obtain the frame PCR, PCR_fr(n), for current temporal frame 'n' (PCR time for the generation of megaframe/T2frame). It also retrieves the associated parameters, such as t2mi_packet_count, framejdx and super_frame_index in case of DVB-T2 or the continuity counter of the MIP packet and the STS in case of DVB-T.

It then extracts the modulation-related marker information as shown in Figure 5. At this moment, the deterministic DTT stream adapter 41 knows the modulation parameters (bandwidth, Guard interval, code, in case of DVB-T2 PLP configuration) and can reconstruct the MIP packet in case of DVB-T, or the T2MI L1 packet in case of DVB-T2.

The operations of the metadata unit 410 are continuously performed as the master input stream is received, in particular to obtain the PCR_fr values.

Based on the PCR_fr values, the re-multiplexing unit 430 is aware of the time boundaries each DTT temporal frame N. In particular, DTT temporal frame starts at PCR_fr(N) and ends at PCR_fr(N+1 ), where N+1 is the temporal frame following frame N as defined in the reference service by the corresponding head-end.

The re-multiplexing unit 430 operates in two steps for a given temporal frame.

In a first step, it retrieves, from the received input stream or streams, all the TS packets to be forwarded for broadcast in the SFN that have an associated timestamp belonging to the time range of the temporal frame. This is made by one or more filters or packet parsers 431. It results that all the TS packets for a DTT temporal frame are retrieved before performing the second step for the same temporal frame.

This is the left part of unit 430 in Figure 7.

In a second step, the re-multiplexing unit generates a remux stream made of remux stream portions from the one or more input streams, each remux stream portion corresponding in time to a temporal frame defined by the retrieved timing information. In particular, it involves a packet insertion module 411 to apply, after all the TS packets to be forwarded have been retrieved for the temporal frame, an insertion scheme to the retrieved TS packets to insert them as packets in the remux stream portion only, regardless of a transmission time of the packets forming the remux stream portion.

This is the right part of unit 430 in Figure 7.

This split into two steps makes it possible to simultaneously generate the remux stream portion for DTT frame N (second step), while the TS packets are filtered from the input streams for DTT frame N+1 (first step). It may happen that reception errors occur on the master stream in such a way a frame PCR, PCR_fr, is not received. As a consequence, the re-multiplexing unit 430 does not receive the indication of a temporal frame boundary, which in turn may break the timing structure of the output SFN stream compared to the other transmitter sites that do not experience such reception errors.

Embodiments of the invention seek to overcome this situation. To achieve that, once the re-multiplexing unit 430 has finished generating a remux stream portion corresponding to a previous DTT temporal frame N (and thus is retrieving the TS packets for DTT temporal frame N+1 ), it may check whether or not all the PCR_fr necessary to delimit the next DTT frame to process have been retrieved.

Indeed, the retrieval of the TS packets from the input streams is made independently of the remux stream generation, although they are generally cadenced with a quite similar rhythm.

If a PCR_fr is missing (generally the PCR_fr defining the ending time of the next DTT frame), the re-multiplexing unit 430 may retrieve the parameters stored in memore from the previous DTT frame (the current values of PCR_fr; MIP CC and STS in case of DVB-T; T2MI superframejdx, T2MI frame_idx , T2MI packet_count and T2 timestamp in case of DVB-T2). Thus based on these values, it may infer the missing values for the next DTT frame.

Indeed, as the DTT frame duration is known, the new STS value may be computed for DVB-T by adding the megaframe duration to the previous value of STS. The MIP CC is increased by 1. In case of DVB-T2, the T2MI packet is increased by the number of T2MI packets in a T2 frame (it is a fixed parameter defined by the L1 parameters). The frame_index and the superframe_index are managed to take account a new T2 frame. The new T2MI-timestamp is computed as the STS.

By using this approach, the deterministic DTT stream adapter still delivers a transport stream containing a timing structure that will allow not disturbing the exciter in the absence of an indication of a temporal frame boundary.

During the first step in the re-multiplexing unit 430, each filter 431 filters the TS packets of a dedicated input stream 20 that belong to the services required at the output and store them in a dedicated memory module 432, preferably of a FIFO type. Each input stream 20 thus has its own memory module 432 for storage of the filtered TS packets.

Figure 9 illustrates the assignment of input TS packets to a given temporal frame whose time boundaries are PCR^N) and PCR_fr(N+1 ). Each filtered packet p belonging to input stream T which verifies the condition PCR_fr(N) < T_in(i, p) < PCR^N+I ) is assigned to DTT frame N. It means that the secondary input streams are split into temporal frames based on the PCR_fr provided by the master input stream.

To allow the two steps of re-multiplexing unit 430 to be performed simultaneously for two successive DTT temporal frames, the dedicated memory modules 432 may be made of two memory banks to store the TS packets, one for each temporal frame. A bank switch thus occurs each time a DTT frame boundary is detected. Thus a first memory bank in a memory module is used to store TS packets being retrieved for temporal frame N+1 , while a second and separate memory bank in the same memory module stores all the TS packets already retrieved for previous temporal frame N and is simultaneously used to generate the remux stream portion corresponding to previous temporal frame N.

Of course, subpart of the components of one service can be filtered. For example, an audio component can be removed, or ECM suppressed. A map of services and components required at the output for a SFN area is known by all the transmitter sites.

A PID remapping can also be performed by filters 431 if appropriate.

In embodiments, the filtering parameters may be provided by the head-end, for instance using an optional in-band configuration. The latter can also be used to set SI tables configurations required for DTT stream. The in-band configuration makes it possible to centrally drive or adjust the transmitter sites with respect to the generation of the SFN compatible DTT streams.

For illustrative purposes, the File delivery protocol (FDP) defined by HbbTV V2.0 may be used to transmit the desired configuration, through an XML-based configuration file.

In addition, signaling tables (e.g. DVB/SI tables) are generated for the current temporal frame, from the received input stream or streams and each table is stored as one or more TS packets in a dedicated memory module.

The table generation may be made locally by a table generator 433 based on the services that are to be forwarded (as defined in local configuration). It thus generates PAT, CAT, and PMT tables accordingly by extracting the required information from the incoming tables (from the input streams).

It then stores them, as TS packets (usually one TS packet per table), in the dedicated PAT or CAT or PMT memory module 434. A similar process is made for the SDT or EIT actual table which may be generated by extracting/filtering the required information from the incoming SDT or EIT tables.

In the particular case of DVB-T2, the SDT table shall be split into one MPEG section per PLP, a section referring to the services contained in a given PLP.

Again, the corresponding one or more TS packets are stored in the dedicated SDT or EIT memory module 434. In particular, a memory module is provided for the EIT table of each input stream 20.

A similar process is also made for the SDT other and EIT other. They can come from different sources. They may be extracted from ghost PIDs contained in the input TS streams 20, or from one of the input TS streams containing SDT or EIT other tables.

If appropriate, a ghost PID is translated into a SDT other PID.

The SDT (or EIT) other tables may be stored, as TS packets, in a number of memory modules 434 corresponding to the number of input streams 40. In a variant, they may be stored in a single SDT (or EIT) other memory module.

One may note that the signaling tables do not need to be generated for each new DTT frame. Indeed, some are (nearly) invariant over time, such as PAT, PMT, CAT, SDT tables.

For instance, these signaling tables are generated once when starting the deterministic DTT stream adapter 41 , and the corresponding memory module 434 stores the table for a long time. Thus a single memory bank may be used in the corresponding memory module 434.

Of course, update of the tables may be performed upon detecting a change in the service configuration for the SFN area.

As the EITs evolve over time, the corresponding tables may be updated regularly, based on a predefined periodicity, and preferably at each new temporal frame.

A particular attention may be provided to the retrieval of the EITs because they should be stored by full EIT sections. This is illustrated with reference to Figure 10 which show a table period pulse (vertical full line) as defined below (which merely corresponds to a multiple of temporal frames).

The EIT memory module is made of two banks, dedicated for two consecutive DTT frames. Figure 10 illustrates when a switch between the banks may be conducted, sometimes with a slight shift from the DTT frame pulse (corresponding to table period pulse) compared to the stream memory modules 432. This is to ensure whole EIT sections are saved in the same memory bank.

In short, before switching to the other memory bank after a new table period pulse, the table generator 433 checks whether or not an EIT section is pending. This check can be made based on the section length defined in the first packet of an EIT section and the amount of TS packets already retrieved for this section.

In the negative, the bank switch can be made immediately as for the other stream memory modules 432. In the affirmative, the table generator 433 must wait retrieving the end of the pending EIT section before allowing the switch.

Other signaling tables (not shown in the Figure) may be processed in similar ways. TS packets for these tables are stored in dedicated memory modules.

For instance, the NIT table may be generated locally as the PAT or PMT or CAT table, or retrieved from one input stream (for instance from a ghost PID, thus translated into the conventional NIT PID).

As another example, the BAT table may be handled like the SDT table.

Back to Figure 7, also for the special case of DVB-T or the like, a Mega- frame Initialization Packet, MIP, per megaframe (DTT frame) is generated by MIP generator 435 and stored in the appropriate bank of the dedicated MIP memory module 436 for insertion in the remux stream.

The MIP packet is generated based on modulation and timing information contained in the private sections of TS packets composing the reference service within the master input stream, i.e. based on the metadata 423 retrieved from the reference service: TPS field (i.e. modulation parameters) is extracted from the T modulation parameters, the continuity counter and the STS are directly obtained from the metadata (see for instance Figure 4a).

At the end of the first step in the re-multiplexing unit 430, all the memory modules 432, 434, 436 are filled with TS packets.

The second step in the re-multiplexing unit 430 consists in generating the remux stream portion for the current temporal frame.

The structure of the remux stream portion is first generated by NULL packet stream generator 440. It is part of a TS stream generated by this generator.

The TS stream is generated with NULL packets only at a bitrate corresponding to the modulation parameters for the output DTT stream to be broadcast in the SFN area. In case of DVB-T, the bitrate of the NULL packet TS stream is equal to the modulation bitrate defined by the MIP packet. The bitrate corresponds to an integer number of TS packets per megaframe.

In case of DVB-T2, the bitrate of the NULL packet TS stream is equal to the sum of the maximum TS bitrate acceptable per each PLP defined in the T2 configuration and modulation parameters. Again, the bitrate is computed in order to have an integer number of TS packet per T2 frame.

A transmission time for each TS packet forming the part of the NULL packet TS stream corresponding to the current temporal frame N can be computed. The transmission time or timestamp T_ou,(0) of the first packet is set to PCR^N) as it starts the current DTT frame.

Preferably, the transmission times of the other packets are inferred from PCRfr(N) and PCR_fr(N+1 ) using a linear interpolation. For instance, the transmission time for packet p is as follows:

PCR_fr{N + l) - PCR_fr(N)

T_aut {_P) = PCR_fr(N) + £ * N,

'"out

where N_ou, is the number of TS packets forming the temporal frame and Nj is the number of TS packets between packet p and the first packet of the temporal frame.

Once the NULL packet stream portion timestamped from PCR^N) to PCR_fr(N+1 ) is available, the packet insertion module 441 can insert the TS packets saved in the memory modules 432, 434, 436 in replacement of NULL packets in order to produce a remux stream portion for the current temporal frame N.

The insertion is made using the same insertion scheme as other transmitter sites of the same SFN area.

The insertion scheme does not take into account the transmission time of the NULL packets in the remux stream portion to allow or not the insertion of a specific TS packet (for instance given its timestamping T_in). This is to allow a higher number of retrieved TS packets to be inserted in the remux stream portion compared to prior art. It also helps to smooth the packet insertion over time, within the DTT frame.

The insertion scheme is implemented by requesters 442, one requester corresponding to one memory module.

The packet insertion module 441 and the requesters 442 are cadenced at a packet rate of the remux stream. It means that a request can be emitted by any requester at the same rhythm as the insertion of a TS packet in the remux stream. They follow the same TS packet synchro.

The insertion scheme follows two insertion periods depending on the TS packets to be inserted.

The DTT frame defines a first period (DTT period) dedicated for insertion of the filtered PID, and possible of some (short) signaling tables such as PAT, CAT, PMT tables.

On the other hand, a table period controls the insertion of some signaling tables, mainly those that are large, such as EIT and SDT tables. The "table period" is preferably defined as a multiple of the time length of the DTT temporal frame (i.e. of the DTT period), usually a multiple that allow an efficient management of the continuity counters of some signaling tables.

In case of DVB-T, the table period may be set to 16 megaframes and it starts with the MIP packet having a continuity counter equals to 0.

In case of DVB-T2, the table period may be set to 16 T2 superframe and the beginning of the table period is aligned with the pulse of a T2 superframe having its superframe index equals to 0. Thus, if we define Nb_T2frame the number of T2 frames per T2 superframe and 'a' the multiplying factor used (e.g. 16), a table period comprises

Nb_T2frame T2 frames.

Thus, each stream requester 442 successively sends, to packet insertion module 441 during the time length of the temporal frame (i.e. the DTT period), one or more requests for successively inserting the stored TS packet or packets in the remux stream portion corresponding to the temporal frame.

In particular, it preferably periodically sends a request for inserting a respective stored TS packet in the remux stream portion, wherein the sending period is computed based on a ratio between a number of TS packets to be inserted and stored in the memory module, and a number of packets composing the temporal frame.

As a maximum of all the N_fi|_tered(N) filtered TS packets stored during the previous period have to be inserted in the current remux stream portion made of N_out NULL packets, the ratio N_ou,/N_fi|_tered is used to define a rhythm of generation of the requests by a stream requester. For instance, starting at the frame pulse PCR_fr, requests for packet insertion may be send every step(i). An algorithm may be used that iteratively calculates step(i) to uniformly spread the Nfj|_tered(N) filtered TS packets over the N_out available packet positions in the remux stream portion. Thus step(i) is mainly equal to the integer part of N_0L,_t/N_f,itered-

However, step(i) is occasionally set to the integer part of N_out/Nfii_tered plus 1 to ensure an improved uniform spreading given the remainder (or rest) of N_ou,/N_fi|_tered- This increased value is used for instance each time i * rest(N_out/ fii_tered) reaches and exceeds each new integer multiple of the integer part of Ν_ου(/Νπι,_βΓβ(ι ('i' is the index to designate the i-th packet handled in the temporal frame).

Of course, the insertion requests process the stored TS packets by age order (the oldest one first).

Thus, the packet insertion module 441 periodically receives insertion requests from the stream requesters 442.

A similar approach is adopted for requesters corresponding to signaling tables to be inserted in the DTT period. The number N_PMT N_CAT N_PAT of times the (PMT, CAT or PAT) table must be inserted during the DTT period is user-defined.

Thus a request for table insertion is sent every N_0U,/N_PMT (or N_out/N_PAT or O_UI/NCAT) by the appropriate memory module 442, starting from the beginning PCR_fr of the DTT frame.

As the SDT and EIT are large (numerous TS packets), preference is given to insert a single occurrence of the tables every table period. Of course, it may be decided to insert several occurrences of them.

If the number of TS packets corresponding to the SDT or EIT table is not a multiple of 16 TS packets, a stuffing table with a discontinuity indicator and with pid 0x11 (for SDT) or pid 0x12 (for EIT) is inserted at the end of the insertion of the SDT or EIT TS packets. In that case, the number N_SDT or Ν_ΕΠ· of SDT or EIT packets (as stored in the corresponding memory module) is incremented by one.

Next, a request for insertion is generated and sent every N,abieperiod* out/N_SDT (or N_EIT) packets in the remux stream. Of course, the mechanism of SDT/EIT requests begins at the table period pulse.

Note that all the EIT packets must be managed together because they have to be inserted in an appropriate order (without mixing EIT from different multiplexes). It means that the total number of EIT packets to insert is computed. It is the value N_Err used above to determine the rhythm of insertion request generation. This value corresponds to the sum of TS packets for each EIT stored in memory modules : EIT source 1 + EIT source 2 + ... + EIT source n + EIT from ghost PID source 1 + EIT from ghost PID source 2 ... + EIT from ghost PID source n.

As the EIT packets are stored in several memory modules and must not be mixed, the memory modules are read one by one. It means that the first insertion requests progressively empty a first EIT memory module. Once the first EIT memory module is emptied, the following insertion requests request insertion of TS packets from a second EIT memory module, and so on.

Thus, the packet insertion module 441 periodically receives insertion requests from the ( P MT/P AT/C AT/S DT/E IT ) requesters 442.

Next and specific to DVB-T or the like, MIP requester 442 corresponding to

MIP memory module 436 sends a single insertion request per DTT frame (megaframe). As the MIP shall be the last packet of the megaframe, the insertion request may be sent for the last packet only. In a variant, the request may be sent at any time, but the packet insertion module manages the request so to insert the MIP only as the last packet of the megaframe.

Thus, the packet insertion module 441 can also receive an insertion request for a MIP packet from MIP requester 442.

As a single NULL packet is available at each TS packet synchro for packet replacement and a plurality of insertion requests may be received, the packet insertion module 441 implements a request priority management.

In embodiments, the packet insertion module gives priority, for TS packet insertion at a current packet position of the remux stream portion, to a memory module or requester for which a higher number of requests has not yet been served compared to a memory module/requester for which a lower number of requests has not yet been served.

For instance, when an insertion request is received from a memory module/ corresponding requester, the number of pending insertion requests is incremented by one.

The packet insertion module 441 thus selects a request, preferably the oldest one, of the requester 442 having the highest number of pending requests.

Of course, it may happen that two or more requesters 442 have the same number of pending insertion requests not yet served. In that case, the packet insertion module may give priority to a memory module/requester storing TS packets of an input stream compared to a memory module/requester storing signaling tables. In an exemplary implementation, priority to the requesters in case of same number of pending requests is as follows: Input stream 1 , Input stream 2, PAT, PMT, CAT, SDT, EIT.

Note that, as mentioned above, the packet insertion module 411 gives the highest priority to the request for inserting a Mega-frame Initialization Packet, MIP, in particular when it is time to insert the last packet in the DTT frame (megaframe).

The execution of an insertion request by the packet insertion module 411 comprises reading the corresponding memory module to extract the TS packet to insert (and thus to delete it therefrom) and then replacing the NULL packet at the current packet position in the remux stream portion by the extracted TS packet. Of course, once a request is processed, the corresponding number of pending requests is decreased by one.

Thus, the packet insertion module 411 progressively modifies the remux stream through NULL packet replacement.

Referring to the signaling tables, the packet insertion module 411 increments the continuity counter for the PAT, CAT, PMT, SDT, EIT tables. At each table period start, the continuity counter of each signaling table is set to 0. There is no discontinuity because the number of PAT, PMT, CAT inserted is a multiple 16 TS packets as mentioned above. For the SDT packets and the EIT packets, the last inserted table may be a stuffing table with a discontinuity indicator, if needed, as also mentioned above.

Once the NULL packet replacement has been done for the whole remux stream portion, the PCR of each TS packet is updated by the PCR update unit 443 since a bitrate adaptation is performed. For instance, PCR for packet 'p' now inserted at packet position 'n' is updated using the following formula:

PCRou,(p) = PCR_in(p)+T_0Ut(n)-T_in(p)

where PCR_in(p) is the PCR indicated in packet p when received from the head-ends.

The remux stream has thus been generated in a deterministic manner, based only on the content of the input streams (and possibly on user-defined configurations known by all the transmitter sites).

This remux stream is ready for inputting the modulator 43 in the specific case of DVB-T, because the MIP packets are already inserted in the stream, thereby defining the megaframe based on which the modulation will be performed. In the case of DVB-T2, an additional processing, DTT or T2MI encapsulation 450, is performed to generate a T2MI stream, including T2 Frames and PLPs. The metadata retrieved from the reference service (Figures 4b and 5b) allow to do that.

An exemplary deterministic method of generating a T2MI stream for such a remux stream is described in WO 2013/010872, being noted that the T2 frame delimitation in the remux stream is already known and corresponds to the remux stream portions progressively generated (from PCR_fr(N) to PCR_fr(N+1)).

The generation of the T2MI stream merely includes:

computing an ISCR for each TS packet of the remux stream,

generating PLP TS streams to be encapsulated in PLPs,

encapsulating the TS packets into BB frames,

adding the T2MI timestamp and L1 packets.

The resulting T2MI-stream can be output to modulator 43.

For illustrative purposes, the ISCR can be computed in a full deterministic manner from the T2 timestamps extracted for the temporal frames (here corresponding to the T2 frames). Indeed a T2 timestamp expressed in absolute time can be converted in ISCR format using the relationship between T_sub used in T2 timestamp and T.

Given T2_timestamp = seconds_since_2000 + subseconds ^*T_sub and given the standard table defining the T_sub-T relationship:

A ratio a between T and T_sub is either 71 (for 1 ,7 MHz bandwidth) or 7 (for other bandwidths). Thus T2_timestamp /T= seconds_since_2000 /(T) + subseconds/a. Moreover, as ISCR is 22-bit counter based on T2 Elementary period, ISCR of the T2 Super frame pulse is as follows:

ISCR(T2 Super frame pulse)= seconds_since_2000 /(T) + subseconds/a modulo (2^Λ22).

Furthermore, as the number of TS packets per T2 frame is known and the

T2 period is also defined, the step of ISCR, ISCR_step, for one TS packet can be inferred as a ratio between the period and the number of TS packets.

The ISCR value for each TS packet p can thus be computed:

ISCR(p) = ISCR(T2 superframe pulse) + Nj *ISCR_step

where U, is the number of TS packets between the packet p and the first packet of the T2 superframe.

Next, the PLP streams can be generated as follows.

The remux stream is duplicated to have as many duplicated streams as the number of PLPs (as defined in the retrieved modulation metadata). Each duplicated stream is assigned to a specific PLP.

Based on the known configuration of the PLP (i.e. which services to be included therein), the services are filtered: the PIDs of the unrequired services and the associated PMTs are replaced by NULL packets. It results that the bitrate remains unchanged, and the T2 frame delimitation is kept.

The SI tables are updated accordingly.

For instance, the PAT table is recomputed for each filtered duplicated stream, i.e. for each PLP: only reference to unrequired services are removed, label TS_ld is updated and the CRC is updated. Each PAT table is replaced by the new one. However the table repetition is kept.

A new CAT table is also computed: some EMMs should be removed, the

CRC is updated. Each CAT table is replaced by the new one.

The SDT table can also be updated. The DTT encapsulator 450 generates a SDT table divided in as many sections as the number of PLPs. Each SDT section declares the services contained in a specific PLP. The following process may be performed, for each section of the incoming SDT table:

The TS_id of the current PLP is taken into account in the field TSjd,

The Section_number and last_section_number are updated,

If the services described in the SDT table are present in the current PLP, the table is declared as a SDT actual, If the services described in the SDT table are present in another PLP, the table is declared as a SDT other, and

CRC is updated.

The SDT tables which references TS stream not contained in the 12 muliplex, the SDT are unchanged.

A similar process is applied to the EIT tables.

Next encapsulation of the filtered PLP streams into BB frames is operated, being noted that the T2 frame delimitation is still kept.

The mode High Efficiency Mode may be used with the NULL packet deletion option being activated.

For each PLP stream, the sync byte is removed if the mode High Efficiency is activated, the NULL packet deletion is activated to suppress the NULL packets, and the resulting data are encapsulated into BB frames.

Note that the T2 modulation parameters (retrieved from the metadata) supply the number of BB frames per PLP. Thus a mere division of the number of computed data per PLP with this number of BB frames per PLP allows the computed data to be uniformly divided in the BB frames.

Of course, this process is fully deterministic and thus can be performed by all the transmitter sites 40 of the same SFN area 4.

It remains for the BB frames to be encapsulated to build a T2-MI stream.

For each T2 Frame, the BB frames are encapsulated in T2MI packets as provided in the DVB-T2 standard. The T2MI packet count of the first T2MI packet of the T2frame is a value contained in the corresponding metadata. The retrieved metadata also provide the super_frame index and the T2 frame index. Furthermore, the packet count of the following packet is incremented.

As expected in the standard, the T2MI timestamp packet (generated from T2MI timestamp retrieved from the metadata) and T2MI L1 packet (generated from T2MI L1 information retrieved from the metadata) are added at the end of the T2 Frame. The packet counts of the two packets are incremented.

The resulting T2MI stream can thus be fed to DVB-T2 modulator 43 for

SFN broadcast.

Although the present invention has been described hereinabove with reference to specific embodiments, the present invention is not limited to the specific embodiments, and modifications will be apparent to a skilled person in the art which lie within the scope of the present invention. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. The mere fact that different features are recited in mutually different dependent claims does not indicate that a combination of these features cannot be advantageously used.

Claims

1. A method for deterministic re-multiplexing of transport streams in Single Frequency Networks, SFN, made of transmitter sites synchronously transmitting on the same frequency, the method comprising the following steps performed at one of the transmitter sites:

receiving one or more input streams;

using the generated remux stream to broadcast an output SFN stream, wherein generating a remux stream portion corresponding in time to a temporal frame includes:

2. The method of Claim 1 , wherein the timing information defining the temporal frames is retrieved from a private section of TS packets composing a reference service within the received input stream.

3. The method of Claim 2, wherein a TS packet of the reference service includes a conventional PCR field and, within its private section, a frame PCR sampling the time at which a temporal frame starts in the input streams.

4. The method of Claim 3, wherein the time range of the temporal frame is defined by two consecutive frame PCRs provided in the reference service.

5. The method of Claim 3, wherein the private section further includes a synchronization timestamp defining a transmission time of a temporal frame for broadcasting in the SFN.

6. The method of any of Claims 2 to 5, wherein a private section of TS packets composing the reference service includes modulation parameters to be used by the transmitter site for broadcasting the output SFN stream.

7. The method of any of Claims 1 to 6, wherein at least two input streams are received by the transmitter site that both include a separate reference service embedding timing information defining temporal frames, and the method further comprises selecting one of the input streams as a master input stream from which the timing information is retrieved.

8. The method of any of Claims 1 to 7, wherein retrieving all the TS packets to be forwarded for broadcast includes storing TS packets to be forwarded that belong to a first input stream into a first memory module and storing TS packets to be forwarded that belong to a second and separate input stream into a second and separate memory module.

9. The method of Claim 8, wherein a first memory bank in a memory module is used to store TS packets being retrieved for the temporal frame, while a second and separate memory bank in the same memory module stores all the TS packets already retrieved for a previous temporal frame and is simultaneously used to generate the remux stream portion corresponding to the previous temporal frame.

10. The method of Claim 8 or 9, further comprising generating signaling tables for the temporal frame from the received input stream or streams and storing each table, as one or more TS packets, in a dedicated memory module.

11. The method of any of Claims 8 to 10, further comprising generating a Mega-frame Initialization Packet, MIP, for the temporal frame from modulation and timing information contained in a private section of TS packets composing a reference service within the received input stream, and storing the MIP as a TS packet in a dedicated memory module.

12. The method of any of Claims 8 to 11 , wherein applying an insertion scheme includes the following step by each of the one or more memory modules storing one or more TS packets retrieved from the input stream or streams for the temporal frame:

13. The method of Claim 12, wherein successively sending one or more requests includes periodically sending a request for inserting a respective stored TS packet in the remux stream portion, wherein the sending period is computed based on a ratio between a number of TS packets to be inserted and stored in the memory module, and a number of packets composing the temporal frame.

14. The method of any of Claims 8 to 13, wherein applying an insertion scheme includes the following step by one memory module storing a signaling table as one or more TS packets for the temporal frame:

successively sending, to a packet insertion module during the time length of the temporal frame, requests for successively inserting a multiple number of the stored table in the remux stream portion corresponding to the temporal frame.

15. The method of any of Claims 8 to 14, wherein applying an insertion scheme includes the following step by one memory module storing a signaling table as TS packets for the temporal frame:

successively sending, to a packet insertion module during a table period corresponding to a multiple of the time length of the temporal frame, requests for successively inserting the stored TS packets in the remux stream portion corresponding to the temporal frame.

16. The method of any of Claims 12 to 15, wherein the packet insertion module gives priority, for TS packet insertion at a current packet position of the remux stream portion, to a memory module for which a higher number of requests has not yet been served compared to a memory module for which a lower number of requests has not yet been served.

17. The method of Claim 16, wherein at same number of requests not yet been served, the packet insertion module gives priority to a memory module storing TS packets of an input stream compared to a memory module storing signaling tables.

18. The method of Claim 16 or 17, wherein the packet insertion module gives the highest priority to insertion of a Mega-frame Initialization Packet, Ml P.

19. The method of any of Claims 12 to 18, wherein the packet insertion module and the memory modules are cadenced at a packet rate of the remux stream.

20. The method of any of Claims 8 to 19, wherein each time a TS packet retrieved from the input stream or streams is inserted in the remux stream portion, it is deleted from the corresponding memory module, and the TS packets remaining in the memory module or modules at the end of the remux stream portion are discarded.

21. The method of any of Claims 1 to 20, further comprising, by the transmitter site, generating a T2-MI stream from the remux stream, wherein T2 frames of the T2-MI stream are aligned with the remux stream portions.

22. The method of any of Claims 1 to 21 , wherein two input streams are received from two separate head-ends.

23. The method of any of Claims 1 to 22, further comprising inferring timing information for a next temporal frame from the timing information retrieved for a previous temporal frame.

24. A deterministic re-multiplexer for re-multiplexing transport streams at a transmitter site in a Single Frequency Network, SFN, made of transmitter sites synchronously transmitting on the same frequency, the deterministic re-multiplexer comprising:

a receiver for receiving one or more input streams;

25. The deterministic re-multiplexer of Claim 24, further comprising a plurality of memory modules to store TS packets

wherein the packet parser is configured to store TS packets to be forwarded that belong to a first input stream into a first memory module and to store TS packets to be forwarded that belong to a second and separate input stream into a second and separate memory module.

26. The deterministic re-multiplexer of Claim 25, wherein a first memory bank in a memory module is used to store TS packets being retrieved for the temporal frame, while a second and separate memory bank in the same memory module stores all the TS packets already retrieved for a previous temporal frame and is simultaneously used by the re-multiplexing unit to generate the remux stream portion corresponding to the previous temporal frame.

27. The deterministic re-multiplexer of Claim 25 or 26, further comprising a table generating unit for generating signaling tables for the temporal frame from the received input stream or streams and for storing each table, as one or more TS packets, in a dedicated memory module.

28. The deterministic re-multiplexer of any of Claims 25 to 27, further comprising a MIP generator for generating a Mega-frame Initialization Packet, MIP, for the temporal frame from modulation and timing information contained in a private section of TS packets composing a reference service within the received input stream, and for storing the MIP as a TS packet in a dedicated memory module.

29. The deterministic re-multiplexer of any of Claims 25 to 28, wherein each of the one or more memory modules storing one or more TS packets retrieved from the input stream or streams for the temporal frame successively sends, to the packet insertion module during the time length of the temporal frame, one or more requests for successively inserting the stored TS packet or packets in the remux stream portion corresponding to the temporal frame.

30. The deterministic re-multiplexer of Claim 29, wherein the packet insertion module gives priority, for TS packet insertion at a current packet position of the remux stream portion, to a memory module for which a higher number of requests has not yet been served compared to a memory module for which a lower number of requests has not yet been served.