Question concerning K resolution and 35mm film

[Erik Lokegaard]

How high K resolution does the full 35mm camera aperture, 1.33 : 1, silent film, amount to? How high K resolution do 35mm negatives have to offer when it comes to digital intermediates? I assume the latter varies depending on which aspect ratio is being used.

 

[David Mullen ASC]

Since "K" describes horizontal resolution only, aspect ratio factors less into it, just the width of the film negative, which is usually either Sound / Academy or Silent / Full / Super for 35mm. 35mm Full Aperture is roughly 24mm wide and generally should be scanned at 4K across (4096 pixels) in order to capture all the detail in the frame, including grain. But there's a slight difference between optimal scanning resolution and measurable scene resolution -- that can vary.

Also, if you're trying to compare this to the RED, remember that a scanner scans each color channel at the selected resolution, whereas a single-sensor camera has to generate three colors from a single monochrome frame with a color array pattern.

Take a look at the examples between a 2K and 4K scan of 35mm in this article:
http://digitalcontentproducer.com/mag/video_digital_cinemas_special/index.html

Remember, film does not have pixels, so you can scan it at 100K if you wanted to, the only question is when does it become a waste of time and money because you've already captured all the detail on the piece of film.

There's also loss of resolution in the generations from printing negative to projection prints, through projection, etc. but to me, this argues with starting out at a higher resolution rather than arguing that you might as well just start with a lower resolution.

But some have argued that a 4K D.I. where the release prints are made from IP / IN dupes has about the same resolution as a 2K D.I. where all the release prints are made from the original filmed-out negative. And also that 2K digital projection is on par with a 35mm contact print off of the original negative, and thus higher in resolution than release prints made off of IP/IN's.

 

[Dan Hudgins]

Sync speed or not

Sync speed or not

The problem with talking about 35mm movie film resolution is that it is distributed over more than one frame when running at full sync speed.

When looking at a freeze frame it is about twice as granny and half as sharp or in resolution viewed as a still frame. So if you look at a scan of a single frame you get a false measurement of what it looks like running through a projector is a darkened room (you persistence of vision is longer on a dim projection and in a dark room, so you intergrate more frames).

On our scanner which works at near 4K the grains in the film are about 3 to 5 pixels wide, so the high MTF resolution in a single frame is about 1280x720. Running in a projector in focus you might get 150% to 200% as much detail, but the projectors are so much out of focus because there is no projectionist any longer that you cannot get even close to 2K on the screen at above 50% MTF much of the time. Not to say that details at 5% MTF do not help make the image look better, but if you sit back from the screen, you could get by without them.

I saw a 2K cinevator print of CHE #1 and #2, and the projector was weaving in and out of focus much of the time, so while it is good to shoot 4K and to scan 4K to hold what the camera lens puts down, the resolution you see on the screen depends to a great deal on if the projector is in focus, and how good a projector lens us used since many might not be showing better than 20lp/mm to 35lp/mm being a little out of focus.

Mostly you want to shoot 4K to reduce sensor aliasing and chroma moire, the OLPF cuts the resolution some, so you do not get single pixel resolution on a Digital Cinema camera or 4K film scan.

 

[David Mullen ASC]

The problem with talking about 35mm movie film resolution is that it is distributed over more than one frame when running at full sync speed.

Yes, I wonder if that contributes to the sense the 35mm film has more resolution than it measures on a single-frame, since detail is stored as grains, which are randomly structured on each frame, so perhaps the perception of detail increases as the images are projected in sequence, whereas the fixed grid structure of a digital sensor may not benefit as much from this "temporal resolution" as some people call it.

 

[Erik Lokegaard]

Thanks.
Does one know approx. which K scan would capture all of the information on a 35mm negative?

 

[David Mullen ASC]

Jeez, didn't I answer that in my first post? Did you read the article I put a link to?

4K is generally considered the optimal scanning resolution for 35mm cine. A few people say 6K, particularly for 35mm anamorphic photography shot outdoors on slower film stock, but that's debatable.

Most scanning these days is either done at 4K but downsampled to 2K immediately, or at 2K.

 

[Erik Lokegaard]

I clearly gave the wrong impression in my previous post, I'm not implying that you didn't. I see I didn't express myself clearly.

Remember, film does not have pixels, so you can scan it at 100K if you wanted to, the only question is when does it become a waste of time and money because you've already captured all the detail on the piece of film.

I just wondered if there would be some image info detectable by, say, a 5K scan, not obtainable by a 4K scan, but I get the point, 4K is the desirable D.I. if cost effective.

Your effort is appreciated. A nice article, thank you.

 

[Dan Hudgins]

"temporal resolution"

"temporal resolution"

Yes, I wonder if that contributes to the sense the 35mm film has more resolution than it measures on a single-frame, since detail is stored as grains, which are randomly structured on each frame, so perhaps the perception of detail increases as the images are projected in sequence, whereas the fixed grid structure of a digital sensor may not benefit as much from this "temporal resolution" as some people call it.

I would think so, I can see the image transform like above 16fps when I use my "bull-s-eye" Moviola, above a critical speed you can see more details in the merging grain. Your eyes track the details, so it is had to simulate since the mis-registration does not let you just fuse frames together, you would loose resolution that way, unless you compensate for the frame to frame image movements.

Also when things move and are shot with a uncompressed Digital Cinema camera your eyes track sub-pixel resolution as small details cross from one pixel to another, works better on a 3-chip camera since the OLPF would reduce the pixel-to-pixel image movements.

The cells in your retina are not in a grid pattern, they are more like film grains in arrangement so do not alias with "runners" like a fixed grid pattern used in the Bayer filter and sensor.

I have thought you could move the sensor around and take many high fps frames to make the grid issues less, but the projector or your eye would need to follow the image details or keep them still while the grid moved around at high speed, then it would be more like film, you could move the CMOS and DLP on Piezo so that they track movements, and shoot at maybe 120fps.

Video had some wiggle in the interlace before digital sensors, and so has lost the video look since compression and sensors came in, analog scanning had some motion of the raster past the details focused by the lens, and your eyes would track the details as the move from one scan line to another. I miss the "window" like quality video had before the digital era...

I think some people have said that watching films is more interesting than watching digital video on a screen, that may be because your eyes need to track the image around, and the trend to shoot video without a tripod?

 

[Dan Hudgins]

Tonal resolution

Tonal resolution

I just wondered if there would be some image info detectable by, say, a 5K scan, not obtainable by a 4K scan, but I get the point, 4K is the desirable D.I. if cost effective.

There is something more important than scanning film at 5K or 6K, it is the tonal resolution. The digital camera used to scan with has various noise that can be averaged out if you make many scans and fuse them to a high bit format like 16bit TIF (48bpp). 10bit DPX (30bpp) do not have enough tonal range to capture everything that a negative shot at high ratio holds, more so for heavy black and white negatives, maybe.

So the type of camera used to scan, and the file type used can impact how much detail you can later extract from the file. My tests have shown that 3 exposures are better than 1 with the camera we are using, 9 are better than 3 for color seperations, and the upper limit of useful gain would be about 27 exposures per frame.

I would think that 16K would be the high limit for 35mm negatives to archive them for all time to come, and at least 48bpp of averaged exposures shot with slight exposure bracket. You would not be able to see much of that in a film print projected, but you could see some of it looking at high resolution digital images with the tonal range expanded. If the idea is to burn all the negatives, then 16K with 48bpp would be good enough. At 16K you image each grain and dye blob, along with the dust and scratches, so you are archiving more than the "pictures" you are making a record of everything that the film holds. You are also recording the shape of the circle of confusion of the camera lens, not just a "point" so you could use computers someday to bring out details that film printing could not.

 

[Erik Lokegaard]

Great, thanks.
Do you know what the upper limit for 15 perf. 65mm film, IMAX, is? And 5 perf. 65mm? What is the highest K scan available today?

 

[Reinhold R.]

There's a great article from ARRI that covers most of the basics about DI and resolution of film stock.

http://www.arri.de/camera/tutorials/4k_systems_theory_basics_for_motion_picture_imaging.html

Check out the PDF at the bottom of the page.

-Reinhold

 

[Radoslav Karapetkov]

A slight off-topic... (but still, connected to the topic of perceived resolution).

How would you comment what James Cameron said about 2K and 4K?

He said that 2K shot and projected at 48 fps looks as sharp as 4K shot and projected at 24 fps.

 

[Dan Hudgins]

Max

Max

Great, thanks.
Do you know what the upper limit for 15 perf. 65mm film, IMAX, is? And 5 perf. 65mm? What is the highest K scan available today?

You can work out the resolution required for film by the square mm, you can look at some Super16 scans I did, they were not noise filtered, so you can see what film looks like in a RAW scan,

http://reduser.net/forum/showthread.php?p=401086#post401086

Note: links may not work because of the crash of reduser, let me know and I might be able to fix it etc.

The max file size for JPG files was reduced (?) to 97.7KB it seems, which it too low for viewing the images, so you can email me if you want to look at them and I can send them to you. Not even 1MB for a 1K image is WTF?

The focus is a bit off on those, but you can see the grain issues. The better you get the color separation the more you can see the grain in color film. If those were for use I would stop the lens down to soften the grain, the printing Nikkor can resolve the grain quite sharp, the lens I used was the one used for JFK to blow up the 8mm Zapruder film (so we were told), if you remember how well the grain was resolved in that.

Our scanner used Oxberry parts. So if you make one like it for IMAX you can use a 4x5 digital back as the camera, or a 2 1/4 camera back and get more resolution, if you make several scans of each frame and fuse them, you can get smoother tonal ranges. My system has such a fuse option that works with color seperation exposures. You should not use the Bayer filter for color scans as the results are not as good as using colored LED or seperation filters like Wratten 98, 99, and 70 along with 2E and IR cut.

Here is a link to high resolution backs,

http://www.photographyreview.com/cat/cameras/digital-cameras/digital-camera-backs/PLS_4287crx.aspx

http://www.photographyreview.com/ca...ra-backs/better-light/PRD_129513_4287crx.aspx

You also need a VERY good lens to get the job done...

You can also use a "line scan" camera like used in a flat bed scanner, some of those are very high resolution, you just use the right lens to focus on the moving line array sensor, and make several passes to get the colors, IR scan for dust removal, and extra scans for tonal range and noise fusing.

 

[David Mullen ASC]

A slight off-topic... (but still, connected to the topic of perceived resolution).

How would you comment what James Cameron said about 2K and 4K?

He said that 2K shot and projected at 48 fps looks as sharp as 4K shot and projected at 24 fps.

Certainly is possible for scenes with movement, not for static shots. Though of course, you then get into the debate as to whether 48 fps motion looks "video-ish" or not for some viewers.

As for IMAX, as a frame of reference, "Dark Knight" scanned their IMAX footage for the 35mm anamorphic release version and for IMAX visual effects at 8K. Now the Cinefex article on the movie said that they felt at IMAX was actually 15K but that wasn't practical.

You could think of it this way, if a 24mm wide piece of film (Super-35) needs to be scanned at 4K, then since a 70mm wide piece of film (IMAX) is 3X wider, it should be scanned at 12K.

I think they used a modified 4K Northlight scanner for "The Dark Knight", which probably means some sort of half-pixel offset dual-scan to build an 8K frame.

 

[Erik Lokegaard]

You people are great for the scope of your posts.

 

[Jonathan Payne]

Dan,

All of your image links on the post linked to are indeed broken it seems.

 

[Dan Hudgins]

Fixed link, email for images?

Fixed link, email for images?

Dan,

All of your image links on the post linked to are indeed broken it seems.

I fixed the link buy looking in my post log and using google to find the post date, but I could not repost the JPG images since the image size is now (reduced?) to 97.7KB each?

It would seem the limit should be more like 1MB for a 1K image to be viewed for critical review of image quality, would it not?

If you want to see the scans you can email me and I can email them to you, they are about 300KB each.

 

[Evangelos Achillopoulos]

It is interesting to read this study also for that subject...

http://www.motionfx.gr/Files/35mm_resolution_english.pdf

 

[Radoslav Karapetkov]

Certainly is possible for scenes with movement, not for static shots. Though of course, you then get into the debate as to whether 48 fps motion looks "video-ish" or not for some viewers.

As for IMAX, as a frame of reference, "Dark Knight" scanned their IMAX footage for the 35mm anamorphic release version and for IMAX visual effects at 8K. Now the Cinefex article on the movie said that they felt at IMAX was actually 15K but that wasn't practical.

You could think of it this way, if a 24mm wide piece of film (Super-35) needs to be scanned at 4K, then since a 70mm wide piece of film (IMAX) is 3X wider, it should be scanned at 12K.

I think they used a modified 4K Northlight scanner for "The Dark Knight", which probably means some sort of half-pixel offset dual-scan to build an 8K frame.

Thank you for the explanation.

 

[Moskito]

where to scan?

where to scan?

Could anyone advise on a good & reasonably priced facility anywhere in Europe to have 35 mm film scanned to 2K or 4K? I have appr. 1 hour of footage, I'm not really in a hurry. Thanks.