ARRI: 16mm film is 3K, 35mm film is 6K...

[John Moores]

ARRI does not say 16mm is 3k.

S16 physical dimensions=12.35mmx7.42mm.

Size of "pixels" on a piece of film=0.006mm.

Number of "pixels" in a S16 frame=2058x1237.

In order to transfer full detail from the randomly placed "pixels" of film grain to the grid pattern of a film scanner's sensor, without introducing aliasing, you need to scan at a resolution at least twice as fine as the original source material.

In order to preserve all the detail of S16, the film is scanned at 3k, which results in a final image size of 2058x1237.

The claim was that S16mm is equivalent to, or better than HD 1080p. In terms of actual information contained in the image, it is.

2058x1237 is a larger number than 1920x1080...right?


Hope that helps.
Jm

 

[Kevin Halverson]

To absolutely avoid aliases, the highest sampled frequency sampled must be <.5 the sample rate (where the phase relationship between the excitation and the sampled system is random). This can be "pushed" a bit (no pun intended) if one is willing to accept small magnitude aliases, but if one wants to ensure that absolutely zero aliases are quantized, the sampled system's response must decrease the magnitude of the input signal to less than 1 LSB by .5 FS.

 

[Dan Hudgins]

grains and detail

grains and detail

We already know grain does not contain details about a scene.

Film layer thickness has more than one grain, grains are not side by side but in a 3D arrangement.

Detail in film is spread over several frames where for any given point in the image grains of different sizes will and will not exist, depending on the frame selected.

When viewed uncompressed at sound speed film resolution as viewed is higher than when looking at a freeze frame since the eyes and brain fuse the image details spanned over several frames.

At the brightness of a large movie theater the persistence of vision is longer than when images are viewed at higher brightnesses such as off a computer monitor, so the frame span is longer.

Since large and small grains can exist at any point in the image detail at low MTF can exist, and since the grain pattern is sort of "random" details too low in contrast in a single frame to see are reinforced when viewed uncompressed at sound speed to the point that the invisible low contrast details become more than just visible. The improvement in visible detail from freeze frame to sound speed looks about double, so if a still frame from a 35mm movie looks about 1280 resolution, when running at speed it would look over 2K in a film to film print.

If the film has been scanned then the "sub-pixel" detail in the grain is in part lost. Also if the frames are compressed for Blu-Ray the images are filtered to remove frame rate and texture, so the details spread between frames are lost, and much of the low contrast detail in general is filtered out, especially low value temporal detail.

The eye-brain views low contrast temporal detail at a greater contrast than it has, so when you view a film to film print details that are too low contrast to see in a freeze frame become enhanced by the uncompressed frame rate near the flicker threshold of the eye and look more intense than their measured delta would suggest.