DPX (Digital Picture exchange) is the format of choice for transferring film to the digital world. The terms DPX and Cineon are sometimes used interchangeably. Cineon was a format that Kodak created and was adapted to a standard by the good folks at the Society of Motion Picture and Television Engineers (SMPTE.) The resolution of the image is measured in thousands of pixels across the frame, or K. 1k is 1024 pixels (why it‘s 1024 and not 1000 is because that is how computers work in multiples of 8, don‘t bother to ask why, just accept it it‘s the same reason RAM comes in 128, 256, 512 and yes 1024.) 2k is 2048, 4k is 4096 and so on.
DPX is an RGB format meaning that each primary color that makes up the image (Red Green and Blue) is recorded on a separate channel , lower end video formats combine colors or sample some colors at different rates. Most film work is done recording each channel as a 10 bit log file. With 10 bit you have 1024x1024x1024=1,073,741,824 possible colors, good but still not good enough to capture the dynamic range of film linearly. The problem is in a linear 10 bit format will burn out bights too soon. It will turn them to pure white rather than resolve all the detail that film can (think trying to videotape your friends on the beach on a sunny day.) 16 bits per channel and usually even 12 bits has enough range to capture it all but that’s a lot more bandwidth and computer storage space and all of that detail is wasted in the low light areas where there is physically not that much data to resolve. To combat this problem it stores the data as a 10 bit logarithmic file. Logarithms are numeric values that more values toward one side than the other. With a logarithmic file you have more values on the bright side where it counts. This also mimics the characteristic curve of film which can capture more detail high light areas than in low light areas.
For the complete area of a film that was shot anamorphically 2K would be 1828x1536. For a film that was shot 1.85x1 2K would be 1828x988. Since it is good to have a little bit of extra area above and below the frame line (not to mention that some places in Europe may still use 1.66x1) it would be more likely that they would render to 1828x1101 for a ratio of 1.66x1.
The advantages to a digital intermediate are the fact that the cinematographer has a greater degree of control over the image and can change the color in particular areas of the frame in ways that can not be done photochemical. We Were Soldiers was shot anamorphically and digitally color corrected at 1828x1556. NOTE many film scanners that use the Cineon format for full aperture 1556 lines of horizontal resolution and not 1536. I have no idea why they adopted a standard that is not 1.33x1 (2048/1556 is 1.316) but it was their standard and they got to make the choice. Cinecite uses the Kodak standard as they are partners with Kodak.
For films that are shot full aperture (super 35 films) 2048 is the horizontal dimension. Lord of the Rings was shot in Super 35 and the film was scanned into the computer at 2048x1536 (1.33x1.) For The Fellowship of the Ring they only imported 75% of the film into the computer, the rest was not touched digitally. In order to cut the two together seamlessly they recorded all digital work back out to super 35 hard matted at a ratio of 1.77x1 (2048x1157), spliced that with the negatives that had not been scanned into the computer and optically extracted the 2.35x1 scope image. For The Two Towers and Return of the King they scanned the entire film into the computer and output a 2048x871 image with a digital anamorphic squeeze already added, saving the optical printing step that added grain to Fellowship’s release prints.
4 perf super 35 has a 1.33x1ratio that’s a awful lot of room that is wasted if you’re just going to extract a 2.35x1 image or a 1.77x1 image (like for HDTV work.) Newer professional cameras have a 3 perf mode that only uses 3 perforations per frame for a 1.778x1 image, perfect for HDTV work just as good to extract a 2.35x1 image from and with 25% less film needed and 25% less to process at the film lab, meaning less money spent. At the same time shooting 3 perf commits the film to do a digital intermediate since they don’t have optical printers that can print 3 perf negatives to 4 perf film prints for general release (and all of the theaters still have 4 perf projectors) meaning they can’t decide half way though the project that they’d rather skip the digital and do an optical blowup instead. Panic Room and Kill Bill were both shot in 3 perf and digitally converted to 2.35x1 scope for release. The advantage of a digital super 35 blowup over an optical one is that optical blowups can introduce a great deal of grain into the final product. Quentin Tarantino shot Reservoir Dogs in super 35 and did an optical blowup and was never happy with the amount of grain that ended up in the final release prints of the film. Only the improved quality of the digital blowup convinced him to shoot Kill Bill in Super 35. Cinematographer Robert Richardson pointed out that the cost of the digital intermediate was offset by the fact that film costs and lab processing costs were reduced by 25% by shooting 3 perf.
The latest thing in digital transfers is the Spirit Datacine. It was built by Thompson media with technology developed by Phillips. It was designed to transfer film into the digital world for HDTV work. Normally the device that does this is called a telecine ( i.e. it transfers cinema to television.) The datacine does more than just turn something into NTSC or PAL video it can output computer data hence the name. The datacine is not as good as a real film scanner; it can scan only 1920 pixels of horizontal resolution on the CCD rather than the full 2048. It is also a YUV format not an RGB. YUV is a format where fine color variances half as often as luminance (brightness). It scans at 8:4:4 (for every 8 times green and overall luminance are sampled red and blue are sampled only 4 times each) while a true film scanner can full 8:8:8. A datacine can output a 2K DPX file but it is not a true one. It is has been up sampled; the electronics in the machine electronically stretched the 1920 pixels of resolution to 2048 and digitally converted the YUV file to RGB format. Whenever you convert an inferior format to a superior one it will never have the real detail that a true one would have, it can only be as good as YUV 1920 can get. The big advantage of the datacine is that it is fast. It can sample in real time, rather than the few frames per second (or seconds per frame) that a true film scanner can do. When transferring footage in an expensive digital editing suit that is charging by the hour the datacine is frequently considered good enough and most movie patrons can not tell the difference. Seabiscuit and Pirates of the Caribbean were both digital intermediates that were transferred on datacines.
