Digital High-Definition Intermediates - the Significance of 24P David J. Bancroft Philips Digital Video Systems United Kingdom [email protected]

Abstract A new high-resolution video format has been proposed which claims to offer a singleformat solution to the problem of performing post-production operations for digital television (DTV) economically, when different DTV distribution outlets will require the finished content in not one, but in several distinctly different formats. The format takes its inspiration from the comparatively universal 24 frame format of motion picture film by modelling film in a video domain. This paper offers a short description of the new format, an illustration of its use in typical operations, including conversions to desired destination formats, international issues and some limitations. Comparisons are made with an alternative method of universal image representation and an assessment is made of the overall value of the proposal to the DTV production community. The viewpoint is directed primarily towards the North American television situation, but it is hoped that the principles will remain valid as other parts of the world become more involved in high-resolution DTV. INTRODUCTION Ever since its introduction over sixty years ago, television has eschewed the use of the universal 24 frames/second rate of motion picture film. It has not even used a consistent, albeit different, frame rate around the world. When productions made on film are to be distributed on television, this has led to the requirement to make multiple copies of film transfers, with associated burdens of extra cost and complexity.

The situation has been particularly problematic in 60Hz television countries such as the United States, where the film versus television frame rate disparity is greatest, and complex calculations are required when going back and forth between the two domains at the picture editing and audio synchronization stages. Recently, the situation has shown signs of getting worse, with the introduction in the United States of digital television (DTV) employing not a single video format, but eighteen, including some permitted variations in frame rate, as well as other major parameters, such as spatial resolution. In addition, DTV requires a widescreen aspect ratio as well as the 4:3 ratio of traditional television. When this new factor was added to the situation already facing major producers of filmed entertainment for global consumption, it prompted several interest groups—in the Hollywood community in particular—to reconsider the whole process of mastering and post-production for both DTV and existing television. A proposal has therefore resulted that the advantageously universal, 24 frames/second rate of motion picture film should be modeled in video, so as to create a single intermediate post-production video standard. Just as it has been possible to make all world television distribution formats from a single film intermediate, so it would be possible, according to the new proposal, to make the same formats from a video intermediate that also used a 24 frames/second picture rate. It

should be stated that the idea of an electronic intermediate is not new; some very effective data representations exist, although as yet these are not yet available in real time versions and require a fairly specialised computer workstation environment for their execution. The new video intermediate proposal reflects the urgent need to get facilities up and running for large-volume, short editing time productions, such as episodic programmes, where time and budget constraints favour a real-time approach, using as far as possible techniques from regular television that will be relatively familiar to production crews. With the added proviso that this new intermediate must convey enough of the original film's information to satisfy the display requirements of the high definition formats (HDTV) within DTV, this new format must also in itself be high definition. Using the latest television engineering techniques, therefore, the term “digital high definition intermediate” is coined to characterise the new format. The format uses a spatial resolution of 1920 pixels horizontally by 1080 active lines vertically, scanned progressively at 24 whole frames per second. It is becoming colloquially referred to as “24P” or as “1080/24P.” ORIGINS OF THE NEW FORMAT Although the recent impetus for this format came from the Hollywood production community1, its origins go back somewhat further. In 1993, there was a proposal from the major Hollywood studios2 for a universal high resolution electronic format for feature film mastering that should avoid existing practices such as 3-2 pulldown addition. In 1994, there was a proposal to add film-based frame rates to the already-published SMPTE 274M standard for high definition studio video. This proposal was in fact adopted by the SMPTE and in 1998 became a revised version of this standard3. A related SMPTE standard, SMPTE 292M, for the carriage of high-definition video over serial links in the studio, was also revised to accommodate the additional frame rates4. 2

Thus the seeds had already been sown for the implementation of the new format. FORMAT DESCRIPTION (Figure 1) The full description is in reference (3), but the salient features are: Scanning: 1920 pixels horizontal by 1080 lines vertical, progressively scanned Analog display raster: 1125 lines, normally with frame rate upconversion (reference 3). Frame rate: 24 whole (progressive) frames per second, with permitted variations of 23.97 frames/second (corresponding to 60.00/59.94 ratio in distribution) and 25 frames/second for international distribution. Digital sampling/interface clock rate: 74.25 MHz (same as other frame rate options within SMPTE 274M and also same as clock rate used in 1280 x 720 progressive scan video (SMPTE 296M)). Serial digital interface rate: 1.485 Gbits/sec. Aspect ratio: 16:9 Colour space: Y'Cr'Cb' (4:2:2 colour subsampling), with ITU-R Rec. 7095 reference primaries and component matrix equations (different from 601 video). Transfer characteristic: ITU-R Rec. 7096, i.e. conventional video “CRT gamma” type characteristic, rather than, say, linear or logarithmic film print density representation. An important criterion that went into the design of the new format was the desire for the maximum possibility of “multi-purposing” equipment, because at high definition

resolutions, production and post-production equipment items are inevitably more expensive than their standard definition counterparts. Thus, for example, video tape recorders, telecines, routing switchers, production mixers and all parallel and serial interface distribution equipment designed for network standards based on 60Hz operation will operate with relatively small modifications at the new 24 Hz rate (and in fact, several manufacturers have already announced support). This important consideration led to the decisions in the 24P format about the number of pixels/lines, colour space, and especially the digital clock rates. Another important point is that because it is anticipated that at least some consumers of DTV will be able to view the end product on a progressively scanned display, the full vertical resolution of the source material should be preserved. This point is made because it has become implicit in television that the display will always be interlaced and that therefore the additional vertical filtering required should be embedded in the signal at its point of creation, i.e. at the sensor, in a telecine or studio camera. It is going to be a requirement that if the full potential of the new format, and its role in better quality television through DTV, is going to be realised, this practice must now be discontinued. Instead, the additional filtering should be regarded as part of any subsequent downconversion process into interlaced video—it should definitely not be embedded into this “digital high definition intermediate.” Another reason is that preserving the full vertical resolution will make the intermediate usable for reconforming to “film finish” for theatrical distribution, as well as to “television finish.” Variation on Format - 24sF (Figure 2) A variation on the new 24P format has been proposed by some users and some manufacturers. This is the so-called “24sF” format7, where “sF” means “segmented frame.” 24sF is different enough from 24P to offer some advantages in equipment in certain 3

situations, yet similar enough that it will in principle achieve the same objectives as 24P. The equipment advantages are that, for example, some video tape recorders designed for 60 fields interlaced operation can be more easily converted to support the new format, if it is represented in a form that looks like a familiar interlaced signal to the recorder (the recorder will see it as 48 Hz interlaced, which, for its electronics, is not significantly different from 50Hz, an established VTR format). Similarly, a 50Hz high definition monitor will lock with little difficulty to such a 48 Hz signal, solving one of the display issues with 24P. However, there is an aliasing issue to be considered (see below). This similarity between 24sF and interlace has led to some confusion. 24sF represents what would otherwise be a 24P signal by rearranging the scanned lines from each frame into a pair of divided parts of the frame. These parts are called “segments” to emphasise that they are not fields, even though they may be handled as fields by certain types of equipment. The time represented by the pair of segments equals the time of the original frame (1/24 of a second) and the two segments can be joined back together at any time to reconstruct the original progressive frame. They are not “fields” for two important reasons: i) the lines in the second segment of the pair were captured at the same moment in time as the lines in the first segment, because the original frame came either from a film camera or an electronic cinematography camera (which would also be 24 frames progressive capture); this distinguishes the lines and their “motion phase” from those produced by an interlaced studio video camera ii) the additional vertical filtering performed in true interlace studio video cameras to avoid small area flicker (also known as “twitter”) in interlaced displays will not have been applied to the 24sF

signal; thus full vertical resolution will be preserved, as it is in 24P; however, if displayed directly on an interlaced monitor running at 48Hz, 24sF signals will exhibit significant aliasing, limiting this display mode to continuity and other, non-qualitycontrol, functions. If anti-aliased interlaced monitoring is required, the 24sF signal has to be reordered back to 24P, have the necessary filtering applied, and then be interlaced (this time in the conventional sense of the term). It is expected that, where they wish to, manufacturers of 24P equipment will be able to add support for the 24sF variation without any particular difficulty. However, users should consider carefully the number and costs of conversions between the two formats if they are contemplating working with both of them within the same production facility. OPERATIONAL USE OF 24P The following description makes some assumptions:

4

•

Fundamental descriptions applicable to conventional production are mostly avoided; instead differences arising from use of the new format are emphasized.

•

Productions are being made within a “60Hz” television environment (e.g. U.S.).

•

Initial acquisition is with conventional 24 frames/second motion picture film.

•

Film camera framing and composition will be for television finish only, and remaining compatible with any subsequent theatrical release will not be a consideration. Shooting for the two aspect ratios now needed for DTV and conventional television will not be examined within the scope of this paper; instead the simpler case of 16:9-only production will be assumed.

Monitoring Just like 24 frame film, 24 frame video cannot be displayed directly—the flicker would be totally unacceptable, even in the dimmest of ambient light and the lowest usable display intensity. Some form of modification to the 24P signal—but only in the monitoring feed, not in the main signal path—is therefore necessary. Just as with 24frame film, one solution is the electronic equivalent of the twin-bladed projector shutter—each frame is flashed twice in succession on the display, taking the flicker rate up to 48 Hz. However, even with this doubling of the refresh rate, a television display will require at least the same restrictions in display brightness and viewing environment as film. Compared to the conditions accepted as customary in 60 Hz (and even 50 Hz ) television production environments, this may be unacceptable for operational reasons. In fact, the situation is even worse with a television CRT at 48 Hz than with a film projector, because of the instantly decaying nature of the CRT phosphor. A variant on this is the 24sF format described earlier, but with its associated aliasing limitation. Another solution that is discussed is simply to add the familiar 3-2 interlace and pulldown sequence associated with 60 Hz television to the 24P signal, to get the display refresh up to this more comfortable frequency. This may in fact be quite appropriate if the target delivery format out of the process anyway is to be a version for a network running 60 Hz interlaced DTV. However, if the target format is to be a progressive scan version, then the image displayed on the monitor will not be representative of the client's end usage for quality control purposes, although it will be satisfactory for continuity. The ultimate solution is to stay in progressive scan, and upconvert, not to 48 Hz, but to a 72 Hz monitor refresh rate. This may be a design and budget challenge for the

horizontal deflection circuitry of a high definition CRT monitor, however.

needed, because the negative cutter cannot cut in the middle of a film frame.

Another approach may be to exploit the non-decay characteristic of several new, nonCRT, display technologies. Devices such as LCD and digital micro-mirrors can either hold the image at full amplitude for the whole picture period or utilise a much higher refresh rate that is not limited to the vertical rate of the input signal. Either way, the flicker problem is solved. However, there is still some work to be done to assure that the colour reproduction accuracy, consistency and stability of non-CRT devices is appropriate to quality control, even considering that CRTs themselves have limitations.

The necessary controls are the familiar ones of:

Film Transfer (Figure 3)

(i) capturing 24 Hz film edge code, adding to it the various 60Hz time codes associated with the video signal and time code associated with lip sync audio, and embedding all this into “3Line VITC”8, then adding the VITC (or a digital ancillary data equivalent) to the offline video source recordings; (ii) creating “sync lists” (phase relationship of film frame to video 3-2 pulldown sequence) for each film camera exposure start on the original camera negatives; these lists are used in combination with the EDL to produce the negative cut list.

Dailies Transfer and Offline Editing Online Transfer It is expected that older, “non-front-line” telecines will continue to be used for dailies transfer, rather than new machines. This means that by default, video cassettes of dailies will be created using 525/60 video, with interlace and embedded 3-2 pulldown, despite their targeting for a 24P production. It will therefore fall on the offline editing system to be “24-frame aware” and create “source bins” or whatever the terminology is that is used, that have recalculated 24 frame boundaries from within the 60 Hz dailies' video, so that a “24 frame clean” EDL will result from the offline session. An even better way is for the offline editing system actually to convert the 60 Hz video back to 24P video as part of its “digitization” process. This will prevent the risk of “non-film-A-frame” offline edits being made (interlaced video frames whose constituent fields are derived from two different film frames, producing a blurred effect). In any event, the controls that need to be applied are not new: when a production was edited in 60 Hz video but a cut negative was also required, the same process of reestablishing 24 frame boundaries was always

5

A telecine equipped for the 24P format is required, together with an online capture system that may be linear tape based, or nonlinear disk-based. The decision as to which is employed is likely to be based on the form factor (short duration or long duration of acquisition material) and available budgets. New versions of video tape recorders for 24P and 24sF have been announced by major manufacturers and are being launched this year. Consideration should be given to the overall complexity of the post-production process at a user's facility: a complex process involving a large number of generations (e.g. complex compositing) may mandate the use of a high-bandwidth recorder that accommodates the high definition signals without the use of lossy compression, so does not add accumulating generation losses and artifacts; on the other hand, a more simple process may be accommodated within the capabilities of a low-bandwidth (standard-definition-based) recorder equipped with lossy compression and even with prefiltering of the luminance and chrominance response before the signal is compressed.

At the end of the post-production process, however, these constraints do not apply to the same extent; one generation of compression in the delivery format to the network will not have a serious impact. Online transfer will result in a high definition video recording having one video frame (or segment pair) for each original film frame. The material will be either one-light, best-light, or fully scene-by-scene colourcorrected, depending on the capabilities of the telecine and the recording format. However, with long-form work, it is more efficient to color correct only the required footage, after online editing. Online Editing and Conforming Effects, Titling, etc. The difference that should be noted is that if the 24sF format is used, the position of moving transition edges, such as moving mattes, wipe edges and scrolling titles, must be updated at 24 Hz, not 48Hz, otherwise some artifacts could be generated on conversion back to 24P. For example, a moving vertical split wipe edge would end up with a “toothed” effect after conversion. This would not be apparent at the time on a 48 Hz monitor. This requirement applies whether effects are added in a real time production switcher or rendered in a computer workstation. Conforming The original online-transferred 24P material should be treated conceptually as if it were an original or duplicate negative that was to be cut. It will have the same requirement in that the EDL must not request a cut to be made at a point in time corresponding to the middle of a 24P frame. The EDL must therefore have been created in an inherently 24 Hz-aware offline environment or be a “cleaned” EDL to achieve the same result. Additionally, management software must keep track of the occurrences of any such 6

shifts in an offline-generated record in or out point needed to resolve such a “mid-frame” edit point request. This tracking will remember the direction of accumulation of these shifts relative to the targeted total programme time, and toggle the adjustment direction appropriately to keep the accumulated programme length on target. Audio Layback The post-production facility must decide whether the video operation is running at an actual “24P” rate or is in fact running effectively at 23.98 Hz, in comparison to the audio, which will most likely have been synced in acquisition to a 24-frame-based reference. This will determine whether audio sync rate resolving is done at this point, or later, when a converted network delivery copy of the content is created. In the former case (true 24P video operation), resolving is not an issue at this stage and audio layback can proceed as if a conformed negative were the picture reference. In the latter case, audio resolving must be done before or at the layback stage because the video and audio sync clock rates will differ. Since the creation of the 24P format was triggered by DTV, the audio requirements at this stage will be more demanding than hitherto. The surround sound audio formats associated with DTV require at least six audio channels, with preferably another two available for a conventional stereo mix that is separate from the surround mix. This may impact the choice of video tape recorder, since very few have this many channels, so that other types will require the addition of a slavelocked audio recorder. Another DTV issue is that video delays in video production are now much greater than before and there must therefore be a correspondingly greater provision for adjusting audio timing to achieve synchronisation. Provision for this is normally made in several equipment items, but most

notably in the playback side of VTRs, where a large range of adjustment can be provided. Formatting for Commercials and Network Delivery It is common to apply variable speed operation (speed up) to an edited production to allow extra content, such as commercials, to be inserted into the production. Compared to 60 Hz television, it should be noted that the increment of “speed bump” in the 24P domain will now be a 41.7 millisecond frame, instead of a 16.7 millisecond field. If the speed change factor is relatively large, this may mean that there are insufficient “hiding” places for the skipped frames, such as shot and scene changes, and it may be preferable to defer this stage until after conversion to the required network delivery format. CONVERSION TO NETWORK DELIVERY FORMATS (Figure 4, (a), (b) and (c)) It is at this point that the advantages of the 24P format become apparent. Familiar and established conversion “tools” are used to convert from the single 24P format of the post-production master to the required network delivery format. These tools perform interlacing, 3-2 pulldown addition, frame doubling and spatial resolution downconversion. They are used in different combinations according to the output format. None of them introduce any new picture artifacts not already accepted by end users.

approximately 4% fast. In other words, 24 frames/second film is transferred at 25 frames a second. The accompanying audio pitch shift is compensated with a pitch corrector. Exactly the same principle can be applied to the new 24P “Digital High Definition Intermediate” since, as stated earlier, it is designed to model the film in the video domain. Instead of the telecine being requested to run 4% fast, it is now the VTR or disk recorder that does the same thing (Figure 4 (d)). This kind of operation on the VTR is referred to as a “crossplay” mode, since the playback speed is deliberately made different from the record speed (the same principle is used, more subtly, in 24.00 to 23.98, or vice versa, crossplay). Crossplay capability in the VTR is therefore important for facilities engaged in international distribution of their productions. Audio Those VTRs so far announced by manufacturers will reproduce digital audio tracks involved in the crossplay operation with the same speed change factor as the video; in other words, the audio sample frequency, normally 48 kHz, will be translated to a different frequency. Although some VTRs are being equipped with on-board sample rate converters, users may prefer to optimise this process with a dedicated outboard converter to restore the rate to 48 kHz.

LIMITATIONS OF THE 24P FORMAT INTERNATIONAL CONSIDERATIONS Video It is established practice when making international video versions of film originals, that the frame rate difference between the video and the film, being smaller (25 Hz to 24 Hz, versus 30 Hz to 24Hz), can be accommodated without the complexity of video field pulldown addition, by the simpler expedient of running the telecine 7

This discussion makes some comparisons between the 24P video representation of filmbased images with a more generic “data” representation. When representing film in an electronic form, it is necessary to consider the amount and type of information in the original film images, and the amount of that information that must be retained for the intended end users of the post-produced content. If the

representation is to be a single format generic to the needs of all possible end users, then the quality contained therein must be sufficient for whichever user or application is the most demanding. Disregarding standard definition television, these quality-defining end-user applications now divide into two types: - those needing high-definition television quality (“DTV television finish”), - those needing restoration to the original film quality, for large-screen cinema or electronic cinema projection (“film finish”). The 24P format has been designed primarily for “DTV television finish,” for the following reasons (not an exhaustive list): i. It is aspect-ratio specific; any image transferred into its inherent 16:9 aspect ratio that is not already composed for this ratio will have to be cropped, letterboxed, “pan/scanned” or “tilt/scanned,” with losses in either horizontal resolution, vertical resolution or scene content. This will not occur if 24P is used to convey film that has been shot specifically for television finish. For film shot for theatrical release, a generic data representation9 will convey any non-television film aspect ratio (e.g. 1.66, 2.39) with complete transparency through to film recording after post-production. ii. Colour information is sub-sampled on a 4:2:2 basis. This can be a limitation in colour correction operations, since the colorist may wish to store a single wide-range untimed version of his source material from which multiple different colour balance versions can be conformed. Less precision (and therefore less time) might be needed in a one-light or best-light transfer if an RGB (4:4:4) representation were available. For film finish, the colour resolution reduction might become visible in the release print under certain adverse conditions.

iii. The transfer characteristic of 24P inherits the “CRT gamma” curve of conventional television. The function of gamma correction in television is not only to pre-correct for the CRT10; it is also a form of analog compression that is vital, because it allows a subjectively acceptable range of intensity values to be conveyed through a transmission and storage system with a smaller bit depth—the same range can be conveyed in 10 bits, CRT gamma corrected, that would need at least 12 bits in a linear transfer characteristic. 24P has to use the CRT curve for compatibility with related television formats. However, other transfer laws are available that offer even more intensity range coding efficiency. For example, a logarithmic law, such as “log print density,” will give the equivalent of at least 14 bits linear bit depth in the same non-linear 10 bits. The advantage of such a wider intensity range in the electronic format is that compensation for film exposure variations can be deferred until after film transfer: a one-light transfer can be done much more quickly (and with less expensive staffing) than a fully corrected transfer; the complex scene-by-scene correction process is then deferred to a more efficient random access stage. Past attempts to achieve the same objective in “tape-to-tape” colour correction have been limited in success by the need to avoid vital film information loss through the limited video dynamic range. In contrast to a video representation, a generic data representation can convey a logarithmic transfer characteristic just as easily as a CRT, linear or any other characteristic. iv. The colour gamut of 24P is again inherited from conventional television, which is in turn designed specifically to be optimal for CRT display (Figure 5). This is expressed in terms of the colour reference primaries, which define the size of the colour triangle in the CIE chromaticity diagram11,12 and therefore the range of colours that can be reproduced. Film does not have specific, definable colour primaries, but it is evident that there are some

8

high saturation colours which cannot be reproduced on a CRT display, but which can be captured on film, restored using correction for unwanted dye absorptions via electronic masking in the telecine process13 and reproduced on projectors using wide band primary light sources in conjunction with optical colour filters. These colours are deliberately sacrificed in conventional television systems so as to maximise the utilisation of the available analogue dynamic range or digital code space for those colours that a CRT can reproduce. This colour gamut limitation will not be a real problem as long as the primary display technology for consumers remains the CRT (and while some other display technologies continue to have their own problems), until displays migrate on a large scale to non-CRT devices with a wider colour gamut. It will be a more immediate problem if the content is required in “film finish” for theatrical projection. Working Around Some of These Limitations It may be that as television settles down to its new widescreen form, because 16:9 is much closer to film composition formats than was 4:3, film directors might actually start to compose generally in 16:9. This would eliminate the fixed aspect ratio problem. For the colour resolution issue, reduced colour space (Y'Cr'Cb'/4:2:2) was originally devised to ease the limitations of transmission channels and tape storage devices, where the difference between 4:2:2 and 4:4:4 has serious cost and complexity impacts. In a restricted post-production environment, on the other hand, disk-based operation in confined “islands” can permit the retention of full R'G'B'/4:4:4 colour space until production is complete, with the “downconversion” to 4:2:2 being deferred to the finishing stage. A number of manufacturers are now adding support for this mode of operation. It is not even an exception to the SMPTE scanning standard, because 274M does include an R'G'B' signal representation as well as 9

Y'Cr'Cb'. For serial interface transmission, digital high definition R'G'B' does not fit within a single SMPTE 292M serial link; however two links can be used in parallel, provided the user has the means to reconsolidate the two serial streams into one at the serial receivers. CONCLUSIONS A new format has been proposed for postproduction of content for DTV distribution. Its parameters have been chosen so that a single version of the format will yield, via acceptable conversions, any of the multiple high definition video formats allowed within the ATSC (US) transmission specification, as well as different frame rate versions for international distribution. In arriving at these qualities, the starting point was to model the desirable universal 24 Hz frame rate and progressive frame capture of motion picture film in the electronic domain, while retaining the maximum compatibility with equipment designed for existing video distribution formats. Other criteria were the need to support real time operation and to facilitate cost-effective and timely implementations in real-world products, so as to get the new format established as quickly as possible. A variant on the format, based on segmenting the electronic frames was also discussed. Some limitations in the proposed “24P” format are recognised, in comparison to alternative formats based on more generic data, rather than video, representations of film-based images. However, it is at the same time recognised that the comparative time scales of the respective technologies currently favour the 24P video solution for system implementations needed on line in 1999.

© Copyright David J. Bancroft 1999

REFERENCES 1

Society of Motion Picture and Television Engineers (SMPTE) Working Group M21.04 (now absorbed into SMPTE Image Technology Committee I23), “Proposed Recommended Practice for High Definition 24 Frame Progressive Post Production Mastering Format,” December 1, 1998. (www.smpte.org) 2 Technology Council of the Motion Picture-Television Industry, Montreux, June 12, 1993 3 SMPTE Standard for Television, 274M-1998, “1920 x 1080 Scanning and Analog and Parallel Digital Interfaces for Multiple Picture Rates,” (revision of SMPTE 274M-1995). Refer to Table 1, Systems 9, 10 and 11 for details of 24P format and its variants. 4 SMPTE Standard for Television, 292M-1998, “BitSerial Digital Interface for High-Definition Television Systems,” (revision of SMPTE 292M-1996). 5 ITU-R Rec. BT.709 “Parameter Values for the HDTV Standards for Production and International Programme Exchange”

The Author

Biography: Dave Bancroft is Business and Technology Development Manager for the Film Imaging Group of Philips Digital Video Systems. His responsibilities include the long-term planning of new production systems for film-originated high-resolution images, and the formulation of standards and operating practices through organisations such as the SMPTE. Prior to Philips, his engineering background included the BBC, RCA and Ampex, working and living in several different continents and countries. He is a member of the Royal Television Society and the SMPTE. Contact: Philips Digital Video Systems Theale Cross, Pincents Kiln, Calcot Reading RG31 7SD, United Kingdom Tel. +44-(0)118-932-5000 Fax. +44-(0)118-932-5050 email: [email protected] 10

6

op.cit. An SMPTE Recommended Practice for the implementation of 24sF, as a variation of 24P, is being developed in SMPTE Image Technology Committee I23. It is intended that the Recommended Practice shall be used in conjunction with SMPTE Standard 274M (ref. 3 above). 8 SMPTE Recommended Practice RP203, “Encoding Film Transfer Information Using Vertical Interval Timecode.” 9 ANSI/SMPTE Standard 268M-1994, “File Format for Digital Moving Picture Exchange (DPX).” 10 Poynton, C.A., “A Technical Introduction to Digital Video,” Wiley, 1996, PP. 94-95. 11 op. cit., P. 135. 12 Publication CIE No 15.2, “Colorimetry,” Second Edition (1986), Central Bureau of the Commission Internationale de L'Éclairage, Vienna, Austria. 13 Hunt, R.W.G., “The Reproduction of Colour,” 5th Ed., Fountain Press, 1995, PP. 177-185. 7

1920 active pixels

9 units

1080 active lines,

16 units

24 progressive frames per second (one per transferred film frame)

Figure 1

24P High Definition Intermediate Format - Basic Parameters

16 units

9 units

Equivalent to 24 progressive frames per second (one per transferred film frame), each divided into two segments Figure 2

11

24sF Format Variation

540 even lines,

540 odd lines,

1920 active pixels

1 pair of segmented frames = 1 original progressive frame

Telecine Telecine Transfer Transfer

Dailies Dailies

Telecine Telecine Transfer Transfer

Digitise Digitisetoto Offline Offlinebins bins

Add Add 3-Line 3-Line VITC VITC

Online Online Video Video

Film Edge Code

Audio Time Code

Add Add 3-Line 3-Line VITC VITC

Video Time Code

Make Make 3-Line 3-Line VITC VITC

Offline Offline Edit Edit

EDL 24P Video Master

Online Online Edit Edit 24-frame Clean EDL (cut the “digital neg”)

Trace Trace Logic Logic

Make Make FLEx FLExFile File Flex File Sync Points

TC TCGen Gen

For comparison:

Cut Cutthe the Negative Negative

Figure 3

interlace

1920 1080P

Cut CutList List

Offline and Online Dailies, Editing and Conforming Processes

add interlace

24P Digital Intermediate

Pull PullList List

add 3-2 field pulldown interla ce

1920 1080i

(a)

60 fields

down-res 1280 720P

add 3-2 frame pulldown interla ce

1280 720P

(b)

60 frames

down-res 720

60 Hz distribution

480P

50 Hz distribution

speed up 4%

12

interla ce

720 480P

(c)

60 frames

1920 1080P (@ 25 Hz)

Figure 4

add 3-2 frame pulldown

add interlace interlace

e.g. CCTV (China), Australia, etc.

1920 1080i

(d)

50 fields

Conversion from 24P Digital Intermediate to Network Delivery Formats

Some additional saturation is available in this zone, at certain L* values, with real film dyes

Figure 5

13

ITU-R Rec. 709 reference primaries shown as R, G and B

CRT-phosphor-based colour gamut

A Partial Comparison of Colour Gamut of Rec 709 based video with the Colour Gamut Available with Film