Thanks Adam, clear and concise as always :-)
So much BS in our business.
To the person mentioning Zodiac, yes DigiPrimes, I visited the shoot to learn the approach before I shot Mutant Chronicles using the same kit!
Once I match angle of view and depth of field, I can't find any evidence of "a more natural perspective". I'm not saying categorically that there isn't such a thing, just that I can't see it myself when I carefully control the experiment to equalize differences between the images. As far as I can tell, when you shoot from the same position, with the same angle of view and the same depth of field, the geometric, size, and focus relationships between foreground and background objects are *identical* whether it's large or small format.
To take an extreme example, have a look at the various "iPhone vs. RED" comparisons floating 'round the 'net. These are invariably shot in bright sunlight, so the iPhone is at low gain (minimizing its noise) and the RED's lens is stopped down (to better match depth of field with the tiny, tiny iPhone lens). The RED's sensor diagonal is in the 25 to 46mm range; the iPhone's is around 6mm (https://en.wikipedia.org/wiki/Image_sensor_format). That's WAY more different than S16mm vs. S35mm vs FF! Yet the images from the iPhone and the RED match up fairly well (well, ignoring color rendering and highlight handling, of course).
I'd suggest the preference for the larger format in these cases is due to (a) shallower depth of field, which does come more "naturally" with larger formats, and/or (b) higher image quality as measured by noise / grain levels and/or resolution, which again are more easily obtained with a larger format, and/or (c) "we couldn't quite match the angle of view of the different lenses, so we moved the camera in a bit, and look how much better the FG/BG separation is on the larger format" (I see this a LOT in large-vs-small comparisons, usually combined with depth-of-field differences. Yes: you change the shot composition, and it'll look different!).
True, a lot of this comes "naturally" with larger formats: higher resolution / lower noise, shallower depth of field (from shooting at the same T stop in large format as used in small format work) and different angles of view (typically wider, as DPs gravitate to the same focal lengths they normally use on a smaller format. Thus the camera is moved closer to the subject to keep the same subject size, and you get a "deeper" perspective in the scene). All I'm saying is that, as best I can tell, the "large format look" can be decomposed into these sorts of easily comprehensible individual factors; there's nothing mystical or ineffable about "large format" vs. "small format", and we don't need to resort to arcana such as "roundness" or "naturalness" to explain those differences.
Yes, I agree. However, what's your understanding of the whole "large format/longer lenses increase magnification and lead to 'more natural' perspective and size relationships between foreground and background" thing? I do agree that I prefer the look of larger format images compared to sub-S16 formats, but when I think about what should look "natural", I recall that the actual focal length of the eye is about 14mm, and since that is obviously considered normal, then that would mean the equivalent diagonal of the eye's retina is about 14mm as well. And yet, people generally prefer the look of a 40mm on a full-frame sensor compared to an 18mm on S16.
It's much better than other Panavision format-size-comparison videos I've seen, but it's still a sales pitch for large format, and it throws in some red herrings and insufficiently-qualified assertions to bolster its pitch. It's pretty good through the first half, but then starts confusing the issue. "But what if you want to keep the same focal length and match the subject size?" In other words, let's change the angle of view and move the camera and recompose the shot. OF COURSE it's going to be different: it's not the same shot! You could do the same thing (on either large or small format) by going to a wider lens and moving closer; there's nothing inherent in format size that forces that other than an arbitrary desire to match the numbers on the lenses used, which is a silly thing to insist on. "What happens if we shoot S16mm with an 18mm lens? See how much deeper the depth of field is?" Well, sure, of course: you need to open the iris to match depth of field. Finally, the "you can match smaller formats using large format but you can't do it the other way 'round" assertion is only true at the extremes: if you can't open that 18mm's iris up enough to match the DoF seen on FF with a 55 at T2.8, for example. I'm not trying to be negative; it's a really good large format sales pitch video. I'm just saying it's a sales pitch video, not an "understand the relationships between format size, focal length, T-stop, angle of view, and depth of field" video.
Great writeup, and I thank you for it! Adam, what do you make of this rather thorough video comparison from Panavision: https://vimeo.com/251731362
Thanks, Bob. I am gradually coming to the same conclusion, having looked for differing "roundness" seven ways from Sunday and not being able to find it (except, by its lack, in the quickie 3D conversions of which you speak. And in computational photography, of course).
As with many things in this discussion, one could put together a perfectly logical case, *not disprovable by the available data*, that it's all a plot by Big Format (like Big Oil or Big Tobacco, only with sensors 'n' lenses) to push expensive upgrades upon an unwitting populace using ill-defined terms, combining FUD with FOMO to better flog their kit. However I am not a conspiracy theorist, so I don't *ever* mention such a thing.
And I agree: that Tim Sassoon is a bit of a wizard, and a very pleasant fellow, too.
Hey, Adam, I really enjoyed your iPhone slates :).
As for the 'roundness' of different sized sensors/lenses...
People see what they want to see based on their preconceived notions, which are based on, usually, very little factual data. When pressed to define exactly what they mean, proponents of 'more roundness' are almost always at a loss for words, and the conversation ends when I hear them utter a Supreme Court Justice's definition of pornography: "I know it when I see it."
The only place where I have seen 'roundness' be a thing is in post 3D conversion, where the crappy 'we charge by the movie minute' conversions all look like the items are two dimensional cardboard cutouts, whereas the properly done ones use shading, reflected light and colors from the surroundings, and other tricks of the Jedi Masters (like Tim Sassoon) to make the items appear as if they have some dimension to them. They're not more 'round', but the eye is pleasantly tricked by post finesse.
Neither sensors nor lenses have that 'luxury'.
You won't get any argument from me on those points, grin!
It's also worth noting that the DigiPrimes from 7mm through 70mm are T1.6 lenses (while the 3.5mm, 5mm, and 135mm are T1.9), designed specifically to be used wide-open to better emulate the shallow depth of field found with 35mm primes shot at T4 – T5.6. So a technically savvy filmmaker (and Fincher is nothing if not technically savvy) could readily get the same look at he'd normally get on 35mm with "normal" apertures by using the DigiPrimes wide open on 2/3" instead.
The more I look at the topic, the more I see the "special look" of large format coming down to two things: more resolution (which normally happens with larger format imagers), and shooting at the same numerical aperture (e.g., f/4) regardless of format size (while maintaining the same angle of view, e.g., using a longer lens). The former gives larger format images more pellucid detail (the "smoothness" that Michael Cioni talks about with 8K acquisition), while the latter gives you shallower focus on larger formats. But if you understand that, you can replicate both those factors on smaller formats (within the limits of what physics and practicality allow, of course!) and get the same darned look.
You'd think so; it makes intuitive sense that a larger diameter aperture would have more look-around. But I just now did that test, shooting that same round can with the same lens at f/1.4, f/5.6, and f/16, and I can't for the life of me see any further 'round the curve of the can (or see anything behind the can more revealed around the edges) at wider apertures than at smaller ones. (I don't know how to add pix to a comment; maybe I'll add them into the article when I get a chance).
In the synthetic-aperture, camera array, and light-field cases, you're looking at the same point on the subject from multiple, spatially-dispersed viewpoints. In the fixed-position, conventional, single-camera case, you're not. That's probably why synth-ap / cam array / light-field imaging has more look-around—just as those methods let you "reposition" the viewpoint in post—while in boring old one-camera photography actual look-around difference are hard to come by—and likely for the same reason you can't reposition the viewpoint in post with just one plain old dumb camera.
In my test, I was instead exploring the effect, if any, of the *front element diameter* on the look-around. It's often claimed that big honkin' lenses show more look-around that their more svelte counterparts, so I was trying to see if that was the case, *holding everything else constant*: subject distance, angle of view, and aperture diameter.
Interestingly, neither the front element size nor the aperture size seems to affect look-around, at least not that I can find.
Thanks for pushing the discussion further!
I've been a proponent of large format and have felt excited about all the large format cameras coming out and the new generation of large format lenses from all the majors... however, I recently watched two of David Fincher's films shot on the 2/3" chip digital cameras a decade ago (Zodiac and The Curious Case of Benjamin Button), and they hold up really well, with perfectly cinematic images with nice depth and dimensionality. So after seeing those, and knowing that, at least in the case of Benjamin Button, lenses in the range of 6-24mm (!!!) were used (Zeiss DigiPrimes according to the article in AC magazine), I feel like "cinematic look" is really all about dynamic range, lighting, color reproduction/skin tones, and camera movement/staging.
Regarding the ability of large lenses to "look around", I was thinking that this phenomenon is indeed true, and it's the fundamental reason why large apertures have a shallow depth of field, and why a shallow depth of field can be synthetically created (via computational photography techniques) from either a multi-camera array, or a single camera moved through space in a circle (in an effort to effectively create a large aperture by drawing it in time, see here for more: https://sites.google.com/site/marclevoy/Tutorial). Anyway, the idea is that blurriness created by large apertures is fundamentally a composite of many slightly different points of view from over the surface of the lens/aperture that are overlayed. When a plane is in focus, each point of view is aligned so object features all fall in the same corresponding place on the image plane, and when that happens, the other planes are out of focus because their features are not aligned.
But here's an image from Lytro that shows how a large lens/aperture should be able to achieve more "look around": https://s3.amazonaws.com/lytro-corp-assets/blog/Living_Pictures.png
The key thing here is that if you match angles of view and depths of field, as you suggest, then the motion will render the same way on both formats. The 24mm lens is a numerically shorter lens than the 35mm, but it's on a smaller format, so in that context it's not any wider: the 24mm is the same "width" and "shortness" on S35 as the 35mm is on FF. It's the Series 1 test above, but with motion. As the geometry of the scene is the same in both cases—same AoV, same DoF, same camera-to-subject distance, same relationship of subject to background in the image—that sameness carries through a motion sequence just as it occurs in each of the individual frames.
Thank you for this comparison, Adam.
One thing i noticed in all CML discussions is almost no one tells about difference of all formats in motion scenes. Cinematography is not only about still frames, of course. So... If you match FF and S35 for example in all terms for a given scene with a person standing in front of a lens with some background or foreground elements (35mm 5.6 for FF and 24 mm 4.0 for S35 for example with same distance to the subject to match all differences in a "still" frame), wouldnt the motion of a persont towards the camera be more exaggerated with S35 because of wider 24mm optics? In FF case a person will make same steps, but after his walk he still will be more "distant" from us, because magnification of his body will be less. Or am i wrong?
That could be also considered as an important part of large format look, if it's the case.
I'm confusing with this question since the very beginning of large format talk. Can you clarify this moment?
Ha! So true on all points. Thanks again for the test.
No argument whatsoever with any of your points from a technical perspective, but implementation could be a nightmare. The active imager size on various "S35mm cameras" is all over the map: Arri isn't the same as Arri open gate, which isn't the same as F55, which isn't the same as Varicam, which isn't the same as EVA1, which isn't the same as EVA1 with EIS. And I haven't even thrown in the various RED active sizes based on the umpteen different resolutions they shoot.
So what you really need is an update to /i and LDS and EOS and other protocols to tell the lens what the camera is "seeing" at that moment, in that mode, and have the lens dynamically update its AoV / FoV labeling to match. Mind you, an OLED or LCD panel needs power, but you could use an e-Ink display: to heck with a Kindle, I want a Cooke S6/i paperwhite for Xmas. ;-) Oh, and existing cameras need to be updated to send that info. For your low-cost Rokinons and other passive (no-electronics) lenses, I suppose you could gaffer-tape a small bit of whiteboard to the side of the lens, and during prep the AC does the math and uses a very fine-tip dry-erase marker or grease pencil to write it in...
It's this sort of ongoing confusion that caused CML founder and DP Geoff Boyle to say a couple of years ago that "we should all just shoot standard S35mm because we all know what focal lengths work", or words to that effect (though with considerably more swearing, of course). Sadly I think the confusion is here to stay, and the best we can do is continuing education.
"As science pushes forward, ignorance and superstition gallop around the flanks and bite science in the rear with big dark teeth."
-- Philip Jose Farmer, "Riders of the Purple Wage"
We just have to be ready and willing to bite back. Think of it as job security.
Thank you for this test. I've been trying to convince people of these points for years. But there's a kind of hard wired correlation in a lot of minds of focal length to perspective. And to a lesser degree angle of view. When perspective is entirely a function of subject to camera distance. And angle of view relative to focal length is entirely dependent on format size.
So when one is familiar with FF 35mm for example, and learns that a 24mm is a "wide angle" with certain charateristics and perspective distortions, they assume that carries over to other formats. But on a MFT sensor it will be a "normal" lens with no hint of the distortions some think are inherent to that focal length. When the distortions were in fact inherent to that subject to camera distance coupled with that AOV. I know I had that misconception when I was starting out and hadn't used multiple image formats.
I actually think that with the use of so many formats a better "descriptor", I guess you'd call it, of different lenses would be the AOV on a given format (for all formats it's image circle covers) rather than focal length. Because focal length is irrelevant without the frame of reference of format size. (I know there are many charts that do these calculations, but it still seems to greatly confuse a lot of folks).
Also, one little thing about this article that I think also often confuses people is the interchangeable use of the terms aperture and f-stop, when they're not the same thing. F-stop is aperture relative to focal length. So the same f number represents any number of different actual aperture diameters. Where actual aperture is just that. The actual size of the opening. So while f-stop/t-stop are useful/necessary to calculate exposure, actual aperture size is what determines depth of field. And again, there are many charts to calculate this. But I think a useful descriptor on lenses would be actual aperture in addition to f-stop to help folks match image characteristics across formats. And to help clear up some of the confusion and misinformation that gets spread around the internet. Match AOV and match actual aperture and what people consider to be inherent format characteristics disappear.
Agreed: it's easier to find wider lenses for larger formats, and if your larger format has higher resolution, you can shoot wider and still preserve the critical details needed to read the scene, like nuances of an actor's expression or the glint in her eyes. And, most of the time, a larger sensor will be better in low light than a smaller sensor.
As to "all of your glass, not a (sub-optimal) portion of it", that assumes two things: you're using the same lens—or at least, lenses designed for large-format imaging—on both large and small formats, and that the center of your lens's image area (the only part used on small-format) is the "sub-optimal" part.
If you use a large-format lens on large format, and a small-format lens on small format, each is optimally designed for the image circle it needs to cover, and all else being equal they will perform the same. And it's more frequently the case that a lens falls down towards the edges of its image rather than the center; a lens (typically a complex zoom) that's merely OK on FF often performs much better on S35mm / APS-C / etc.
However, a large-format lens may not be held to as high a tolerance as a comparable small-format lens: to resolve 4K on FF requires only half the resolving power of a 4K lens for MFT, for example. In that case, yes, you'll see poorer performance moving the large-format lens to a small-format sensor. This is more likely to be an issue with older, "vintage" glass than with more modern designs, but as we move into ever higher resolutions, resolving power becomes ever more important and even minor differences will become more apparent.
Thanks for the kind words, and the discussion!
Thanks for taking the time to do these elaborate tests - very insightful.
The key reason why I would prefer an FF sensor over an S35 has to do with flexibility, although both formats have its merits.
With FF, you can shoot under more difficult circumstances
- tight spaces
- dark locations
With S35 (or MFT even more so), you can reach further
- crop factor (1.6 for S35 or 2.2 for MFT)
Also, with FF, you get to use all of your glass, not a (sub-optimal) portion of it.
Thanks for your thoughts, Daniel. I agree: even if you can replicate "the look" on smaller formats, it comes more easily and naturally on larger ones.
First thanks again for taking the time to make these articles. I find them to be very helpful and insightful.
You nailed the factor that I think is just not being explored enough about big formats - Specifically test #8. Look at the background magnification between the A7 and a6300 and the GH5. The A7S's background is less magnified and feels less compressed and gives me a better sense of depth to the room. I feels more "normal".
This is something that I really love about the larger format's inherent shallower depth of field and larger field of view - you can get this beautiful close ups (or wide shows like in test #7) with a 35mm that don't distort the background the way a telephoto or a wide angle lens would while still getting a very pleasing shallow focus. As a result the image looks more "natural" and pleasing to my eye.
You can shoot a wide two shot with a 35mm vs having to break out an 18mm lens due to physical restrictions of a location or have the camera 5 miles away from the talent. And while you can achieve the same thing with a smaller format, I'd argue it's just more difficult and often times I don't have the budget/clout/time in this world of always having to work faster, fit more pages in a day, and camera dept budgets that are a race to the bottom.
You can live in this great world of a "normal" lens without super flat backgrounds of a telephoto lens or the very wide distorted and compressed backgrounds of a wide angle lens.
Of course that is an aesthetic choice that is not always correct for the story (and the story is king) and there are plenty of examples I can think of where I'd prefer S35 or even smaller formats....but you know blah blah blah tools, professionals, horses...courses, etc.