A Digital Imax Camera

[Antti Näyhä]

I like how the test made sure all the IMAX material was 4th generation; as we are well aware most IMAX content is 2nd generation contact printing.

I for one wasn’t. That would certainly explain some of the results.

 

BTW, I saw Wild Ocean in 15/70mm, and it was far from being the sharpest example of IMAX I’ve seen. While watching it, I actually thought that some of it must have been 8/70mm and/or 10/70mm. For what it’s worth, IMDb says it was indeed shot on 15/70mm but went through a DI (4K?), which is probably the reason why it was softer than usual.

[David Mullen ASC]

The first true 15-perf 65mm IMAX movie I saw that went through a D.I. was "Hurricane in the Bayou" - it was a bit disappointing, like the edges had been dulled.

[John Holland]

DI what ever the capture image is sucks still big time .

[Antti Näyhä]

It says in the article:

 

The Wild Oceans footage was filmed out from a 4K DI using a CRT film recorder (the only type that is available for 70mm film)

Now, does that imply that 4K is the highest film-out resolution for 70mm?

[K Borowski]

why do people get so upset over discussions about resolution? cant anyone calmly disagree?

 

DAMIEN: I admit I'm biased. I work in a film lab. I roll 35mm film every day, and I am addicted to it. When 35mm film gets surpassed by digital in resolution, dynamic range, versatility, I will cry, drink a fifth, and convert.

 

Until then I shall passionately fight against any step backwards in quality that we take, regardless of this cost, ease of use BS.

[K Borowski]

I for one wasn’t. That would certainly explain some of the results.

 

 

Even with an extra two generations, that would cost you, what 20-30% resolution extra? That should still provide ample resolution over 4K from an IMAX frame.

[georg lamshöft]

Until then I shall passionately fight against any step backwards in quality that we take, regardless of this cost, ease of use BS

 

I think this is what causes the "enthusiasm" and "emotion" of these discussions - I feel the same way.

 

It's not about "K" but MTF and other artifacts. A bayer-sourced image is unusable in it's original size for the big screen. In cinematography we cannot carefully sharpen every single frame and remove all the artifacts caused by color interpolation. We have to downsample it, even if it reduces MTF or countable lines on a test pattern near extinct resolution.

A film negative on the other hand has to be scanned and processed professionally to minimize the quality loss - I personally think that an oversampled Arriscan-scan (?) which is carefully degrained looks better than a normal 1st gen optical copy. Here you can see two nice samples: http://www.filmaufzeichnung-hamburg.de/media/pdf/Testscan.pdf

I think everybody has to agree that the cherry-tree 100% crop (2k) looks way better than the still-photography-samples done with a consumer-scanner.

4k is not tack-sharp anymore, because MTF is too low close to extinct resolution - a 4k-image from 65mm will look clearly superior while having a not-too different extinct resolution when shooting a test-chart.

Is 4k with high-MTF and little artifacts (sourced from an hypothetical 65mm-Alexa or 6k->4k 65mm Arriscan) comparable to IMAX with optical prints? In terms of extinct resolution - I don't think so. But on the big-screen, without looking at still wide-angle-shoots of landscapes all the time, with a next-gen DLP-beamer, our eyes might be the limitation...

 

The fact is: our technology is not ready yet, we cannot process sufficient data-rates for uncompressed 6k-RAW practically and we cannot squeeze this many photosites on a small sensor

without loosing the just desperately achieved "looks-nearly-like-film" IQ from Alexa. Using a larger sensor? And managing to get new lenses for this format? Bigger and heavier again? Even if this would happen, we still cannot manufacture sensors beyond 30mm wide without "stitching" - which leads to further problems, which might be manageable in still photography, but not for cinematography, not without "stepping backwards" in many ways again.

 

I think we should be happy with the 2k-digital-quality we just achieved and the options of 35mm/65mm/Imax - keeping them alive and fight for them. Digital 70mm-IMAX-quality will be possible, but not yet. When we scream for "more K" any longer, we will propably end up in "megapixel race" again: large, unhandy images without justifiable IQ-gain.

Edited July 23, 2011 by georg lamshöft

[K Borowski]

I think this is what causes the "enthusiasm" and "emotion" of these discussions - I feel the same way.

 

Georg :-)

 

 

 

As far as an OPTICAL copy, are you talking about a contact print, or an actual optical (through a lens) copy? I'd agree with you a 4K, even a 2K will be better than optical, but NOT with a contact print. Those are very sharp.

 

As far as my percentages, that is a guess. I don't have access to the equipment to measure myself. I do a lot of 1:1 optical printing and I know that this method looses a lot. I actually HATE optical printing, and wishI could transition everything to optical; the equipment just isn't available.

[georg lamshöft]

I have to admit that I only had the chance once to see a 1st gen copy - I think it was contact printed - but I'm not sure. As far as I understood it, a 1st contact print was the benchmark for the development of the ARRISCAN, not just choosing a DI because it's simpler or fits into the workflow but to actually maintain as much possible from the original in comparison to an optical/analogue process.

But don't get me started about real-world prints in movie theatres...

Edited July 23, 2011 by georg lamshöft

[David Mullen ASC]

There are a lot of conflicting ideas going on regarding IMAX resolution, but keep in mind that the 15-perf 65mm negative itself is 2.9X wider than Super-35 film. So even if you picked a really low value for Super-35 image resolution, like 2K, you'd have to accept that IMAX should be able to deliver nearly 6K. At the high end, if you believe Super-35 can resolve 4K, then you'd think that an IMAX negative was resolving nearly 12K. I tend to be in the middle, I think Super-35 resolves around 3.5K at best but should be scanned at 4 to 6K therefore... which suggests that an IMAX negative resolves around 10K but probably should be scanned at 15K or so. On the other hand, large format lenses often do not have as high an MTF (because they don't need to) and other real-world factors may reduce IMAX negative's resolution downwards to a potential 8K to 9K.

 

Now at the other end is printing and then projection... generally we accept that about half the resolution is lost in projecting even a contact print, hence why today most 2K projection feels about on par with the best 35mm print projection, and decent 4K projection of 35mm material feels like the old 70mm blow-ups we used to see.

 

John Galt at Panavision once mentioned putting a 4K chart on a piece of IMAX print film and projecting it in an IMAX theater and most people viewing his test said they saw a uniform grey field, they couldn't see the lines. Now of course I'm skeptical but on the other hand, I'm sure John is reporting the results he got accurately.

 

But if IMAX is about 1.4X the size of regular 70mm, and 4K projection "feels" similar to 70mm projection, you'd want to believe that IMAX projection should be 6K if done digitally, not 4K as IMAX has been planning. 4K projection for IMAX sort of falls into "what can we get away with minimally" line of thinking which goes against the whole purpose of IMAX, which is, frankly, resolution OVERKILL. That was its whole point, excessiveness!

 

Anyway, if we are going to eventually push for a 4K standard for mainstream digital feature production / post-production / projection -- which I think would be ideal -- then digital IMAX has to clearly be much better than that or else what's the point of the IMAX label? It has to feel twice as good, twice as big, as what you are getting in a normal theater space. Which to me means a 6K standard at the minimum for production / post / projection (I know... 6K isn't twice 4K), 8K would be even better, whether or not that is possible or practical. IMAX was never really about being practical afterall.

[K Borowski]

David: I have to disagree with you here. An OPTICAL PRINT looses 50%. Since we all know contact prints are of a higher quality, I can't see, worst case you loosing more than 20% of the resolution. You'll loose values in the D-min, D-max of the neg, but if we are talking about spatial resolution, there's no way a CP's losses are that high.

 

Aren't IMAX prints also printed frame-by-frame, as opposed to being run on a continuous motion contact printer? Where you'd conceivably loose a little bit more is with one of these machines, cranked up to maximum speed to churn film out. I've read an article, I think one posted by someone on here, that shows the horizontal resolution is significantly higher than vertical resolution on a 35mm release print due to these practices.

 

Maybe I am wrong and IMAX prints are churned out as fast as possible as well.

 

 

 

Then again, this is the fault of the industry, not a technical problem with the medium itself.

 

As far as 35mm goes, even at 10% loss per generation, the ONLY form of 35mm content that has conceivably gained resolution since the advent of the DI is the Super 35 blowup process. Scope and Flat movies have lost near 50% of the release print resolution. . . due ENTIRELY TO THE 2K DI. Let's give "credit" where credit is due: poor laboratory quality control, equipment cranked up way past its recommended speeds, and lazy DI, low-res. files.

 

 

I need to look up the image area of IMAX again. I know it is smaller than 58x65mm, which is what I quoted. So my numbers may be a bit high, if only because I forgot the exact negative dimensions. Even 10K with the best glass (say rehoused Hasselblad or Fujinons) may be a bit high. At 4x3 dimensions 8K is 48 megapixels, 6000x8000. I think this agrees with 85lp/mm. I forget if you multiply this number by two for calculations, as there are PAIRS of lines. Then again, plugging that into a 22x18mm scope frame gives a number that is so low it can't be right, think might be less than 2K, so I think in some ways the resolution is adequate, but in other ways it is far too conservative an estimate.

 

I think 8K is *good enough* but agree with you on the excess factor. My experience watching the film "The Dark Knight" were that the camera (or contact print of the camera negative ;-)) was out-resolving what my eye could've seen were I up in the helicopter instead of the IMAX camera.

 

 

 

I want this experience to remain if they expect me to continue to travel to other states to see Chris Nolan movies. . .

 

 

 

EDIT: Missed your comment about 2K. From a 2K file, a *Blu-Ray* is significantly better than a 35mm print, even with just 10% loss per generation, you are down below 1500 lines starting with a 2K master. Using 10% per generation and starting I think it was 3.2K for negative, you were down around 2.5K as the best possible scenario for a 4th generation release print.

 

Still don't understand why 35mm RPs are 4th generation, there should be a high-speed laser film recorder that goes straight onto ECP right now to combat the digital trend in the lab industry. They have just allowed it to die, they aren't doing anything to fight back. At most why not make only 2nd gen copies from 2K files. Even there you are worse than an HD movie delivered on a harddrive, but at least it is close.

Edited July 23, 2011 by K Borowski

[K Borowski]

Hmm, I've found data that suggests 63lp/mm is really really pushing it, but 40lp.mm is a reasonable resolution for 35mm (of course, the things I have read take into account more the THICKNESS of the emulsion than its speed, so clearly there is more going on than just resolving power in determining useful resolution)

 

For some reason the conversion I found is FOUR pixels per line pair. IDK why, reading it five times, but that puts 40 just slightly less than the 85 I saw quoted. Maybe it is pairs of pairs, as I've seen lines per mm quoted as well.

 

 

 

Anyway 40lp/mm x4 = 160 pixels per mm equivalent that comes out to 10,400 on a 65mm horizontal plane. Lens resolution for some reason is 0.7 but as we take digital sensor data as the full number, you'd have to take only two thirds of what any digital sensor going through a lens resolves as well.

 

10K with a slow film, good lens stopped down 2 stops is a pretty good number. So that comes out to 96MP, which I think is at the resolution of the human eye. That agrees with my observation; my eyes are worse than your average viewer's ;-)

 

 

Here's the article, an interesting, comprehendible take on the very complicated physics involved: http://www.theimage.com/photography/photopg1.htm

 

So, by this count, IMAX is selling 4 TIMES as much bullsh**, to offset the 25% resolution of 4K "replacements." I find this upsetting.

[David Mullen ASC]

David: I have to disagree with you here. An OPTICAL PRINT looses 50%.

 

I'm not just talking about generational loss but also loss of resolution from projection -- keep in mind that some people have measured on-screen resolution for 35mm movies be around 1.2K on average (around 750 lines). That's pretty abysmal and probably represents worst-case scenarios which is why I'm more prone to believe it is only a 50% loss by the time a contact print hits the screen, i.e. 2K for 4K material. But this is just a layman's guess, not scientific.

[Phil Rhodes]

Anecdote:

 

I went to see something at the local googol-plex, and it was digitally projected.

 

Between the parts of the screen that were bright, the parts that were dim, the parts that were pink, and the parts that were a sort of fungal shade of bilious green, there may have been some resolution.

 

Odeon, change your lamps more frequently!