2.0 Overview / Background
Abstract
The purpose of this report is to provide scientific image analysis on two motion picture capture technologies: film and digital. Both cameras have been tested using qualitative and quantitative techniques that explore critical image quality characteristics. The evaluations include each camera’s capabilities and limitations in dynamic range, color reproduction, noise, sharpness, and compositing.
The results and calculations are used to form meaningful comparisons and conclusions as to which camera technology produces better image quality. It should be noted that this report has based its conclusions from scientific and mathematical analysis; therefore a person’s aesthetic preference may be different than final results.
Introduction
In recent years, digital cinema camera technology has advanced enough to be on an almost equal footing with film cameras. The objective of this shootout is to compare the capabilities of film and digital capture in motion picture cameras. The cameras that will be used for this comparison are the Arriflex BL-III 35mm film camera and an Arri D-21 digital cinema camera. In order to make a fair comparison to a RAW workflow, a 4K scan of the film will be used instead of the cheaper 2K scan.
The shootout consists of a series of scenes shot over the course of 3 shooting days. These scenes are designed to push the cameras to the limits of their abilities. Image quality will be compared between the two cameras by assessing dynamic range, exposure latitude, MTF, resolving power, noise, and color reproduction. These tests should show how close digital cameras have come to achieving the “film look”. Alternatively, the 4K film scan should prove a worthy adversary to the RAW workflow and demonstrate why film is still a popular choice amongst cinematographers in the film industry.
Background
The Arriflex D-21 was the second camera to be produced by Arri as an attempt to create a digital solution that would provide an image that would have similar qualities to film. Arri had succeeded at this and had a fairly successful trial with the D-21 where it was used in quite a few features. This camera was the precursor to the modern day Arri Alexa which built upon the Arri “look” that captivated so many DPs. As part of this comparison the Arriflex D-21 will be analyzed to see if the D-21 can hold up to film.
For Arri to create the film look, they started by designing a sensor that was only 1mm shy of being a full 35mm film frame (23.76mm x 17.82mm). This provides a depth of field that is extremely close to that of film when using the same lenses. This allows DPs to use the lenses that they are familiar with for film and get similar results on the D-21. The capture resolution of 2880 x 2160 also makes for a fair comparison to film when transferred at 2k to 4k resolution. The camera also provides a mechanical shutter with variability between 11.2 and 180-degrees, similar to a film camera. The ISO ranges of the D-21 also provide a similar range to film with 100 ISO to 500 ISO.
There are also the color space options of Rec. 709 or logC to add to the film like feel. The camera is capable of recording both raw and video signals using the 3G-SDI out of the camera. Raw requires dual link to support the data rates while video only utilizes one of the SDI outs for signal. In our case we will be using the raw capabilities of the camera. To capture the raw footage, an external recorder must be used. For the shootout, the Stwo DFR2K will be used to capture the .ari raw files. The recorder is controlled using a laptop that communicates via ethernet.
In an article by ARRI, they tout how the D-21 has successfully intercut with 35mm film. Even David Mullen ASC has commented on the the abilities of the D-21 against past competitors, as it is a capable camera that has the feel of film on set and allows the usage of traditional film lenses, even anamorphic with its 4:3 sensor. For comparison Kodak Vision3 500T will be used. This was chosen because of the common usage of this particular film stock due to its ability to provide decent low light shooting and a nice grain structure that is appealing for many DPs that enjoy the grittiness of film.
This will be used in the Arri 35 BL-III which has a gate of ⅝” by ⅞” providing for a full 4:3 frame. A Zeiss 25-120 zoom lens is intended to be shared between the two cameras allowing for a better comparison of specifically the capture media; and not the media and lens combination. Vision3 film stock has shown to have improvements in the grain structure that are designed to decrease the amount of noise visible using a newer t-grain that is significantly smaller. David Mullen ASC, commented on this in a test where the new abilities came out when the film was underexposed and pushed a stop. According to him, the result between Vision2 and Vision3 showed that the Vision3 stock was better at keeping the noise down in the shadows, but still suffered from the lowered contrast. Mike Sowa, a colorist at Technicolor, also comments on how the 5219 stock holds up better in the highlights providing for more information in the DI process.
For the comparison both the D-21 and the Film will be at resolutions much higher than the final exhibition source. Film will follow the 'tradition film-workflow with a data scan at 4k 10-bit color. The D-21 will be a 2.8k raw file. The point of this is that even though the capture formats will be significantly higher than the exhibition final format, research that has been published by SMPTE shows that starting with a higher capture format and scaling it down to the lower exhibition format results in an image that appears sharper to a human observer. For this, color correction must be taken into account to bring the large amount of data from the data files into a video standard. Things that need adjusting such as: the gamma, contrast, and gain, to get the image looking like an image you would expect to see on a television; requires bringing everything into a video space and verifying that quantization errors are not occurring due to the dithering of the signal from a bit depth higher than 8 down to 8.
“research that has been published by SMPTE shows that starting with a higher capture format and scaling it down to the lower exhibition format results in an image that appears sharper to a human observer”
Editorial process will work off of technically graded shots that are brought to a pleasurable level and exported to proxy files that are ProRes 422 to allow the editor to work off of files that will not load the system down. The edit will be reconformed in DaVinci Resolve to the source resolution openEXR files that are in the ACES color space.
The final export to an H.264 file will be completed in Resolve to meet the proper Rec.709 specifications. This is where the down-res will be done to ensure that data is not lost through the post production pipeline.
Overall the comparison should provide for a fair testing of film vs digital and to see if Arri’s motto of “less pixels, but better pixels” holds up to film that is processed through the normal route for feature films. Also this comparison will allow us to test and see whether or not the D-21 is capable of its claims of being able to intercut with film. With both touting their high dynamic range, high latitude, and great color reproduction, the test should give a great insight into the abilities of both systems.
Our guess would be that the film will have an advantage over the D-21 with respect to both dynamic range and latitude due to the design of the film. Color reproduction might go to the film as the winner but there is the chance that when properly graded, neither will have a significant difference. Both noise and sharpness should yield significantly better results for film, due to the newly designed t-grains which will probably have a great advantage over the D-21 throughout the over under scene where it has been noted that the new grain structure greatly improves the noise during underexposure.
