24P and Digital Film by Dave Bancroft

*

When “DTV” started in North America in 1998, things became a lot more complicated. North American DTV allows: •

Film is global. Television is local. A filmmaker can distribute his work on film confident that it can be projected anywhere on earth. A television director, on the other hand, has to think very carefully at the beginning about his choice of medium. If he shoots on video, he is almost certainly confining distribution to the country of origin (and its former colonies). If he wants to be sure his work can play everywhere, he will shoot it on film, and even though he transfers it to video for his home market, he will certainly keep the negative for the international distribution deals made in the original coproduction agreement. How did this situation arise? First, film arrived at its own standards through purely mechanical considerations, so its almost universal capture rate of 24 frames per second derives from the limits at which early cameras could be hand-cranked and from the limits on reliably imparting intermittent motion to the film and holding it steady in a gate for each exposure. Television, in contrast, was an inherently electrical medium from the start, so it was likely that some parameter or other would be related to the local AC mains frequency. Thus was created the great transatlantic 50Hz/60Hz field rate divide. There was, too, a cultural contribution. Making a film was always such an indirect process - always waiting for developing - that it seemed quite natural having gone to such trouble that the finished product should enjoy some longevity. So film was made for the long term. That in turn, led to it being made for wide consumption, especially for the output of Hollywood. Television, on the other hand, was seen as an eyewitness that could eavesdrop on events, which the public could share. With the near impossibility of recording it directly for the first twenty years of its existence, or of disseminating it globally for the first thirty years, television was essentially a live experience for a local audience. Standards, or the lack of harmonisation of them, did not matter that much. The film world was a consistent 35mm, 24 frames/second, while the television world had split itself into permutations of two frame rates and line counts, and three colour standards. In fairness to television, the receiver had to operate under domestic ambient light conditions, not in a controlled, darkened auditorium, and had to convey rapid “unscripted” action such as sports events. In short, it needed a higher temporal frequency than 24 frames/sec anyway, to give adequate motion portrayal and reasonably flicker-free displays in bright rooms. So the convention emerged: shoot on video for home consumption here and now, but use film for the world and for posterity.

• • • •

three picture rates (24, 30 and 60 complete pictures per second), progressive or interlaced scanning of those pictures, “exact” picture rates or the above frequencies slowed down by 0.1% (for NTSC compatibility), four image formats (1920 x 1080, 1280 x 720, 704 x 483 and 640 x 480 active pixels) two aspect ratios (4:3 and 16:9)

Not all combinations of these are permitted as inputs to the ATSC compressor – only about 36 are! When this matrix of formats was first announced in 1993, it was expected that the transmission system would adapt flexibly to whichever format the original material had been produced in, so programme makers could just “do their own thing” and send it to a network for transmission. Instead, things evolved to the cold realisation that a given network would of necessity adopt a so-called “plant native format,” running the master control area on just one standard, anything more complex than this being a nightmare beyond contemplation. And, of course, each network chose its own, different “plant native format,” CBS and NBC plumping for a 59.94 Hz 1080-line interlaced version, ABC choosing 720 lines at 59.94 Hz progressive and Fox just doubling up classic NTSC to 60 progressive frames per second. (This is a lot of discussion about America, and we are in Europe, so why should this matter? Please bear with me, and all will be revealed!)

Making Programmes for DTV Perhaps you can guess what the first conclusion was for the people making programmes for DTV. Yes, we’ll have to shoot most things on film! It’s a truly universal format that gets us into any of those multiple transmission formats (plus the 50 Hz standards for the rest of the world). Who said film was dead! However, although film is universal while it’s still film, we don’t do post-production of film within film (optical effects and all that) if it’s primarily destined for television – we use telecine and do post-production in the video domain, so we can enjoy real time effects, editing and signal distribution. But the moment we go from film to video, we lose our universal 24 frames/second and are stuck with one of those specific video transmission formats. For an independent facilities house in Hollywood that wants to be able to deliver to any DTV network, the conclusion is grim: † build a complete “1080i” post-production facility for CBS and NBC, a replica in “720P” for ABC, and yet another replica in “480P” for Fox (and don’t forget international distribution!). Financially, this just doesn’t fly.

The Stupidly Simple Solution HDTV and DTV Stir the Pot When NHK started their pioneering work in 1964 on HDTV, they had the theoretical opportunity to start afresh with television and possibly make a single world standard. This did not happen for a number of reasons, one of which was the hope of achieving interoperability between the new standard and existing television during the inevitable transition period. Thus the old frame rate divisions were perpetuated, with Europe unable to accept 60Hz and the United States and Japan unable to accept 50 Hz.

*

“DTV” in North America currently means digital terrestrial television that has the capacity to be HDTV if the broadcaster so chooses. † This notation abbreviates the description to just the number of active scan lines and whether interlaced or progressive.

35 mm 24 fps film universal format in ACQUISITION

Incompatible, specific, video formats in POSTPRODUCTION

Incompatible, specific, video formats in DISTRIBUTION

The Role of 24P What are we really doing when we use 24P in this way? We have at last learned to represent the image source, rather than the transmitter input. We’ve made a “universal mastering format.” We should be able to take that 24P signal and use it anywhere in place of the original 24 frames/sec film.

Figure 1 (a): Conventional Transfer and Production Figure 1(a) shows a line of demarcation. To the left, we have universal 24 frame film. To the right, we branch out into the “localised” world of multiple video formats. That’s because we’ve always thought that we had to produce and post-produce video in the same format as its transmission target. What if we could move that line to the right, so that the fragmentation into differing transmission standards could be deferred until production was completed, yet with conversion from film to video still at the same point?

35 mm 24 fps film universal format in ACQUISITION

Incompatible, specific, video formats restricted to DISTRIBUTION

universal format extended into POSTPRODUCTION

24P 24P Universal Universal Mastering MasteringFormat Format

Simple Conversions

Figure 1 (b): Universal Transfer and Production Fig. 1 (b) shows that all we have to do is scan each frame of film into one frame of video. It’s called 24 frame video! No speed up by 4% to make 50 Hz video, no interlacing with 3-2 pulldown sequences to make 60/59.94 Hz video. We can do all these things later, from the same, single, 24 frame video copy that emerges from our single post-production process. One process and one set of equipment instead of four. The complexity is confined to the conversions at the end, and I’ll show you later that even there, nothing new has to be invented. The full title of this new format would be something like: “1920 x 1080 video in an 1125 raster, captured at 24 frames per second, with progressive scanning, compliant with SMPTE Standards 274M and 292M (1998 revisions).” You and I can just call it 24P.

Philips Spirit DataCine Film Scanner

In fact, it’s not a new thing to be doing that, and 24P is not the only way. Back in 1993, Philips Digital Video Systems (known then as BTS) started responding to a request from the Hollywood studios for a better electronic format to represent film. The first fruit of this work appeared in 1996 – the Spirit DataCine film scanner that has since become world renowned. But why was it called a DataCine? To answer this, let’s take the principle – that we are concentrating on representing the information in the film frame and not yet thinking about the transmitter – a little further. Imagine we are measuring the density of the three film dye layers with a real time RGB densitometer, and recording each frame’s numbers in a computer file. This is a universal “data” representation of film rather than a video representation, hence the DataCine name. The Spirit maintains this data representation internally on its ultra-high-bandwidth digital processing buses and makes it available on a HIPPI port in DPX format for export to other devices. But Spirit also converts the data on the fly into real time SDTV and HDTV video (all world standards supported). Bottom line: the video outputs are all real time; the export data output is about one-quarter speed. Back in the Hollywood studios and facilities once again (Autumn 1998) the next message was: real time is important for longform work, but we like the universal format idea; we do need lots of resolution, because we’re mastering for American DTV now, so let’s just do it with the existing 60 Hz interlaced HDTV studio format but run it at 24 Hz progressive instead. 24P is therefore a substitute for a literal data format in fulfilling this role; it’s still modelling the film frames on a one-to-one basis. It’s got certain limitations compared to data (another article!) but they don’t have a significant impact when the final product is television and not film.

Philips Support for 24P in the Spirit DataCine

It took just a few milliseconds for Philips to decide to add support for 24P to their existing support for DPX data. To output 24P instead of 60 Hz interlaced HDTV, all that is needed is to “refrain” from the normal telecine practices of adding interlace and pulldown in the output stages and, lo and behold – we have 24P! So in 1999 Philips launched 24P output in Spirit, to add to the existing 1080/60i and 720/60P HDTV options and the DPX data export option. For existing users, every Spirit ever delivered can be retrofitted for 24P.

The Voodoo Multi-Frame Rate HD Recorder In 1993, a new HDTV recording format appeared that became standardised by SMPTE as D-6. This remains the only recording format providing uncompressed, full-bandwidth HDTV recording with a tape cassette. The 1993 D-6 machine from Philips supported the HDTV transmission formats of the time: 1125/60 and 1250/50.

Support in Other Film Imaging Products The Spirit DataCine Film Scanner is the front end member of a family of products designed to maintain film’s universal format qualities through sophisticated post-production until distribution.

Philips Voodoo Media Recorder This year, with the considerable interest in North America in 24P, Philips has relaunched this machine in a multi-frame rate version, the DCR-6024 Voodoo Media Recorder. The Voodoo retains support for the 1125/60 transmission format, but the 1250 mode was commandeered to support 24P.

Philips Specter Virtual DataCine A companion product is the Specter Virtual DataCine. The essence of this product is threefold: first, a file of data is exported from a Spirit that represents all the useful film image information, as previously discussed. This file is stored in a computer server’s random access disk array. Second, we provide a real time playback path from the disk array through the server and out to the world, with conversion on the fly into the video format of our choice. Third, we provide full emulation of the Spirit physical DataCine by including exactly the same output processing electronics in the real time playback path, giving us the same pan/scan, zoom, rotation, colour correction and other functions. To complete the emulation, the colorist controls Specter from the same telecine controller that he would use for the physical DataCine. We call Specter a Virtual DataCine because you get all this without the film actually being present. Oh, there’s a couple of differences – with disks, the shuttle is better and all your shots are accessible without changing reels.

The machine supports other vertical rates; in the USA there is always the minus 0.1% variation needed for compatibility when downconverting to NTSC. But we also felt that since the universal post-production format idea was such a good one, the 50 Hz world should enjoy it too, so the machine also offers “25P” (25 frame progressive). The final finesse is “crossplay.” If you think about it, how do we handle frame rate changes in telecine transfers? If we’re transferring 24 frame film for 50 Hz television, we just run the telecine 4% fast. We might have to correct the audio pitch, but it’s just a simple change of capstan speed. Using 24P to represent the original film requires deferring this speed-up to the VTR playback stage, and that’s exactly what the Voodoo does; it allows us to play back a 24P recording at 25P.

23.98P/ 23.98sF 24P/ 24sF

Since the Specter uses Spirit’s output processing hardware, exactly the same output options are available, including 24P. I referred to film image information coming into Specter from Spirit as data, not as video. There will be provisions for converting 24P video from Spirit into data for Specter in the near future, but, again, that’s another discussion.

25p 50i

60i

60i

Figure 2: Voodoo Crossplay Modes

Re-purposing Voodoo’s 1250/50i format to record 24P and 25P also freed up some capacity for increasing the luminance component dynamic range from 8 bits to 10 bits. This improves the pre-correction handling of film-originated images with wider than normal dynamic range due to large exposure variations.

Conversion from 24P to Transmission Formats 50 Hz distribution (25 Hz playback) 1920

24P/25P D-6 Master

interlac e

1920 1080i 50 fields

down-res 1080P

720 576P

add interlace

60 Hz distribution (24 Hz playback)

Standards Conversions from 24P/25P Master

interlac e

interlac e

ARC

720 576i

add 3-2 field pulldown

inter lace

1920 1080i 60 fields

add 3-2 frame pulldown

down-res

1280

1280 720P

720P 60 frames

add 3-2 frame pulldown

down-res 720

720 480P

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

540 odd lines

9 units

1 pair of segmented frames = 1 original progressive frame

Figure 4 (b): 24sF Format

add interlace 1920 1080P (@ 25 Hz)

16 units 1920 active pixels 540 even lines

In fact, Fig. 2 shows that “crossplay” between any two vertical rates is possible, with the added touch that “interlacing” and “deinterlacing”∗ can be thrown in easily too (with the exception of 60i, which is a transmission, not production format in this context).

480P 60 frames

Figure 3: Output Standards Conversions

A 24sF signal has to have been derived from a 24P (progressive scan) signal. It cannot come from an interlaced signal (e.g. conventional video camera). In a frame captured progressively, all the scan lines represent the same motion phase, while in a frame captured as two successive interlaced fields, half the lines represent one motion phase and the other half represent a motion phase occurring one field time later. The important point is that a progressive frame can be reordered into two fields, recorded and sent over interfaces as two fields, but can be “glued back together” again when it’s needed once more in its original progressive form (for display, recording back to film, MPEG encoding, etc.). The reordering back and forth is completely transparent, which is not the case with a true interlaced signal. This distinction about the origin of the signal is marked by the use of the term “segments” as opposed to “fields,” in the 24sF signal.

Fig 3 shows that only three kinds of conversion process, in differing combinations, are needed to create any world TV standard from 24P. None of these require any new techniques or cause unfamiliar artefacts, because they replicate what telecines do playing the original 24 frame film. The processes have simply been deferred until post-production is finished.

[There is another subtle difference in 24sF with regard to vertical filtering which is beyond the scope of this article.]

Segmentation

The other reason is that some equipment designed for interlaced signals can handle 24sF with less modification than for 24P. This is especially true for VTRs.

You may have heard of a format called “24sF” or some similar name. It’s a variant of 24P. Figs. 4 (a) and (b) show the difference between 24P and 24sF. It’s different from interlace and here’s why. 16 units 1920 active pixels 9 units

1080 active lines

So, what is the purpose of 24sF? One reason is for monitoring: splitting the frame up raises the vertical refresh rate to 48 Hz. Displaying 24 Hz video directly would produce a terrible flicker, but in a fairly dark room, 48 Hz flicker is tolerable (like the cinema projector, with its two-bladed shutter).

Starting with a film orientation, Philips had no difficulty supporting 24P, but decided for the sake of users’ choice to support both 24P and 24sF. It’s important to remember that if all other factors are equal, they will give identical results in the final output.

24P is Pragmatic 24P came into being to solve a problem: how to produce programmes for multi-format DTV economically via a common mastering format that models the universal format nature of film. 24 progressive frames per second (one per transferred film frame)

Figure 4 (a): 24P Format

∗

This is not the tricky deinterlacing performed in sophisticated standards converters - see “Segmentation.”

A pure data representation does this best of all, and remains the method of choice for cinema rather than television productions, but by making some reasonable compromises, 24P allows some savings and some speed-ups. First, the format is a derivative of existing HDTV standards, so HDTV equipment designs can be modified with very little trouble. Second, because it’s video, it is inherently real-time. Third, also because it is video, it requires a less steep learning curve for operators than data.

24P and Europe I have said 24P is a solution for producing programmes for American DTV/HDTV. We don’t have HDTV in Europe, and we don’t seem to want it anytime soon, so why should we be interested in 24P? One reason is internationally co-produced and distributed programming. For example, the UK is known for period drama. We can assume that the productions will continue to be shot on film (whether 35mm or super 16mm) for some time to come, but the edited video transfer in 625/50 really never did convert all that well to 525/60 and it certainly won’t be adequate for US HDTV. Making a conformed cut negative version of the production for export is an expensive addition to the home market production costs, so it is time to bring out the calculators and see whether it might not be more economical to transfer the production to 24P right from the start. The single 24P master will, after all, yield a beautiful home market version as well as all the export derivatives, while the negative doesn’t get touched once transferred. European facilities equipped with 24P will also be able to participate in American DTV productions when they have some particular in-house skill that is known and appreciated.

After all is said, 24P is as universal as film.

Originally published in “Broadcast HARDWARE International” Number 81 – December 1999