Note: The summaries and opinions expressed in this section are not intended to provide complete or accurate answers to the processes and complexities associated with the video to film conversion, and are not to be reproduced or reprinted in whole or in part without express written permission of VTP, Copyright 2000 - 2005 by Videotape Products, Inc.


What is the spatial resolution of 35mm Motion Picture film?
The spatial resolution of the active area of standard 35mm Motion Picture Camera Color Negative was, prior to Kodak's introduction of Vision stock, measured as 80 line pairs per millimeter, or approximately 167 pixels per millimeter in any direction.  The horizontal and vertical pixel count depend upon the camera aperture (e.g., there is no increase in horizontal pixel count due to anamorphic lens image compression onto the camera negative). 35mm film negative can easily be rated at as high a resolution as 4096x2987 for Academy Aperture.   The "2K" level is considered half resolution of 2048x1493 for Academy Aperture.  The visual effects industry has been working at the "2K" level for years, usually with less vertical resolution due to the masked horizontal aspect ratios the filmmakers have chosen of 1.85 or larger (a horizontal rectangle).  Because of film to film generation loss, poor projection quality (from such factors as simple as turning up the lamp brightness on the projectors), and forced limited print contrast range of from 6 to 8 stops, the 1920x1080 8 or 10 bit capabilities of high definition video converted to 35mm print film can be indistinguishable and thus quite useful for mainstream filmmakers who still want a feature film print release. 

What is the color space of film compared to video?
In simplistic terms, Component Digital Video generally represents 8 or 10 bit linear Y and 8 or 10 bit weighted subsampled U and V linear color values (a sampling process resulting in separate "color difference" Y/R-Y/B-Y or Y/Pb/Pr channels of 8 or 10 bit depth each, for a combined 24 or 30 bit weighted color space).  DV, DVCAM, DVCPRO(25), DVCPRO50, HDCAM/CineAlta, DVCPRO-HD, IMX, Betacam SX, and HDV are all 8 bit formats, however specialized videotape recorders such as Sony's DVW Digital Betacam camcorders and VTRs, Sony's SRW-5000/SRW-5500 HD Mastering VTRs (in SRW 4:2:2 Mode) and Panasonic's D-5 HD VTRs are capable of recording at 10 bit luma Y and weighted subsampled U and V linear color values in a combined 30 bit weighted color difference color space.  Sony's SRW-5000 VTR is uniquely capable of recording (in 10-bit) in either the YUV or the RGB mode, with the RGB mode offering the best recording at 30-bit full band (non-weighted) 4:4:4 color space.  In terms of today's camcorders, the Sony Digital Betacam camcorders offer 10-bit recording onto tape (even Sony's HDCAM CineAlta is an 8-bit system, although it records about twice the spatial resolution of Digital Betacam), yet the new Technicolor Sony Genesis camera docks to the SRW-1 portable VTR which allows for either the standard quality YUV mode or the high quality RGB mode in HDCAM-SR to be recorded on small HDCAM-SR tape cassettes.

Composite Analog Video represents luma Y and weighted subsampled linear color values (for example, resulting in separate IQ channels) on a modulated subcarrier frequency or phase alternate line frequency based upon the "color envelope" of the scanning standard used (NTSC presumes a display gamma of 2.2, whereas PAL presumes a display gamma of 2.8, High Definition television has an altogether different presumed display gamma).   Film generally represents up to 12 or 13 bit log color and luminance values (10 bit log is generally a sufficient representation as long as the software weights the values properly, compared to say 16 bit where the weighting may not be efficient) somewhere between RGB (30 bit color space at 10 bits per channel, 36 bit color space at 12 bits, etc.) and CMYK (40 bit color space at 10 bits per channel, 48 bit color space at 12 bits, etc.), with a film negative gamma of slightly over 2 (e.g. 2.048 for some stocks) and a print display gamma of about 1.5 with an inherent projection print contrast ratio of up to 800:1 and a displayed print contrast ratio of from 500 to 600:1.  Although in limited use today, Kodak Cineon software was developed as a 10 bit log file system for Eastman Kodak motion picture negatives with the assumption at the time that the remaining 2 bits of information were deemed unrecoverable and unuseful given practical image scanning and recovery methods. 

In 2004, the DCI (Digital Cinema Initiative) essentially recommended the equivalent of 12 bit log RGB (as a minimum "deliverable") to facilitate better gradations in computer generated images so that the transitions between light and dark can occur without adverse or notocable concantination.  Also recommended was 4096x1080 scanning and for file server playout, encoding with Motion JPEG2000 compression at either 60Mb/s or 120Mb/s.

For the purposes of understanding exposure latitude (e.g., the difference between the maximum value of light and the minimum value of light),  at normal gain setting, 8 bit video can be generally representative of up to 8 exposure stops of value, each exposure stop representing twice the amount of light than the exposure stop preceding it, with the exceptions of reserved values for timing (using traditional video processing, which include a minimum 4 bits at the top and the bottom of the scale, plus a DMin and a DMax which allow some room for undershoot and overshoot.  Camera film negative can represent up to even 13 exposure stops of value, so it can be fairly easy to expose film for the light levels inside a room with a window and not worry too much if the exposure of the image outside the window exceeds the normal print range of the negative (with film, you usually still get some kind of an image in the highlight areas to print down with if the exposure falls within the range of the camera negative). 

It is fairly easy, however, to change the extreme values of a digital video camera head, such as to set minimum black levels below the nominal pixel values usually assigned to represent black (also referred to as "black stretch").  The same can be said for the digital video camera head in representing highlights as either clipped and compressed lower values or as higher values.  Today's best digital video cameras are very sensitive to red light, offer very poor performance in blue light, and may now offer varying levels of color value indexing capability depending upon the manufacturer and model (e.g., Sony CineAlta, Panasonic Varicam, and Sony's Digital Betacam Camcorders which offer a color value indexing system allowing for film log color and exposure latitude emulation based upon special color index tables read in real time off of camera memory card or memory settings while photographing).

Why does film get scanned and printed back to film?
To mention a few reasons, film is scanned into digital data files for the purposes of restoration in digital paint systems such as from Quantel, film is scanned for color correction, and film is scanned for the purpose of adding special digital effects and animation.  Digitally Preserved or digitally retouched Estar base safety film color negative is used for archive storage (color separation negatives may shrink unevenly), and the digital backup tape formats available today are not expected to be available in 15 to 20 years, whereas mechanically simple film projectors and simple optical recovery methods are expected to be available.  Film is also of course scanned for conversion to television and video.  A high profile digital effects house may not have a high definition video recorder but may output film material which may be re-scanned back into high definition at another facility. 

There are different scanning methods to choose from based upon the desired end result.   A telecine is usually used to scan film for conversion into video.  Three basic kinds of telecine systems are in use today:  the older film chain systems, the older analog telecine systems that employ precision scanning tubes, and the newer digital telecine systems that employ CCD technology.  Of the new CCD HD telecine systems, large format scanning is considered 65mm, and Super 35mm scanning based on 4096 and regular 4-perf 35mm based on 2048 horizontal resolution is common.  High Definition video processing (as it is described as video data) can support up to a horizontal resolution of 1920 pixels in the 1080i/P/PsF formats, and 1280 pixels in the 720P formats.  High Definition video in the data mode can be a non-real time extended raster image.  There is a picture safe area and there are legal limits with HD video data to be measured and considered.

With film, what are some of the scanning options?
To mention a few options, currently, the daVinci Color correction system works with telecine "best light" scanned images of  film at 10 bit log color, and stores the data as a 1920x1080 High Definition Video file on a Disk Recorder system.  Material can be scanned off a Philips Spirit DataCine and color corrected off the Disk Recorder in random non-linear order much faster than starting, stopping, and backing up the data telecine (thereby risking the negative to undo stress, dirt, and scratches).  However, if higher resolution scanning is needed, you could use something like an 8K resolution Kodak Lightning Film Scanner (there are a few around in Los Angeles) but scanning would be non-real time process.  When you make a film to data file transfer, you have to consider the image file format and the operating system you will be using.  Also, software, hardware, time, and storage costs can force you to make choices or to defer choices in the way the original film color values are preserved as data (e.g., Kodak Cineon software supports 40/10 bit log, as does FIDO, whereas many common image file formats only support 16/8 bit linear).

A single frame of 35mm motion picture film scanned full Academy Aperture at half resolution 2048x1556 with 13 bit luminance values and 10 bit log color  (CMYK) would represent a data file of approximately 11 Megabytes (a 100 Megabyte Zip drive could store just 9 complete frames).  The data rate for displaying this frame in real time at 24 frames per second would be 264 Megabytes per second (2,112 Megabits per second, twice the rate of high definition video).  It is important to understand the precise calibration testing a professional motion picture film scanner undergoes and the particular file formats used make a difference in how easily you can store, recall, and interchange material.  Common data storage systems include Ampex DST, Quantum DLT, Sony DTF-1 and DTF-2 (with ultra-wide SCSI or a fiber channel interface), and Exabyte.  Other data storage systems include Sony ID-1, and Cybernetics M2 AME.

What is the process of converting video to film?
In brief, interlaced video is turned into digital video, transferred into a digital server, and de-interlaced.  Next, the 30 frame NTSC odd/even fields are compared for sophisticated temporal reduction to 24 frames (25 frame PAL avoids the temporal reduction and causes the film to be 1 frame per second slower), a three color separation black and white negative is exposed to indexed pixel values with an electron beam recorder system.  The exposed separation negative is combined and step printed into either camera color negative or an interpositive "answer print" (at Sony High Definition Center, the finished resolution was measured as 1920x1035 on the old-style Electron Beam Recorder). 

Other types of film recorders may be used such as 9" High Resolution CRT systems, and newer, more sophisticated Laser Beam Recorders.  Many film production companies own film recorder devices, and use them to output their composited digital effects and animation.  Examples of film recorders include the "Arri Arrilazer" and the "Celco Xtreme Nitro".   There are productivity issues to consider as well.  The film frame exposure time used to take approximately 1 second per frame with the Sony Electron Beam Recorder.   Other systems and technologies can be more resolute, but are slower, with the next closest time at approximately 5 seconds per frame at what is considered full 2K (2048, but 35mm half-film) spatial resolution.  It is highly suggested to consult with experts in these processes prior to production since many factors affect the up-conversion process including the amount of processing you may have engaged in the camera (it turns out that extraneous camera processing adds to the loss of resolution or blurring of the film image). 

Audio can be a complex matter, being resampled if necessary and/or speed/pitch corrected to match the film frame rate.  Theatrical audio mixing is usually done with consideration to surround sound systems and performed carefully to preserve phasing in multi-channel effects and music tracks, and in any stereo emulation. 

The final post production process is the same as the normal feature film process.

Will Progressive Scan Video help in the Video to Film Process?
Progressive scan effectively eliminates having to deal with "temporal motion blur" between video fields (an interlaced television frame is traditionally written in two video "fields", a first field of 1/2 resolution, followed by a second field with 1/2 resolution, so in the time it takes to write the first video field, a fast moving object moves to a different position so that when the second field is written, the combined two fields depict an inter-field motion blur).  Yet, there is still motion blur from moving objects to be considered, which is partially a function of shutter and CCD clock speed. 

However, trying to get 30 frames progressive to 24 frames progressive is a serious challenge.  In NTSC interlace, you have 30 frames to work with also, but each frame is made up of 2 fields.  Because 60 fields already have a temporal interfield blur, and there are 60 field choices to combine and eliminate 12 fields for a total of 48 fields (12 fields are removed out of 60 compared to 6 whole frames removed out of 30), the frame reduction process is not as severe with interlaced material.  However, with enough time, effort, and money, software exists such as from Realviz which can use pixel image recognition from two consecutive frames of a sequence and create extremely accurate in-between frames, thus allowing you to expand the timing of a 30 frame progressive clip to a more useful 60 frames.  At the 60 frame temporal level, the "M" film conversion process that used to be used at Sony High Definition Center would have treated your material similarly to 60 field interlace, and temporally combine images back down to 24 frame film.  The "M" process was apparently under Patent which was due to expire somewhat recently.

Therefore, for the easiest conversion back to film, one would use a progressive scan camcorder that could support 23.98 frames per second imaging, even if it meant taking a step or two at the beginning of the post production process to properly remove the pull-down repeated frames or fields.   It should be noted that Sony has introduced expanded 23.98 HDCAM recording capability in the HDW family of VTRs, meaning that HD frame rate conversion may not be necessary if working 1080 23.98 or 24 fps through the Post Production Process.

Will PAL Video help in the Video to Film Process?
PAL video (if it was photographed as PAL video) effectively eliminates any image temporal combining process from 25 frames (50 fields) down to 24 frame film since the process would be too severe.  Rather, film is exposed to PAL video frame for frame, effectively slowing down the running time of the film by 1 frame per second (60 frames per minute).   Consequently, the audio tracks must go through a matching speed and pitch conversion.   PAL also increases the vertical resolution of the image since there are more scan lines used, although it has been said that for many reasons, progressive scan has a potentially bigger impact on the perception of spatial resolution, followed by horizontal resolution, and then vertical resolution.

PAL is limited to 720x576 resolution (in it's optimal PsF form), but can be up-converted in post production (resolution is not added or increased) to 1920x1080PsF 25, the European HD Standard, and then "slewed" to 24 frame playback (with a matching 4% audio pitch correction).

What about consumer or low priced pro digital camcorders?
If you are attempting to use a consumer or professional camcorder to ultimately convert a project to film, know that there is a large difference in the quality of the sampling you will be making in your camera compared to a broadcast quality digital camcorder, and even more of a difference compared to a high definition digital camcorder (all consumer cameras and many professional camcorders have less than 600x400 overall resolution and have less resolution than the old Super 8mm film format cameras used years ago, whereas the broadcast standard definition digital camcorders may offer up to 700x480 or slightly more resolution).  For the record, there were some forgettable films shot in Super 8mm that were blown up to 35mm, as the images were blurry.  There were even more forgettable films shot in 16mm and blown up to 35mm, and even more shot in Super 16mm that have been forgotten.  Know that camcorders differ in price because of a number of things, including the quality of the software they use internally to derive their digital sample from the CCD chips and in creating analog filtered outputs.  Non optimized imaging systems and related camcorder internal software can derive substantially less resolution from the same size imaging system in an optimized camcorder, especially when the actual grade quality of the CCD chips are taken into account.

How do I rate my professional digital camcorder?
It is not just the active picture area measured in pixels that derives a good image suitable for enlargement, rather it includes how that pixel sample was originally obtained with respect to such things as edge detail, ailiasing, image enhancement, filtering, and processing prior to recording (also see the section above on consumer camcorders).  You must also know where the limiting resolution is within the camcorder system.  For example, the limiting resolution may be in the lens, in the imager, in the digital processing sampling rate prior to compression, within the compression itself, and even in the tape format or analog output.  You cannot measure this resolution by the number of active elements within the imager, nor can you assume that your digital tape format will magically supply you with the same resolution it is capable of storing. 

In any case, the simple test of how good your camcorder video looks on a monitor or small projection system does not really correlate with the resolution necessary to resolve a large image up on a large film screen for closer viewing. Imagine 720x486 (4:3 or 16:9 aspect ratio) standard definition video expanded on a large theatrical sized screen. Know that it takes at least a 900 line resolution video monitor with proper decoding and filtering to show off 720x486 interlaced resolution (do in part to things like the "Kell" Factor).  Even projecting the image with special processing to eliminate visible video lines (de-interlacing, or line quadrupling, and interpolation) the image will only resolve detail to the limit of what was originally recorded at best. Now, imagine 1920x1080 high definition video (16:9 aspect ratio) on the same screen, using similar image processing to remove visible lines. There is a significant difference in capability to resolve detail.  For direct television monitor viewing, high definition video is intended to be viewed in closer proximity (at 3 times or less picture height) than standard video (up to 7 times picture height is acceptable), and is more similar to film resolution than standard video (even broadcast uncompressed component digital standard video) is to high definition video.

You can measure your camcorder's resolution by subjecting it to recording camera line pair charts, reading the digital file into a computer, importing it into a program such as Photoshop or After Effects at full resolution, and counting the line pairs that are actually resolved in the digital file.

Updated 05/03/05