Three Paths to Film

Three Paths to Film

Byline: Steve Mullen

Now that shooting video with the intention to go to film has become commonplace, we are confronted with a large selection of camcorders with which we can shoot the video.

Sony offers solutions at various price points. At the high end, you can shoot at 24p with a resolution of 1080×1920 using Sony’s CineAlta solution. Of course, 1080×1920 HDCAM can also be shot at 23.98p, 25p, 30p, 50i, 59.94i, or 60i. Moving down in price, Sony’s NTSC MSW900 IMX camcorder shoots at 29.97p and 59.94i. Sony’s PAL version of the MSW900 can shoot at 25p and 50i. Sony DV and DVCAM camcorders shoot NTSC at 59.94i and PAL at 50i.

JVC offers both DV and D-9 camcorders running at 50i or 59.94i. JVC now has a low-cost SD/HD camcorder, the GR-HD1/JY-HD10U, that can shoot MPEG-2 at 480/59.94p and 720/29.97p. It also shoots NTSC DV at 59.94i.

Panasonic offers DV, DVCPRO, DVCPRO50, and DVCPRO HD camcorders that can be used for filmmaking. Most of these camcorders record NTSC at 59.94i or PAL at 50i.

However, Panasonic offers two special camcorders. Last year’s phenom, the AG-DVX100, is available in PAL and NTSC versions. The NTSC version shoots at 23.98p, pseudo 24.00p (24p ADV mode), 29.97p, and 59.94i while the PAL version shoots at 25p and 50i. Panasonic’s AJ-HDC27V Varicam shoots 720p at frame rates from 1fps to 60fps. The camcorder’s “field clock” can be set at 59.94Hz or 60Hz. Panasonic’s AJ-FRC27 Frame Rate Converter can process Varicam 720/59.94p and 720/60p video shot at 24fps at output this video at 23.98fps or 24fps.

Three Target Film Rates

For those of you who think of video running at either 30fps (NTSC) or 50fps (PAL), the frame rates above may seem to confuse the issue of transferring video to 24fps film. Thankfully, exactly the opposite is true. Upon organizing these disparate frame rates, we see three clear patterns emerge (see table on p. 20).

The three categories include a target rate of 24fps and two other options – one slightly lower and one higher. The lower rate, 23.976fps, is 0.1% slower than 24fps. The higher rate, 25fps, is 4% faster.

The first fraction is known to all of us in NTSC land because it represents how much 30fps video was slowed to enable color information to be carried compatibly with a monochrome picture. It is also the difference in program durations between dropframe and non-dropframe timecode. The second fraction is familiar to those who use PAL equipment. It is the difference between the frame rate for PAL video and the frame rate of film.

Many of us imagine video being transferred to film using the Kinoscope technique, where a film camera shoots video from a high-quality CRT. Current transfer processes are very different. Video is digitally captured or digitized to hard disk. Alternately, you can send your hard disk to the transfer house. During a transfer, several video frames are pulled from disk and used to calculate the frame that will be recorded on film. Current recording technology requires a second or two to expose each film frame.

Each transfer-house has a proprietary algorithm to generate the frame to be recorded. These algorithms have been designed to handle interlace scanned video. In fact, because so much R&D has been expended on these algorithms, some houses are not yet ready to handle progressive video.

Interlaced material looks good when played because each field is removed from a frame and presented on its own at a presentation rate twice the frame rate. Thus, you see NTSC video at 60 images per second. The eye integrates these rapidly appearing fields into smooth motion. Film, of course, does not work this way. A film projector’s rapidly spinning shutter only creates a 48Hz/72Hz flashing that raises the flicker-fusion frequency high enough so most see a steady image. The projector’s shutter does not alter film’s dramatically lower 24fps temporal rate.

Going to film requires satisfactory solutions to three problems. First, an interlaced frame with its motion artifacts should not be transferred to film. Second, a field, which has only half a video frame’s vertical resolution, should not be transferred to film. Third, interlaced video’s higher temporal sampling must be “resampled” to a lower temporal rate.

The solution to all three problems is to calculate the points in a sequence of video fields where a film frame needs to be generated. With NTSC, film frames are mapped to 60 fields. The result is a pattern where one film frame lies upon a video field and, two fields later, one film frame lies exactly between two video fields. Then, two fields later, the pattern repeats.

Advanced math is used to construct a 480-line frame buffer at each of these two points. At the first point, the field prior to the target frame (and possibly the field after it) is combined with the target field to generate a frame. At the second point, the two fields involved are used. In both cases, the motion of pixels is analyzed to determine movement over time. (With NTSC video, the time between samples is 1/60th of a second.)

A pixel that does not move between two fields is placed in its proper place in the frame buffer being generated. With static video, all pixels from two fields will be loaded into the new frame, which now will have 480 lines of resolution. When motion is detected, three options are open: the pixel in the early field is loaded into the buffer, the pixel in the latter film is loaded into the buffer, or two pixels from both frames are blended with the result being loaded into the buffer. The result of this process is lower vertical resolution for objects in motion.

It might seem that progressive video would be easier to transfer because motion artifacts do not need to be removed. In fact, the temporal resampling process can be easy (see green text in table on p. 20), difficult (blue text in table), or very difficult (red text in table). Progressive video with a frame rate that matches one of the three target rates converts on a 1:1 basis. When the ratio of the target film rate to video frame rate is greater than two, a pulldown process can be used to resample the temporal rates. Unfortunately, when the ratio is less than two, conversion is very difficult.

Audio is transferred from a hard disk, videotape, or a DAT tape. Using a variety of techniques, the soundtrack is married to the film, both for preview viewings and for the final print. Titles from the video may be used or may be created on film.

After the transfer from video to film, when we project the film at 24fps, only the conversion done with a target rate of 24 will be spot on. Audio will play in perfect sync and at the correct pitch. Transfers with a target rate of 23.98 will be visually perfect. Voices and sound effects will also sound fine. Most audience members will not be able to detect the higher pitch of music. Of course, the film will run 0.1% shorter than expected, which is easy to compensate for during the edit when you know it’s going to happen.

When, however, we project the film at 24fps after a conversion with a target rate of 25, all will not be well. The film will run 4% longer than expected – again, easy to compensate for during the edit. The original soundtrack cannot, because it runs 4% faster than the visuals, be used without complex processing.

Dialog and sound effects can be slowed by 4% and then pitch corrected. Music, however, is a problem. If you don’t have music as a separate element, you’ll have to live with the distortion created by the pitch-raising process on sustained tones. Before deciding to work in PAL, you should read D.W. Leitner’s “Shooting for Sundance” (Millimeter, January 2003), an article that describes the complex process by which 25fps video was transferred to film.

Progressive video shot at 24fps – with the greatest possible resolution – will transfer at the highest quality with the least transfer complexity. The next best transfer option is progressive video shot at 23.98fps. However, image resolution issues make shooting at 25p a viable alternative if the budget doesn’t permit shooting in HD.

feedback

To comment on this article, email the Video Systems editorial staff at vsfeedback@primediabusiness.com.

COPYRIGHT 2003 PRIMEDIA Business Magazines & Media Inc. All rights reserved.

COPYRIGHT 2003 Gale Group