By Jim Mendrala
December 5, 1998
The new ATSC DTV standard permits the transmission of basically 5 video formats. 3 SDTV formats (704x480, 16:9, non-square pixels, 704x480, 4:3 non-square pixels and 640x480 4:3 square pixels) and 2 HDTV formats (1920x1080 and 1280x720, 16:9, square pixels), along with their respective variations of fields per second (Fps) and/or frames per second (fps). This seems to be creating an impossible scenario for the production and post production industry. In the past, the production and post production industries have delivered their products in either NTSC, PAL, or component at either 30 fps interlaced with 60 Fps, or 25 fps interlaced with 50 Fps.
With the introduction of DTV, the networks and independent television groups are specifying formats listed in the ATSC Table 3. Some want 1920x1080 at 30 fps interlaced (30i), some want 1280x720 at 60 fps progressive (60p), and others want 640x480 at 30i fps, some want 704x480, at 24 fps progressive, and so on. This is creating an impossible financial and operational burden on the production and post production industries.
Note: All reference in this article to 24 fps and 30 fps also applies to 24/1.001 and 30/1.001, except where noted.
A solution to the problem has surfaced and a 24 fps progressive Mastering Format in 1080x1920 has been proposed to simplify the process of converting to the various video standards which the networks and broadcasters worldwide will want and need in as short a time as is possible. It doesn't make sense to transfer a film, for example, in HDTV then change the telecine to output NTSC, then change the telecine again to output PAL, etc. Since the new HDTV and data scan telecines are very expensive, a way to bring everything down to a simpler format is required.
On December 3, 1998, a SMPTE meeting was held at Sony Studios' Tri-Star building to discuss what SMPTE can do to put some sense and order into the matter.
In addition to the 24 fps progressive recommendation, 24 fps progressive format conversions and display devices were also discussed. The proposed documents introduced at the meeting were finished in the shortest possible time ever for a SMPTE working group and feedback from the production and post production communities is drastically needed. SMPTE Engineering Vice President William C. Miller, said "We would like to have several recommended practice documents in place by April of 1999, in time for NAB (National Association of Broadcasters)."
Equipment manufacturers say they will be delivering 24 fps progressive types of equipment after NAB '99, so producers will be able to start turning out high end product by the start of the fall 1999 season.
What has been proposed is:
1. A straight forward 24p fps Mastering Format standard for mastering of television images that accommodates conversion to both progressive and interlaced delivery formats and will includes how timecode, audio and metadata necessary to supplement the video will be handled.Since telecines today all progressively scan film at 24 fps, the image can be output as 1080x1920 at 24 fps progressive. Video tape or disc recorders to do this, while possible, have not been "invented" yet and will not be ready for the fall 1999 season due to product development time.
2. A new 24p fps/48sF standard referred to as 48sF where each frame is progressively scanned but read out as 2 segmented frames (sF). Segmented frame 1 with all odd lines and segmented frame 2 with all even lines derived from the progressively scanned frame. (Note: Small " f " for field in an interlaced system. " sF " for segmented frame in a progressive system.)
The second suggested mastering format has come about because of the need to meet a shortened product development time in order to have product in time for the fall 1999 season. Since a lot of equipment is now dealing with interlace or fields, a modification to convert the operation to a slower field rate would be easy to do.
The desire though, is to have progressively scanned images. Progressively scanned images when run through an MPEG encoder will cut the bit rate down anywhere from 25 to 35% compaired to 30 fps interlaced. With adjacent picture elements, motion vectors are easier to calculate, and since the frame rate is lower, there is more time per frame to perform computations.
So to make the 24 fps progressive more compatible with today's interlaced equipment, a second recommendation has been put forward, namely 48sF. In 48sF, the 24 fps progressive image would be converted to two segmented frames, with the first segmented frame composed of all the even lines of the progressively scanned frame, and the second segment all the odd lines of that same progressively scanned frame. The frames would be captured at 24 fps progressive but would be converted to 48 segmented frames each with half the number of scan lines.
The argument presented by two of the equipment manufacturers is that 24 fps progressive equipment does not exist yet so why not use 48 segmented frames, 48sF, to be able to get product out ASAP, with minor or slight modifications to existing equipment. So it is hoped that SMPTE can provide the common grounds and interface of ideas to come up with a working solution.
In a typical telecine suite, an average one hour movie requires approximately 40 hours of actual telecine machine time. Due to reel changes and color correction considerations, a week is usually dedicated entirely to the production of just one Telecine Transfer Color Master.
During the second week, with more film handling required for the many and varied broadcast formats, pan scan, letterbox, anamorphic or squeeze, NTSC, PAL in 4:3 and 16:9, and HDTV in 16:9, are done for worldwide distribution.
All NTSC, PAL and HDTV distribution masters made from that Telecine Transfer Color Master take anywhere up to 3 to 5 times the length of the source to process.
In order to simplify the process and make it more economical in the post production area, a simpler way has to be found.
1920x1080 24 fps progressive seems to be the format of choice. It is progressive. It is at the same frame rate as the film. It will compress more efficiently than other HDTV formats.
1920x1080 24 fps progressive can easily be up converted to a higher frame rate such as 72p, 60p or 30p. Interlace can be introduced as well as 3:2 pull down. Time code is the same as the film and is non-drop frame.
1920x1080 24 fps progressive can also be played back at 25 fps for those countries that require it. A speed change of 4%.
1920x1080 24 fps progressive should become the universal standard of choice for all film-originated video.
But there are some problems that rear their ugly head. For example: How do you monitor 24 fps progressive . If you've ever watched a movie at 24 fps with a single-bladed shutter in the projector, the flicker would drive you nuts. 25 fps is no better. Then how do you monitor 24 fps progressive?
In the film world about a hundred years ago, the use of a two-bladed shutter and later limiting the light on the screen to 16 ft-Lamberts helped reduce the flicker. Almost all movies today use this technique. Viewing a movie at 24 fps with a three-bladed shutter has less flicker but more light is required because of the smaller amount of openings in the shutter. The flicker at 72 times a second (24x3) is imperceptible.
To view a 24 fps progressive video, the display device needs to display the frame captured in less than 1/24 of a second and be updated in a very small amount of time to the next frame. This is how some of the new micro-display products, such as Digital Micromirror Devices (DMD), Grating Light Valves (GLV), Organic Light Emitting Diodes (OLED), Liquid Crystal Displays (LCD) and flat panel plasma displays work. In a CRT display the phosphor doesn't stay lit after it has been excited and has a brightness decay time of x number of milliseconds. The phosphor has to be refreshed continuously. For a 24 fps progressive video to be displayed with only a small amount of flicker, the display would have to display the frame two or three times for each frame. In other words, the monitor would be reproducing the 24 fps progressive at either 48 or 72 times a second. Each CRT monitor then would require a frame buffer that would be filled in 1/24 second but read out at two or three times during the next 1/24 of a second. The price of the monitor would probably be a little more expensive initially because of the added frame buffer. To make the monitor work at other standards would add to the cost and would still only show the same image, no matter what frame rate (interlaced or progressive) it is.
Since in the DTV standard the display is handled in the digital/receiver, it is no problem in the home. All incoming ATSC video standards will be converted to whatever the receivers native display requires. In the production and post production areas, a need to view the images before they are compressed and transmitted is critical. Engineering and creative judgments have to be made to evaluate these images and clients will want to see how their images are going to look.
A few years ago, people who use surveillance cameras to record images of bank robberies and holdups asked for a way of improving the clarity of images that are displayed in a still-framed on playback in a pause mode. A few cameras, developed by Sony and others, captured the images with a 30 fps progressive camera and output the frame as a 60sF frame. In those countries that are at 25 fps, the camera output 50sF from the 25 fps progressive images. These images were easily recorded on conventional equipment and worked quite well. (As a matter of note, all film images scanned today, on most telecines, are scanned progressively.) The progressively scanned image is read out as two fields. Those two fields in reality are two segmented frames, segmented frame one containing all the even lines and frame two containing all the odd lines. 3:2 pulldown is added for the conversion to 30 fps. In 25 fps countries, the image is scanned progressively and read out as two segmented frames. There is no need for a 3:2 conversion.
So the decision that has to be made is:
1. Do we come up with another video standard of 24p/48sF, orOther display devices like Fujitsu's flat panel plasma display, Texas Instrument's DMD (Digital Micro mirror Device), or the LCD and JVC's D-ILA (Direct Image Light Amplifier) do not require any conversion from the 24 fps progressive, as their images are updated almost instantly. There is no fading of the image over time. The image remains there until each pixel is updated, unlike the decaying phosphor image. (The DMD device actually delivers pulses of light for each pixel at around 50,000 times per second. Each pixel is updated once every frame.)
2. Do we stay at 24 progressive (25 progressive) and convert at the display device (if it is a CRT) using 48 Hz (or 50 Hz) with some flicker, or a flicker-less display at 72 Hz (or 75 Hz) rate.
Another matter discussed was the need for a Format Conversion Standard from the 1920x1080 24 fps progressive or the 1920x1080 48sF to those frame rates that are not simple multiples of the 24 or 25 frame rate. An example is 24 fps to 30 fps.
In Hollywood, 3:2 can be a nightmare if the sequence is interrupted. Today there are many DVDs out in the market place that do not have a continuous 3:2 sequence. Some have intermittent 3:2, 2:3, 3:2, etc. sequences, due to careless video editing. True, one can be meticulous in keeping an eye on the 3:2 sequence, but "Murphy's Law" says it will happen and it does. To extract 3:2 for MPEG encoding (be it SDTV or HDTV) when the sequence is out of order, will cause problems.
In conclusion, I believe that at 1920x1080 24 progessive should be the Mastering Format Standard. Any conversion for processing and/or display should be done within those devices. Otherwise there will be a proliferation of seguencers and de-seguencers. (Is this segmented frame 1 or 2? Did the sequence get upset in editing?) 1920x1080 24 frame progressive is the K.I.S.S. (keep it simple stupid) approach without playing the numbers game and giving the production and post production industries a robust standard that does not have any of the pitfalls that segmented frames have and will work nicely. Adding another video standard, 48sF, to the 18 ATSC standard is not wanted nor desired.
The URL for this page is http://www.tech-notes.tv/Jim/Articles/24vs48sF.html
Please use this URL in all links or references to this page
© 1998-2001 by James A. Mendrala