Abstract

This report describes the results and achievements of Eureka VADIS project (VADIS: Video-Audio Digital Interactive System) which started in 1991 and finished in 1995. More than thirty companies have participated in the VADIS project.

The goals of the VADIS Project were:

• To coordinate input from Europe into the MPEG2 standards committee and to promote European functionalities and specifications within the MPEG2 standardisation process.
• To promote and encourage the use within Europe of the emerging MPEG-2 digital television standard for TV and multimedia distribution.
• To initiate the development of appropriate technology which would allow the rapid exploitation within Europe of the new standard.
• To promote the growth of MPEG2 expertise within Europe.

VADIS has achieved its goals. For example:

• the MPEG-2 standard has been adopted by ETSI in a number of its standards;
• the contributions made by VADIS have strongly influenced the functionalities and definition of the MPEG2 standard;
• MPEG2 decoder VLSI has been developed within the project and is commercially available;
• prototype MPEG2 coding and decoding equipment has been constructed and many field trials have been carried out to gain experience of the use of MPEG2 over satellite, cable and ATM links.

This document reports on the work and results of the individual workpackages within the Project.

Contributions from: L.Chiariglione (CSELT), G.Nocture (LEP), H.Hoffman (IRT), N.Tanton (BBC), J.Minot (LEP), J.P.Henot (CCETT)

Editor: N.Wells (BBC)

An introduction to VADIS and its achievements 

1. A bit of history 

The European commitment to digital television is now almost complete. However, in 1990, when VADIS was first discussed as a European project to develop digital TV standards and technology, the idea of digital television broadcasting and distribution appeared to many as revolutionary even though collaborative projects in digital video for the mass market were not new in Europe. For example, in 1988, a small number of European industries interested in MPEG-1 had already cooperated in the ESPRIT COMIS project.

In July 1990, at the MPEG meeting in Porto, the first discussions on the requirements for an MPEG-2 standard were made in the presence of a few experts. At the next MPEG meeting in Santa Clara, the first demonstrations were shown of what could be expected in terms of bitrate and quality for conventional definition pictures coded using state-of-the-art algorithms.

In September 1990, the proposal was made (at a meeting of the COMIS project) to establish a project for developing a European technology for MPEG-2. As there was no RACE or ESPRIT call for new projects at that time, the participating companies decided to establish a collaborative project within the framework of EUREKA and the name given to the project was VADIS (Video-Audio Digital Interactive System). The main goal of the project which was to help to establish a European technology for MPEG-2 was to be reached through the following activities: video coding algorithm development, system level studies, construction of demonstrators, VLSI design and fabrication, and field trials.

The appearance of the word “Interactive” in the VADIS project title was motivated by the fact that the project did not wish to polarise the debate that was current at that time concerning the issue of whether or not HDTV/TV broadcasting should be based on digital or analogue techniques. This was very much in line with the MPEG approach to develop a generic technology for digital television which would be usable in a number of different application domains and not only for broadcasting

Not without resistance from some countries, the project obtained the EUREKA label in April 1991, but the technical work had already started in November 1990 and had continued through a series of meetings. The project was initially intended for a duration of 30 months until September 1993, but was subsequently extended to December 1994 in order to align the project with the achievement of “International Standard” status for MPEG-2 (November 1994). Two further extensions, until June and December 1995 were agreed mostly for the purpose of completing and assessing the field trials.

The VADIS membership increased from an initial 10 companies to more than 30 and encompassed the major sectors interested in digital television applications; i.e. broadcasters, telecom operators, and consumer electronics manufacturers. The list of all companies which have participated in the project is given in Annex A.

2. Project organisation

The project was organised in six workpackages (WP’s):

• Algorithms
• Systems
• Demonstrator
• VLSI
• Field Trials
• Conformance

Algorithms initially coordinated the development of several candidate “European” video coding algorithms for submission to MPEG. This task was basically completed in November 1992 when VADIS submitted 11 different proposals of algorithms to MPEG. A total of around 30 different algorithms were submitted to MPEG from different parts of the world. After this, followed the optimisation phase of the MPEG-2 Video Test Model, done collaboratively with VADIS and MPEG. In the latter stages of this phase, VADIS work was predominant in driving the selection process of tools to support scalability, both the SNR and the Spatial Scalable profile. The third phase of work consisted in the development of strategies to optimise MPEG-2 Main Profile Main Level encoder performance and in verification tests of MPEG-2 Video quality.

Systems had the task to coordinate VADIS participation in the development of what eventually became “Part 1” of the MPEG-2 standard (ISO/IEC 13818-1). Work in this area began within VADIS much before the importance of systems aspects was recognised in MPEG, where real work only began in November 1992. VADIS also played an essential role in achieving European consensus on the support of MPEG-2 Systems as the generic way to interface MPEG-2 data to physical channels. In the last stage of the project this WP contributed to syntax verification by producing and exchanging bitstreams.

The Demonstrator workpackage had the task to develop hardware for the purpose of demonstrating real-time performance of the standard. The first stage of work targeted the development of systems to record and replay MPEG-2 data stored on a variety of digital storage media. Next followed the development of an encoder and a decoder. In parallel with the codec development, software for synthesising and analysing MPEG-2 streams was developed. Finally the basic building blocks of a transmultiplexer were designed . Availability of such equipment has been essential for successful execution of field trials.

The VLSI workpackage had the task to develop the strategic European VLSI technology of MPEG-2 Video decoding. This workpackage achieved the first large-scale European-wide collaborative VLSI design using the VHDL language. One company is using the results of this workpackage in the production of their MPEG-2 Video decoder chip.

Field trials had the task to prepare and execute field trials to assess the performance of the MPEG-2 standard for a wide range of applications.

Conformance had the task analyse the requirements for the conformance testing of MPEG coders and decoders and to contribute to the drafting of MPEG-2 “Part 4” on “Conformance testing”.

In summary it can be stated that in its 5 year duration the project successfully accomplished a number of important objectives. For example:

• VADIS was a major influence within Europe towards its conversion to the benefits of digital for future TV and HDTV distribution bypassing any intermediate analogue step. This process culminated in February 1993 with the “European Symposium on Digital Television” organised by VADIS and hosted by the EBU.
• VADIS succeeded in convincing Europe that it was in the common interest to adopt a generic digital video technology instead of a set of application-specific technologies. In the area of system multiplexing this was started in January 1993 when VADIS convened a meeting of several projects hosted by Deutsche Telekom in Darmstadt with the purpose of jointly checking that MPEG-2 Systems did indeed support all requirements issued from the different application areas.
• VADIS provided the coordinating forum through which Europe could have a strong and unified voice in the development of the “Systems” and “Video” parts of the MPEG-2 standard. VADIS succeeded in encouraging disparate European companies to cooperate in the development of generic MPEG-2 technology. This was particularly true for the strategic area of VLSI where the project succeeded in the (until then unachieved) difficult goal of collaboratively designing a chip of the complexity of an MPEG-2 Video decoder.
• VADIS developed prototype equipment directly based on the MPEG-2 standard which included a real-time encoder, real-time decoders, bitstream sources and analysers and a number of channel adaptors for satellite, cable and terrestrial broadcasting.
• VADIS provided its partners with equipment and technical know-how to enable them to carry out the first comprehensive set of field trials to assess the performance of the MPEG-2 technology across a wide range of applications such as client-server performance in ATM networks, terrestrial and satellite broadcasting, Cable TV etc.
• VADIS also provided some “firsts” such as implementation of a complete chain of satellite broadcasting according to the appropriate ETSI specifications (HDTV Workshop, November 1994) and retrieval of MPEG-2 data via 1,000 km ATM link (Montreux Conference and Exhibition, June 1995).

2. Algorithms Workpackage

The work of this group falls into distinct phases:

• Generation of initial proposals of candidate coding schemes for consideration by MPEG-2 at Kurihama.
• Work towards the improvement and optimisation of the MPEG-2 Video Test Model.
• Promotion within MPEG of MPEG-1 compatibility and hierarchical/scalable coding modes.
• Bitstream exchange as a method of testing MPEG-2 syntax.
• Encoder optimisation.
• Subjective testing.

1. Preparation for the “Kurihama” tests

At the very beginning of the VADIS project, that is during end 1990 and 1991, the goal of the Algorithms workpackage was to prepare proposals for MPEG evaluations which were to be held in November 1991 in Kurihama. The workpackage was organised into three subgroups which concentrated on different coding algorithms:

• Subgroup 1 concentrated on MPEG-1 compatible coding schemes
• Subgroup 2 concentrated on non-compatible DCT based coding schemes
• Subgroup 3 concentrated on (non-compatible) subband coding schemes for hierarchical representation of video sequences.

The results of this collaborative work were 11 complete proposals (including VLSI implementation studies for each proposed scheme) and these were submitted to MPEG and evaluated by subjective testing at an MPEG meeting in Kurihama in November 91. In these subjective tests, the VADIS proposals were ranked very honourably in competition with other proposals from all around the world.

2. Improvement of the MPEG Test Model

After the subjective testing of the proposals submitted in November 91, MPEG set up an MPEG-2 Video “Test Model”. This was based on the MPEG-1 standard and also on a few good technical tools which were identified from coding proposals which did well in the subjective tests. The purpose of the Test Model was to provide a good reference coding scheme to which refinements and extra tools could be added if they proved to lead to improvements in coding efficiency or if they could provide additional functionality.

VADIS was active both in the improvement and refinement of the video Test Model and active in defining tools to provide for additional functionality (in particular algorithmic features which enabled efficient scalable coding).

Also, thanks to a close relationship between the Algorithms workpackage and the VADIS VLSI workpackage, VADIS contributions always took into account the impact on decoder VLSI of their proposals.

3. Improvement of the coding efficiency of the Test Models

A few important topics for which significant effort were deployed by the VADIS Algorithm group are recalled in the following:

Motion compensated prediction: The efficiencies of new motion compensated prediction modes such as “FAMC”, “dual-field” mode and dual-prime” were tested. The final choice by MPEG of dual prime between successive P frames (only) was the one supported by VADIS partners for implementation reasons.

Quantisation issues: Significant contributions were made by VADIS in the following quantisation-related areas:

• The use of linear or nonlinear quantisation step scales
• Down-loadable matrices
• Programmable accuracy for the DC coefficient of Intra coded blocks

The contributions made were largely taken into account in the final MPEG-2 specifications.

Variable length coding issues: All aspects of variable length coding were looked into by VADIS, from the general issues related to the choice of syntax to the very detailed issues such as the choice of escape codewords in the coding of DCT coefficients.

4. Securing features for European applications in the MPEG-2 standard

Two main features of the Video standard were of particular concern for VADIS partners. These were:

• Compatibility with the existing MPEG-1 standard
• The existence of efficient tools to enable scalable/hierarchical coding of HDTV/TV for possible broadcasting applications in Europe.

MPEG-1 compatibility: For the Kurihama tests, one VADIS proposal described a scheme which generated a two-part bitstream, with one part decodable by an MPEG-1 decoder. However, compared to single-layer schemes, such proposals were shown to bring a slight decrease in coding efficiency and a significant increase in decoder VLSI implementation cost. Consequently, this approach was not retained within the Main Profile of MPEG-2.

However, care was taken to keep the MPEG-2 video standard close to the MPEG-1 standard so that, in terms of implementation, the tools required to decode MPEG-2 are close to those required to decode MPEG-1 Video bitstreams. Also, an MPEG-2 compliant decoder must decode an MPEG-1 video bitstream.

Hierarchical coding: Until recently, it was considered important in Europe that any digital HDTV broadcasts should simultaneously allow decoding on conventional-definition receivers. Also, there are other applications, such as error resilience on ATM networks, that may require two or more quality levels within a single bitstream. Therefore, a subgroup was set up with the specific goal of developing hierarchical coding schemes suitable, in particular, for digital HDTV/TV broadcasting. Another goal was to develop such schemes in sufficient detail that they could be accepted into the MPEG-2 standard.

The main areas of work of this new subgroup were:

• Development of three candidate scalable coding schemes (SNR, Spatial, Frequency).
• Direct comparison of the performance of frequency- and spatially-scalable schemes.
• Comparison of candidate variable-length coding schemes.
• Investigation and optimisation of up- and down-conversion processes for spatial scalability.
• Definition of syntax for scalable schemes.
• Bitstream exchange for scalable syntax verification.

European effort was pre-eminent within MPEG in the investigation, comparison and optimisation of scalable coding schemes. The final scalable schemes adopted for inclusion in the MPEG-2 standard (spatial-scalable and SNR-scalable) can be clearly identified as originating from European companies participating in the VADIS Project.

5. MPEG-2 bitstream exchanges and bitstreams for MPEG-2 tests

In order to test the validity of the MPEG-2 syntax it was essential that bitstreams generated by individual companies could be decoded by separate, independent companies. VADIS contributed extensively to such bitstream exchanges.

Another activity within MPEG was the verification of picture quality that could be obtained from MPEG-2 at different bitrates. For this, the MPEG “Test Group” requested encoded sequences to be generated which could be independently decoded and the picture quality subjectively tested.

VADIS contributed in a number of coded sequences and for the Main Profile and the SNR Profile. The bitstreams were sent to the MPEG Test Group chairman and distributed to other companies for decoding. Decoded pictures were sent to MPEG for subjective testing.

Participation in this activity was a very efficient way for partners to fully check the conformance of their software coders and decoders.

6. Encoder optimisation and subjective testing

The MPEG-2 standard specifies a decoding process and does not specify how to build the most efficient or highest quality encoder. Therefore, an “encoder optimisation” activity was initiated within the Algorithm workpackage. At the conclusion of this activity, subjective tests were done to assess the picture quality obtained using an optimised encoder.

The subjective tests compared the optimised encoder against the MPEG-2 “Test Model” (TM4.2) at 4 Mbit/s and at bitrates of 6, 4 and 2 Mbit/s against analogue PAL. The sequences used for both tests were: “mobile and calendar”, “flower garden”, “table tennis”, “horse riding”, “cactus with noise”, “basket ball 2”. The picture material was encoded by several VADIS companies active in the encoder optimisation workpackage. The editing of the tapes used in subjective assessment, as well as the subjective assessment itself were done by CCETT.

The testing conditions were in accordance ITUR Rec. 5005 and the viewing distance was 6H only (conventional definition sequences and displays). The observers (15 minimum) were non expert in the field of television.

In the tests at 4Mbit/s, perceptible improvements were achieved for the optimised encoder only for sequences with a high motion content. On other sequences an improvement was more difficult to observe.

In the comparisons with analogue PAL, the optimised picture quality at 6Mbit/s was perceived to be exactly the same as PAL for the sequences tested which were not critical for PAL coding/decoding. At 4Mbit/s the optimised coder quality was slightly lower than PAL quality.

7. Conclusion of algorithm workpackage:

The work of the Algorithm workpackage was successful in many areas:

• In the proposal of candidate coding schemes at the beginning of the MPEG-2 process which performed very well in comparativesubjective tests.
• In bringing European requirements and solutions successfully to MPEG-2.
• In working to maintain a maximum compatibility of the MPEG-2 standard with MPEG-1.
• In working to maximize MPEG-2 coding efficiency for a reasonable decoder complexity level. In the proposal, optimisation and adoption of schemes for hierarchical coding within MPEG-2.
• In helping to develop optimised and conformant software for MPEG-2 coding and decoding. In developing expertise about how to best use the complicated MPEG-2 tool kit.
• In promoting good cooperation between European companies.

3. Systems Workpackage

Contributions from the VADIS Systems workpackage gave a lead in many areas of the MPEG-2 systems-layer definition. All of its outline proposals were accepted by MPEG although discussions within MPEG often yielded improvements or produced better generality than the original proposals.

For example, at the London MPEG meeting in November 1993, there were moves towards developing two further levels of multiplex, ‘supermux’ and ‘mastermux’, above that of the basic functionality of an MPEG-1-type of multiplex. The European position reached at a VADIS meeting in Darmstadt was not to support the additional second level as it appeared to be rather specific to requirements of the North American cable TV industry. The Rome MPEG meeting subsequently dropped the mastermux and decided to attempt to define the supermux such that the combination of ‘progam multiplex’ (equivalent to the MPEG-1 multiplex) plus supermux could be viewed and implemented as either one or two layers of multiplexing/demultiplexing.

1994 was a particularly active year for the VADIS Systems Work Package. During this period, the MPEG Systems sub-group met in March/April, July, and September, and additionally held ad-hoc meetings in May, June, and July. Between April and September the Working Draft of the Systems specification was progressed from little more than a Table of Contents to a detailed document. The VADIS project played an important role in the definition of this specification at all of the MPEG meetings, with a number of contribution documents tabled at each meeting.

VADIS has played a seminal role in the MPEG Systems work in several ways, namely:

• The project partners have influenced the MPEG committee away from a multi-layer multiplex structure originally proposed by US Cable TV companies towards the adoption of a “two-layer synchronous” approach. (This can be implemented in either one or two layers of coding.)
• VADIS has played a particularly important role in identifying the requirements and solutions for Conditional Access and for Control and Indication. MPEG has decided to standardise only the hooks for adding Conditional Access facilities to the Systems layer, and the specification for these hooks follows VADIS proposals.
• VADIS has made significant contributions to the specification of Service Information (“Programme Specific Information”) and Control commands for digital storage media (“DSM-CC”).

The VADIS Systems group encouraged the adoption of the generic MPEG-2 Systems layer for digital video broadcasting in Europe. A joint meeting of the VADIS Systems Workpackage and EBU V4/MOD-A was held in June 94. This group had been given the task by the Technical Module of the European project on Digital Video Broadcasting (TM-DVB) to propose a multiplexing solution for European Digital TV Broadcasting. The meeting reviewed the alternative methods (e.g. a scheme based on ATM, DAB or on the MPEG Systems layer). It came to the conclusion that the generic MPEG transport multiplex was well suited to the requirements. It was noted that MPEG does not specify the data-link and physical layers, i.e. the error correction and modulation schemes that should be used. The MPEG-2 approach is that the optimum techniques for these layers are dependent on the medium and it is therefore appropriate that a generic standard for multiplexing does not specify these layers.

The final work of this workpackage was to generate, by simulation, MPEG-2 bitstreams which included the full multiplexing Systems Layer information. This work enabled independent checking of the Systems Layer syntax and was preparatory to the interconnection of hardware encoders and decoders.

4. Demonstrator Workpackage

The specific goals of the Demonstrator Workpackage within VADIS were:

• To realise a generic codec in hardware once the MPEG-2 coding algorithm had been defined.
• To service the requirements of the Field Trials workpackage by constructing this codec and other necessary equipment.

Both of these goals have been achieved as a result of the considerable and spirited collaborative effort from the partners working in the Demonstrator WP. Enabling technology and “know-how” developed within or deriving from the work of the workpackage is in use or for sale throughout Europe and field trials using from the VADIS codec continue.

Initially the effort offered by VADIS partners to the Demonstrator workpackage was largely focused on decoder design with some interest in the development of bit-stream sources and recorders. During the first part of the Project it became clear that additional and realigned effort would be required if a real-time coder was to be built and this was reflected in the revised work plan of the project extension.

A further and extrinsic influence on the work of this workpackage was the relatively slow convergence firstly of the MPEG2 video syntax and then of the system syntax. The MP@ML video syntax became largely stable in April ’93 whilst the system syntax only became sufficiently stable for hardware design safely to commence in March ’94. Although this meant that detailed design for much of the demonstrator could not commence until Spring 1993, architectural, systematic and interfacing studies could be undertaken at a comfortable pace and in greater detail. As a result, the system integration phase of our work was much simplified by the high level of preparation which this delay provided. In the event successful real-time coding and decoding by the VADIS demonstrator was achieved in Autumn 1994 less than 15 months after coder development commenced in earnest.

The real-time coder codes a single programme as MP@ML and was developed by BBC, CCETT, IRT and Telefonica I+D with IRT supervising the system integration. The decoder, which is based around DSPs, was developed by CSELT, Italtel and Philips Italy with CSELT supervising system integration. After minor optimisations this highly flexible demonstrator is capable of coding and decoding with a picture quality which appears at least as good as any seen from other commercial equipment at recent exhibitions (e.g. IBC ’94, Montreux ’95).

Meanwhile bit-stream sources and/or recorders have been developed by CCETT, CSELT, IRT, MATRA Communications, NTL and Philips Italy and used in this and other projects whilst MATRA have also developed channel adaptors for satellite broadcasting field trials, PC-based stream synthesis and analysis tools. Thomson/LER has designed ASICs for MPEG2 Transport Stream MUX and DEMUX functions. The Workpackage has concentrated more on the MPEG2 transport stream rather than the programme stream because the commercial and professional interest of workpackage partners has predominantly been in the Transport Stream form.

A second strand of decoder development has involved Nokia Research in developing a board-level decoder based on the video decoder chip developed by the VLSI Workpackage and manufactured by Philips LEP.

The work of the Demonstrator workpackage was spread over 15 meetings commencing in June 1991 and finishing in October 1994. During this period 189 input documents were submitted and studied and 22 output documents generated. Close interaction with the other workpackages, particularly Algorithms, Systems, Field Trials and VLSI, with joint or sequential meetings has been an important factor in productivity. The momentum of WP work was largely due to the positive spirit of collaboration and participation exhibited by almost all the partners involved and maintained by prompt and accurate action minutes.

In addition to direct application of hardware and “know-how” to other projects or products the work of the workpackage has been influential in initiating discussions on and providing a working model for practical interfacing at the transport stream level.

Additional equipment design and construction : In addition to the coder and decoders, the following equipment was produced within the project.

• Bitstream sources: Bitstream recorders/player have been developed and built by several partners providing data rates in the range of < 24 Mbit/s in play mode and < 19 Mbit/s in record mode. For the data transfer, serial as well as parallel interfaces have been implemented.
• MPEG-2 data stream generator and analyser: Equipment for the generation of an MPEG-2 transport stream as well as an MPEG-2 transport stream analyser has been developed by MATRA.

5. Field Trials Workpackage

The Field Trials workpackage was set up with the following objectives:

• To provide the necessary infrastructure to connect the MPEG2 coder/decoder (VADIS demonstrator) being developed within the VADIS project to the target transmission channels.
• To test the applicability of MPEG2 coding schemes (video/audio/systems) against the requirements for different applications and networks.

The trials fell into the following categories:

• terrestrial broadcasting,
• satellite broadcasting and SNG,
• cable distribution
• ATM-based distribution.

The specific goals of each trial were:

• To check whether MPEG-2 and associated interfaces will work on different networks.
• To test and demonstrate MPEG-2 operation for different applications.
• To test interworking with networks for distribution.
• To examine the behaviour of the system under different traffic conditions and parameter settings.
• To test the effectiveness of the chosen AAL and end-to-end synchronisation.
• To examine the protection ratios for interference situations.
• To cascading MPEG-2 with other BRR systems and digital processing algorithms.
• To gain experience with Service multiplexing requirements.

1. Field Trials Organisation 

Digital Terrestrial Broadcasting : For digital terrestrial broadcasting field trials, interfaces were constructed (by MATRA and CCETT) to connect the VADIS demonstrator to channel modulation equipment provided by the dTTb Project (RACE R2082). A block diagram of the adaptor is given in Figure 1.

Figure 1: Block diagram of the channel adaptor for digital terrestrial television

The adaptor box adds an outer FEC together with the interleaving needed for the transmission over non error-free channels. Also, it provides an optional energy dispersal scrambler. At the receiving side a box providing the inverse functions was provided. The partners that carried out field trials for this application were IRT, RAI and CCETT Satellite Distribution and SNG: One major task in this area was again the provision of channel adapters and bitstream “translators” for existing systems. A block diagram of the adaptors is very similar to that given in figure 1. The partners that carried out field trials for this application were: ESA, FI/DBPT and NTR.

Cable distribution : An important item for studies and trials was the interworking between terrestrial distribution systems and cable networks. Partners studying this application area and providing the infrastructure for field trials were: CCETT and FI/DBPT.

Multimedia applications and MPEG-2 TS over ATM-based Networks: The ATM part of Field Trials aimed at validating the transport over ATM-based Networks of MPEG2 Transport Streams, multiplexed with other types of services, for example, computer interconnections for file transfers. The standard chosen for adapting the transport stream packets to ATM packets was AAL1 (ATM adaptation layer 1). To allow these field trials, a number of further channel adaptors (of the type shown in Figure 2) have been built by partners to interconnect the VADIS demonstrator to different types of networks such as LANs (Tribune) or WANs (European Networks). At the same time, an integrated chip set has been designed by Thomson as a first level of integration of AAL type1. Partners carrying out field trials for this application, have been: CSELT, PTT Research, TELEFONICA, TELIA, INTRACOM.

Figure 2: Channel Adapter for Multimedia Applications

2. Field Trials Participation

The following companies participated in the field trial experiments:

• Terrestrial broadcasting: CCETT, RAI, RETEVISION, TELEKOM, IRT, BBC
• Satellite broadcasting: ESA, NT, RAI, RETEVISION, TELEKOM, TELIA/TID
• Cable distribution: CCETT, TELEKOM
• ATM interactive networking: INTRACOM, CSELT, PTT Research, TELEFONICA, ITALTEL

For the ATM trials the networks used were the “TRIBUNE”, “RECIBA” and “EUROPILOT” networks.

6. VLSI Workpackage
7. Introduction 

The VLSI workpackage had the task to develop the VLSI for MPEG-2 video decoding. This workpackage achieved the first large-scale European-wide collaborative VLSI design using the VHDL language. One company is using the results of this workpackage in the production of their MPEG-2 Video decoder chip.

Another important task of the VLSI group was to liaise closely with the Algorithm workpackage in order to ensure that algorithm proposals made during the development of the MPEG-2 standard could be efficiently implemented in decoder VLSI.

The work of the group fell into three distinct phases:

• Specification and behavioural VHDL modelling
• Gate-level chip design and chip construction
• Design studies of VLSI for HDTV MPEG-2 decoders

2. Activities 

Specifications and VHDL modelling: A specification of an MPEG 2 system video decoder was first produced and this led to the joint definition and design of a behavioral VHDL simulator among partners in the workpackage. The initial specifications took into account inputs from the Demonstrator workpackage since it was planned that they would be early users of the chip for construction of decoders for use in the Field Trials activities. For the VHDL simulator, each partner was responsible for the delivery of an assigned VHDL simulation block and these blocks were assembled together by the workpackage leader. The VHDL simulator was then used successfully for the validation of the retained architecture of the IC.

Implementation : The functional model described by the behavioral VHDL model was redesigned in synthesizable VHDL by LEP and ENST. The resulting synthesizable VHDL was used to generate an IC lay-out. The IC has been manufactured in 0.5 micron 3.3 V Philips/ST technology. The IC has been made available at the beginning of June 95 to the Demonstrator group.

Studies of HDTV architectures: Starting from the MPEG 2 means to accommodate HDTV for instance for terrestrial applications, the VLSI group has studied possible architectures for a Main Profile at High Level Video Decoder.

3. Conclusions of VLSI activities 

The existence of the VADIS project has enabled fruitful discussions between algorithm and VLSI experts to develop. Also, collaboration between the different VLSI partners was very positive in the development of a common VHDL model of a decoder IC. The IC was available from early June, and was tested successfully (internally to Philips) in a complete set top box environment at the end of June.

Through this activity, MPEG 2 video and transport VLSI know-how has been developed and is now available in Europe. This will be valuable expertise since the consumer set top box market is currently exploding in Europe.

The VLSI and models developed in VADIS are already planned to be used in a fruitful manner in the next generations of European decoders.

7. Conformance Workpackage

The goal of conformance testing is to define methods which can be used to test and ensure the interoperability of various MPEG-2 coders and decoders. VADIS set up this workpackage in order to assist MPEG in the definition of a specification of conformance for both MPEG-2 coders and decoders. A Draft International Standard for video conformance was first published in March 1995 (ISO/IEC DIS 13818-4).

For verifying the compliance of decoders, the procedure defined includes the use of reference bitstreams which can be used to compare the output of the decoder under test with a reference decoder. By analysing any differences in the outputs, the conformance of a decoder can be tested. VADIS partners (HHI, Nokia, NTR, CCETT, LEP) have participated actively in the generation of such bitstreams for the Main, SNR and Spatial Scalable profiles. These bitstreams have been made available to MPEG.

For verification of encoders (or bitstreams) the main test is to verify that the syntax generated stays within the limits of the MPEG-2 specifications.

VADIS partners continue to work in this area, in particular in the generation of bitstreams to test the Systems-Level conformance of MPEG-2 decoders and in the construction of equipment for real-time testing of MPEG-2 coders/decoders.

8. VADIS Demonstrations
9. Demonstration of spatial scalable coding for HDTV at London MPEG meeting, Nov. 1992

This demonstration showed standard-definition television (SDTV) compressed to 6 Mbit/s and HDTV compressed to 14 Mbit/s using pictures derived from software simulations using the MPEG test model. The picture quality achieved was remarkably good considering the relatively low bit rates involved.

The SDTV and HDTV pictures were shown in two modes, “simulcast” and “compatible”. In the simulcast mode, the SDTV and HDTV images are coded independently. In the compatible mode, the SDTV signal is used as one prediction option for the HDTV encoder, and is therefore part of the overall data making up the high-definition picture. This approach offers the potential for improved coding quality and efficiency. Importantly, it also allows for the HDTV picture to degrade gracefully to standard definition in non-ideal reception conditions rather than suffering the total loss more generally associated with digital systems. One possibility offered by this approach would be to use the same signal to provide HDTV to an appropriate receiver using a fixed external aerial, whilst enabling standard-definition pictures to be received on portable receivers using simple set-top aerials.

The simulation work was carried out jointly by workers at BBC Research Department and BT Laboratories.

The HDTV sequences to be displayed were replayed from an HDTV recorder at BBC Research Department in Surrey and were transmitted via satellite (Eutelsat-II F3) to the MPEG meeting at the British Standards Institute conference centre in London hosting the MPEG meeting. For transmission over the satellite the HD signal was coded using 140 Mbit/s video coding equipment developed within the RACE HIVITS Project. This provided a virtually transparent path for the HDTV signals.

2. HDTV Workshop, Turin 1994

A complete digital television chain using the MPEG-2 audio-video coding standard and a satellite transmission link was shown in this demonstration. The chain was compliant with the “DVB” ETSI standard for multi-programme TV by satellite.

The digital TV source was a hard disk with SCSI interface housed in a commercial PC platform. By off-line running software programs, the audio signal had been coded at 256 kbit/s, the video signal at about 5.6 Mbit/s and both were multiplexed into a Transport Stream at 7.7854 Mbit/s.

The Transmission Channel Adapter added FEC redundancy by RS(204,188) with an output bitrate of 8.448 Mbit/s. The delivery chain was composed by a QPSK modulator with rate 3/4 convolutional coding and a portable up-link station to access the satellite in the FSS Ku band. The receiving installation included the antenna, a QPSK demodulator with Viterbi decoding and the RS(204,188) FEC decoder.

The digital TV receiver included the VADIS MPEG-2 Transport Stream demultiplexer and audio/video decoders, with monitor and loudspeakers. A PC intelligent platform was used to handle the presentation parameters and to input commands.

3. Montreux ’95

An experimental system for retrieving remote multimedia information over the Pan European ATM network was demonstrated at Montreux 1995. This demonstration also included prototype MPEG-1 and MPEG-2 based set-top units as well as video servers (see figure 3).

Figure 3: Set-up of Montreux ’95 demonstration

Video at 1.5 Mbit/s (MPEG1) and at 8 Mbit/s (MPEG2) was stored on a video server at CSELT premises. The video server was connected to a local ATM network featuring ATM FORUM 100 Mbit/s links. A number of 34 Mbit/s PDH links connected the CSELT local network to a local ATM switch in Montreux; the AAL5 adaptation layer was used to convey user and control data over a few PVC’s, with a overall bandwidth occupation of about 10 Mbit/s.

The set-top unit (STU) platform was a PC with dedicated hardware for bitstream handling and ATM network interfacing. The VADIS Transport Stream demultiplexer and audio/video decoders was used for MPEG-2 real-time decoding. Software modules running under Windows 3.1 presented the user with a user-friendly service shell, allowing access to emulated interactive applications like Movies On Demand and Teleshopping.

Appendix 1

List of companies participating in the VADIS Project.

Aspex Microsystems Ltd.	ASPEX	GB
Bilkent University	BILUN	TR
British Broadcasting Corporation	BBC	GB
British Telecommunications	BT	GB
Centre Commun d’Etudesde Télédiffusion etTélécommunication	CCETT	F
Centro Studi e Laboratori Telecomunicazioni S.p.A	CSELT	I
Deutsche Thomson-Brandt	DTB	D
Ecole Nationale Supérieure des Télécommunications	ENST	F
Ecole Polytechnique Fédérale de Lausanne	EPFL	CH
Ente Publico Retevision	RETEVISION	E
European Space Agency	ESA
Forschungsinstitut der Deutschen Bundespost Telekom	FI/DBP	D
Heinrich Hertz Institut für Nachrichtentechnik Berlin GmbH	HHI	D
Instituto de Engonharia de Sistemas e Computadores	INESC	P
Institut für Rundfunktechnik GmbH	IRT	D
Intracom S.A. – Hellenic Telecommunincations and Electronic Industires	INTRACOM	GR
Italtel	ITALTEL	I
Katholieke Universiteit Leuven	KUL	B
Laboratoires d’Electronique Philips	LEP	F
Matra Communication	MATRA	F
National Technical University of Athens	NTUA	GR
National Transcommunications Limited	NTL	GB
Nokia Research Center	NOKIA	SF
Norwegian Telecom	NT	N
Olivetti Systems and Networks	OLIVETTI	I
Philips	PHIT	I
Philips Research Laboratories	PRL	GB
RAI – Radiotelevisone Italiana	RAI	I
Royal PTT Nederland N.V., PTT Research	RNL	NL
Siemens AG	SIEMENS	D
Swedish Telecom	STA	S
Telefónica Investigación y Desarrollo	T I+D	E
Telenorma GmbH	TELENORMA	D
Thomson-CSF/LER	LER	F
Universidad Politécnica de Madrid	U.P.M.	E

Project organisation

Project Director	Leonardo Chiariglione	CSELT
Strategic Advisory Group	Cesare Mossotto	CSELT
Technical Coordination Committee	Ken McCann	NTL
	Nick Wells	BBC
Algorithm WP	Gilles Nocture	LEP
Hierarchical WP	Nick Wells	BBC
Systems WP	Geoff Morrison	BT
	Antonio Gaspar	INESC
Demonstrators WP	Nick Tanton	BBC
VLSI WP	Thierry Fautier	LEP
Field Trials WP	Herbert Hofmann	IRT
Conformance WP	Jean-Pierre Henot	CCETT