35 mm film resolution

[JD Holloway]

A couple of years ago a number of indepentent testers attempted to determine the actual resolution of a typical "theatrical distribution" film.

http://www.cst.fr/IMG/pdf/35mm_resolution_english.pdf

Would somebody like to comment on 750 lines average they appear to have measured? I'm interest in assessing a number in % that a super-clean 4K digital aquisition/projection would be in comparison.

Or is this apples to oranges.

Thanks in advance.

J.

[Petr Dvorak]

from 4K digital data you get direct 4K resolution on 4K digital projector, pixel per pixel

[Eirik Tyrihjel]

from 4K digital data you get direct 4K resolution on 4K digital projector, pixel per pixel

Not necessarily, I am not into all the tech stuff, and spesifications (and probably shouldn´t post this...), but as far as I understand the way of the CMOS censor (in the RED) also has some compromises to it, and depending on how you judge it, if it was put to a similar test - it would come out with less resolution than what is claimed by the RED team (I hope not, but I would be suprised if it wasn´t so).

Having said that, I am sure that RED will deliver an image with a resolution that could compete (and probably do better) with 35mm anyday, then there is all the other factors...

RED will provide me with a camera that will do most of the things a 35mm film camera will do (some slightly better, some slightly worse), but now I don´t have to buy expensive stock, and develop it, and then have it run through an ungodly expensive telecine to work with it. I am ready for RED!

[Brook Willard]

This (http://www.reduser.net/forum/showthread.php?t=1487) has some resolution comparisons from a fresh negative... no prints, though.

[JD Holloway]

Agreed,
I love your FAQ.

But the point of the test was to determine exactly what the heck you're seeing in the theatre. Your test, as brutal as it is, is almost a "best case scenario". I hate most of the projected prints I see, and I live in a part of the world that has decent projectors/projectionists.

My point in linking this article is to have people examine the methodology of the test and make sure its on the level....and draw some conclusions...maybe start some fires.

[J. Bernard Vallon]

I've heard someone say that Kodak says their sharpest stock is currently somewhere between 80lpmm and 100lpmm (if i remember correctly, the post said the card they tested with didn't have anything between those values, 80 and 100, and it must be somewhere in between).

Therefore, if you do the math, a S16 frame has a horizontal width of 11.1mm, the horizontal res would be between 888 and 1110. Most s16 is scanned at 2k, because of nyquist's theorem (being that to transfer mediums, you must double the scanned resolution to not loose data in the transfer).

s35mm has a horizontal frame of 24.4, so the math says 1952 to 2440. Again, 2k scan is enough for a beautiful image, but 4k is theoretically lossless. As far as this 6k and 8k stuff is concerned, i'm pretty skeptical.

[Brook Willard]

Oh, of course, it's actually a very interesting report and I'd love to do tests on it personally. I just wanted to pop that info in here in case people were wondering about negative resolution.

I'd love to get my hands on a 35mm print from a photochemical/optical finish... I could get it scanned and check it out up close.

[Alexander Black]

I wish I could remember where I read this, but I do recall reading that prints are roughly 1.5k if done properly. I know Pixar records sth like 1600x900 to film after exhaustive tests at resolutions I believe as high as 6k. They just couldn't see a significant improvement in image quality. Of course, they have absolute control over the incoming signal, so they can probably afford the lower resolution and still get gorgeous prints. From "real" footage I don't believe (for example) 6k would be useful unless you were recording to a large format like IMAX.

It's apples to oranges - but you can measure effective resolution accurately.

Anyway, I think resolution is really beside the point once you get past 2k. 4k is great, don't get me wrong, but past 2k my concerns shift toward highlight handling, sensor size (red has this down nicely), color, dynamic range, etc.

That is actually my only "disappointment" at the footage I saw - which was otherwise incredible - the highlights still had that nasty saturation halo around them. Alpha cameras, sure, but I have seen that on every other sensor except the foveon (though I haven't done any extensive testing since I don't own one of those Sigmas).

That's actually the only negative thing I have to say about the entire demo, which, overall, is extremely minor.

It would be very good fun to do a comparison between a red and any-old film still camera to have a look at the way highlights are handled. I've never done a real (nerdy) comparison. Would be interesting.

_a

[david farland]

J.
This doc has been raised previously here (http://www.reduser.net/forum/showpost.php?p=15204&postcount=22).

This doc goes into a little more/clear detail using same/related data here (http://www.reduser.net/forum/showpost.php?p=15267&postcount=28).

And there's another thread on Reduser that talks about actual resolutions across theatre auditoriums, somewhere?

Cheers,

[GlennChan]

Perceived/subjective resolution is determined by factors other than MTF (modulation transfer function)... in particular, contrast plays a big part in perceived resolution. AFAIK, film currently has a higher contrast ratio than digital projection... so it may appear as slightly sharper than it actually is.

Noise and film grain also increase perceived sharpness.

Movement in the film may decrease perceivable resolution.

2- A great article on film resolution can be found at:
http://etconsult.com/papers/Technical%20Issues%20in%20Cinema%20Resolution.pdf

[JD Holloway]

J.
This doc has been raised previously here (http://www.reduser.net/forum/showpost.php?p=15204&postcount=22).

This doc goes into a little more/clear detail using same/related data here (http://www.reduser.net/forum/showpost.php?p=15267&postcount=28).

And there's another thread on Reduser that talks about actual resolutions across theatre auditoriums, somewhere?
Cheers,

Yup, I had to dredge it up again. Something about the Film vs Red post bonanza.
I posted on it way back on the old forum too or CML (cant remember). If you read my link it lists areas they tested in by city. Several independant groups performed similar tests at the same time, testing around the globe I believe.

But is 750 Lines/PH= 750P? in this test?

[JD Holloway]

I've heard someone say that Kodak says their sharpest stock is currently somewhere between 80lpmm and 100lpmm (if i remember correctly, the post said the card they tested with didn't have anything between those values, 80 and 100, and it must be somewhere in between).

Therefore, if you do the math, a S16 frame has a horizontal width of 11.1mm, the horizontal res would be between 888 and 1110. Most s16 is scanned at 2k, because of nyquist's theorem (being that to transfer mediums, you must double the scanned resolution to not loose data in the transfer).

s35mm has a horizontal frame of 24.4, so the math says 1952 to 2440. Again, 2k scan is enough for a beautiful image, but 4k is theoretically lossless. As far as this 6k and 8k stuff is concerned, i'm pretty skeptical.

Stock is one thing. Projection is everything. Many factors/processes before the light hits the screen.

[JD Holloway]

Perceived/subjective resolution is determined by factors other than MTF (modulation transfer function)... in particular, contrast plays a big part in perceived resolution. AFAIK, film currently has a higher contrast ratio than digital projection... so it may appear as slightly sharper than it actually is.

Noise and film grain also increase perceived sharpness.

Movement in the film may decrease perceivable resolution.

2- A great article on film resolution can be found at:
http://etconsult.com/papers/Technical%20Issues%20in%20Cinema%20Resolution.pdf

1) Agreed about contrast but film neg higher than digital? I'm guessing you must mean high con projection print.

2) noise and grain...counter intuitive but I've heard this before, at least on edges.

3) For sure! And dont drop that hard drive, the images will lose focus!

[Jorge Díaz-Amador]

I've heard someone say that Kodak says their sharpest stock is currently somewhere between 80lpmm and 100lpmm (if i remember correctly, the post said the card they tested with didn't have anything between those values, 80 and 100, and it must be somewhere in between).

Therefore, if you do the math, a S16 frame has a horizontal width of 11.1mm, the horizontal res would be between 888 and 1110. Most s16 is scanned at 2k, because of nyquist's theorem (being that to transfer mediums, you must double the scanned resolution to not loose data in the transfer).

s35mm has a horizontal frame of 24.4, so the math says 1952 to 2440. Again, 2k scan is enough for a beautiful image, but 4k is theoretically lossless. As far as this 6k and 8k stuff is concerned, i'm pretty skeptical.

I belive that was me (sounds like what I said). Some negative stocks might go higher than 80 lp/mm but probably not more than 100 lp/mm.

Also, you are making a very common error: Film resolution is measured in LINE PAIRS per milimeter, whereas digital resolution is measured in individual PIXELS. It takes TWO pixels to image a line pair. Also the width of the full Super 16 image area is 12.35mm (I should know having spark eroded many stock 16SR gates to that size).

So the full width of an S16mm negative, at 80 lp/mm resolution would equal 1976 pixels:
80 lp/mm = 160 pixel/mm x 12.35mm = 1976 pixels
Actually a digital scan width would probably be 12.00mm wide so:
80 lp/mm = 160 pixel/mm x 12.00mm = 1920 PIXELS... haha! Super 16mm = HD resolution!
(I'm just laughing because for years I claimed on CML and elsewhere that S16mm was equal to full uncompressed 1080p HD and I got flamed).

As for 35mm, lets assume a scan width of 24.4mm so it's the exact same size as RED (although Super 35mm width can be between 24 and 24.9mm)
80 lp/mm = 160 pixel/mm x 24.4mm = 3904 pixels

SO... why do you think hard core film guys like me are excited by RED? Because it's the first time we've seen real 35mm equivalent resolution from a digital camera (among other reasons).

Also, as far as the troublesome Mr. Nyquist, that theorem is related to digitally sampling waveforms. You do not have to double the sampling. The problem with CCD/CMOS sensors is, if a detail meets or exceeds the pixel pitch of the sensor, you will get a false visual artifact (moire pattern).

Most digital cameras (definitely the Sony CineAlta) use an optical low pass filter (OLPF). This reduces the response of the chip at maximum detail (1920 horizontal) to ZERO. I don't know if RED uses an OLPF, but since their pixel pitch is equal to about 90 lp/mm, I don't think one would be needed at that high a frequency.

Jorge Diaz-Amador
CinemaTechnic, Inc.
Miami, FL USA

[Bruce Allen]

Red uses an OLPF. Very-high-megapixel still cameras use one too, so I guess it's necessary.

Bruce Allen
www.boacinema.com

[GlennChan]

JD:

1- Hmm I may have butchered the term contrast... sometimes it means different things.

1a- Ratio between brightest output to darkest output = contrast ratio. Film projection is better than digital AFAIK. The Sony 4k is apparently worse than other projectors (I personally wasn't too impressed with the milky blacks on the Sony 4k).

1b- Perceived contrast. Things like unsharp mask in Photoshop increase this.

2- Grain demo:
http://www.cambridgeincolour.com/tutorials/sharpness.htm

[Andrew M.]

So the full width of an S16mm negative, at 80 lp/mm resolution would equal 1976 pixels:
80 lp/mm = 160 pixel/mm x 12.35mm = 1976 pixels
Actually a digital scan width would probably be 12.00mm wide so:
80 lp/mm = 160 pixel/mm x 12.00mm = 1920 PIXELS... haha! Super 16mm = HD resolution!
(I'm just laughing because for years I claimed on CML and elsewhere that S16mm was equal to full uncompressed 1080p HD and I got flamed).

As for 35mm, lets assume a scan width of 24.4mm so it's the exact same size as RED (although Super 35mm width can be between 24 and 24.9mm)
80 lp/mm = 160 pixel/mm x 24.4mm = 3904 pixels

Hey! again you forgot that your above data is for 35mm negatives, not projections prints.
Also most digital specs we quote here is for MTF=50
AT MTF=10 give me a break, who wants to count lines at such low contrast of 10%
If your system can’t pull off at minimum 40% then don’t count these lines.

600 lines per 35mm is the maximum you can get on the projection screen (not projection film)
Remember, projectors lenses and projection jitter are lowering resolving power as well.


Excellent articles posted here from international French and US/Montreal tests.
It is how we should do the test, on the projection screen not on negatives or even projection film.

RED data will be 100% the same on the non projection screen so don’t compare apples to oranges. Even if we use Sony projector we already loose 30% of resolving power.
(Sony is ~3.7K plus loss on the lenses)

[Martin Drew]

Jorge

The RedOne definitely uses an OLPF, Graeme has confirmed that a number of times, eg here:

http://www.reduser.net/forum/showpost.php?p=4190&postcount=5

M

[Andrew M.]

Yes, and this OLPF will bring the resolving power of the sensor quite considerably.
If we get 1900 (950 white and 950 black) horizontal lines resolution taken from chart test in worse conditions and at MTF 50% I will be very happy.

[Jorge Díaz-Amador]

Hey! again you forgot that your above data is for 35mm negatives, not projections prints.
Also most digital specs we quote here is for MTF=50
AT MTF=10 give me a break, who wants to count lines at such low contrast of 10%
If your system can’t pull off at minimum 40% then don’t count these lines.


Andrew,

I didn't forget, but I did jump in to this thread rather than reading from the beginning so I must have missed that...

Traditionally 35mm prints are 4th generation so there are losses in each of the three contact print generations. And if they do a bad job of making the prints it can really deteriorate. But today more features are digital inermediate so theoretically each print can be an original off the laser recorder. Obviously this is cost prohibitive today.

I agree that you have to set a cut-off. 40% MTF makes sense, although it's a tough standard. Sony uses 35% for the HDCAM CineAlta (only 1550 horizontal lines). Where did you post figures for 50% MTF?

Yes, 10% response is very low, although I understand the human eye can detect as low as 2%. The ARRI scans that I saw (trying to find the URL for those) were not MTF tests so I can't say what the resonse was other than it was visible. I'd guess the 100 lp/mm was responding around 2% as you could just barely see the lines.

It seems that a lot of this has been discussed before. Could you direct me to the relevant posts on these issues?

Jorge Diaz-Amador
CinemaTechnic

[Andrew M.]

We tested Zeiss and Canons with 50% MTF cut off.
use search option->show posts
resolving power
resolution
You will get all posts out there about it here:-)

[oldphart]

...
Also, you are making a very common error: Film resolution is measured in LINE PAIRS per milimeter, whereas digital resolution is measured in individual PIXELS. It takes TWO pixels to image a line pair.


Actually, it takes four - unless you happen to be perfectly aligned. Friend Nyquist again. But you will be very hard put to come up with a camera/lens/film combination that actually resolves 100 lp/mm at any reasonable contrast.

The conclusion, however, is that perfect S16 will be noticeably degraded by transfer to 1080P.