This is the second part of our report on the Camera-Assessment Series that was recently conducted by the American Society of Cinematographers and the Producers Guild of America. The series, detailed in the June ’09 issue of AC, assessed seven digital motion-picture cameras, comparing each of them to 35mm film, the benchmark standard for theatrical motion-picture quality.
This article contains no judgments or conclusions about individual cameras or how they performed. Instead, we focus on the participants’ general thoughts about the tests, what they learned, and what still needs to be explored. Although this article discusses the overall post workflow, the details of that work — for example, how much time was spent color-correcting specific scenes shot by specific cameras — will be reported at a later date.
There are many ways for digital cameras to be integrated into a production, but the CAS was designed to test them within a commonly used film/digital post workflow designed for theatrical releasing on print film and digital cinema (DCP). “As far as we know, this is the first time anyone has done a photographic performance assessment of the seven selected digital cameras using a common, film-centric, digital-intermediate workflow finish,” says Curtis Clark, ASC, chair of the Society’s Technology Committee. “We wanted to know how these seven cameras would fit in that workflow without putting the footage through a series of proprietary, ‘secret sauce,’ post-facility contortions.”
“We also wanted this assessment to be applicable to production and post now,” adds producer Lori McCreary, chair of the PGA Motion Picture Technology Council and president of Revelations Entertain-ment, which funded the CAS. “Because most feature films are finished at 2K resolution, we chose a 2K post workflow.”
Many people were eager to suggest alternate approaches based on their experiences with custom workflows. “As a result of their individual experiences, filmmakers develop highly personal and frequently conflicting perspectives on digital-camera image quality, corresponding with their varied results,” says Clark. “To get an effective and manageable handle on the digital cameras’ image performance, we realized it was necessary to eliminate the ‘wild card’ variability of multiple workflows by selecting a commonly deployed workflow for all the cameras. We’re talking about a theatrical-motion-picture post workflow — not DVD, and not HD broadcast. Television production will require a different assessment series based on an HDTV Rec. 709 finish, a different deal. The digital-camera workflow finishing you can do on a MacBook Pro in Apple ProRes HQ is not consistent with the DI workflow that is used for most studio productions.
“The manufacturers of most of the current digital motion-picture cameras who participated in the CAS have adapted their HD image-capture technologies to be more compatible with the prevailing, film-centric DI workflow by adopting a Log-mode encoding of their image capture as opposed to gamma-encoded, ‘linear’ HD video,” he continues. “These Log modes attempt to emulate the characteristics of film negative, reproducing a wider dynamic range of scene tones [from highlight to shadow] within a wider color gamut that is closer to film than the conventional HD Rec. 709 color-space gamut.”
To briefly recap our June ’09 report, the CAS shoot took place over two days in January 2009 at Universal Studios. The digital cameras were the Arri D-21, the Grass Valley Viper, Panasonic’s AJ-HPX3700, Panavision’s Genesis, the Red One and Sony’s F23 and F35. The same tests were also filmed with an Arri 435 using two tungsten stocks (Kodak Vision2 200T 5217 and Kodak Vision3 500T 5219) and two daylight stocks (Kodak Vision2 250D 5205 and Vision3 250D 5207). One cinematographer was assigned to each camera and stayed with it throughout the tests; each camera manufacturer helped choose which cinematographer would be paired with its camera. The cameras all passed through six different test scenarios, each with its own on-set cinematographer.
After the shoot, each camera manufacturer was responsible for converting its own images to 10-bit log DPX files and delivering those files to LaserPacific, which joined Deluxe Laboratories in donating post services to the project. (All of the 35mm processing was done at Deluxe.) The 35mm footage was scanned at 4K on a Spirit 4K scanner and then downconverted to 2K 10-bit DPX files. All of the files were then ingested into an Autodesk Lustre for color-correction. The entire post workflow, which had to accommodate the schedules of people who were volunteering their time, lasted five months. The results were a film print and DCP for public presentation. “We cut out most of the editorial phase in order to get the material ready to screen as quickly as possible,” says Dave Stump, ASC, chair of the Technology Committee’s Camera Subcommittee. “It was really an exercise in seeing how ingest and color-correction worked rather than how an entire editorial workflow worked.”
Because the goal was to make the images suitable for printing to film, a film-print-emulation display look-up table was applied to all of the footage, simulating the look of the images when recorded to film. However, “it was discovered soon after we started our work that a lot of the digital cameras delivered images that didn’t give me enough range to work with in a traditional Cineon workflow,” says LaserPacific colorist Mike Sowa, who graded the tests. “When I put up some of the images and looked at them with the print-density LUT applied, the information in the lowlights appeared to be clipped.”
To expand that image area and gain access to the digital cameras’ full dynamic range, Sowa had to apply an Input Device Transform (IDT), which adjusted the gamma slope of the cameras’ Log-mode images to more closely match film negative. “Although they’re designed to emulate film-tone scale, the Log-encoded gamma slopes of the various digital cameras are not congruent with the Log gamma slope of film negative,” explains Clark. “Applying the IDT ensured that the final color-grading accurately predicted the filmout results.”
Doug Jaqua, who works in color science at LaserPacific, notes that each of the cameras has its own method of encoding the digital signal to record the largest dynamic range the camera can capture. “The problem is that they all do it differently,” he says. “None of these things play well together out of the box in a common workflow.” Stump adds that the digital cameras are linear recording devices, whereas film is an analog medium characterized in log terms, and because the DI/DCP pipeline was designed for film, it is built around film’s log signal. “When you plug a digital camera into a film workflow, inevitably the signal from that camera has to be transformed into the Cineon log signal space that accommodates film,” says Stump. “That doesn’t just happen; it requires a significant amount of effort.”
At the moment, each post facility creates its own input processes to convert the raw material from an electronic camera into a form that will work within that facility’s infrastructure. “Every post house has to create its own input matrices,” says Stump. “That input-matrix data is the special sauce of every lab for every camera.”
The CAS guidelines prevented LaserPacific from applying its own special sauce because the workflow had to be platform- and facility-independent, able to be reproduced anywhere. Fortunately, the ASC Color Decision List includes a Power function — approximating the traditional Gamma function of color correctors — that could be used as an IDT, offering Sowa the full dynamic range of the digital cameras’ Log-encoded images for color-correcting.