The Odyssey – DCP/Format Distribution by Territory vice

[Christos Gartaganis]

The approximation of 4K resolution of 35mm film was based on the 1.85:1 ratio. So, a Scope format is using more film area than that in the vertical dimension.

i've seen demonstrations of 35mm film going through the various stages where they shot a target and projected the negative, then the resulting inter-positive (or EK print), then the resulting internegative and finally the release print and they were shown back-to-back-to-back through the same projector. One could, indeed watch the degradation.

What I have yet to see is a conversion of 35PA to digital and then projected where the line pairs are any good.

Here you are. Enjoy !
Smpte_RP40_Director_Full_Frame_web_02.jpg

[Steve Guttag]

The image I downloaded shows 24ish, line pairs (depending on the target). But this is, no doubt, a compressed jpg.

This is also not a projected pattern but is likely a scan of an original. At what resolution does one need to scan it to get the 85 line pairs as distinct lines (and note, that is not the limit of film projection. NTAV's target went up to 100 and I had no problem resolving those.

Then, put the scan through the whole DCP process (make a DCP of it and what is the maximum line pairs the projector can resolve (2K and 4K)?

-Steve

[Marcel Birgelen]

Comparing film resolution with a fixed pixel raster remains particularly difficult, it feels like comparing apples to oranges, while the concept in itself clearly has merits.

If we look at the pure technical details, then a 4K DCI projector can show 2048 horizontal line pairs of distinctive colors (for the sake of the discussion: black and white also being a color)

Looking at the height of a full 35mm frame 17.5 mm:

single line width: 0,00427246 mm or about 4,3 microns
pair line width: 0,00854492 mm or about 8,5 microns

Those details are smaller than the average grain clumps found in most 35mm emulsions, which range between 10 to 30 microns, but individual grains can be as small as 0.2 microns. Also, you're trying to map a random grain structure onto a fixed pixel grid, which will cause aliasing at least somewhere in the process, even if you'd scan the source material at e.g. something maniac like 64K, the presentation format will be 4K, some information will be lost, some aliasing will occur.

So, there may be more detail in 35mm than we could practically resolve with a full 4K sensor, but this is highly dependent on the type of emulsion used and also obviously bound to generation loss, unless you're scanning directly from negatives.

In the end where it matters is if it actually shows on screen. Do people see more detail in a pristine 35mm projected film than in a good 4K scan of the same print?

We've seen that DCI projection isn't without its own challenges. Older mediablocks for example, struggle rendering complex details in 4K and will often soften part of the image. JPEG2000 compression isn't perfect and introduces artifacts, especially with "mechanical" patterns like hard black and white lines.

What digital projection probably gives a higher perceived resolution compared to 35mm is the lack of any jitter in the frame and the lack of any full-frame flashing (at least as long as not using stereoscopic 3D).

[Stefan Scholz]

In real film, we are not dealing with playback of camera originals. We are not dealing with direct reproduction of grains.
It is a copied grain structure at least. Even if printed in contact printing, there is no such a thing as direct contact between 2 emulsions. Even a table with 4 feet is only standing on 3, and this more or less applies to any materials in contact. This is, why we use thermal compound mounting high power semiconductors to the heat sing, to fill the gaps, and thereby increase thermal conductivity, even though it is lower, than a 100%, direct metal contact would be.
In film printing often optics are used, and diffraction limits resulution quickly. Think of a smart phone camera with 120 mega pixels. On a sensor surface as large as a pin head? Impossible, so the image you see on screen is mostly AI created.
35 mm and 70 mm film images offer a limited "sensor" area. Color film stock has a vastly limited resolution over low sensitivity document film Black and White negative. Best seen here, in optical sound.
A copied sound negative struck on the document film used and printed to color stock redeveloped black or silver sulphide tracks is limited to 10, max 12 kHz upper limit, if the slit width allows. On Black and White 16 kHz can be copied.
As I pointed out earlier, optics do have a limited resolution, which limits resolution of the photographed negative, the print, and finally projected image on screen.

Luckily, our human eyes resolution in center vision is around 60 pixels per degree of the vision field. With maximum allowed theatrical viewing angles of 90 degrees, a merely 5400 pixels can be separated in center vision in the front row. Going for the smallest allowed viewing angle at 36 degrees. Just a little above 2k resolution remains. That in mind, we find the reason for mandated 4k theatrical projection.

In 35 mm and 70 mm projection, there is no automatic focus system present. Focus is adjusted in most cases from the worst possible position in the house, within the booth, next to the machine.Beside some 1960s or 1970s road show theatre installations, which eventually featured a control console in a sweet spot of the room to fine adjust focus and framing remotely, these controls are rarely seen (or even used) in today's operation. Best I know, some kind of operator adjusting the focus after starting the plattered movie. Binoculars to assist? Permanent control? Non existant.

It is a discussion that leads nowhere. Film has a specific look, the image on screen is created totally different, as is the function of the light intensity curve. Film following more or less a triangular pattern, twice per image. Digital differs on the type of projection system used. There can be absence of frame flashing, but also sequentially R-G-B-W flashing of the same image. Depending on technology and number of imagers used. All types exist in legitimate DCI systems.

We can't take away the magic, and as long as that magic sells for premium, we will still get film prints in limited quantity.

[Steve Guttag]

The study of just how images are perceived is an ongoing and likely never fully understood. A digital picture will be perceived as "sharper" than a film one, not from more detail but from less. That is, if one gets a well focused digital image, the pixels are going to be very sharp. You could reduce your resolution to just a 2x2 grid with a checkerboard pattern, the resulting "image" will be perceived as very sharp with distinct lines. However, the detail is just 2x2 still. One is handcuffed if they want to represent anything not on the grid.

A 2K picture can look quite sharp and full of detail, until you realize that everything is hyper-flat. It has the appearance of an image made from paper cutouts laid on top of each other (think Southpark). The detail is lost in a 2K image but I'll bet the perception of a 2K image versus a 35mm image of the same movie is a positive one. With a 4K image, there is less of the hyper-flat issue and it has, to my eyes, a more-filmic look, with respect to details, at least compared to 35mm. There are visual cues to give the image depth that is absent from 2K.

While there is no direct comparison between digital resolution and film's resolving power, there are points where the differences in perception diminish. Trying to rationalize or computer what a person can possibly see as a distinct pixel from a seated position is a poor means of determining how people perceive images, particularly color moving images. There is more to how we perceive depth as well.

16PA, 35PA and 70PA are designed to show the various parameters of the projector/lens itself, exclusive of the process that got a print to the projector (negative, IP, IN, Release Print (plus any optical effects done alone the way). Resolving power (line pairs) is but one part of it. Vertical and Horizontal motion come into play too. As mentioned, I've seen the degradation of the same shot film from negative all of the way through to final print. It is notable. Heck while doing some premiers, one screen might get an "EK" print (off of the (OCN or as close to it) and a "Select" print (considered the best of the standard release prints...a print that was viewed and determined to be the best representation) and difference is stark, and not just clarity. The line pairs do not degrade as much as some might think, on duplications but motion and color get more away from the EK version as you go. Release prints also are very dependent on if IP to IN was step printed or contact printed. If contact printed, it will look worse as there will be slippage using the same pitch perfs on top of each other.

[Marcel Birgelen]

In real film, we are not dealing with playback of camera originals. We are not dealing with direct reproduction of grains.
It is a copied grain structure at least. Even if printed in contact printing, there is no such a thing as direct contact between 2 emulsions.

We can scan the camera negative, if present. Then there will be no generation loss.

In film printing often optics are used, and diffraction limits resulution quickly. Think of a smart phone camera with 120 mega pixels. On a sensor surface as large as a pin head? Impossible, so the image you see on screen is mostly AI created.

To be fair, if you buy an ARRI Alexa, you also buy into part of their software magic, that makes those images look like they do.

Luckily, our human eyes resolution in center vision is around 60 pixels per degree of the vision field. With maximum allowed theatrical viewing angles of 90 degrees, a merely 5400 pixels can be separated in center vision in the front row. Going for the smallest allowed viewing angle at 36 degrees. Just a little above 2k resolution remains. That in mind, we find the reason for mandated 4k theatrical projection.

Nobody mandated 4K as the bottom-line for digital cinema, that's why most went ahead with 2K, it took ages before even 4K DCPs became the norm rather than the exception. IMHO, it should've always have been 4K though, as I think the DCI rollout was extremely rushed by the studios. Their focus was to get rid of film distribution costs, not on maintaining adequate quality.

It is a discussion that leads nowhere. Film has a specific look, the image on screen is created totally different, as is the function of the light intensity curve. Film following more or less a triangular pattern, twice per image. Digital differs on the type of projection system used. There can be absence of frame flashing, but also sequentially R-G-B-W flashing of the same image. Depending on technology and number of imagers used. All types exist in legitimate DCI systems.

What discussion exactly leads to nowhere, I'm a bit confused here? While your average DCI picture is created entirely differently than that of a 35mm film, a properly projected DCI image still can look very film-like.

What DCI projectors do use color flashing? AFAIK, with the exception of those Chinese Espedeo SupraSomething boxes, all are using a discrete RGB imager without flashing in non-stereoscopic modes?

[Christos Gartaganis]

The image I downloaded shows 24ish, line pairs (depending on the target). But this is, no doubt, a compressed jpg.

This is also not a projected pattern but is likely a scan of an original. At what resolution does one need to scan it to get the 85 line pairs as distinct lines (and note, that is not the limit of film projection. NTAV's target went up to 100 and I had no problem resolving those.

Then, put the scan through the whole DCP process (make a DCP of it and what is the maximum line pairs the projector can resolve (2K and 4K)?

-Steve

This is the original SMPTE PA-35 test film (aka RP-40) digitized using a Lasergraphics Director film scanner featuring a full bandwidth 6.5K black and white sensor. Even though this a B/W film it was scanned in full color via RGB sequential exposure. You can download the uncompressed TIFF image file using the link below:

https://we.tl/t-5yyWmgGWX3uA6QLD

[Steve Guttag]

Well that is a lot better. If I was judging that image in a theatre, I would give it a 56 lp/mm rating (barely). So, how much further in resolution does one have to go to get the 85 lp/mm? And, if you were to make a 4K DCP of it...what would it look like?

[Christos Gartaganis]

You could simply crop the TIFF file at exactly 4096 x 2160 pixels, convert it to a png file and then import the png file as a test pattern on a 4K digital projector. Then select it from the test pattern menu and walk down the auditorium to check the actual resolution limit. In fact this is a test I have never done before and I am seriously thinking of trying this tomorrow on a DP4K-18BLPHC projector.

I am also impressed by the scanner's optics since there is no any trace of chromatic aberration on the scanned image file.

[Marcel Birgelen]

Yeah, but that's faking it.

I'm really too lazy and burned out, otherwise I'd made a DCP of a few minutes from it... but it will be interesting to see what JPEG2000 compression and several mediablocks do with it before it hits the light engine.

[Ryan Gallagher]

If I hadn't just shown our last DCP before closure I would have been tempted to see how that scan looks in our 2K with periscope mirrors setup.

Something to make for seeing how well our new 4k setup can be overlaid in our 35mm masking sizes at minimum. Same test pattern on both systems would be a "nice to have".

[Christos Gartaganis]

For those interesting in creating their own DCPs for further testing on both 4K and 2K digital projectors you may download a 30 sec. Prores 4:4:4 clip created by the original 6.5K TIFF files using the link below:

https://we.tl/t-exzya831VYgeBiG7