Process Designs Applied in Production of Media Content ...lib.ugent.be/fulltxt/RUG01/001/418/476/RUG01... · A major investment was designing and testing the ... processes rely heavily

Process Designs Applied in Production of Media Content

Bart Bogaert

Promotor: Prof. Dr. Ir. Rik Van Landeghem

Begeleider: Hendricus W. Koopman

Masterproef ingediend tot het behalen van de academische graad van

Master in de ingenieurswetenschappen: industrieel beheer

Vakgroep Technische Bedrijfsvoering

Voorzitter: Prof. Dr. Ir. Rik Van Landeghem

Faculteit Ingenieurswetenschappen

Academiejaar 2008–2009

Preface

In 1989, two decades ago, I completed my master of Industrial Sciences with a thesis about interactive variation techniques in computer aided design and manufacturing. The main focus of this paper was the usage of computer graphics and simulation techniques in the design of electric circuits. Immediately after the completion of this paper I joined Siemens’ team to investigate visualization and simulation techniques in different business areas. I worked for nearly five years in the Siemens Laboratories in Munich doing investigations, building prototypes in computer visualizations, simulations, and animations for manufacturing, cartography, electric circuit design, optical circuit design, and civil engineering in 2D, 2½D, and 3D. I travelled around the world to work with colleagues, customers, and partners to achieve new challenges by applying the basic research findings into problem solving new designs.

Although computer simulations, animations, and graphics were truly interesting to me, I moved back to Belgium in 1994 and started working in information technology for the telecommunications industry. A major investment was designing and testing the integration of business applications in a new manner. Instead of building point-to-point integrations based on program or database interfaces, I designed the implementation of a message based interfacing technique according to the hub-and-spoke network topology. Applying this in a telecom environment was even more challenging given the volume of transactions and availability requirements. These topologies and developments were the first steps of what nowadays is more common and called ‘Service Oriented Architecture’ integration.

I decided to leave Siemens in 2001 after several projects in the telecommunication industry and explored new horizons that were offered to me by IBM. I started a new project to design market wide processes orchestrated by process servers on top of a service oriented architecture, which integrated applications across different enterprises.

Later IBM offered me the opportunity to move back to the world of visuals, graphics and animations. I started to develop and manage the media and entertainment industry across Europe. Combining past experience in graphics and experience in service oriented architectures enabled me to put forward new ideas. As the whole media and entertainment industry is now going through a wave of digitization, the first adoption of service oriented architecture starts and leading enterprises are moving to the next step of operational efficiency after digitization. Improvements in the fundamental design of the production process remain the major challenge.

Late 1997 I started the Master after Master program in Industrial Management, and after completion of all courses and examinations I restarted travelling. That delayed writing this thesis.

For the accomplishment of this work I received support from many professional colleagues, customers, partners and co-competitors all over the world, besides leveraging the know-how and experience gathered during my professional career. Many of them are now also good friends next to being a peer in the industry we are working in.

A special accomplishment for the guidance obtained from H. W. Koopman during the writing of this thesis.

Finally I wish to express my sincere gratitude to my family for the continuous understanding and encouragements while working and finishing this thesis.

Bart Bogaert, May 2009

Toelating tot bruikleen

“De auteur geeft de toelating deze scriptie voor consultatie beschikbaar te stellen en delen van de scriptie te kopiëren voor persoonlijk gebruik.

Elk ander gebruik valt onder de beperkingen van het auteursrecht, in het bijzonder met betrekking tot de verplichting de bron uitdrukkelijk te vermelden bij het aanhalen van resultaten uit deze masterproef.”

Bart Bogaert, mei 2009

Process Designs Applied in Production of Media Contentdoor Bart Bogaert

Masterproef ingediend tot het behalen van de academische graad van

Master in de ingenieurswetenschappen: industrieel beheer

Academiejaar 2008–2009

Promotor: Prof. Dr. Ir. Rik Van Landeghem

Begeleider: Hendricus W. Koopman

Faculteit Ingenieurswetenschappen

Universiteit Gent

Vakgroep Technische Bedrijfsvoering

Voorzitter: Prof. Dr. Ir. Rik Van Landeghem

Samenvatting

De productieprocessen in de media industrie veranderen onder invloed van veranderingen in de markt, van het gebruik van digitale technologie, en van standaard IT componenten. In dit werk gaan we na waar de grote veranderingen plaats hebben in deze processen en gaan we na welke invloed de keuze van een mediastandaard heeft op deze bedrijfsprocessen. Vervolgens worden de verschillende procesmodellen in kaart gebracht en geanalyseerd, rekening houdend met de invloed van deze mediastandaarden. Tenslotte lichten we enkele potentiële toekomstperspectieven toe.

Trefwoorden

Media, Content, Video

Process Designs Applied in Production of Media ContentBart Bogaert

Supervisors: Rik Van Landeghem, Hendricus W. Koopman

I. INTRODUCTIONThe content industry undergoes drastic changes due to the

adoption of digital technology and from the use of IT based technologies. This has an impact on the core content production processes and results in the design and implementation of new process models. This article analyses the differences between these models.

II. PROCESS STEPS IN NEW IT BASED PRODUCTION MODELS

Analysis of the different process steps used in content production showed that six steps are crucially influenced by the content essence and require attention during process design. These steps are: ingest, editing, post production, transcoding, storage / archive, playout / distribution. The new processes rely heavily on the use of metadata, and therefore accurate metadata becomes crucial.

III. DIGITAL CONTENT FILE FORMATSProcessing of digitized content uses different formats for

audio- and video-essence, metadata, and wrapping of essence and metadata. Compared to the traditional tape based production environments far more combinations are possible and used for file based formats. Today we distinguish fifteen combinations used commonly. Analysis shows that the choice of the format for essence has a drastic impact on bandwidth consumption when moving content between different systems, and on storage consumption when storing the content. Using uncompressed HD material in a production environment requires careful design of the network topology as the number of concurrent streams is limited, if streaming is possible at all.

IV. SYSTEMS DESIGN METRICSThe six crucial steps that touch the content essence as

mentioned previously can be put into the following model.

Figure 1. Abstract system design

With this abstract model the storage and bandwidth requirements are calculated and a simulation is done for the different content formats. We derived from this model that

archiving of content in an uncompressed format requires enormous volumes of storage and that compression prior to archiving can save a factor 10 on the archive storage but requires additional transcoding capacity (processing and bandwidth) and makes editing from retrieved archive material less convenient. We also derived from the model that 90% of the bandwidth requirements come from three process steps: ingest, editing and proxy creation. Moving from SD to HD has an important impact on these requirements. Upgrading to the 720p50, 1080i50 or 1080p25 formats requires five times more resources compared to the SD format and even ten times when using the 1080p50 standard. Higher compression ratios save storage and bandwidth, thus reduce this impact. The final choice of a format is determined by the type of content. For example a news production commonly uses a lower bitrate compared to a sports or long form production.

V. CONTENT PRODUCTION PROCESS DESIGNS

Analysis of the different workflows currently used or in development in file based content production resulted in a cataloguing into five different types of workflow: single work centre workflow, digitized workflow, workgroup model, central media asset management model, and the Service Oriented Architecture integrated enterprise model.

A. Single work centre workflow

This workflow is typically implemented as a first step in digitization; it is the replacement of the physical tape by files without a major change in the production process.

Figure 2. Simplified file based production workflow

The production process is not automated or system managed, leaving a lot of flexibility to the users. This process is widely used in small environments or when recurring programming occurs as required manual operations are limited. The model has typically low storage and network bandwidth requirements.

B. Digitized workflow model

A second model is a managed environment, where content is managed actively by a workflow automation layer to support larger operations resulting in an increase of

efficiency. Two different flavours are used: a first implementation focuses on tight integration of components from a single vendor using proprietary interfaces, a second focuses on efficiency through workflow automation, integrating components from different vendors.

Figure 3. Managed file forwarding production workflow

The production process becomes a high efficient workflow, but tight integration between components results in a low flexibility. This model is applied in mid-sized operations, and storage and network bandwidth requirements can be kept moderate as managed by the automation.

C. Workgroup model

A third file based production workflow is a grouping of different isolated environments; frequently used to meet the different requirements of functionality, and/or formats, or due to historical reasons.

Figure 4. Workgroup model – ‘production islands’

The workflow characteristics of each production island are comparable to the previous model. The lack of sharing possibilities results in a low utilization of resources and limited content exchange.

D. Central media asset management model

The forth model combines the benefits of previous models into a single environment with the use of a central media asset management. It is one of the most advanced models available nowadays and can combine technologies from different vendors.

Figure 5. Central media asset management

The storage and network bandwidth requirements are high; a careful design of network and storage topology is required. This architecture is not widely used in large scale operations today, although this is a very promising architecture.

E. Service Oriented Architecture integrated workflow model

The fifth model integrates production environments with business functions and provides elementary services that can be shared between production environments.

Figure 6. SOA integrated enterprise

File based content production workflows improve quality, efficiency (productivity) and cost compared to the traditional tape based workflow using workflow automation with underlying integration of components. A rather negative aspect is the lower flexibility of the workflows, especially in creative program making where new formats are continuously created. This can be an important point of attention. Flexibility can be improved using a service oriented architecture approach.

VI. CONCLUSIONSThe IT industry evolves towards further use of service

oriented architectures and cloud computing will further influence the content production processes. Standardization of essence formats and metadata within the media industry will positively influence the evolutions towards service adoption and will lead to the use of cloud computing technologies, bringing efficiency and flexibility to a new level. Further research in this area is required to fulfill the industries’ needs.

REFERENCES[1] J. FOOTEN, J. FAST, The Service Oriented Media Enterprise:

SOA, BPM, and Web Services in Professional Media Systems, Elsevier Science (2007).

[2] S. J. BERMAN, N. DUFFY, L. SHIPNUCH, End of Television as we know it, IBM Institute for Business Value, Somers, USA (2007).

[3] E. SANTOS, Operational Patterns, the MXF Flavours?, MOG Solutions, Portugal (2007).

Table of Contents

Table of Contents............................................................................................................iAbbreviations.................................................................................................................ivTable of figures..............................................................................................................viTable of tables..............................................................................................................vii1. Introduction................................................................................................................11.1. Content industry turns to digital technology.............................................................11.2. Content industry turns to information technology....................................................22. Scope of thesis..........................................................................................................42.1. Scope of investigation.............................................................................................42.2. Approach to resolve................................................................................................52.3. Structure of the thesis.............................................................................................53. The content industry...................................................................................................73.1. The content lifecycle...............................................................................................73.1.1. Idea creation step.................................................................................................83.1.2. Pre-production step..............................................................................................83.1.3. Production step....................................................................................................93.1.4. Post-production step............................................................................................93.1.5. Play-out / distribution step....................................................................................93.2. Traditional production process................................................................................93.2.1. Physical content production workflow...................................................................93.2.2. Video formats in content production...................................................................103.3. Focus area of the thesis........................................................................................114. Process steps in new IT based productions models.................................................124.1. Content ingest.......................................................................................................134.2. Content cataloguing..............................................................................................144.3. Content search, browse and retrieve.....................................................................164.4. Desktop editing on low resolution content.............................................................174.5. Post production / NLE editing................................................................................184.6. Content transcoding to other formats....................................................................194.7. Online – nearline – offline – archive content storage.............................................204.8. Subtitling...............................................................................................................21

4.9. Graphics / effects..................................................................................................214.10. Audio...................................................................................................................224.11. Playout-automation.............................................................................................224.12. Content repurposing............................................................................................224.13. Newsroom integration.........................................................................................244.14. Rights management............................................................................................244.15. Content watermarking.........................................................................................254.16. Content quality monitoring...................................................................................264.17. Other data...........................................................................................................274.18. Content production steps conclusion...................................................................274.19. Video sampling formats.......................................................................................304.20. Video compression formats.................................................................................314.20.1. MPEG standard................................................................................................324.20.2. H264 standard..................................................................................................324.20.3. Motion JPEG 2000 (MJ2) standard..................................................................334.20.4. DIRAC standard...............................................................................................334.20.5. MPEG Material eXchange file Format (IMX) standard......................................334.21. Audio standard formats.......................................................................................334.22. Metadata standard formats.................................................................................344.22.1. P/Meta Metadata Exchange Standard..............................................................354.22.2. SMPTE Metadata Dictionary Structure (SMPTE) standard..............................354.22.3. Descriptive Metadata Scheme-1 standard........................................................354.22.4. Standard Media Exchange Framework............................................................354.23. Content wrapping format standards....................................................................364.23.1. Advanced Authoring Format (AAF) standard....................................................364.23.2. Media Exchange Format (MXF) standard.........................................................374.23.3. Operational Pattern Atom (OP-Atom) standard................................................374.24. Common commercial formats used.....................................................................374.24.1. Standard definition formats..............................................................................374.24.2. High definition formats......................................................................................384.24.3. Archive formats................................................................................................394.24.4. Distribution formats..........................................................................................394.24.5. Summary overview common commercial formats............................................394.25. Content digital file formats conclusion.................................................................424.26. Abstract content production model......................................................................434.27. System design metrics........................................................................................444.28. Assumption and simplifications...........................................................................454.29. Storage capacity calculations..............................................................................46

4.30. Bandwidth requirement calculations....................................................................484.31. Storage and bandwidth requirements conclusion................................................504.32. Traditional workflow model..................................................................................524.33. Single work centre workflow model - unmanaged...............................................554.34. Digitized workflow model – managed file forwarding...........................................584.35. Workgroup model – production ‘islands’..............................................................604.36. Central media asset management model............................................................624.37. Service Oriented Architecture integrated workflow model...................................644.38. Content production process designs conclusions................................................685. Next evolution in content production models............................................................715.1. A future enterprise – SOA business orchestration.................................................715.2. A future outlook: cloud integration – enterprise virtualization................................726. Conclusion...............................................................................................................74References...................................................................................................................75Samenvatting...............................................................................................................76

Abbreviations

AAC Advanced Audio Coding

AAF Advanced Authoring Format

ALE Apple Lossless Encoder

AMWA Advanced Media Workflow Association

AVC Advanced Video Coding

B2B Business-to-Business

Content Essence plus metadata

DCT Discrete Cosine Transform

DMS-1 Descriptive Metadata Scheme-1

EBU European Broadcast Union

EDL Edit Decision List

EPG Electronic Program Guide

Essence Essence is the audio, graphic or text itself – the physical output which can be heard or seen by the consumer.

FLAC Free Lossless Audio Codec

GOP Group of Pictures

HD High Definition

IPM Intellectual Property Management

JPEG Joint Photographic Experts Group

Key frame Single frame in a video selected as important to identify the beginning of a new sequence

MAM Media Asset Management

Metadata Metadata is the information or data which identifies and describes associated essence

MOS Media Object Server Communication Protocol

MP3 MPEG-1 Audio Layer

MPEG Motion Picture Experts Group

MXF Material eXchange Format

NLE Non Linear Editor

NRCS NewsRoom Computing System

OP-Atom Operational Pattern Atom

P/Meta Metadata exchange format defined by workgroup organised by the European Broadcast Union

RSS Really Simple Syndication

S2S System-to-System

SD Standard Definition

SMEF Standard Media Exchange Framework

SMEF-DM Standard Media Exchange Framework Data Model

SMPTE Society of Motion Picture and Television Engineers

SOA Service Oriented Architecture

UHDTV Ultra High Definition Television

VCR Video Cassette Recorder

VOD Video On-Demand

VTR Video Tape Recorder

WAV Waveform audio format

WMA Windows Media Audio

Table of figures

Figure 1. Content lifecycle steps....................................................................................8Figure 2. Traditional tape based production process....................................................10Figure 3. Manual content ingest...................................................................................13Figure 4. Low resolution and key frame extraction.......................................................15Figure 5. Search, browse and retrieve content.............................................................16Figure 6. Desktop editing.............................................................................................17Figure 7. Storage hierarchy..........................................................................................20Figure 8. Content repurposing......................................................................................23Figure 9. Rights management......................................................................................25Figure 10. Watermark encoding...................................................................................26Figure 11. Watermark detection...................................................................................26Figure 12. Quality monitoring.......................................................................................27Figure 13. Quality monitoring with reference content...................................................27Figure 14. Abstract system design...............................................................................44Figure 15. Storage example comparison......................................................................47Figure 16. Bandwidth example.....................................................................................50Figure 17. Simplified tape based production workflow.................................................53Figure 18. Simplified file based production workflow....................................................56Figure 19. Managed file forwarding production workflow..............................................59Figure 20. Workgroup model – ‘production islands’......................................................61Figure 21. Central media asset management..............................................................63Figure 22. The workflow automation architecture.........................................................65Figure 23. The service oriented architecture integrated enterprise...............................66Figure 24. The SOA orchestrated enterprise................................................................71

Table of tables

Table 1. Research methodology....................................................................................5Table 2. Traditional video formats................................................................................10Table 3. Essence and metadata in process steps........................................................29Table 4. Bitrate calculation in common used standards...............................................31Table 5. Overview formats...........................................................................................40Table 6. Number of streams for Ethernet and Fibre channel........................................41Table 7. Major system design metrics parameters.......................................................45Table 8. Example storage requirement calculation.......................................................46Table 9. Example bandwidth requirement calculation..................................................49Table 10. Content production process designs conclusions.........................................69

1. Introduction

Media companies such as TV broadcasters, radio broadcasters, publishers, press agencies … are facing unprecedented challenges today. Viewing and listening habits are changing as consumers are bombarded with new choices through new distribution channels, new formats and new business models. The rising use of personal media devices is fragmenting the industry and putting pressure on revenue and costs. Additionally, governments are mandating the move to digital, for example mandatory move to digital TV broadcasting by EU legislation. Flanders switched off the analogue distribution end of 2008. The main driving force behind this changing environment is the digitization of sounds, voice, and pictures, allowing companies to transmit and manipulate their rich media assets in entirely new ways. The entire digitization can be distinguished in two steps:

• The content lifecycle uses content in a digital format.

• Information Technology is used in the content lifecycle.

We first take a closer look to these both steps.

1.1. Content industry turns to digital technology

The entire content industry migrated over the last decade from using analogue technology to digital technology, mainly using proprietary solutions. The adoption of different technologies resulted in a very diversified landscape, and publishers, television and radio broadcasters are confronted with issues when trying to effectively leverage their media assets including:

• Managing islands of content.

• Managing multiple proprietary file formats.

• Minimal ability to scale as traditional archives and content grow.

• Difficulty in finding specific content segments as traditional archives grow, e.g. retrieval of an audio or video segment from a tape archive.

• Holding operational costs in line.

• Managing change in skills.

Moving from an analogue, non-integrated environment to a digital infrastructure based on open standards technology can reap many benefits and provide options that were hard, or even not, to achieve in older business models.

1

1.2. Content industry turns to information technology

Digital media technology based on standard information technology is fundamentally reshaping the way content is produced, managed, stored and distributed, especially in the audio and video industry. By embracing this new technology, the industry is able to establish parallel process flows and drive productivity through automating repetitive or formerly-manual steps within them. It reduces production costs, shortens time to air, and provides an ever-expanding palette of media treatments and content formats. In a content world, fighting for audience share, digital technology in the content production is helping companies to meet their competition head-on.

Taking a look at the broadcast industry, as broadcasters turn to IT systems to store and manage their digitized content, they begin to accrue the benefits from it. The implementation of an open digital framework results in a simpler way to content acquisition, content editing, content distribution and the storage of content. An open solution always makes the most sense whenever possible to implement. When proprietary components are still in place, or still have to be used then the use of an open solution can help streamline the collaborative sharing of video inside and outside of the media enterprise. Open solutions enable using ‘best of breed’ applications from different vendors. The power of ‘going digital’ lies in competing effectively and adapting to a swift-changing environment.

The media industry is also challenged seriously by new market entrants, using new technologies aggressively and applying new business models. Kids and youngsters become heavy users of new media content and new content providers. The use of IT technology in the content production and content distribution lifecycle opens the possibility to explore the new business models. The entry barriers lowered significantly, content can be created at affordable cost and distribution is now possible directly from the creator to the end-user.

Another new trend is content creation by consumers. New digital capturing technologies combined with editing facilities on standard PC’s and content aggregation platforms on the internet allow creative people to distribute personal content. Massive growth of content aggregation platforms such as Flickr for still images or YouTube for video content are well known examples.

Technology also increases the demand for new content formats. Where historically content was created for a dedicated device, TV screen for example, and the entire chain was controlled; we see now a diverging landscape of new devices. Device manufacturers focus on feature rich devices, manufactured in very large volumes at low cost for the consumer market. End-users adopt them rapidly and explore the possibilities to consume content on new ways, having access to content anytime and anywhere and any devices becomes important. Traditional media companies are faced with these new behaviours and demands, so a repurposing of content becomes critical in the current business environment.

Last but not least is the new interactive possibility. Where communication was historically a one-way channel it is now a continuous complex interaction between the

2

content creators, the consumers and all market intermediates. Interactive content becomes a new source of revenue for traditional media companies.

Adoption of new technology was historically a slow process; assets were procured and used over a long period, people used new technology step by step, working processes changed slightly, working habits and organizational structures remained. The use of information technology based components in the core production processes of traditional media companies becomes common business nowadays; the use of assets over a longer period becomes uncertain. Technology differs largely from previous ones and has a steep learning curve; working processes change drastically and require new working habits. On top a lot of media companies use the transformation to change the organizational structures to gain efficiency and productivity.

3

2. Scope of thesis

The content industry is going through a transformation due to adoption of digital technologies and due to the usage of standard information technology components. The challenges occur in the different steps of the content lifecycle and for the different content formats. If we take a closer look on the content production itself we will see different approaches and models applied through the production phase. Hence the investigation on the different production models applied in the content industry.

In this chapter we outline the scope of this thesis. Chapter 2.1 defines the scope of what is investigated and in which area of the content industry. Chapter 2.2 describes the approach on how the investigation is done. Chapter 2.3 describes the structure of the thesis.

2.1. Scope of investigation

Understanding the challenges the content industry is currently facing includes multiple aspects, such as market changes, technology changes, culture changes … etc. Within the content lifecycle the production of content is undergoing drastic changes from digitization and from the adoption of new IT based technologies. Hence the major topic of the thesis is the analysis of the different models applied in the production of content.

Within the industry different models are implemented, or being implemented. Also in literature various ways for content production are recommended. There is a lack of a clear cataloguing; required to understand the current evolution with the associated current problems and benefits of each model applied. A crisp overview derived through literature study and field inventory will be the base to identify tangible benefits from these different models applied.

The entire content industry uses different formats, such as printed publications, music, audio, video, TV …. etc. The focus of this thesis is on broadcast content production, as this part of the industry undergoes drastic changes and management of rich content through complex workflows has some specific attention points.

Prior to investigate the new and future model it is also important to understand the traditional production process model. The old model was based on physical manipulations of content stored on tapes; it was used intensively and includes a series of benefits. Traditional workflow served during decades and has proven values incorporated in the way of working. Understanding these prior to starting the investigation in new process designs for digital media content production will help understand the advantages and disadvantages of some new process designs.

4

Understanding the processes starts first with the definition of the elementary steps performed in a content production process. These different steps are first catalogued. As the processes are now based on the exchange of digital information it becomes important to understand the different formats used as these formats have an impact on the process steps and the entire process design. In a second step these content formats are studied. In a third step the impact of the choice of format on the entire process design is studied, with a special focus on bandwidth and storage usage. After these three steps the different models applied can be inventoried and we can start the analysis of the impact of choice of format on the different components. As the different steps can be combined into different processes used, we can conclude the analysis of advantages and disadvantages of each process model.

A glimpse towards the future is given by outlining how new technologies up till the business level can potentially build new integrated virtualized enterprises, meeting new business demands in a flexible manner.

2.2. Approach to resolve

In order to answer the topic mentioned above three methodologies will be applied in four stages:

• We will identify the process steps used in content production used today and those under implementation through literature study, also leveraging industry insights gained during my professional life.

• Studying the content formats will be based on a literature study.

• Followed by the creation of a model to determine the system metrics for bandwidth and storage consumption.

• Then analyse the findings by distinguishing the different content production models, categorize them and make for each model the analysis benefits and disadvantages.

Table 1 gives an overview of the methodologies used for each stage:

Process Steps

Content Formats

System Metrics

Content Production

Models

Literature study √ √ √ √

Market survey √

Analysis √ √ √ √

Table 1. Research methodology

2.3. Structure of the thesis

The thesis is structured in different chapters; Chapter 3 provides selected background information on the entire content lifecycle to position the content production workflow.

5

In the same chapter the traditional production process is briefly explained to understand the benefits and disadvantages of it.

The next four chapters include the findings of the previously mentioned four stages. Chapter 4 explains the basic process steps used in the content production process. This includes the result of the literature study and how new standard IT based technologies are applied. Chapter 4.18 explains the different formats used in the content production processes. In chapter 4.25 the model created is explained for calculation of bandwidth and storage usage. Finally the results of the study on the different process models are outlined in chapter 4.31. The chapter contains new production processes based on the virtualization of the physical tape assets as studied and are catalogued as follows:

• The digitized traditional workflow - File forwarding

• Work centre workflow model – unmanaged

• Digitized workflow – managed file forwarding

• Workgroup model – production ‘islands’

• Central Media Asset Management model

• SOA integrated enterprise

For each of these processes design the possibilities, the benefits and the performance are studied.

To conclude the thesis a glimpse on future models is bundled in chapter 5 and the conclusions and areas of future research can be found in the conclusion chapter 6.

In the appendices the literature references are listed together with a brief summary.

6

3. The content industry

This chapter provides some background information on the content industry. The first chapter 3.1 provides a high level overview of the content lifecycle to position the production process, the process that is further analysed in this thesis. This is also the process step that drastically changed the insides of the content industry, due to the adoption of new technologies, resulting in technical and non-technical changes.

The second chapter 3.2 outlines how content is produced in the traditional way, using a physical workflow. Understanding this ‘old’ work process provides useful insights for potential benefits and disadvantages of new working processes.

3.1. The content lifecycle

When we talk about content we mean the product created by the media industry for consumption by the audience. These products are created in different formats (audio, video, text …) and distributed through different distribution channels (paper, CD, on-line, TV …). The lifecycle of the ‘content product’ can be divided into 5 major steps, regardless of the format such as music, drama, film... etc. These 5 steps are:

• (1) Idea creation

• (2) Pre-production

• (3) Production

• (4) Post-production

• (5) distribution

These 5 steps are depicted in Figure 1.

7

Figure 1. Content lifecycle steps

Each of the five steps is explained briefly hereafter:

3.1.1. Idea creation step

Media industry is an extremely creative business; driven by creative people inventing continuously new formats (fiction, drama …) and/or making combinations of different media (TV, cinema, radio, on-line, publishing …). The whole business model starts with a new creative idea, mostly quite unstructured, that needs to be translated into a production of the content asset. The ideas and concepts are mostly described in text (scripts, scene descriptions …) and images (sketches, pictures …) and are used to do the marketing research and fund rising. Although we see already some IT based tools able to gather the creative elements of an idea and transform them as input for the next steps, the creativity of the artist / creator remains crucial and most people do not want to tight to much to systems and structures in this phase. Systems and tools can be supportive with simulations of characters or scenes, rough animations using virtual reality and avatars that allow expression of the thoughts.

3.1.2. Pre-production step

Once the ideas have been formulated the pre-production process step can start. This step defines the planning details, budget details, and quality criteria such as video and audio quality expectations. Although several applications have been available for a long time in the market to manage this part of the process, such as SchedulAll, the

8

adoption of such systems is just taking of. Therefore we see now a variety of new market entrants with such applications like S4M, GMedia … Also traditional ERP providers like SAP bring industry focused modules for production planning. The creative element of the industry makes it often quite difficult to shape each media production into an IT system, especially due to the fact that quite often media productions are one-time operations and therefore we see a lot of spreadsheet management around. For the production of drama series or recurring shows it is more doable to develop a re-usable template into an IT system to manage the production.

3.1.3. Production step

The production phase itself is very resource consuming in terms of people and equipment. Traditional production environments are based on integration of the different process steps with forwarding of assets (physical boxes). This is now being replaced by forwarding of digital stored content (network stored files). The production starts with capturing of the content (video, audio, still images, text), includes the whole processing of rough editing, transitions, animations, special effects, graphics, banners, music and sound overlays … until the finished product is created.

3.1.4. Post-production step

Once the production is completed the post-production step finishes the media product as required for a distribution channel. Depending on the distribution channel and the target market items such as subtitling, banners, advertisements, watermarks, content access rights … are added. Physical distribution (cinematic tape, DVD, tape) requires additional packaging.

3.1.5. Play-out / distribution step

Media products have a specific value-cycle since value is digressive over time. For example the value of news items is extremely digressive (high value when the news is hot, and very low value shortly afterwards), whereas the value of films or culture programs is much more slowly digressive compared to news. Media companies distribute content via multiple channels. The content will be adapted depending on the target market and distribution channel. Frequently the distribution on different channels is also shifted in time.

3.2. Traditional production process

3.2.1. Physical content production workflow

The traditional broadcast production process is based on forwarding of content stored on a physical transportable tape or cassette. This medium is used in each step of the production until the final product is also stored on tape/cassette and used in the distribution/playout. After distribution the physical medium is moved to an archive for preservation purpose.

9

Figure 2. Traditional tape based production process

3.2.2. Video formats in content production

A major characteristic of the traditional ‘physical’ workflow is that a limited number of standards are used. The traditional video standards used in most countries around the world were one of the three main standards: PAL, SECAM and NTSC.

PAL (°) SECAM NTSC

Lines/Field 625/50 625/50 525/60

Horizontal Freqency 15.625 kHz 15.625 kHz 15.734

Vertical Frequency 50 Hz 50 Hz 60 Hz

Video Bandwidth 5.0 Mhz 5.0 MHz 4.2 MHz

Territory Europe, Latin America (°), Africa

France, Middle East

Northern America, Japan

(°) 2 additional variants PAL-N and PAL-M are used in Latin America

Table 2. Traditional video formats

As these standards were not compatible with each other, conversion from one standard to the other was required when content was exchanged between different geographical areas. The majority of the exchanges was in the same area and did not require conversions.

Early broadcast video was stored on tape. In 1969 the first video cassettes (U-matic tape) were introduced by Sony. The introduction of the successor Betacam in 1975 and its competitor JVC’s VHS in 1976 changed the industry landscape, including the way of working. VHS became the leader consumer VCR format, and Betacam the professional VCR format. These VCR where used massively and in 1996 the digital standards were introduced for the professionals, Sony’s DVCAM and Panasonic’s DVCPRO. Both

10

formats store content on tape digitally and were used for content acquisition, for editing and for play-out.

The description of the traditional workflow model is only limited since the focus of this paper lies on file based workflow models.

3.3. Focus area of the thesis

The focus of this thesis is on the production process itself, which is heavily influenced by new technologies from standard information technology. Also the way of working is drastically changed, requiring new skills, organisational changes … etc. All this is not considered in this thesis. The focus remains on the impact on the production model itself.

11

4. Process steps in new IT based productions models

The traditional content production environments are under pressure to shift towards file based production processes due to cost pressure, competition of new market entrants, new media standards, new distribution channels, new business models … etc. The entrance of IT technology into the production environment shifts the industry away from tape based production to file based production. This means that the physical tape is virtualized and that media content is stored on a file. It also means that the different steps of the production workflow are done file based, fully digital.

Implementation of IT based production systems in the content industry has multiple benefits. A major characteristic is that the content can be handled in a non-linear way. Browsing through tapes is a linear way of working, requiring moving backward and forward on the physical medium, which is very time consuming especially during edit operations when multiple tapes must be handled. Digital stored content on an IT platform enables moving backward and forward in the content by jumping around on the files stored, a non-linear operation.

In this chapter we discuss the different process steps used in the digital content production workflow. The digital content production workflow can be divided into the following steps:

• Ingest

• Cataloguing

• Search, browse and retrieve

• Desktop editing on low resolution content

• Post production / NLE editing

• Transcoding to other formats

• Online – nearline – offline – archive storage

• Subtitling

• Graphics / effects

• Audio

• Playout-automation

• Repurposing

12

• Newsroom integration

• Rights management

• Watermarking

• Quality monitoring

• Other data

Each step is described in the following subchapters, followed by a short conclusion.

4.1. Content ingest

The first step is the ingesting of existing material from different kind of sources. For each acquisition workflow the operator has to control the video content and the associated metadata. In most cases and ingest application is used that controls the VTR’s, grid and video servers and has also a basic metadata entry screen. Once the content is selected based on time codes and the metadata has entered, the life cycle of the digital assets starts for production and archiving. The operator can perform further cataloguing once ingested by entering additional metadata associated with sequences, shots or single frames. These kinds of operations are typically performed on a low-resolution version, also known as browse video with available key frames.

The figure below shows how tape and manual ingest is typically performed

Figure 3. Manual content ingest

Following acquisition workflows are used:

• Manual ingest from existing video material from a VTR (analogue).

• The operator selects a video tape, the VTR and optionally transfers the timecodes to the ingest application manually. The ingest application controls the

13

VTR, the video router, the video server and the files on the video server. In addition it allows visual feedback and entering associated metadata to the video tape.

• Batch ingest from existing video material using the Sony Flexicart (VTR robot).

• Similar to the manual workflow, the ingest applications also manages the robot in the Flexicart, enabling ingesting of a larger amount of video tapes in an automated manner.

• Scheduled ingest from a live feed, e.g. from event location, newsfeed, …

• Instead of digitizing from a tape medium it is also common to ingest live feeds or news feeds. Ingest planning is created from the feed planning and entered into the ingest application together with the associated metadata for each feed. The ingest application controls the grid and ensures the feed is directed to the videoserver where ingest is started. This includes scheduling, monitoring and administration of a large number of channels. Typically also a calendar view schedule display with file status and full recording history is available.

• Ingest from digital video material.

• Ingest of video material stored on digital video tape is identical to manual ingest from analogue video tapes.

• Ingest from other assets (audio files, graphics, text files) .

• Together with the video data the audio (e.g. second language), still graphics (e.g. covers, highlight pictures) and text tiles (e.g. subtitles, dope sheets from press agencies) are ingested. This is typically done in the ingest application where these files are associated with the video content.

• A specific way of ingest is the immediate ingest from an incoming live feed, in this case the operator uses a button and the ingest application immediately routes the feed to an available ingest server and initiates live capturing. This is done for unexpected occurrences (e.g. incoming newsfeed) and called ‘hot ingest’.

The VTR and robot are typically controlled by a RS422 serial interface.

Beside adding metadata of the entire piece of content it is frequently useful to extend the metadata by adding descriptive metadata to individual sequences, or to single frames.

4.2. Content cataloguing

Adding metadata is a very important part of the workflow as it is crucial to enable retrieval of the available content in all other stages of the content lifecycle. Historically the cataloguing was done on tape level for the entire piece of content. Now it is common to do this on a detailed level such as sequence or single frame.

Most of the cataloguing is done during ingest of content, it adds metadata:

14

• To available content on clip level (e.g. a soccer game).

• On a sequence in a clip (e.g. name of football player who made a goal during a soccer match).

• A specific frame in a clip (e.g. a specific event).

There are a number of ways of entering metadata, for both mandatory and non-mandatory fields:

• Metadata can be pre-registered; this is frequently done when content is expected to arrive e.g. from an incoming live feed, or when available content is ingested manually. Metadata is stored, and a placeholder for the content is created. The archivist / librarian receives a notification upon arrival of the content and he can validate the correct assignment of the metadata to the content. If required he can add adjustments.

• Metadata can be created and adjusted after ingest. This is typically done by an archivist / librarian; he will be notified upon arrival of new content and then manually enters the metadata by watching the content.

• Metadata can be created through an interface, commonly through an XML template provided together with digital content. This is used for ingesting of content delivered in a digital format.

• Metadata can be adjusted by authorized users of the system.

As browsing and cataloguing of high resolution material is quite time- and resource consuming, it is common to generate a frame accurate low resolution version of the content from the high resolution content automatically and to derive key-frames based on either scene changes or time based.

Figure 4. Low resolution and key frame extraction

The bundled content enables now manual content cataloguing to modify and update the catalogue metadata for assets that are being ingested or already present in the archive. This process step takes place after the content has been ingested and the automatically created metadata is filed.

The archivist / librarian retrieves a particular video segment and he can use the thumbnail image and browse tools with the low resolution proxy content to display the video and move quickly through it using VTR emulation functions of the browse tools.

15

At the same time the operator can type annotation data into defined fields in the application for the entire content, also by scene or by frame. Mostly also free-format text metadata can be added.

4.3. Content search, browse and retrieve

One of the major benefits of storing content digitally is the easiness to search, browse and retrieve content, regardless of the type of media or its state (ongoing ingested material, edited material for distribution, archived material …). A range of operations is performed typically to search, browse and retrieve the content.

Figure 5. Search, browse and retrieve content

The first step of each search is the definition of the search criteria; this can be either a relatively non-complicated task by a structured search on certain metadata fields or an unstructured search throughout the entire metadata structure. Typical search functions provided are:

• Free text search– ability to search all fields using a single operation

• Sectioned free-text search – limit free text search to a specific metadata field such as “creator”

• Indexed field search on key database fields such as title, creation date etc.

• Masking, left and right truncation options

• Boolean operators, proximity operators and interval searching is available

• Support for thesaurus

• Restricted input and search of keyword fields using a controlled vocabulary

• Search individual fields

• Combine free text search and field search

• Run combined searches on several individual fields

• Display search history and combine and edit previous searches

• Search templates with guidance/ help functions set up for different user groups

• Updating of metadata quickly reflected in the search engine

16

• Advanced search interface enables the user to build search statement interactively (expression by expression)

The result of the search operation is a series of search result, most frequently presented to the user with important metadata fields (status, date, duration, quality …) and a sequence of key frames.

The user can select in a next step a result and request the low resolution content for browse preview. The low resolution content is returned to the user, and with tools on his desktop he can emulate key VTR functions.

One of the major functions used during browse is the selection of content. In the last step the ‘In’ and ‘Out’ points are defined on the low resolution material and the user then requests the material to the content repository. The conformation is done in the background and the high resolution content is returned to the user.

4.4. Desktop editing on low resolution content

Another major benefit of file based production is the possibility of story creation and editing based on the low resolution content. The user (journalist) can use a simple editor for clip generation (and simple logging) and trimming of material, or an extended editor for story assembly, voice-overs and relatively simple video manipulations.

Figure 6. Desktop editing

Content is retrieved by the user and based on the preview of the low resolution content the different materials are selected and integrated in the newly assembled clip. In its simplest form the material is selected and posed on a new timeline, also called rough-cut assembly. Multiple browse clips can be loaded and trimmed. The result is always an Edit Decision List (EDL) describing how the assembly is created. In a more advanced mode the user can use a number of basic editing features such as a timeline with multiple video tracks, voice-over and audio mixing enhancements. Sometimes this includes the following functionalities:

• Browsing of raw, edited, rendered or EDL material

• Searching for whole clips or parts of by metadata

• VTR emulation: Play, stop, fast forward, go to start, go to end, jog, shuttle, in-mark, out-mark

17

• Time line-based and split-track editing

• Voice-over recording, voice-over importing and basic audio mixing (server-side)

• Split video/audio edit in own timelines

• Shift of timeline for voice-over

• Audio level adjustment on audio timeline

• Frame accurate video editing

• Multiple audio tracks on timeline

In addition, the user has metadata entry/update facilities similar as described in previous chapters.

Once the editing is finished the EDL can be returned to the back-end for conformation. In this process the decisions / operations described in the EDL are performed on the high resolution content. Different operations are possible:

• The EDL is conformed and a new item is created and posted into the repository. This invokes the creation of a low resolution version and the key-frame extraction.

• The EDL is copied together with the associated original high resolution material to a Non Linear Editor (NLE) for further post-processing.

• The EDL is conformed and the content is copied to a play-out server.

4.5. Post production / NLE editing

Desktop editing on low resolution content with server based high resolution conformation is a very powerful and fast way to create a clip. It is frequently used for news production where time is crucial and for fast trimming of films or drama series (called rough-cut). The functionality of the desktop tools is rather limited to standard / simple operations and to limited numbers of video / audio tracks that can be handled simultaneously. The creation of complex transitions between video tracks, adding special effects to the video, complex quality corrections, handling a multitude of video / audio tracks … requires the use of a NLE. The content is used as input for this process in its original high resolution format, and the result is also a high resolution format. This requires sufficient processing capacity and bandwidth and therefore this kind of operations are performed on a graphical workstation.

NLE’s can also import EDL’s, this facilitates the NLE user browsing through the high resolution material as he can start directly with the prepared NLE and the appropriate high resolution content.

Examples of functions:

• Working directly with the high resolution images

• Support of multiple formats, resolutions, encoders …

• Special effects, special transitions

18

• Video correction (colours, light …)

• Multiple video and audio tracks

• Direct conformation

• …

4.6. Content transcoding to other formats

Ingested material should be available in low-res and in key-frame format for browsing purposes. The transcoding workflow converts high resolution to low resolution video and does the key-frame extraction. The encoding of low resolution is a conversion from high resolution content file and done on a server farm. Immediately after the start of ingest or file transfer session, the system should invoke the transcoder to create low resolution video and key frames. In the case of available transcoding resources, transcoding can start as soon as the first high-res GOP/frame has reached the transcoder or has been made available to the transcoder on a shared storage pool. The transcoder provides the following benefits:

• Creation of a frame accurate copy.

• The browse processing is done independently of file origin – a media file exported from a NLE or transferred from a production system is processed exactly the same way as ingested video.

• Support for transcoding of video being ingested can be faster than real time (depends on the codecs used and processor speeds of the processors deployed in the encoders/transcoders).

• In the event of browse encoder shortage, available resources can be grouped into pools, and lower priority media is automatically queued for later processing – less transcoding units are required.

• Generation of thumbnail images for the key frames based on shot detection, scene changes or maximum duration during transcoding operations.

Ingested contents sometimes have another format than the one used internally. This requires a transformation from one format into another. The transcoding farm can be used to perform this operation.

Content for distribution regularly requires a transformation to another format, several examples where transcoding is required:

• High resolution content that needs to be distributed to an internet platform or a mobile-TV platform requires rescaling to a lower resolution, lowering the bitrate, and commonly transcoding to a H264 format. In most cases “cropping” of the media content is also required due to the resizing of the presentation surface.

• High resolution content where the audio tracks are stored in separate files needs a wrapping prior to distribution to the target video server.

• Downscaling from a high resolution format to a lower resolution.

19

• Upscaling from a low resolution format to higher resolution.

• Transcoding from one bit-rate to another bit-rate.

4.7. Online – nearline – offline – archive content storage

High resolution content stored digitally needs adequate management; otherwise systems will be flooded with data comparable to desks flooded with video cassettes / tapes in the past. However, digitized content can be managed actively with systems, and force users to accept policies and rules. Once the media is digitized it can be managed as files, like any other file, by storage management solutions.

In the media industry a wide variety of abbreviations are used when talking about storage of content; following classes can be distinguished:

• Online storage: dedicated to contain the materials, those being produced (working materials) or are ready for transmission or playout (finished materials). Typical characteristics of this kind of storage are the high performance and high availability. The simultaneous read/write access of this storage enables for example ingest (write) and browse or playout (read) simultaneously. Given these requirements, the storage is frequently limited in storage capacity due to budgetary constraints.

• Nearline storage: dedicated to materials that were used in the near past and may be further needed for other productions (after transmission or playout), or to materials that will be used in the near future, not immediately (ingested for later use). Typical characteristics of this kind of storage are a good performance and a large capacity.

• Offline storage: dedicated to those materials that will be kept for long-term storage for potential re-use in the future. Typical characteristics are the extreme high storage capacity and a slower retrieval time.

• Disaster Backup: dedicated to store all materials for long-term preservation. This storage environment is physically separated from the other environments to ensure that the content can be retrieved in case of a major disaster on the primary site. The disaster backup also hold backups of metadata repositories and browse proxy files.

Figure 7. Storage hierarchy

20

Users intend to leave their content on the on-line storage, as this is not complicated and makes their live easy. Key factors limiting the breakthrough enhancements to the production workflow are the serial movements (often physical) of materials between these storage layers. To realize these movements a storage hierarchy is defined and managed by a hierarchical storage manager. Based on policies and rules the media is moved between the different storage layers back-and-forward transparently for the end-user. Materials required in short term (as known from the rundown programming for news or from the playout automation that controls the playlist) are moved upwards and those not needed moved downwards.

The browse content in low resolution is always kept online, enabling the user to search and retrieve content. Content requested that is kept on the near-line or on the off-line storage will be moved automatically upwards by the hierarchical storage manager. Depending on the detailed storage architecture this might take between tens of seconds to a few minutes. No manual interaction is required.

Building a hierarchical storage environment has multiple advantages. Beside the automatic & transparent hierarchical space management it has unlimited scalability across systems, possibility for easy operational backup and integration of a disaster backup.

A particular way of retrieval is partial retrieval: content is partially retrieved from the offline or the backup environment. This is commonly done by defining the in- and out-points of the required content. The system retrieves only the required sequences.

4.8. Subtitling

Subtitling has always been a worry in the content production process. Even in a digital workflow it remains a major concern. The subtitling text is stored in a text file, for each string with the associated timecodes. Following approaches are commonly used:

For linear transmission the traditional way of working is to transpose the subtitles during transmission on the video content. This can be done after the video server by mixing the signals from the video server with the signal from the subtitling engine.

An alternative for linear transmission is to burn the subtitles into the video stream prior to distribution by an encoding engine. The result is a new content file with embedded subtitles. This way of working is also frequently used for distribution towards those media without subtitling support, for example content repurposed via mobile TV or via internet.

If digital distribution is possible then the subtitling texts go along with the content, for example used in DVD distribution of content

4.9. Graphics / effects

Graphics / Effects are banners, logos (brands, symbols, flags …) and explanatory elements (figures, statistics ….) transposed on the video stream are handled quite similar as subtitles; either they are transposed on the video signal or burned into the

21

digital video file. In case of digital distribution the graphics are also burned into the digital video file due to the lack of standardization.

4.10. Audio

Audio follows the same workflow as the video content, going from ingest to the transmission. Depending on the formats used audio tracks can be embedded into the video file. Sometimes audio tracks are stored separately; this enables bundling of video and audio content depending on the target distribution. Content can have for example different audio tracks depending on the geographical area where the content is transmitted to, or different audio tracks can be burned on DVD dependent on the target market.

4.11. Playout-automation

In linear transmission playout-automation is the management and automation of playing the right sequences. To do so the rundown is defined in detail and for each slot in the rundown the correct video files are associated, if needed together with the subtitling and / or graphics files. Playout-automation typically interconnects with the video servers and the broadcast management systems (that handles advertisements, EPG ...). It also automates different scenarios in case of problems and for important operations it has control over secondary video servers. Also alternative programming is defined in the playout-automation. The content repository typically monitors the rundown available in the playout automation to determine the files needed for playout. If content is missing the automation has to ensure the movement from the content repository to the play-out video server in due time.

For non-linear content distribution the content placed on the distribution platform has also a separate management environment. From this environment the required content has to be managed and communicated to enable the content repository to prepare and provide the content in due time.

4.12. Content repurposing

Reuse of content has become crucial in the media industry to remain profitable. Repackaging and redistribution of content through new channels became a common practice, according to Rupert Murdoch: ‘Distribution is King, Content is the Crown Jewel’.

Disney continuously redistributed his cartoon catalogue to generate additional income. As production costs increased for large productions it became a necessity, the primary distribution channel targeted for production was money loosing business. The production of ‘The Lion King’ in 1994 had a production cost of $55 million, with a release weekend gross of $40,9 million. The film alone made $763 million, becoming one of the top-grossing animated films ever. By repurposing the film Disney was able to add initially an additional $500 million to the return on this initial story. This was just the beginning; additional $3 billion was generated through repurposing and merchandising.

22

Repurposing content efficiently gains importance and becomes tightly integrated within the production environment. As each distribution channels has its own specific format the production workflow must take this into account.

Figure 8. Content repurposing

Examples of repurposing of content:

• Web / Internet distribution requires creation of smaller clips and transcoding into a lower resolution, lower bitrate and a new format. Subtitles are commonly burned into the image and multiple language support requires the creation of multiple versions with different audio tracks. Also cropping of the content is done to fit the targeted presentation surface.

• Mobile TV distribution has the specific requirement of cropping the content to increase visibility / readability. Additionally the content is repurposed into smaller clips, then transcoded to a lower resolution, lower bitrate and a new format.

• DVD distribution requires transcoding to the appropriate format, creation of the subtitle files and audio tracks. Additional menu and navigation information is created. Content can also be bundled, for example a sequence of drama series.

• VOD-casting has similar requirements as Mobile TV.

• Narrowcast distribution requires creation of smaller clips and transcoding into the resolution of the target equipment.

• Wholesale distribution is distribution to other parties; content is proposed and requires packaging / bundling as requested by the other party.

• …

23

4.13. Newsroom integration

For the production of news the journalist uses a NewsRoom Computing System (NRCS). It consolidates the scripts, media and graphics of the stories to the final rundown; it is the core working environment for news journalists covering various functions from planning to finishing the story.

Information is traditionally available from agency wires, now extended with information from web sites, blogs and RSS feeds. The NRCS aggregates content sources, has an embedded sophisticated cross-media searches and a flexible notification engine. Journalists can create new stories by combining own information and information available from the sources. New digital NRCS include exchange of picture and video information and allow creation of entire stories with rich content.

NRCS integrates with different other components to have an end-to-end workflow, some examples: archive information must be made available, story subtitling information must be provided to the subtitling engine, content must be delivered to the distribution environment … etc.

The market is dominated by two major NRCS solutions, Avid's iNEWS (formerly named AVStar until this company was acquired by Avid) and the Associated Press' (AP) ENPS (developed by AP together with BBC), both supporting the Media Object Server Communication Protocol (MOS) to exchange information. According to the technical specification; MOS is an XML-based protocol for communication between newsroom computer systems and media object servers. The protocol covers the exchange descriptive data for media objects, exchange of playlists and exchange of status.

4.14. Rights management

Digital Rights Management (sometimes called Intellectual Property Management (IPM)) is a set of solutions designed to streamline the process of acquiring, researching, clearing, granting, selling and accounting for rights across the content lifecycle and to allow the business to manage this information on enterprise level. Managing rights is a separate solution, tackling the increasing transaction volumes and growing complexity of managing intellectual property in the new digital world. Beside the aspect of management the tight integration of the content handling must provide visibility to the end user on the rights prior to further usage of the content partially or fully.

24

Figure 9. Rights management

The example above is a clip C composed from the clips C1, C2 and C3, where C2 is a purchased clip with associated rights. At the time T of transmission the rights are cleared and the content can be used. At a later stage T+x the rights on the purchase content C2 is expired and when clip C is retrieved by the user from the archive. It is crucial to retrieve the rights of usage of Clip C from the rights management system and draw the user’s attention to the fact that no rights are available to repurpose the entire clip due to the lack of rights on the part coming from clip C2. If the content repository knows how clip C was created it can show the user the available rights for re-use for each sequence.

A tight integration streamlines a complex process and allows a better management and exploitation of Intellectual Property.

4.15. Content watermarking

Watermarking is a technique for protecting content against piracy by embedding a still image watermark or a motion vector watermark into the content images. One of the primary requirements of watermarking schemes is the robustness to unintentional attacks while avoiding the introduction of disturbing artifacts. The most important unintentional attack on video is "lossy" compression. In the content production chain, compression is usually applied before broadcasting or before transferring the video to other devices.

25

Figure 10. Watermark encoding

To recover the embedded watermark, the content is first decompressed and then each individual frame is passed through the watermark detection system. If the detection is perfect all the bits of the retrieved signature are correct and the same signature is found in each frame. This also means that only one single frame is required to recover the embedded information. However, practice has shown that in order to keep the watermark invisible, only very slight modifications of the frames are acceptable.

Figure 11. Watermark detection

4.16. Content quality monitoring

Quality has been monitored during decades by human eye and ear using a reference monitor (color calibrated) and a quality headset. Quality monitoring is done for ingest content (arrival inspection), sometimes during the production process and at the transmission.

Nowadays new systems are developed to automate this monitoring process. The transmitted content is returned to a quality monitor process, decomposing the signal back into individual frames and audio. Quality analysis is done on a series of criteria, such as black frames, dropped audio signal, contrast analysis, color variation analysis … etc. At the current state of development major improvements are still required before deployment on larger scale, as frequently to many anomalies are detected that are either invisible to the human eye or the results of special effects introduced in the scenes.

26

Figure 12. Quality monitoring

Another technique to monitor the quality of transcoding is quality monitoring that also uses the reference material as input. In this case the content is compared to the original quality and deviations are reported. This technique is more frequently applied and delivers good working results.

Figure 13. Quality monitoring with reference content

4.17. Other data

Beside storing the content as described in the chapters above it is common to extend the repository with additional information associated to the content, such as still images (pictures), text documents (scripts, reports), graphical documents (flyers, brochures) … etc. The benefit of storing content digitally is that all these pieces can be associated to each other, enabling the user to retrieve the entire bundle swiftly.

4.18. Content production steps conclusion

Digital media content management requires handling multiple files associated with the same content in a bundled approach:

• High resolution content

27

• Low resolution browse content, frame accurate with the high resolution content

• Key-frames

• Audio tracks

• Subtitles

• Still images

• Text / Documents

• Graphics

• Rights (or reference to rights)

• Transcoded versions for transmission / distribution / repurposing

• …

The importance of the essence and metadata format varies depending on the step in the production workflow. For each of the 16 production steps outlined in chapter 4 we determine the impact on storage and bandwidth from the essence (we focus on high resolution essence and not the browse proxy material) and main usage of metadata, an overview is given in Table 3.

Process step Metadata Essence

Ingest either write or attached to pre-created metadata

write (important streams for new ingested content)

Cataloguing Read / write / update -

Search, browse and retrieve

read - (unless read for preview high resolution)

Desktop editing on low resolution content

read read and write, when conforming an EDL after low resolution editing

Post production / NLE editing

read read and write, during editing and when conforming an EDL.

Transcoding to other formats

- read / write, read only when creation of browse proxy, write also when transcoding or conforming

Online – nearline – offline – archive storage

read read (when archived) / write (when retrieved)

Subtitling - - (as this is an overlay at distribution)

28

Graphics / effects - - (as this is an overlay at distribution)

Audio read / write read and write, to be paid attention when multiple audio channels are used

Playout-automation read read (to distribution)

Repurposing read read / write, similar to transcoding

Newsroom integration read -

Rights management read / update -

Quality monitoring - read only, unless comparison with reference material.

Table 3. Essence and metadata in process steps

We can distinguish six steps that have a crucial impact on the process design, especially on the storage and bandwidth requirements:

• Ingest: new essence that arrives in the process.

• Desktop editing on low resolution content: conformation of edited essence requires reading of input essence and creation of new essence.

• Post production / NLE editing: editing requires reading of essence during editing and creation/rendering of resulting new essence.

• Transcoding to other formats: reading of essence and in case of transcoding new essence is created.

• Online – nearline – offline – archive storage: essence is moved, requires reading and writing of essence.

• Playout – distribution: requires reading of essence for sending towards the playout / distribution platforms.

These six steps will further be used in the different content production models.

29

Digital content file formats

Content stored in a digital format using information technology results in content stored in files. Digital content stored in file formats contains two major parts: the essence and the metadata. The essence is the media content and can be a still image, audio, video … etc. The metadata describes the essence. We discuss therefore in this chapter the essence, the metadata, the bundling of essence with metadata, and the formats commonly used.

In this chapter we discuss the content formats used for video and audio. Chapter 4.19 discusses the digitization (sampling) formats for video, followed in chapter 4.20 about the compression of digitized video. We also discuss the compression of audio in chapter 4.21 similar to video.

Beside the essence we also discuss in chapter 4.22 the different metadata formats used for video content.

In chapter 4.23 we discuss the bundling of essence and metadata into a single file, called wrapping formats.

Finally, in chapter 4.24 we discuss the common formats used in the industry for production, archival and towards distribution and in chapter 5.7 how these formats are related to the different production steps.

4.19. Video sampling formats

Video images are sampled one by one; several parameters determine the quality of the video stream and the data volume. Influencing factors are:

• Aspect ratio: is the width of the image divided by its height, common aspect ratios in broadcast video are 4:3 and 16:9 (cinema common projection ratios are 1.85:1 and 2.40:1).

• A frame rate is the frequency at which the image device produces unique consecutive images, called frames per second (fps). Common rates are 25, 30, 50 and 60 fps.

• Interlaced (I) or Progressive (P): an interlaced scan is carried out on alternating even and uneven lines of each image, in order to fill each time the gaps from the previous image. A progressive scan is carried out on the entire image.

• Resolution or the number of pixels per line.

30

• Traditional Standard Resolution (SD) formats: The American standard NRSC in 4:3 has a resolution of 640 pixels width x 480 pixels height; whereas the European standard PAL/SECAM has 768 pixels width x 576 pixels height.

• New High Definition (HD) formats are 720 lines (1280 x 720 pixels) and 1080 lines (1920 x 1080 pixels).

• Pixel colour information, encoded using Chroma subsampling, common sampling at 4:4:2; 4:2:2 and 4:2:0 depth (high end cinema uses 8:4:4).

• We can now compare the data volume of the raw video material for the popular HD formats used compared to the SD format:

FormatResolution Framerate I/

PColour coding

Bitrate

H pix V pix fps Mbit/s Gbit/s

SD 768 576 50 I 4:2:2 (10 bit) 222.184.000 0,205

HD 720p50 1280 720 50 P 4:2:2 (10 bit) 921.600.000 0,858

HD 1080i50

1920 1080 50 I 4:2:2 (10 bit) 1.036.800.000 0,966

HD 1080p25

1920 1080 25 P 4:2:2 (10 bit) 1.036.800.000 0,966

HD 1080p50

1920 1080 50 P 4:2:2 (10 bit) 2.073.600.000 1,931

Table 4. Bitrate calculation in common used standards

From Table 4 we see that upgrading from SD to HD implies that the amount of raw information increases with a factor 5 for the first three HD formats up to a factor 10 for the 1080p50 format.

The combination of resolution, framerate and colour coding influences the production and distribution environments in terms of bandwidth and storage requirements.

If we add the overhead from essence packaging down to network package information then a realistic required bandwidth of 3Gbit/s is needed for the 1080p50 raw data format.

Cinematic formats used are 2K (2048 x 1080 pixels) and 4K (4096 x 2160 pixels). A potential future standard under development by NHK is Ultra High Definition (UHDTV) or also called Super Hi-Vision (SHV) with 7680 x 4320 pixels; this is 4 times wider than and 4 times as high as the HD standard.

4.20. Video compression formats

Video data is compressed using different techniques to reduce the amount of data in such a way that data not necessary for a good visual perception is removed. Compressing video is always a trade-off between quality (low compression factor) and data reduction (high compression factor). Some compression formats do not remove

31

any information, called lossless compression. They are rarely used in video as high compression ratios can be achieved without reducing the visual quality of the images.

An important difference between two techniques needs to be understood, several compression techniques use intraframe compression, others interframe compression. In an intraframe compression the image of the frame is compressed, using only that current frame. In an interframe compression the image of the frame is compressed using one or more earlier frames, frames are grouped in a Group of Pictures (GOP), and interframe compression is done within the GOP. Interframe compression requires the reconstruction of each image and therefore uses the previous frames. A major difference comes when editing both formats, in intraframe compression the video can be cut at any point in time as easy as in uncompressed video. Editing an interframe compressed video requires reconstruction of the frames at the cutting point.

Compression has a positive impact on the requirements for bandwidth and storage. Attention has to be given to editing of interframe compression techniques especially when using long GOP formats.

Following video compression techniques are commonly used:

• MPEG

• H264

• JPEG 2000

• Dirac

4.20.1. MPEG standard

MPEG is the standard design by the Moving Picture Experts Group (MPEG). MPEG-1 was first developed for medium bandwidth and is optimized for CDROM usage. MPEG-2 and MPEG-3 have higher bandwidth requirements, they are used in broadcast video. MPEG-4 was designed initially for very low bandwidth requirement and used for low resolution applications such as videophones. The MPEG standard includes compression in three kinds: intraframe (called I-frame), predictive-coded frames (P-frames), and bidirectional-predictive-coded frames (B-frames). Typically the images are grouped by 15 (in PAL, or 18 in NTSC) in a GOP, where the first frame is intraframe I and then P- and B-frames resulting in a sequence IBBPBBPBBPBB for example.

Using MPEG-2 reduces the raw data, a typically stream HD 1080i50 is reduced to 110 Mbit/s or even with long GOP to 50 Mbit/s.

MPEG-2 Long GOP is widely used in the broadcast industry and has become a proven standard with good quality results. However, the first disadvantage is that the use of interframe compression gives issues during editing and secondly the MPEG standard is not license free.

4.20.2. H264 standard

H264 is another standard compression technique and equivalent to MPEG-4 AVC, or Advanced Video Coding. The standard includes advanced compression using more

32

variability, resulting in higher computing complexity for encoding and decoding. MPEG-4 AVC has the potential to replace MPEG-2 over time as it reaches comparable quality at half the bitrate, requiring more processing capacity.

MPEG-4 AVS is a newer standard, already used in broadcast video, but less still less used than MPEG-2.

4.20.3. Motion JPEG 2000 (MJ2) standard

The JPEG standard was designed by the Joint Photographic Experts Group committee in the year 2000, it supports lossless and lossy compression. Each frame is independently encoded, either lossless or lossy. Its physical structure does not depend on time ordering; it employs a separate profile to complement the data.

JPEG 2000 has major advantages compared to the MPEG compression because it is a scalable coded, with no visual artefacts and it is also license free. It is also the standard used in digital cinema for storage and distribution.

The downside today is the lack of support by various systems and applications.

4.20.4. DIRAC standard

Dirac is an advanced royalty-free video compression format, originally created by the BBC and used in open-source development groups.

Compression has a positive impact on the requirements for bandwidth and storage. Attention has to be given to editing of interframe compression techniques especially when using long GOP formats.

4.20.5. MPEG Material eXchange file Format (IMX) standard

The MPEG IMX compression format was designed by Sony and is derived from the MPEG standard. The standard specifies that the GOP must start with an I-frame followed by a series of P or B frames. The standard does not specify the number of frames per GOP. The GOP of the IMX format contains only 2 frames, first an I-Frame followed by a single P or B Frame. The higher quality IMX format has only one frame in the GOP - a single I-Frame.

The IMX format does not have an issue with editing as only I-Frames are used. The compression factor is not as high as the MPEG format that takes advantage of P and B Frames and it is fully compliant with the standard.

4.21. Audio standard formats

Also the audio that goes along with the video can be stored in raw format or can be compressed and, similar to the image compression, both lossless and lossy compression techniques are used. The following common formats are used:

• Uncompressed formats: Waveform audio format (WAV)

• Lossless compressed formats: Free Lossless Audio Codec (FLAC), lossless Windows Media Audio (WMA), Apple Lossless Encoder (ALE) …

33

• Lossy compressed formats: MPEG-1 Audio Layer 3 (MP3), lossy Windows Media Audio (WMA), Advanced Audio Coding (AAC)

A single audio channel has lower bandwidth and storage requirements than the visual content. A new trend in video is to include multiple audio channels such as in Dolby 5.1 and 7.1 audio. Additionally, different language channels are also sometimes added, resulting in a non-neglectable volume of data where compression of the audio becomes important.

4.22. Metadata standard formats

The quality of image and audio essence is very important to gain the consumer and these poses several requirements to the production infrastructure. However, the entire production process relies on the metadata associated with the digitised content. Metadata identifies and describes the essence and is the unique way to manage and control the process in Business-to-Business (B2B) and in System-to-System (S2S) transactions. Retrieval of essence without metadata would be a monkish work, resulting in useless and worthless essence.

Description of the metadata contains typically the identification of the essence, descriptive information about the content, the rights of usage and the format. Proper metadata makes essence accessible, useable and valuable. The exchange of essence is realised with the use of metadata, some examples in:

• Business-to-Business transactions: procured content is delivered electronically, content needing translation or subtitling is sent to a translation agency, finished content is provided to the distributor … etc.

• System-to-System transactions: ingested content forwarded is to the archive, content is moved to an NLE, content is received from the NRCS …. etc.

Metadata can be transported either attached to the essence (see later) or separately. The metadata is described in a scheme to facilitate the exchange of information between systems. The scheme suggests a series of attributes and definitions, where the attributes are independent, e.g. ‘Title’, from each other and the definitions are a logical group of attributes, e.g. ‘Contract’.

Different initiatives were started and resulted in the development of different standard schemes that are now used in the broadcast industry. The four frequently ones used are:

• P/Meta Metadata Exchange Standard

• SMPTE Metadata Dictionary Structure

• Descriptive Metadata Scheme-1

• Standard Media Exchange Framework

• Many broadcasters create their own metadata model as a subset derived from the models mentioned.

34

4.22.1. P/Meta Metadata Exchange Standard

The P/Meta scheme is the result of a workgroup from the European Broadcast Union (EBU). It is a scheme developed with the objective to support the exchange of logical information and the meaning of information, regardless of the protocol or coding used. The definition of the scheme contains:

• A flat list of attributes complete with semantic definitions.

• A list of transaction sets, each of which is built from attributes and other sets; each set has its own definition of purpose and content.

• A list of reference data (also known as "enumerated values", "code values" or "controlled value sets") for appropriate attributes.

• A syntax and notation for set construction which supports members' requirements for the assembly of a logical set.

4.22.2. SMPTE Metadata Dictionary Structure (SMPTE) standard

The Society of Motion Pictures and Television Engineers (SMPTE) worked on the definition of a metadata scheme at the same time as the EBU. The urgent need for precision in handling and interchanging technical metadata across systems and organisations when using digital systems pushed the definition of an agreed standardisation to enable a correct exchange. The entire metadata scheme is known as SMPTE Standard 335M. The definition of the scheme contains:

• A list of uniquely registered entries.

• Entries are grouped under a number of nodes.

• A list of uniquely registered groups of metadata elements used in systems.

• A controlled list of vocabulary registered, a list of terms that can be used with metadata elements where the values are permitted or an enumerated list.

4.22.3. Descriptive Metadata Scheme-1 standard

The Descriptive Metadata Scheme-1 (DMS-1) standard was constructed using the SMPTE guidelines and relies heavily on the MXF file format structures. It contains many of the descriptive metadata elements used at the exchange of content, both in business-to-business and in system-to-system transactions. The entire scheme is very rich, and 3 variants exist, each with a series of registered groups and each group with an extensive list of registered entries.

4.22.4. Standard Media Exchange Framework

The BBC defined it’s own metadata scheme to support and enable media exchange end-to-end across its different business areas. The Standard Media Exchange Framework (SMEF) was designed and the scheme called SMEF Data Model (SMEF-DM) contains a set of definitions required from production up to the distribution and management of media assets.

35

The intention of the BBC was to use the SMEF Data Model to harmonise the landscape with the objective to enable integration of key information across systems and improve exchange of information. The SMEF Data Model was used to manage the definition of data across different applications. The BBC still continues to develop the scheme, mostly enriching details, and uses it as the baseline for integration.

The BBC made the SMEF Data Model available to enquirers without charge, subject to a no-signature licence incorporated on their website.

4.23. Content wrapping format standards

Essence and metadata can be exchanged between systems / businesses separately. In these cases naming conventions are frequently used, for example video content is stored in a MPEG-2 file and the metadata in a XML file with the same filename but with different extensions. To ensure consistency and to be able to support the exchange of grouped essence another approach is used: essence and metadata are bundled into a ‘content wrapper’. Following three wrapping formats are commonly used:

• AAF: Advanced Authoring Format

• MXF: Media Exchange Format

• OP-Atom: Operational Pattern Atom

4.23.1. Advanced Authoring Format (AAF) standard

The AAF format was created by the Advanced Authority Format Association, since 2007 called the Advanced Media Workflow Association (AMWA), to enable content creators to easily exchange essence and metadata across platforms and between systems and applications across multiple vendors. The format was designed for post production and authoring environments. It includes:

• Bundled exchange of essence and metadata.

• Metadata for identification & location (how the item is uniquely identified), administration (rights, access, encryption & security, etc), interpretive (names, artists, etc), parametric (signal coding & device characteristics) and process (editing & compositing data).

• An object model describing relationships between different pieces of essence and pieces of metadata; the model can handle complex relationships.

• The ability to track the history of a piece of essence, down from the source element up to the final product.

• The possibility to perform the conformation / rendering downstream (if functionality available).

The advantage of the AAF format lies in the information available, such as what the original file format was and where the original piece of media resides, what shows (and how many shows) it has been used in before, who shot a particular piece of footage and when, and where … etc.

36

4.23.2. Media Exchange Format (MXF) standard

The MXF was created by the Society of Motion Picture and Television Engineers (SMPTE) as the standard for exchange of essence and metadata. The MXF format can be described as a subset of the AAF format, with emphasis on straightforwardness for exchange of essence and the associated metadata. It includes:

• Bundled exchange of essence and metadata

• Header and footer information and how the metadata is packed in the format

The MXF does not standardise the way metadata is packed into the format. The MXF format is designed for completed content that needs no further rework or a simple cut only assembly of material.

4.23.3. Operational Pattern Atom (OP-Atom) standard

A derived format is the MXF Operational Pattern Atom (OP-Atom); this format has a tightly defined structure for a single item of essence described by a single essence track. OP-Atom is designed for applications where each essence track is held separately. A clip with visual and audio information will reside in this case in two separate files and the MXF metadata can link the files together belonging to the same clip. A derived version of the OP-Atom format is the OP1A; it enables having both pieces of content (visual and audio) in the same file.

4.24. Common commercial formats used

The above listed overview of video sample formats, compression formats, audio formats, metadata formats and wrapper formats can be combined in huge number of different combinations. Various vendors used over the time several of the possible combinations resulting in common formats used in the industry that can be catalogued in the following four categories:

• Standard Definition (SD) formats

• High Definition (HD) formats

• Archive formats

• Distribution formats

4.24.1. Standard definition formats

Following common formats are used for standard definition:

• DV25 / DVCPRO25 / DVCAM

DV25 / DVCPRO25 is the most common format used to store essence on digital video tape supported by different vendors; it uses a compression called Intraframe Discrete Cosine Transform (DCT) (similar to MPEG) with a fix compression ratio of 5:1 and a colour sample rate of 4:1:1 (NTSC) or 4:2:0 (PAL).

37

The format has a fix bitrate of 25 Mbit/s, making it easy to predict the bandwidth and storage capacity required when the amount of essence is known. Editing is straight forward as intraframe compression is used.

DV25 is the abbreviation used by Sony and DVCPRO25 used by Panasonic.

Digitisation to file based production is called DVCAM and uses DV25 encoding.

• DV50 / DVCPRO50

The DV50 format is identical to DV25 with the exception of the colour sample rate, where 4:2:2 is used. The bitrate is fixed to 50 Mbit/s.

DV50 is the abbreviation used by Sony and DVCPRO50 used by Panasonic.

• IMX

A later development from Sony was to use the IMX standard with standard MPEG compression. The compression is applied at three different levels, IMX-30 (6:1 compression with 30 Mbit/s), IMX-40 (4:1 compression with 40 Mbit/s) and IMX-50 (3.3:1 compression with 50 Mbit/s) using a 4:2:2 profile.

4.24.2. High definition formats

Following common formats are used for high definition:

• XDCAM-HD

XDCAM is the file based video format introduced by Sony, using the MPEG-2 long-GOP compression with a maximum bitrate up to 35 Mbit/s.

The last generation is XDCAM-HD442 which uses the 4:2:2 profile and maximum bitrate up to 50 Mbit/s.

The essence is wrapped in an MXF format, a bundle of the full resolution version and the metadata. Also an MXF with lower-resolution browse format can be made available in MPEG-4.

• AVC Intra

AVC Intra is a fully compliant implementation of the H.264/MPEG-4 AVC standard by Panasonic using a 10-bit intra-frame only compression.

Two different bitrates are used: AVC-Intra 50 with 50 Mbit/s uses a 4:2:0 sampling and the horizontally scaled to ¾ and AVC-Intra 100 with 100 Mbit/s uses a 4:2:2 sampling without scaling.

• DNxHD

DNxHD is a lossy compression format created by Avid Technologies for post production (e.g. post production software suite), usually stored in MXF format (can be others). The compression is intraframe and comparable to JPEG image compression. Typical characteristic is the choice of three selectable bitrates: 220 Mbps (depth of 10 or 8 bits), and 145 or 36 Mbps (depth of 8 bits). The standard is proprietary and owned by Avid Technologies.

38

• ProRes 442

ProRes 442 is a lossy compression format created by Apple for post production (e.g. post production software suite), comparable to the DNxHD format. It is an intraframe compression, using 10 bits color depth, 4:2:2 sampling at full-width 1920x1080 and 1280x720 with a bitrates: 220 and 145 Mbit/s.

• Motion JPEG 2000 (MJ2)

The MJ2 JPEG2000 is used at tree different bitrates: 100, 75 and 50 Mbit/s using 10 bits color depth and 4:2:2 sampling, wrapped in either MXF or AAF. As mentioned before it is a license free format, currently weakly supported.

The MJ2 format is for example used by Thomson.

4.24.3. Archive formats

Archiving of content is commonly done in a long GOP encoding format as it gives a far better compression ratio then intraframe only compression. In most cases the finished product is archived with the associated source content, and editing is not the primary operation on archived content.

Some broadcast organizations, like the BBC, are now investigating the option of archiving digital media files in uncompressed formats.

4.24.4. Distribution formats

Distribution of content is done in a multitude of different formats; a common practice is the conversion of content to the format supported by the target platform. The conversion includes downscaling, resizing, transcoding … etc.

4.24.5. Summary overview common commercial formats

Short overview Table 5 of the above listed formats to compare the different bitrates used:

FormatBitrateMbit/s

Colour sampling

Colour depth# bits

DV25/DVCAM 254:2:0 or

4:1:18

XDCAM-HD420 35 4:2:0 8

DNxHD36 36 4:2:2 8

DV50 / DVCPRO50 50 4:2:2 8

XDCAM-HD442 50 4:2:2

MJ2-50 50 4:2:2 10

AVC-Intra 50 50 4:2:0 10

MJ2-75 75 4:2:2 10

AVC-Intra 100 100 4:2:2 10

39

FormatBitrateMbit/s

Colour sampling

Colour depth# bits

MJ2-100 100 4:2:2 10

DNxHD145 145 4:2:2 8

ProRes 442-145 145 4:2:2 10

DNxHD220 220 4:2:2 8 or 10

ProRes 442-220 220 4:2:2 10

HDCAM SR 4404:2:2 or

4:4:410

Table 5. Overview formats

By looking to the bandwidth required we can determine the number of streams that can be handled for the different technologies, taking into account that maximum 80% of the available networkbandwidth can be used effectively for content.

40

Bitrate Fibre Channel Ethernet

Format Mbit/s2 Gbit 4 Gbit 1 Gbit 10 Gbit

(*)

DV25/DVCAM 25 64 128 32 320

XDCAM-HD420 35 45 91 22 228

DNxHD36 36 44 88 22 222

DV50 / DVCPRO50 50 32 64 16 160

XDCAM-HD442 50 32 64 16 160

MJ2-50 50 32 64 16 160

AVC-Intra 50 50 32 64 16 160

MJ2-75 75 21 42 10 106

AVC-Intra 100 100 16 32 8 80

MJ2-100 100 16 32 8 80

DNxHD145 145 11 22 5 55

ProRes 442-145 145 11 22 5 55

DNxHD220 220 7 14 3 36

ProRes 442-220 220 7 14 3 36

HDCAM SR 440 3 7 1 18

uncompressed SD 222 7 14 3 36

uncompressed HD 720p50 921

1 3 0 8

uncompressed HD 1080i50 1.036

1 3 0 7

uncompressed HD 1080p25 1.036

1 3 0 7

uncompressed HD 1080p50 2.073

0 1 0 3

Table 6. Number of streams for Ethernet and Fibre channel

From Table 6 we can extract the low number of streams that can be handled simultaneously in uncompressed format, an important factor in the content production process. The evolution of Ethernet towards 10Gbit over low cost UTP becomes an attractive alternative to fibre channel in the core processes of content production. The topology of the Ethernet network requires important attention, as network collisions of long lasting video streams have a serious impact on performance. Recurring collisions of a stream will abort the stream and will trigger a retry.

Working with uncompressed HD material is in reach for the compressed 720p50, compressed 1080i50 and compressed 1080p25 formats using Fibre Channel 4 Gbit

41

and Ethernet 10 Gbit connectivity, but is critical (even problematic when editing multiple streams simultaneously) when using the uncompressed formats.

(*) The number of concurrent files that can be transferred is limited when taking into account that the effective bandwidth of 10 Gbit Ethernet is far inferior to the theoretical figure when transferring large media files.

Another technology already used in high performance and supercomputing is Infiniband. This is a serial bidirectional connection based on a switched fabric topology such as Fibre Channel. The base bandwidth is 2,5 Gbit/s in each direction, and it supports double (5 Gbit/s) and quad (10 Gbit/s) data speeds. Infiniband links can be aggregated in units of 4 or 12; a bundle of 12 quad connections comes up to 120 Gbit/s raw bandwidth.

4.25. Content digital file formats conclusion

In this chapter we discussed the different formats used for audio- and video-essence, metadata and wrapping of essence and metadata. Compared to the traditional tape based production environments we notice that far more combinations are possible and used for file based formats.

The choice of essence format has drastic impact on bandwidth consumption when moving content between different systems.

In parallel to the bandwidth requirements we also notice the impact of essence format on the consumption of storage to store content.

42

Systems design metrics

In chapter 4 the process steps as used in new IT based production models were discussed and in the conclusion the following six steps touch the content essence:

• Ingest

• Desktop editing on low resolution content

• Post production / NLE editing

• Transcoding to other formats

• Online – nearline – offline – archive storage

• Playout

In chapter 4.18 the different essence formats were discussed. In this chapter we create a model to calculate the storage and bandwidth requirements for content production, based on the 6 steps as determined in chapter 4. Therefore a generic model is created in chapter 4.26, followed by the determination of the design metrics in chapter 4.27. In the following chapters we list the assumptions used in chapter 4.28 prior to start with the calculation of the storage requirements in chapter 4.29 and the calculation of the bandwidth requirements in chapter 4.30.

4.26. Abstract content production model

Prior to the different workflow alternatives, we discuss the system design metrics used. A production system must be designed in such a way that no visible latencies occur between content arrival, content manipulation and the content transmission during peak-time. The network bandwidth is one of the important design metrics to determine. On the other hand the system should be able to handle the volumes of data, thus storage capacity is the other important design metric to determine. For simulation we assume an abstract content production model with following parameters:

• content arriving as input (ingest)

• content transmitted as output (distribution)

• content searched and retrieved in low resolution browse mode

• content edited in high resolution on NLE’s

• content transcoded, conformed, watermarked …

43

• content that arrives might require transcoding, also content that is distributed might require transcoding

The model for simulation purpose can be depicted as follows:

Figure 14. Abstract system design

4.27. System design metrics

From this model we can now list the parameters that influence the design of the system. Table 7 lists the parameters and the abbreviations further used for our calculations. If the parameter influences the network bandwidth calculation then a tick is set in the column ‘Bandwidth’. If the parameter influences the storage capacity calculation, a tick is set in the column ‘Storage’.

Parameter Metric Abbr. Bandwidth Storage

Browse video data format Mbit/s bb √ √

Estimated number of users that use the system simultaneously at peak time in browse mode

users ub

√

Estimated number of NLE users that use the system simultaneously at peak time

users ue

√

Averge number of files used during NLE operations (read and write)

files fe

√

44

Parameter Metric Abbr. Bandwidth Storage

Preferred archive video data format

Mbit/s bv √ √

Preferred archive audio data format

Mbit/s ba √ √

Number of audio channels channels na √ √

Simultaneous video streams fed into the system at peak time

feeds ni √

Simultaneous video streams going out of the system at peak time

feeds no √

Estimated number of hours ingested per day that will be stored for a limited period of time

hours hi

√

Estimated amount of archive material produced per day (ingest + production)

hours ha

√

% of ingested material that needs transcoding to central format

% pi √

% of streams going out that needs transcoding from central format

% pa √

Desired period that ingested content must remain available online

days to

√

Desired period that the browse space must be able to manage

days tb √

Desired period that the archive must be able to manage

days ta √

Table 7. Major system design metrics parameters

4.28. Assumption and simplifications

Prior to start the calculation of the required network bandwidth and storage capacity, a series of assumptions are made to simplify the model:

• the stored content and the NLE use the same format and that the content exchanged between the production environment and the NLE does not need transcoding.

• Access to content from components like encoders, NLE’s … occur once.

• the file system adds no additional overhead to the network bandwidth requirements. Although it is known that the file system adds approximately 15% to the bandwidth requirement we do not take it into account. The correction factor of 15% can be added to the final result.

45

• during peak no network retries occur. The effect of retries can later be corrected with a correction factor derived from best practices and experience.

4.29. Storage capacity calculations

We first start to determine the capacity of the storage. The total storage capacity can be divided into the following sub-capacities:

• Cb: Storage capacity required for on-line browse format content (proxy format)

Calculation Cb in TB: Cb = bb * ha * tb * 3600 / 223 TB.

• Co: Storage capacity required to for online content for standard-quality and high-quality productions, both ingested and own produced during the time required.

Calculation Co in TB: Co = (bv + (ba * na)) * hi * to * 3600 / 223 TB.

• Ca: capacity required to archive all standard-quality and high-quality productions, both ingested and own produced.

Calculation Ca in TB: Ca = (bv + (ba * na)) * (hi + ha) *ta * 3600 / 223 TB.

Remark: the factor 3600 is the conversion from sec to hours, and the factor 223 the conversion from Mbit to TB.

We take for this example an HD production environment with the following parameters (storage capacity for 3 years):

Parameter Metric Abbr. Value

Preferred browse video data format Mbit/s bb 1,8

Preferred archive video data format Mbit/s bv 100

Preferred archive audio data format Mbit/s ba 5

Number of audio channels channels na 5

Estimated number of hours ingested per day that will be stored for a limited period of time

hours hi

50

Estimated amount of archive material produced per day (ingest + production)

hours ha

60

Desired period that ingested content must remain available online

days to 20

Desired period that the browse space must be able to manage

days tb 1095

Desired period that the archive must be able to manage

days ta 1095

Table 8. Example storage requirement calculation

Results in the following storage requirements:

46

• Cb = 50 TB

• Co = 118 TB

• Ca = 3609 TB

The same calculation can be done for different video formats as listed in Table 6 using the same parameters as in Table 9. This results in the following storage requirements as shown in Figure 15:

0

5000

10000

15000

20000

25000

25 35 36 50 75 100 145 220 222 440 921 1036 2073

bitrate Mbit/s

TB

ArchiveOnlineBrowse

Figure 15. Storage example comparison

We notice that the storage requirements for this example were Co and Ca are with the same compression:

• Cb represents approximately 1% of the total storage volume. As the storage is used for browse content continuously by several users, the availability and bandwidth are important design parameters.

• Co represents approximately 9% of the total storage volume. As the storage is intensely used for production and as a working buffer towards the archive the quality of storage (MTBF of the disks) must be taken into account.

• Ca represents approximately 90% of the total storage volume. As the storage is accessed less frequently the majority can be implemented using an automated tape robot.

A small difference in storage capacity is required for using HD 720p50, 1080i50 or 1080p25 formats. Using the 1080p50 doubles the storage capacity requirements.

A feasible option is to transcode the content in Co to a compressed format with high compression ration prior to archiving on Ca. This can save up to a factor 10 on the large

47

storage Ca but requires additional transcoding capacity (processing and bandwidth) and makes editing from retrieved archive material less convenient.

Finally we remark that the storage capacity multiplies when moving from SD to HD with a factor 4,30 when using 720p50, 1080i50 or 1080p25 and even with 8,5 when using the 1080p50 format.

4.30. Bandwidth requirement calculations

We determine now the bandwidth requirements within the model. The total bandwidth capacity can be divided into following sub-capacities:

• the bandwidth required during content ingest which has the maximum bandwidth requirements when simultaneous video streams are fed into the system at peak time or (bv + (ba * na)) * ni Mbit/s.

• the bandwidth from browsing the low resolution proxy content has the maximum bandwidth requirement when the maximum number of users use the system simultaneously at peak time or bb * ub Mbit/s.

• the bandwidth required during distribution of the content has the maximum bandwidth requirements when simultaneous video streams are going out of the system at peak time or (bv + (ba * na)) * no Mbit/s.

• the bandwidth required during editing content using an NLE has the maximum bandwidth requirements when the NLE users that use the system simultaneously at peak time, using a number of files during the NLE operations or (bv + (ba * na)) * ue * fe Mbit/s.

• the bandwidth required to transcode ingested content into the format used in the production environment, where the content after ingest needs to be send to the transcoder and is returned from the transcoder or 2 * (bv + (ba * na)) * ni * pi

Mbit/s.

• the bandwidth required to transcode the content for distribution that requires another format for playout than the format used in the production environment, where the content needs to be send to the transcoder and is returned from the transcoder or 2 * (bv + (ba * na)) * no * po Mbit/s.

• the bandwidth required to create the proxy content ingested and created using the editors, the content is send to the transcoder and the low resolution proxy content is returned or (bv + (ba * na) + bb) * (ni + ue) Mbit/s.

The worst case situation is where all these requirements for bandwidth occur simultaneously, resulting in the total bandwidth required is the sum of all above requirements.

We take for example an HD production environment with the following parameters:

Parameter Metric Abbr. Value

Preferred browse video data format Mbit/s bb 1,8

48

Estimated number of users that use the system simultaneously at peak time in browse mode

users ub50

Estimated number of NLE users that use the system simultaneously at peak time

users ue3

Averge number of files used during NLE operations

files fe 3

Preferred archive video data format Mbit/s bv 100

Preferred archive audio data format Mbit/s ba 5

Number of audio channels channels na 5

Simultaneous video streams fed into the system at peak time

feeds ni 9

Simultaneous video streams going out of the system at peak time

feeds no 2

% of ingested material that needs transcoding to central format

% pi 0

% of streams going out that needs transcoding to central format

% pa 0

Table 9. Example bandwidth requirement calculation

This results in the following bandwidth requirements:

• the bandwidth required during content ingest: 1125 Mbit/s

• the bandwidth from browsing the low resolution proxy: 90 Mbit/s

• the bandwidth required during distribution of the content: 250 Mbit/s

• the bandwidth required during editing content using an NLE: 1125 Mbit/s

• the bandwidth required to transcode ingested content: 0 Mbit/s

• the bandwidth required to transcode the content for distribution: 0 Mbit/s

• the bandwidth required to create the proxy content: 1521 Mbit/s

Total bandwidth required is 4111 Mbit/s or 514 MB/s

The same calculation can be done for different video formats as listed in Table 6 using the same parameters as in Table 9. This results in the following storage requirements as shown in Figure 16:

49

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

25 35 36 50 75 100 145 220 222 440 921 1036 2073

bitrate Mbit/s

Ban

dwid

th M

bit/s

Figure 16. Bandwidth example

We notice in this example that 90% of the bandwidth requirements comes from three steps: ingest, editing and proxy creation. This is comprehensive taken that the number of ingest channels is a multitude of the number of playout channels, additionally each ingest triggers the creation of low resolution proxy material.

A small difference in bandwidth capacity is required between using for HD the 720p50, the 1080i50 or the 1080p25 formats. Using the 1080p50 doubles the bandwidth capacity requirements.

A working practice is to have high compression ratios during ingest. This reduces the bandwidth requirements drastically in the entire model. Content that needs no editing, is best suited to be ingested directly in a compressed format with high compression ratios.

Finally we notice that the bandwidth capacity multiplies when moving from SD to HD with a factor 4,25 when using 720p50, 1080i50 or 1080p25 and even with 8,4 when using the 1080p50 format.

4.31. Storage and bandwidth requirements conclusion

In this chapter we have seen the impact of choice of bitrate on the storage capacity requirements and the bandwidth requirements. Knowing from experience a 15% overhead to storage and bandwidth capacities, we can conclude that moving from SD to HD has an important impact. Theoretically:

• 720p50, 1080i50 or 1080p25 formats requires five times the storage and bandwidth requirement as those of the SD format.

50

• 1080p50 format requires ten times the storage and bandwidth requirements as those of the SD format.

Higher compression ratios save storage and bandwidth. Moving from SD to HD also commonly implies the use of higher compression ratios. The final choice for a format is frequently determined by the type of content. For example news production uses a 50 Mbit 8bit format, resulting in an increase by a factor three when moving from SD to HD in production. Another example is sports or long form production, where DNxHD145 becomes popular, resulting in an increase by a factor 4 when moving from SD to HD in production.

51

Content production process designs

The core process that handles the content essence is implemented differently at each content producer for the key steps in the essence flow from ingest, storage, and editing to archiving and distribution. Studying these differences resulted in a cataloguing of these workflows into five types. In a first step the traditional tape based workflow is studied briefly to understand the improvements by using a file based workflow and to understand the potential benefits lost when using a file based workflow. The traditional workflow is explained in chapter 4.32.

The next 5 chapters from chapter 4.33 to chapter 4.37 discuss the file based workflows distinguished. These workflows are catalogued into:

• Work centre workflow model - unmanaged

• Digitized workflow model – managed file forwarding

• Workgroup model– production ‘islands’

• Central media asset management model

• SOA integrated enterprise model

The conclusions of this chapter can be found in chapter 4.38

4.32. Traditional workflow model

The traditional workflow based on tape as a medium to integrate the different steps in the production of content is depicted in Figure 17. Tapes from different inputs are used to create the necessary content sequences for transmission. Both the input and the result of editing is stored on tape, given the program maker the flexibility to design his own workflow as he is not tied to a predefined sequence of steps. He can move back and forward, and integrate content available on his own desk or from the archive, until the desired result is produced. Different editing tools can be used, as long as the tapes are carried from one editing tool to the next one and as long as they are compatible. The finished content is then forwarded to the playout center for transmission and finally handed over to the archivist for archival.

52

Figure 17. Simplified tape based production workflow

The playout automation manages the rundown for transmission by controlling the VTR’s, the subtitling engine, the graphics engines and the video grid. Integration of the entire workflow is done on an SDI signal. Interoperability of the different components is assured by support of the same video signals. As all major vendors do so, the flexibility to replace components is very high. Upgrading an environment is quite easy; step by step the different components can be replaced.

The benefits of a tape based production workflow are:

• The formats and standards used for storage of content on video cassettes lasted for a long time, even the digital standards were used for more than a decade. These long lasting standards were used by a wide variety of proprietary and expensive equipments with a high degree of interoperability. This created a high form of flexibility in the workflow. Content can be easily transferred from one step in the production process to the next step by moving the video cassette physically. Also the sequence of the steps can be inverted in several cases, or additional steps can be performed on other equipment as needed.

• This high degree of flexibility stimulated and supported the creativity; if some editing or effects were not possible in one step of the production then it was easy to realise this on another workstation.

• The integration of the value chain by physical forwarding of video cassettes also created the assurance of compatibility between different steps, as long as the vendors used the same intermediate well defined storage format on video cassette.

53

• As video cassettes get copied across the production chain, a redundancy was created, each manipulation was stored on a new tape to ensure the original material was not lost. This working habit created very high redundancy of material; going backwards or restarting a task was possible as long as the original input video cassette was not lost.

• The journalist / program maker can manage his own ‘private’ collection of tapes in his office.

The disadvantages of a tape based production workflow are:

• Reading from and writing to video cassettes is a linear processing step, even some equipments allow reading and writing faster than real-time it remains very time consuming due to the linear storage format.

• Manipulation of the tape requires specific tools and skills. As these tools are expensive, they are not available to a large audience and these are heavily solicited.

• Each step can only start upon completion of the previous step, as the result has to be written to tape and physical media needs to be moved.

• Moving the media from one step to the next requires handling, with risk of errors and low efficiency.

• Although tapes are carefully labelled and indexed, mistakes occur frequently, especially in handling the different intermediate versions of the same content.

• Storing media content on video cassettes is a non-neglectable cost, especially because in each intermediate step another video cassette is used.

• Retrieval of content is very time consuming as in most cases the meta-data associated with the video cassette is related to the overall content, description of scenes or even single frames is non-existing.

• Retrieval of content stored in an archive room requires human intervention, including risks of errors.

• Tape is known for the deterioration of the material (aging) and requires thus good environmental storage conditions as well as making an exact duplicate of the original after a few years to ensure quality. Management of the deterioration is time consuming; each tape needs individually to be loaded and analyzed. Content will be of inferior quality and even frequently lost due to the lack of careful management.

The lead time, or the elapsed time between ingest and playout, is long for a tape and real-time for a live feed. In the best case the tape is manually forwarded for direct playout, something that takes a long handling time. Commonly the content is verified, edited and forwarding before playout, a linear operation that takes at least tens of minutes for a small piece of content.

54

The procurement cost of the environment is very high due to the use of dedicated infrastructure, and the operational costs are also high given the use of expensive tapes and all the manual handling of tapes.

The interoperability of equipment has a positive influence on the excessive flexibility of using different components. Workflows are manually carried out and can be adapted and required without technical design changes.

4.33. Single work centre workflow model - unmanaged

A first step in digitization is the replacement of the physical tape by files. Digitization brings several benefits, regardless of the workflow model implemented:

• Reading from and writing to files is a non-linear processing step, saving time for partial content access for example.

• Manipulation of the files can be done on standard platforms, using affordable user-friendly tools that are available for a large audience.

• Under certain conditions each step can start before completion of the previous step, accelerating the entire process.

• Moving the media from one step to the next requires no physical handling, reducing the risks of errors and increasing the efficiency.

• Storing media content on file has a significantly lower cost compared to tape.

• Retrieval of content is very fast when the associated metadata is available. Metadata can be related to the overall content, describe scenes or even single frames.

• Retrieval of content from an automated digital archive requires no human intervention, reducing risks of errors.

• Digital content is stored for long-time archive on data-tape. Video tape is known for the deterioration of the material (aging), one of the reasons is the physical contact between the tape and the reading head. Data tapes are far less vulnerable for deterioration as there is no physical contact between tape and reading unit. Additionally, data tapes are stored in an acclimatised robot, the robot inspection routine takes care of the tape aging and automatically duplicates on a new tape when necessary.

The disadvantage of digitization is the lack of standardisation. As we outlined in 4.25 we noticed that far more combinations are possible and used for file based formats.

This brings us now to the first file based workflow model.

Incoming tapes are digitized using a VTR connected to a video server, which still is a linear process, resulting in files stored on the video server. These files can now be previewed and manipulated in a non-linear way of working. A straight forward way to do this is to connect the NLE editing stations to the storage of the video server. Now

55

the editor can create content and store it on the video server. The finished content is now immediately available for playout. Remark that the entire workflow is nearly identical compared to the tape based working, except that tapes are virtualized to files, a very high degree of flexibility remains. A sample workflow is depicted in Figure 18.

Management of the content is commonly done manually using enhanced file manager functionalities of the video server. A common practice is the use of naming conventions with characters and timestamps to indicate the difference between raw materials and finished content.

Figure 18. Simplified file based production workflow

The playout automation now controls the video server instead of the VTR’s; the integration of the other component remains unchanged.

The benefits of a single workcenter unmanaged production workflow are:

• Within a single environment a single standard is used to store the content. Given the loose integration with the editing environment the high degree of flexibility in the workflow remains; steps in the production process can be inverted or intermediate work can be transferred to other equipment as needed.

• The degree of flexibility ensures support of the creativity; if some editing or effects were not possible on one editing station then it can be easily performed on another workstation, as long as both are connected and compatible.

• The integration of the value chain is very high, neither physical forwarding of video cassettes nor physical forwarding of files between systems is required as content is stored and managed in a single environment.

• Simple EDL’s as supported by the video server can be created to trim the content, thus to determine the in- and out-points.

56

• This kind of entire setup is commonly provided by the same vendor, this assures compatibility between different steps as long as the content formats remain unchanged.

• As content is managed manually on the storage a degree of redundancy is created; each manipulation is stored on a new file to ensure the original material is not lost. This working habit creates some degree of redundancy; going backwards or restarting a task is possible as long as the original input files are available.

• The journalist / program maker can still manage his own ‘private’ collection of files as long as storage space is available.

The disadvantages of a single workcenter unmanaged production workflow are:

• Editing of the content is done in high resolution, requiring sufficient infrastructure capacity.

• Each step starts upon completion of the previous step as the integration is done manually on file level.

• Although files are carefully labelled and indexed, mistakes occur frequently, especially in handling the different intermediate versions of the same content.

• Management of the content is done manually, this requires discipline otherwise the storage space will be flooded by content creating a continuous demand for extending the storage capacity of the system.

The storage and network bandwidth requirements are low for the following reasons: Content is not moved between ingest and playout, no transcoding to create low resolution browse content, using simple NLE’s to determine the in- and out-points. The parameters bb, ub, ha, pi, pa, tb, ta are equal to zero. For storage we have to consider only the Co requirements, and for bandwidth we can gain more then 50% of the bandwidth calculated in chapter 4.30. The major bandwidth requirement comes from the interaction from NLE operations, and as content is mainly read on the NLE for the creation of an EDL that is returned to the video server, the bandwidth that is required is for transfer of content to the NLE.

The lead time can be very short as ingested content is available for playout after the first block of data is written, typically a few frames. This means that content can be transmitted after a few seconds. Live feeds are typically forwarded directly to playout. If the content needs to be edited then the lead time remains very short, the content is only copied once towards the NLE and the EDL returned.

The procurement cost of the environment is low (< 100K) but the need for manual management creates important operational costs, unless used for those channels where need for manual operations and management is limited.

These kind of environments are typically provided by a single vendor, limiting flexibility of component choices and design changes. The workflow is managed manually and allows flexibility in the operations.

57

These kind of environments are suitable for either very small scale operations or where limited editing functionality is required and housekeeping can be done manually. We find this workflow at the majority of the current implementations, here several examples:

• A local TV channel with a limited amount of air-time that creates a few programs weekly.

• A movie channel. The content is delivered, needs only to be ingested and trimmed. The amount of content is limited due to recurring programming.

• A home shopping channel, content is delivered and loaded. The volume of content is limited due to recurring programming.

• Theme channels also use existing content in a recurring program and have limited editing requirements.

The model discussed has no archive functionality. In many cases the content is not produced and the ingested content is not owned, and thus can be removed after the rights of distribution are expired. If necessary the model from Figure 18 can be extended with a disk or tape archive. Archiving has no impact on the bandwidth requirements when transfers are done outside the off-peak period.

4.34. Digitized workflow model – managed file forwarding

The second file based production workflow we can distinguish is a model adopted by larger operation. This model is a managed environment, where content is managed actively by a workflow automation layer. Different workflows are defined, in most cases depending on the format being produced. The overall automation layer takes care of moving the content within the environment. Ingested content is replicated to a central repository or moved directly to the playout environment. The workflow automation also handles the creation of low resolution browse content and key frame. Those are stored for browse and retrieval with the associated metadata into a content store. A tight integration with the editing facilities enables swift manipulations and easy content creation. Content is offloaded toward a content archive after transmission. A sample workflow is depicted in Figure 19.

Two different flavours of these implementations can be distinguished:

A first implementation of the model lays the focus on tight integration of components, providing a very high efficiency in moving content through the entire workflow. This is commonly achieved with a set of components by a single vendor (video server, local content store, NLE, encoders, automation), tightly integrated through proprietary interfaces. Several vendors provide tightly integrated environments focused on and optimized for production of a single program format, for example live sports.

A second implementation of the modal lays the focus on efficiency through workflow automation, integrating components from different vendors.

58

Figure 19. Managed file forwarding production workflow

The benefits of a managed file forwarding production workflow are:

• A highly efficient workflow, fully automated by format type, moves the content duly on time back and forward between the different components (video server, local content store, NLE, encoders) in a multi vendor environment.

• Single content repository in a single vendor environment avoids moving content and reduces the need for local storage in each component (video server, local content store, NLE, encoders). Several solution designs allow start of a step before the previous step is completed, for example streaming playout while ingest is running with a small latency.

• Content is available in low resolution browse format with key frames and medata, enabling efficient search and retrieval.

• The integration of the value chain is very high, neither physical forwarding of video cassettes nor physical forwarding of files between systems is required.

• This kind of entire setup is commonly provided by the same vendor; this assures compatibility between different steps as long as the content formats remain unchanged.

• Automation enables proper housekeeping of content.

The disadvantages of a managed file forwarding production workflow are:

• Reduced degree of flexibility due to predefined workflows; reduction of the creativity as alternative flows to perform additional steps can only be done if foreseen during the design of the workflow.

59

• In a multi vendor environment the content is moved between the different components and this requires local storage capacity at the different components.

• Proprietary storage solution in case of a single vendor environment.

• Low flexibility in choosing the components; compatibility is mostly only ensured when procured from a single vendor.

• Frequently difficult mixing of multi-format productions.

In terms of process design is this model similar to the model used in chapter 4.26 resulting in the fact that the content and archive store and the bandwidth requirements of the model can be derived from the conclusions in chapter 4.31.

The content is moved forward by the workflow and this mechanism causes significant lead times even when the content is moved during ingest. Each copy in the workflow takes a few seconds and this cumulates to a total lead time to at least tens of seconds.

The procurement cost of the environment averages (500K – 1Mio) but the operational costs can be limited upon the condition that the implemented workflow takes control of repetitive work.

These kinds of environments are typically integrated for a given number of workflows; changes in the workflow require design changes, and changes in components also require design changes.

These kinds of environments can be found frequently, and the workflow is typically configured for a well defined format and purpose. Several examples:

• Sports channel production environments, with high efficient live workflow between ingest and playout and special functions such as highlights creation, slow-motion … etc.

• News channel production environments, with tight integration between produced and live content and dedicated workflow.

• A film production environment with a set of high performance, feature rich NLE’s.

The storage and bandwidth requirements are demanding in these situations, but not to the extreme as the environment serves a single or a limited amount of channels.

4.35. Workgroup model – production ‘islands’

A third file based production workflow is a grouping of different isolated environments. These multiple environments are sometimes implemented either to meet the different requirements of formats or due to historical reasons as different systems have been procured over time. Each production ‘island’ has its individual workflows and individual formats. This model is also applied in decentralized production environments or co-located production environments. Exchange of content is limited and probably not desired given the different formats handled in each ‘island’. A loose integration is realized at transmission. A sample workflow is depicted in Figure 20.

60

Figure 20. Workgroup model – ‘production islands’

The benefits of a workgroup production workflow are:

• The process flow within each environment can be properly configured, adjusted and optimized to the production needs of the format handled.

• Content formats can be configured and optimized independently between the different environments.

• The loose integration creates independent operability.

• Parallel islands create a degree of redundancy, as far as the other system is able to handle the formats and the workflow.

The disadvantages of a workgroup production workflow are:

• Reduced degree of flexibility due to predefined workflows, reducing the creativity as alternative flows to perform additional steps can only be done if predefined into the workflow and if components available in the single production environment.

• Duplication of functionality increases the total cost.

• Utilization is low, no optimization of resources between the environments, for example for encoders, graphics engines … etc.

• Inflexible storage capacity; spare capacity can not be assigned temporarily or shifted between environments. High cost of managing the different capacities.

61

• Low rate of exchange of content between the different environments especially when these are provided by different vendors or when other formats are used because content can only be exchanged after transcoding towards the destination format.

The storage and network bandwidth requirements are spread over the different production islands; each island has its own storage and bandwidth requirement and operates fairly independent from the neighbour islands.

The lead time, the costs and the flexibility of each individual island is comparable to the previous model.

These kinds of environments are typically used for companies handling different formats simultaneously; they typically create environments dedicated for each format. Examples are

• Broadcasters with a mixed rundown, having a production environment for News, one for Sports production, one for drama and series … etc.

• Theme channels with different formats, each format created in another production environment.

4.36. Central media asset management model

The next file based workflow model combines the benefits of previous models into a single environment. One of the major challenges is the definition of the central media asset management in the model; it is one of the most advanced models available nowadays as it combines technology from different vendors and different formats. Major characteristics of such environments are:

• Integration of technology of multiple vendors.

• All content is stored centrally in an independent neutral format (if content is ingested in another format it will be first transcoded to the format stored) or in multiple formats.

• The storage space in the different components is reduced to the absolute minimum. All content is immediately forwarded to the central storage platform. If possible the content is even streamed to and from the central storage platform directly.

• Loose coupling of components.

• Bundling of content and associated data into a single repository.

• A sample workflow is depicted in Figure 21.

62

Figure 21. Central media asset management

The benefits of a workflow with a central media asset management are:

• High exchange of content between all users, as search and retrieve is done on a single repository covering the entire content catalogue.

• Single hierarchical storage management can be implemented, reducing spare capacity between user-groups and application of a single policy.

• The storage architecture can be implemented with standard IT components, although special attention must be given to the design.

• Efficient repurposing of content as grouped together, for example a web site can be fed with different kinds of content like News, Sports … etc. Users with skills in repurposing content to a specific medium can now browse the entire catalogue and start repurposing right away.

• NLE’s from different vendors connected increase the support of creativity, users can joggle around to perform their preferred operations until completion of the product.

• Single view on the availability of finished content that is available for transmission and playout.

• Increased flexibility in the exchange of components or in adding components although attention must be given to the integration and compatibility.

63

The disadvantages of a workflow with a central media asset management are:

• Different ingest technologies might provide different input formats, requiring additional transcoding right after ingest to the target storage format.

• Integration of components from different vendors might reduce efficiency in case of live ingest – playout through the entire environment.

• The centralised system can be heavily solicited and becomes critical in the operations, requiring attention to availability and redundancy in the design.

The storage and network bandwidth requirements are high; a careful design of network and storage topology is required. Also here the conclusions of chapter 4.31 are valid, but with a high number of ingest channels, users and playout channels the total requirements can be extremely high.

The content is forwarded to the central media asset manager by the workflow and this mechanism causes lead times of at least tens of seconds; however once ingested the content becomes available to all users. A major benefit comes from the possibility to perform simple editing on low resolution material; content can be selected rapidly, and an EDL conformation can be done on a server.

The procurement cost of this environment is high, but the operational costs can be low upon use of the benefits, such as the enterprise wide availability of content upon ingest, the possibility to easily re-use content … etc.

Design and workflow changes are limited, although these can be done once centrally. Flexibility of component choice will largely depend on available interfaces from the central media asset management system supplier.

This architecture is not widely used in large scale operations today, although this is a very promising architecture. The severe constrains on storage and network requirements have an impact on project budget and potential risks. An intermediate practise is to integrate this model into the model from chapter 4.35 where independent production ‘islands’ are integrated to exchange assets between islands and to archive assets centrally using the central media asset management system. This combination gives the benefits from both models and reduces constrains on storage and network requirements. Storage volume requirements can be optimized on the central media asset management system by using a transparent storage hierarchy management policy as described in chapter 4.7.

4.37. Service Oriented Architecture integrated workflow model

Regardless of which model described in the previous chapters, it is always a tight integration between a series of content systems linked to content services. The business process requires additional integration with business systems. A schematic sample diagram is depicted below in Figure 22, where the numerous amounts of links are visualized. The entire environment is embraced by a workflow engine to facilitate the user and enabling a better monitoring. The workflow is an efficient implementation, but interfacing lacks flexibility. Changes in the workflow might require changes in the interfaces. Changing components remains critical and is only possible when

64

compatibility is assured or when all interfaces are reviewed and tested. The major drivers to change towards a new architecture are needs for flexibility and responsiveness

Figure 22. The workflow automation architecture

The current workflow automation environments show a lack of flexibility, they are unable to implement continuous changing business needs. New business models and new creative ideas require flexible changes in the processes, deep down the smallest function in the entire system. The lack of responsiveness and to monitor and measure process flows leads to a slow decision making.

Content production is in the early stage of adopting IT based tools and components, and still uses proprietary formats and interfaces. Only recently open standards based tools became available on the market. Production environments still remain tight islands, integration and workflow automation with business applications become possible in a service oriented architecture. We distinguish the following types of components:

65

• Content systems are systems that store and manage content, such as a video server, a production island, an archive …

• Content services are systems that modify, manipulate content without storing or managing the content. Content services receive content as input and produce output. Such services are transcoders, watermarking embedding services, content protection services …

• Business systems are systems related to content, without handling the content itself; examples are right management systems, ERP systems …

Systems are not tightly connected to each other, but an intermediate adapter takes care of the interfacing between the system specific interface and the enterprise service bus.

Figure 23. The service oriented architecture integrated enterprise

Service oriented architecture implementations in the content industry require the following two specific attention points compared to implementations of service oriented technology between business systems:

• Volume handling: Transfer of content between different systems means transfer of larger amounts of data. To optimize the data transfer on the enterprise service bus, intelligence is put on the level of the bus. Content is stored in a generic place and references through the entire bus are used to reduce bandwidth requirements.

• Format management: Content format requirements might be different in two different systems. To facilitate the workflow transparency and make it independent from the content format used, intelligence is put in the level of the bus. The bus becomes “media aware”, it recognises the content format of the

66

requesting service and knows the format of the requested service. It will call an intermediate content transcoding in between when necessary.

From Figure 23 we see that the content production islands remain independent operations, integrated through the enterprise service bus for generic business and content services. The benefits of the service oriented architecture integrated workflow are:

• Each system has a single interface to the enterprise service bus for the integration towards all other systems. This reduces the number of interfaces to develop and to maintain, in this case only one single point of maintenance exists.

• Progress of enterprise processes can be monitored on the level of the enterprise process bus; this increases the transparency of performance of the enterprise process and the performance of each individual service attached to the bus.

• The single interface to the enterprise service bus increases the flexibility in replacing the subsystems, as the interface is specifications are known and transparent for the other services using the service through the enterprise service bus. The subsystems can be replaced as long as the interface specifications can be maintained.

• The business orchestration is done on the level of the enterprise service bus using the process orchestration engine as a workflow engine. Enterprise processes can be modified or extended without impact on the subsystems.

• The increase in the responsiveness of business workflow supports better business management and control, supporting the business decision making process.

The disadvantages of the service oriented architecture integrated workflow are:

• Integration of subsystems using an adaptor to integrate them in the enterprise service bus is dependent of the subsystem’s interface capabilities.

• the support by vendors of subsystems to support the adaptor technology is lacking, mainly due missing standardization in content and messaging formats.

The storage and network bandwidth requirements are spread over the different production islands; each island has its own storage and bandwidth requirement and operates fairly independent from the neighbour islands.

The content is moved forward by the workflow, and this mechanism causes significant lead times.

The procurement cost of the environment is high due to the additional costs of the enterprise service bus and the orchestration tools. The major benefit comes from the flexibility in workflow changes and the possibility to measure and manage the use of each connected subsystem. The operational costs of the workflow can be calculated and the cost of each subsystem benchmarked.

67

This kind of environment supports workflow and design changes in a flexible manner. Flexibility of component choice will largely depend on available interfaces from the component supplier to integrate with the enterprise service bus.

These kinds of environments are currently under development and test, deployments can be expected in the near future, mainly at companies that handle different formats simultaneously and who are seeking for integration of the operations, improved collaboration and maximising asset valorisation.

4.38. Content production process designs conclusions

In this chapter we discussed five content production process designs, each of them with the given benefits and disadvantages. We compare these with the traditional workflow and give a summary in Table 10:

68

Characteristic

Trad

ition

al w

orkf

low

Wor

k ce

ntre

wor

kflo

w

Dig

itize

d w

orkf

low

Wor

kgro

up m

odel

Cen

tral m

edia

ass

et

man

agem

ent

SOA

inte

grat

ed e

nter

pris

e

Number of standards used

Low High High High High High

Flexibility supports creativity

High High Low Low Low Average

Compatibility between components different vendors

High Low Low Low Low Average

Data redundancy Average Low High High High High

‘private’ collections Yes Yes No No No No

Non-linear editing No Yes Yes Yes Yes Yes

Browse proxy editing

No No Yes Yes Yes Yes

Expensive tools Yes No No No No No

Parallel processing No Yes Yes Yes Yes Yes

Physical moving of assets

Yes No No No No No

Cost of tape High Low Low Low Low Low

Retrieval of content Slow Average Fast Fast Fast Fast

Retrieval of content from archive

Slow Slow Fast Fast Fast Fast

Deterioration of tape

Yes No No No No No

Lead time High Low Medium Medium High High

Cost – CAPEX High Low Medium Medium High High

Cost – OPEX High High Medium Medium Low Low

Workflow flexibility High High Medium Medium Medium High

Component choice flexibility

High High Medium Medium Medium Medium

Table 10. Content production process designs conclusions

69

File based content production workflows improve quality, efficiency (productivity) and cost compared to the traditional tape based workflow using workflow automation with underlying integration of components. A negative aspect is the lower flexibility of the workflows, especially in creative program making where different formats are continuously created; this can be an important point of attention. Flexibility can be improved using a service oriented architecture approach.

70

5. Next evolution in content production models

Content production models evolve over time and will be influenced by the trends in the Information Technology industry. In this chapter we discuss two potential trends that might further influence the production models. First the trend of using service oriented architectures will gradually sneak into the core production process. Secondly the IT industry approach of cloud computing technology will also influence the production process. Both potential evolutions are briefly described.

5.1. A future enterprise – SOA business orchestration

Today’s first approaches of service oriented architecture integration as described in chapter 4.37 integrate business systems and production environments where the integration within the production environment remains in most cases an integration using proprietary standards.

Figure 24. The SOA orchestrated enterprise

71

A future enterprise implementation might unbundle these larger islands into smaller components integrated through the use of service oriented architectures. This creates a loose, flexible and measurable integration among the elementary services into seamless enterprise-wide single business process.

Integration of small or niche applications to fulfil specific needs becomes feasible. Todays large monolithically production islands have the tendency to do internal integration of the different components for technical and commercial reasons, externalising the integration of the different components creates independency and flexibility.

Standardisation is the major driver for service integration as it is the basis for interoperability and avoids difficult conversions through adaptors. Market accepted standards stimulate additional developments of new innovative functionality and consolidated deployments of the basic services.

The benefits of using service oriented architecture integration at a lower level are:

• Better utilisation of shared resources through the entire enterprise, as different production environments can share resources.

• Increase of flexibility in the choice of components and in the design and execution of the business process.

• Increase in responsiveness, fast adaptation possible to changes in business environment as process changes are implemented on the level of the process orchestration engine.

• Business functions implemented as a service can be benchmarked with alternative offerings, and changes in the orchestration layer can balance the function from an internal service to an external service transparent for the end-user.

The disadvantages of using service oriented architecture integration at a lower level are:

• Requires more resources on the level of the enterprise service bus compared to the tight integration used today.

• Additional latency might be introduced, something unacceptable for live productions.

• Main concern today is that the lack of standardisation prevents vendor neutral deployment of this architecture.

5.2. A future outlook: cloud integration – enterprise virtualization

Business functions implemented as a service are also available on internet using standardised interfaces. Service providers focus on a set of business functions

72

implemented on a shared platform and run as an individual business model. These kinds of services are called open cloud services. These new open cloud services are commonly adopted by consumers and by small businesses. Larger enterprises and institutions adopt and integrate services for non-core or uncritical business functions.

Integration of cloud services in the content production model is a first trend that can be expected. For example transcoding services or indexing services can be offered as a service and integrated into the business process.

The IT industry adopts cloud computing infrastructure solutions to satisfy internal needs from the IT department to meet business flexibility, growth, and cost requirements. Enhanced infrastructure virtualisation combined with automatic provisioning and management is used for a wide range of applications, a trend that can be expected to be used within the content production model. The services available in a cloud computing environment can scale according to business needs, a flexibility beneficial to the content industry in several cases. Large events such as sports, shows, elections, etc. require temporary additional resources during the events, where today these requirements can only be met with the use of expensive spare capacities. A cloud computing environment would enable a planned provisioning of capacity for the duration required and balanced towards other business functions afterwards.

73

6. Conclusion

The content production process is undergoing drastic changes from digitization and from the adoption of new IT based technologies.

Analysis of the different process steps showed that six steps are crucially influenced. These are: ingest, editing, post production, transcoding, storage / archive, playout / distribution.

Processing of digitized content uses different formats for audio- and video-essence, metadata and wrapping of essence and metadata. Compared to the traditional tape based production environments far more combinations are possible and used for file based formats. The choice of the format for essence has a drastic impact on bandwidth consumption when moving content between different systems and on storage consumption when storing the content.

Moving from SD to HD has an important impact on the storage and bandwidth requirements. Upgrading to the 720p50, 1080i50 or 1080p25 formats requires five times more resources compared to the SD format, and even ten times when using the 1080p50 standard. To reduce the impact higher compression ratios save storage and bandwidth, and the final choice of a format is determined by the type of content.

Analysis of the different workflows used in file based content production resulted in a cataloguing of five different types of workflow: work centre workflow (unmanaged), digitized workflow (managed file forwarding), workgroup model (production ‘islands’), central media asset management model, and the SOA integrated enterprise model.

File based content production workflows improve quality, efficiency (productivity) and cost compared to the traditional tape based workflow using workflow automation with underlying integration of components. An aspect negatively influenced is flexibility of the workflows, especially in creative program making where different formats are continuously created, this can be an important point of attention. Flexibility can be improved using a service oriented architecture approach.

The IT industry evolves towards further use of service oriented architectures and cloud computing, and will further influence the content production. Standardisation of essence and metadata within the media industry will positively influence the evolutions towards service adoption and lead to the use of cloud computing technologies, bringing efficiency and flexibility to a new level. Further research in this area is required to fulfil the industries’ needs.

74

References

[1] N. WELLS, O. MORGAN, B. DEVLIN, J. WILKINSON, M. BEARD, P. TUDOR, The MXF Book – Introduction to the Material eXchange, Format Society of Motion Picture and Television Engineers (2006).

[2] J. FOOTEN, J. FAST, The Service Oriented Media Enterprise: SOA, BPM, and Web Services in Professional Media Systems, Elsevier Science (2007).

[3] R. HOPPER, EBU Technical Review report, European Broadcast Union (2000).

[4] EBU, High Definition (HD): Image Formats for Television Production, Geneva, European Broadcast Union (2004).

[5] B. DEVLIN, MXF— the Material eXchange Format, European Broadcast Union (2002).

[6] W. SIMPSON, Video over IP – A practical guide to technology and applications, Elsevier Science & Technology (2005).

[7] M. COX, L. TADIC, E. MULDER, Descriptive Metadata for Television, Elsevier Inc (2006).

[8] S. J. BERMAN, N. DUFFY, L. SHIPNUCH, End of Television as we know it, IBM Institute for Business Value, Somers, USA (2007).

[9] V. MEGLER, L. ESTRADA, From tapes to bits: The changing landscape of broadcasting, IBM Digital Media, White Plains, USA (2005).

[10] E. SANTOS, Operational Patterns… the MXF Flavours?, MOG Solutions, Portugal (2007).

[11] Avid white paper, Avid DNxHD Technology, High definition without the high overhead, Revolutionary Avid HDxHD encoding, Avid Technologies, USA (2008).

75

Samenvatting

I. Inleiding

De media industrie ondergaat een drastische verandering door het gebruik van digitale technologie en door het gebruik van nieuwe IT gebaseerde technologieën. Deze veranderingen hebben ook een invloed op de productieprocessen, waardoor het mogelijk wordt nieuwe procesmodellen te implementeren.

Gedurende het vorige decennium migreerde de media industrie van analoge naar digitale technologieën, vaak gebaseerd op specifieke oplossingen wat resulteert in een gedifferentieerd landschap van diverse technologieën. Een migratie van een analoge of niet geïntegreerde omgeving naar een digitale infrastructuur gebaseerd op open standaarden kan vele voordelen bieden die moeilijk, of zelfs helemaal niet mogelijk waren in de vroegere business modellen.

Het gebruik van IT-gebaseerde technologieën verandert fundamenteel de manier waarop media wordt geproduceerd, beheerd, opgeslagen en gedistribueerd, vooral in de audio- en video-industrie op basis van bestanden. Door de nieuwe technologie wordt het mogelijk om parallelle processen uit te voeren en het manuele repetitieve werk te automatiseren. Dit vermindert de productiekosten, verkort de doorlooptijden en maakt het mogelijk om de diversificatie van mediaformaten te ondersteunen.

Met media bedoelen we de producten geproduceerd door de media industrie voor consumptie door het publiek. Deze producten worden vervaardigd in diverse formaten en verdeeld via diverse distributiekanalen. De levenscyclus van media kan men indelen in 5 belangrijke stappen:

(1) Idee creatie: de media industrie is een zeer creatieve business. De creatie van een nieuw product begint vaak op een niet gestructureerde manier. De ideeën en concepten worden beschreven en geschetst in een vroeg stadium zonder gebruik te maken van specifieke tools, zodat de creatieve vrijheid niet wordt begrensd. Soms worden animatie technieken of virtuele werelden gebruikt om bepaalde concepten te simuleren.

(2) voorbereiding productie: alvorens de productie te starten worden de verschillende stappen in detail uitgewerkt, inclusief planning, budget details en kwalitatieve verwachtingen. Tools voor productieplanning worden courant gebruikt.

(3) Productie: de eigenlijke productie is zeer arbeidsintensief en middelenintensief. Het proces omhelst de opname, het bewerken van het beeldmateriaal, animaties, effecten, muziek … tot een compleet product.

76

(4) Afwerking: het product moet nadien nog afgewerkt worden volgens het specifieke distributiekanaal. Dit omhelst het toevoegen van de verpakking, reclame, watermerken … etc.

(5) Distributie: het uiteindelijke product wordt via diverse distributiekanalen verdeeld naar het publiek.

Het onderwerp van deze thesis omvat het productieproces die sterk beïnvloed wordt door het gebruik van standaard IT technologie. Hierdoor wordt ook de manier van werken drastisch veranderd zodat nieuwe vaardigheden en organisaties vereist zijn, maar dit maakt geen deel uit van deze thesis.

II. Proces stappen in nieuwe IT gebaseerde productieomgevingen

Uit de analyse van de verschillende productieomgevingen kunnen we de volgende processtappen herkennen:

• Acquisitie: de media wordt gedigitaliseerd of digitaal aangeleverd

• Toevoegen metadata: aan elk stukje media wordt de metadata toegevoegd

• Zoeken, doorbladeren, ophalen: op basis van de metadata is het mogelijk om media met zoekcriteria terug te vinden, te bekijken en op te halen

• Bewerken van media in lage resolutie op een PC

• Post-productie: het bewerken van media in hoge resolutie op een werkstation

• Transcoding: Omzetten van media van het ene formaat naar een ander

• Opslag van media in een hiërarchie van online, near-online en archief

• Ondertiteling

• Toevoegen van grafische elementen en effecten

• Audio toevoegen, bewerken

• Automatiseren van het afspelen voor distributie

• Hergebruik van media

77

• Integratie met een nieuwssysteem

• Beheer van rechten

• Toevoegen van watermerken

• Beheer van de kwaliteit

• Beheer van aan de media gerelateerde gegevens

Uit de analyse van de verschillende processtappen blijkt dat zes stappen cruciaal worden beïnvloed door de essence van de media en bijgevolg bijzondere aandacht vereisen tijdens het ontwerp van het productieproces. Deze zes stappen zijn: acquisitie, bewerken van media, post-productie, transcoding, opslag / archief, distributie. De nieuwe procesmodellen maken sterk gebruik van metadata en bijgevolg is het cruciaal dat de metadata accuraat aanwezig is.

III. Digitale media file formaten

In de productieprocessen van digital media worden verschillende formaten gebruikt voor audio en video essence, voor de metadata en de manier waarop essence en metadata worden gebundeld. In vergelijking met de traditionele tape gebaseerde systemen zijn er meer combinaties mogelijk en in gebruik bij bestand gebaseerde productiesystemen. Tegenwoordig zijn er een vijftiental verschillende combinaties courant in gebruik. Analyse toont aan dat de keuze van bestandsformaat een belangrijke invloed heeft op de netwerkbandbreedte en de opslagvolumes. Het gebruik van niet gecomprimeerde HD formaten in een productieomgeving vergt een gedetailleerd netwerkontwerp omdat het aantal parallelle stromen beperkt is en in sommige gevallen zelfs niet mogelijk.

De volgende tabel is een voorbeeld van het theoretische aantal parallelle stromen voor de courante formaten die mogelijk zijn bij het gebruik van Fibre Channel of Ethernet.

78

Bitrate Fibre Channel Ethernet

Format Mbit/s 2 Gbit 4 Gbit 1 Gbit 10 Gbit (*)

DV25/DVCAM 25 64 128 32 320

XDCAM-HD420 35 45 91 22 228

DNxHD36 36 44 88 22 222

DV50 / DVCPRO50 50 32 64 16 160

XDCAM-HD442 50 32 64 16 160

MJ2-50 50 32 64 16 160

AVC-Intra 50 50 32 64 16 160

MJ2-75 75 21 42 10 106

AVC-Intra 100 100 16 32 8 80

MJ2-100 100 16 32 8 80

DNxHD145 145 11 22 5 55

ProRes 442-145 145 11 22 5 55

DNxHD220 220 7 14 3 36

ProRes 442-220 220 7 14 3 36

HDCAM SR 440 3 7 1 18

uncompressed SD 222 7 14 3 36

uncompressed HD 720p50 921 1 3 0 8

uncompressed HD 1080i50 1.036 1 3 0 7

uncompressed HD 1080p25 1.036 1 3 0 7

uncompressed HD 1080p50 2.073 0 1 0 3

IV. Systeem ontwerpparameters

De zes cruciale processtappen die de essence behandelen en die een belangrijke invloed hebben op het procesontwerp zoals vermeld in het vorige hoofdstuk worden gebruikt in een model. Op basis van dit model wordt een analyse uitgevoerd op de benodigde netwerkbandbreedte en de benodigde opslagcapaciteit voor de courant gebruikte bestandsformaten.

Voorbeeld van opslagcapaciteit benodigd voor de verschillende bestandsformaten:

0

5000

10000

15000

20000

25000

25 35 36 50 75 100 145 220 222 440 921 1036 2073

bitrate Mbit/s

TB

ArchiveOnlineBrowse

Voorbeeld van bandbreedte benodigd voor de verschillende bestandsformaten.

79

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

25 35 36 50 75 100 145 220 222 440 921 1036 2073

bitrate Mbit/s

Ban

dwid

th M

bit/s

Van het model kunnen we afleiden dat niet gecomprimeerde formaten een enorme bandbreedte en opslagcapaciteit vergen. Door compressie kan een factor 10 bespaard worden, maar bijkomend transcoding capaciteit is nodig, en het bewerken van de media essence verloopt veel minder vlot. Tevens wordt afgeleid uit het model dat 90% van de bandbreedte gebruikt wordt door drie processtappen: acquisitie, bewerken en het maken van een lage resolutieversie. De overgang van SD naar HD heeft een belangrijke invloed op de systeemvereisten. Een overgang naar de 720p50, 1080i50 of 1080p25 formaten vergt vijf keer meer middelen in vergelijking met het SD formaat, en zelfs tot tien keer meer middelen bij een overgang naar het 1080p50 formaat. Om de impact te beperken gaat men uit praktische overwegingen kiezen voor hogere compressieverhoudingen om zodoende de bandbreedte en opslagcapaciteit te beperken. De keuze van het uiteindelijke formaat zal sterk afhangen van het type product dat men maakt; bijvoorbeeld voor nieuwsproductie wordt courant een lagere bitrate gebruikt, en bij sportproductie of film een hoge bitrate.

V. Productieprocesmodellen

Een analyse van de verschillende bestandsgebaseerde productieprocesmodellen die vandaag gebruikt worden (of die in ontwikkeling zijn) resulteert in vijf verschillende types: het enkelvoudige werkcentrum model, digital workflow model, werkgroep model, centraal MAM model en het Service Oriented Architecture geïntegreerd model.

A. Het enkelvoudige werkcentrum model

Dit model is typisch de eerste stap in de digitalisatie. Het is het vervangen van de fysische tape door bestanden zonder belangrijke wijzigingen in het productieproces.

80

Het productieproces is niet geautomatiseerd of gecontroleerd door een systeem en laat bijgevolg een grote flexibiliteit en vrijheid aan de eindgebruiker. Dit soort omgevingen wordt frequent gebruikt in kleine omgevingen of bij herhalende programmatie waar de manuele tussenkomst van nature uit beperkt is. De eisen aan bandbreedte en opslagcapaciteit zijn gering.

B. Digitaal workflow model

Een tweede model maakt gebruik van een actieve automatisatie van het productieproces door middel van een workflow systeem om aldus de efficiëntie te verhogen. Men kan twee varianten onderscheiden: een eerste variant legt de nadruk op een intense, performante integratie van componenten van éénzelfde leverancier door middel van proprietaire interfaces, een tweede variant legt de nadruk op een efficiënte workflow door integratie van componenten van verschillende leveranciers.

Het resultaat is een zeer efficiënt productieproces, maar de manier van integratie van diverse componenten beperkt de flexibiliteit. Dit model wordt vaak toegepast, de vereisten qua bandbreedte en opslagcapaciteit vereisten blijven beperkt.

C. Werkgroep model

Een derde variant is een productieomgeving bestaande uit verschillende geïsoleerde omgevingen, frequent geïmplementeerd om aan de verschillende business vereisten tegemoet te komen zoals functionaliteit of formaat. Soms ontstaat dit eiland model (of werkgroep model) ten gevolge van historische redenen.

81

De eigenschappen van elk productie-eiland kan men vergelijken met het vorige model. Het gebrek aan de mogelijkheid om media te delen en om capaciteit van componenten te delen resulteert in een beperkte uitwisseling van media en een lage gebruiksefficiëntie van de componenten.

D. Centraal MAM model

Het vierde model, Centraal Media Asset Management model, combineert de voordelen van het vorige model in één omgeving door gebruik te maken van een central media archief. Dit is één van de meest geavanceerde modellen die men vandaag terugvindt, en het combineert technologie van verschillende leveranciers.

De vereisten aan de bandbreedte en opslagcapaciteit zijn hoog, bijgevolg is een precieze netwerktopologie en opslagtopologie noodzakelijk. Deze architectuur is nog niet erf verspreid maar blijft een beloftevol model.

E. Het SOA geïntegreerd model

Het vijfde model, het Service Oriented Archtecture geïntegreerd model, integreert de productieomgeving met de business applicaties en voorziet ook elementaire componenten die gedeeld worden tussen productieomgevingen.

82

De implementatie van dit model vergt twee bijzondere aandachtspunten: Ten eerste de transfer van media impliceert transfer van grote datavolumes, en om dit te realiseren met een enterprise service bus moet intelligentie toegevoegd worden op het niveau van de bus. Ten tweede zijn de formaten die door de diensten gekoppeld aan de enterprise service bus vereist zijn niet steeds compatibel en moet intelligentie toegevoegd worden op het niveau van de bus om deze incompatibiliteit op te lossen.

De kostprijs van dit model is hoog, temeer omdat de enterprise service bus een bijkomende investering is. Het grote voordeel komt van de flexibiliteit van het proces en van de mogelijkheid om de efficiëntie van elke gekoppelde component te meten.

Dit model is momenteel in volle ontwikkeling en is vooral interessant voor omgevingen waar flexibiliteit belangrijk is.

F. Conclusies procesmodellen

Bestandsgebaseerde productiemodellen verbeteren de kwaliteit en de efficiënte en verlagen de kosten in vergelijking met tape gebaseerde productiemodellen door gebruik te maken van automatisatie en integratie van de verschillende componenten. De flexibiliteit wordt in de meeste gevallen negatief beïnvloed door de automatisatie, in het bijzonder de creatieve mogelijkheden die voortdurend gebruikt worden bij de productie van verschillende formaten. Het gebruik van een SOA architectuur kan de flexibiliteit verhogen.

83

Characteristic Trad

ition

al

wor

kflo

w

Wor

k ce

ntre

w

orkf

low

Dig

itize

d w

orkf

low

Wor

kgro

up m

odel

Cen

tral m

edia

ass

et m

anag

emen

t SO

A in

tegr

ated

en

terp

rise

Number of standards used Low High High High High High

Flexibility supports creativity High High Low Low Low Average

Compatibility between components different vendors

High Low Low Low Low Average

Data redundancy Average Low High High High High

‘private’ collections Yes Yes No No No No

Non-linear editing No Yes Yes Yes Yes Yes

Browse proxy editing No No Yes Yes Yes Yes

Expensive tools Yes No No No No No

Parallel processing No Yes Yes Yes Yes Yes

Physical moving of assets Yes No No No No No

Cost of tape High Low Low Low Low Low

Retrieval of content Slow Average Fast Fast Fast Fast

Retrieval of content from archive Slow Slow Fast Fast Fast Fast

Deterioration of tape Yes No No No No No

Lead time High Low Medium Medium High High

Cost – CAPEX High Low Medium Medium High High

Cost – OPEX High High Medium Medium Low Low

Workflow flexibility High High Medium Medium Medium High

Component choice flexibility High High Medium Medium Medium Medium

VI. Toekomstige evolutie van productiemodellen

De productiemodellen binnen de media-industrie worden verder beïnvloed door de trends binnen de IT industrie. Twee potentiële trends zullen naar verwachting doorbreken. Een eerste trend is de verdere integratie van SOA architectuur binnen de lagere niveaus van de productie. Een tweede trend is de evolutie naar technieken afkomstig uit IT cloud computing.

De huidige integratie binnen de productieomgeving gebeurt door middel van proprietaire interfaces. Verdere standaardisatie zal het mogelijk maken om de integratie los te koppelen en de kleinere componenten te gebruiken in een SOA infrastructuur. De integratie van elementaire diensten in een business proces door middel van SOA is een flexibele en meetbare integratiemanier.

Meer en meer business functies worden geïmplementeerd als een dienst via het internet met standaard interfaces. Dienstverleners op het internet bieden gespecialiseerde diensten aan die gedeeld worden met andere gebruikers op éénzelfde platform. Dit soort diensten wordt open cloud diensten genoemd. Diverse diensten op het internet worden reeds courant gebruikt door consumenten en kleine bedrijven. Grote bedrijven en instellingen integreren deze diensten voor niet kritische business functies. De integratie van open cloud diensten is een trend die men kan verwachten voor bijvoorbeeld formaatomzettingen, indexatie … etc.

84

VII. Eindconclusies

De productie processen in de media industrie veranderen drastisch door digitaal te werken en door het gebruik van IT-gebaseerde technologie.

Uit de analyse van de processtappen blijkt dat zes stappen cruciaal worden beïnvloed, namelijk acquisitie, bewerken van media, post-productie, transcoding, opslag / archief, distributie

Voor het bewerken van digitale media worden verschillende formaten gebruikt voor audio- en video-essence, voor de metadata en voor het bundelen van essence en metadata. In vergelijking met de traditionele tape gebaseerde productieprocessen zijn meer combinaties mogelijk bij het werken op basis van bestanden. De keuze van het essence formaat heeft een drastische invloed op de benodigde bandbreedte tussen de verschillende systemen en op de opslagcapaciteit.

De migratie van SD naar HD video heft ook een belangrijke impact op de vereisten van bandbreedte- en opslagcapaciteit. Een upgrade naar de 720p50, 1080i50 of 1080p25 formaten vergen vijf keer meer capaciteit dan een SD formaat, en zelfs tien keer meer als men het 1080p50 formaat gebruikt. Het gebruik van hogere compressie beperkt de bandbreedte- en opslagcapaciteit, maar in de praktijk bepaalt het type product de keuze van het formaat.

Het resultaat van de analyse van de verschillende modellen die gebruikt worden voor productie op basis van bestanden zijn vijf verschillende modellen: het enkelvoudige werkcentrum model, digital workflow model, werkgroep model, centraal MAM model en het Service Oriented Architecture geïntegreerd model.

In vergelijking met tape gebaseerde productiemodellen verbeteren bestands-gebaseerde productiemodellen de kwaliteit en de efficiënte en verlagen de kosten door gebruik te maken van automatisatie en integratie van de verschillende componenten. Een nadeel van de automatisatie is dat de flexibiliteit negatief wordt beïnvloed, hierdoor ontstaat een belemmering in de creatieve mogelijkheden die voortdurend gebruikt worden bij de productie van verschillende producten. Het gebruik van een SOA architectuur kan de flexibiliteit verhogen.

De huidige integratie binnen de productieomgeving gebeurt door middel van proprietaire interfaces. Verdere standaardisatie van formaten zal het mogelijk maken om de integratie losser te koppelen en kleinere componenten te integreren in productieprocessen door middel van een SOA infrastructuur. De integratie van elementaire diensten in een business proces door middel van SOA is een flexibele en meetbare integratiemanier.

De integratie van open cloud diensten is een trend die men kan verwachten voor bijvoorbeeld formaatomzettingen, indexatie … etc.

85