Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

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

IMAX is in the process to convert all their theaters from mechanical large format film to digital 2+2K.
 
Which Kurosawa-san picture????! :w00t:
 
Tom said:
I can guarantee you one thing: cost was not the reason for choosing a 4K scan over a 6K for The Dark Knight.

Now, I have no idea if there might be other, DMR-related reasons for scanning 35mm at 4K rather than 6K. Perhaps it has something to do with the de-graining process? I don't know.
Click to expand...

I can't guarantee anything.

My opinion is based on experience and from those I talked to on the studio side.

The reality may be different.
 
Can you post or PM some Yojimbo scans? :) plz.
 
Tom said:
I highly doubt that 35mm is 6K.
If it is, then why was Dark Knight's 35mm anamorphic IP scanned at 4K for the DMR to IMAX 70mm?
Click to expand...
They used the IP and not the ON. IP has less detail than ON.
 
OK guys, I am uploading one frame, but just a small warning - it is 130 MB... The scan is anamorphic 4:3 full frame right of the scanner. I have only added a copyright and use disclaimer. This is small JPEG version:

0000400.jpg


Once the original 16-bit TIFF is uploaded I will post a link here...

EDIT: Here You go:

http://www.ramjetfilms.com/peter/ARRI6K/0000400.tif

My upload is really slow, sorry about that...

:) Peter
 
S16mm S35mm Resolution

S16mm S35mm Resolution

ARRI's test showed that the resolution of existing filmstocks was at least 80 line pair per millimeter, which results in 4k resolution for S35mm, and 2k resolution for S16mm. The was no target for 90 lp/mm, and the 100 lp/mm target was too poorly resolved to count.

Please note that the test was shot with a Zeiss Superspeed Planar (I believe 85mm at T2.8). Shooting with Ultra Primes or Master Primes might have pushed the results to 100 lp/mm.

Those that have been following this debate from years ago may remember that I predicted about the same results on my website. I got a few people over the years e-mailing me saying "No way Super 16mm can be 2k or even 1080p".

As Graeme mentioned, to improve response, oversampling by 50% means S16 should be scanned at 3k and S35 at 6K. This gives a better MTF response from the film negative and captures more detail.

Guys, film still works. And there are plenty of people still using it. That shouldn't bother you. You have chosen your camera, which works quite well and has some very compelling advantages.

ARRI is doing quite well selling new equipment. What has been effectively killed is the low-end film market. The high end film market is alive and well (for how long?)

As for the low end, I will probably not be able to continue supporting the older film cameras (for repairs and modifications). I had already discontinued support for the ARRI 16St and 16BL. The 16SR, 35IIC and 35-3 are probably next. Currently, older film camera owners do not want to spend any money on their cameras, not even routine maintenance. I know I'll get a lot of whining when I make the final decision, but it just does not make sense anymore.

If anyone has one of these older cameras and needs to get it service prior to sale (no one will buy a camera with problems these days - too many to choose from), you can PM me to discuss.
 
Who is bashing film?

But when you have people at cinematography.com claiming (with a straight face) that S35mm is "8K"... reasonable people are going to disagree.
 
Yojimbo vs. RED ONE...

Yojimbo vs. RED ONE...

Tom said:
But when you have people at cinematography.com claiming (with a straight face) that S35mm is "8K"... reasonable people are going to disagree.
Click to expand...

Thanks Peter Majtan for posting an actual ARRI 6K scan from a good frame from an actual feature motion picture, Yojimbo. Since this was a Master Positive rather than the camera negative we can assume it is only 66% to 50% of the camera negative detail, if it was made on a continuous contact printer rather than a ACME step contact printer it could be lower, but better than the final print anyway.

I made a quick side by side with a sample RED ONE (tm) frame from a R3D file on REDRELAY, I made two of the Yojimbo since I wanted to see the eye and teeth at 6K. The "Euro-Guy" is smaller in the frame, so the pixel count for the same face area is smaller, even so to my eye the RED ONE (tm) shot seems to resolve a little more detail (see the teeth in both) and have lower noise, even though it is only a 4K image and the OLPF takes a heavy toll on the resolution. In both images there are no details at the single pixel level that are 80% MTF, that I can see, to the lower left (Yojimbo eye closeup) there is a dust spot that is about 2 1/2 pixels wide.

I had a scan of a pictorial negative that we made on Ortho sound recording film, it can resolve 300lp/mm to 1000lp/mm. Our scanner lens can resolve 200lp/mm to 250lp/mm maybe. I cannot find that file now, but even in that the edges were not computer graphics sharp. One thing I do notice is that in the ARRI 6K scan the film grain looks quite low, so there may be some software filters in the ARRI scanner that blur the image to reduce the grain, I am sure in our scanner (made out of Oxberry printer parts) the grain from the Yojimbo frame would be much more pronounced. I have even taken to stop the lens in our scanner down too far in order to REDUCE the resolution so that the film grain is not clear, otherwise the digital artifacts on the grain make it look even worse. In no image I have been able to scan were details on the film one pixel wide on the 35mm camera negative near 4K digital resolution, most are 3 to 5 pixels wide. If someone has a sharp scrap of 35mm film they could lend me to scan I would like to see what a DIY scanner can capture. The KODAK LAD film I have seems to be just plain out of focus.

I would like to remark that you cannot judge film scans by just one frame since when you watch film in a theator you see three frames superimposed, when film is moving your eyes track the mis-registered images and fuse them, so you see more detail in the running film than you see in a "freeze frame". I can see that in my edit system viewer even at 640x350 resolution, when the uncompressed frames run at 24fps you see almost twice as much detail as when stopped. This improvement when running is more for high grain film.

The attached images are for scientific and educational function and their copyright belongs to their owners.

The Yojimbo being 'Scope would project on a larger wider screen, so the 256x256 zoom sample area would be about the same size on the movie screen in both the flat projected RED ONE (tm) image and the Yojimbo image, maybe the Yojimbo a little smaller on the screen, but even so this is not too far off to make a rough comparison, maybe...

I did not remove the color shift in the Yojimbo frame, you can see it is a little off equal gray, the screen shots are for viewing about gamma 2.2 to 2.4, there is a gamma reference along with the two near white and near balck boxes in the lower right to adjust your monitor and image viewer. On a LCD monitor you may have a hard time holding the highlight and shadow boxes, but on a CRT monitor they should be easy to get... All three images suffer a little from going to JPG...
 
The hardest thing with film is determining what is real detail and what is grain.
 
Compression and true frame rate...

Compression and true frame rate...

Graeme Nattress said:
The hardest thing with film is determining what is real detail and what is grain.
Click to expand...

When images get compressed the true frame rate is less than the fussion frame rate for vision in all parts of the frame if the compression works between frames, so the eyes ability to fuse moving uncompressed film images is lost.

People in the future will not be able to know what film looked like projected as film, just like people today cannot see what "live" video looked like since the frame rate in compressed and filtered video does not look the same as true frame rate in all parts of the frame uncompressed (e.g. image dissector).

One rememberes, if one is old enough, what uncompressed non-digital DI film to film printing looked like, but if you see a frame blow up now, it looks unsharp and has more grain, because it is not running at speed. So many films today have a compressed DI that they look more like digital images so the difference between what non-film-to-film printing looks like and compressed digital projection is less notable, that is the filmout has digital artifacts also, just blured digital artifacts, so the digital projection looks sharper than film, in part because the digital film recorder has blured and artifacted the film projection.

And do I need to add there is no projectionist to keep the film projector in focus...
 
I know what you mean about some compression that makes different areas of the screen look like they're moving at different frame rates. Indeed, static grain patterns were one of the things I noticed first on bad MPEG2 compression when DVD first came out.

Graeme
 
Interesting, DANCAD3D.com.

And let's keep in mind that Red One is not even a true 4K camera - not even close. You would need a 5- or 6K Bayer-pattern camera to produce "True 4K". Let's see how S35mm scans hold up against Epic. :wink:
 
Wobble plate...

Wobble plate...

Tom said:
Interesting, DANCAD3D.com.

And let's keep in mind that Red One is not even a true 4K camera - not even close. You would need a 5- or 6K Bayer-pattern camera to produce "True 4K". Let's see how S35mm scans hold up against Epic. :wink:
Click to expand...

There is a trick that can imporve the digital cinema camera, if you put a wobble plate in front of the sensor and project uncompressed then the image details move from one sensor to two a second then to another and so on, you eye follows the image detail around and you see more detail than in a static image.

Film is always moving around, you cannot scan film and have the details stand still even when you use an optical printer movement like we do.

If you put this weave into the digital cinema it will be more like film.

But if you use compression, it will try to filter out the image weave and hold the details on one pixel, and so reduce the apparent resolution.

If you weave too fast the benefit is lost since you eyes cannot track the moving image, it just looks like motion blur, so there is a critical speed that the image should wobble around near pixels, something about an 8 frame cycle, with 24 frames per second... maybe?

---

The bigger issue is can the camera lens resolve a single pixel at 4K on a 6K sensor without producing aliasing artifacts, that is why you need the heavy OLPF. With a Bayer filter you should have 8K to 12K in order to get 4K after using a OLPF heavy enough to remove all aliasing and chroma moire artifacts, so it is not so much resolution that needs more pixels, but artifact reduction. You cannot get by without the OLPF when you use a sharp lens, or you will get too many artifacts, in other words, the lens used on a Bayer sensor SHOULD not be sharp enough to light up just one of the four pixels in each four pixel group, hence the resolution of the lens should be 1/3 the target pixel count, so to get 4K you use a 12K sensor...

As you see in the sample images above, the camera lens used does not come close to a single pixel resolution, the general rule is that a lens can focus on 1/1000 its focal length, so a 50mm lens makes a spot about 1/500 inch. And the daylight shot may have been stopped down to the point where diffraction limits the resolution to less than 2K? With super good lenses on film you should never stop down past f/11 and probably ONLY shoot at f/6.3 or so.

----

For archive use, it is good to scan the film at 6K pixels though, so that you know that you are holding as much as you can within reason...
 
The real problem is associating number of pixels with resolution -- film doesn't have pixels for one thing so saying it is 3K or 4K or 6K is rather misleading. And more subtly, there is some argument to be made that there is a difference between the measurable resolution of a single frame and the perception of resolution in a moving image where there is some averaging of frames in the human eye/brain.

Not to mention that real-world examples in photography contain all sorts of resolution levels.
 
besides of what is the real resolution of a film scan...
I can remember when I asked arri at the IBC why their digital camera is named D-20.
They said that D stands for digital, and the 20 is because HD is between 16 and 35...
so... now the 16mm is better than the d20?
 
Vision III vs. Vision II?

Vision III vs. Vision II?

polispol said:
now the 16mm is better than the d20?
Click to expand...

Each batch of film is a little different. The layers keep getting thnner as dye couplers get more efficient and less silver is put in to reduce costs, this makes the resolution a little higher. Sound recording film is very sharp (250 to 1000lp/mm) and has a micro thin emulsion layer. If you shoot sound recording film in super 16mm you get:

12.4mm x 7.5mm = 12,400 by 7,500 lp (max) or 3100 x 1875 lp (low contrast), maybe double that value in pixels at low MTF.

We shot some 35mm (processed for low contrast) and it had more detail than could be projected, even if blown up to IMAX it would be hard to see the smaller details that could be seen with a microscope on the negative.

No lens can resolve that much, but if you use the best lens at the best stop, you might get close to the low contrast value.

The lens on a 4K digital cinema camera does not resolve anything close to 4K at 80%MTF unless you sharpen the image, and the OLPF softens the image down to 1920 or less at 80%MTF when a fast lens is wide open. A 1920x1080 Bayer camera needs a OLPF to reduce the resolution to avoid chroma moire and aliasing so it will only produce maybe 960pixels or 480lp/image_width of "real" 80%MTF resolution, so using a 4K camera for HD is a real benefit if they are both Bayer filter cameras. If the HD camera is 3 chip there are other issues, if no OLPF is used you get aliasing artifacts even though the chroma moire may be less.

With improvments to their film scanner and the thiner film layers it is possible that the relative resolution of Super16 to the D-20 Digital cameras could swap. But these differences are small in the final film print projected in a movie house with the projector out of focus... It is hard to keep film flat enough in the gate to get the focus that fine at large stops, and the OLPF reduces the resolution of fast stops, so for real shooting would they (film vs. digital) may look about the same, except the film would have dust, water marks, and scratches, and the digital color/range issues and aliasing?
 
You must log in or register to reply here.
Back
Top