FF35 Cine Glass: Get Ready

[David Mullen ASC]

With those factors considered, I expect that FF35 will have only a 1/3-stop loss in control over DOF at the deep end (for a given image quality / noise level), and not the 1.3 stops that one would get with film (due to non-linear response preventing underexposure).

A digital sensor does not break the classic rules regarding depth of field; being able to underexpose is only a workaround solution to compensate for the inherent difference in depth of field due to the size of the sensor, focal length chosen to match field of view, distance focused, f-stop, and Circle of Confusion figure used.

Besides, we're comparing the depth of field difference between the S35 sensor of the RED versus the FF35 sensor of the EPIC, so film's response to light is not a factor. That difference will be, practically speaking, about 1.5 stops once you match f-stop, field of view, and distance focused. You may be able to compensate by rating the larger sensor faster and thus stop down the lens, but that doesn't mean that the true depth of field difference is less than 1.5 stops, just that there may be solutions to work around that.

1/3 of a stop falls within a margin of error in exposing, so you could practically say that a 1/3 difference is ignorable, which means that a 36mm wide sensor and a 24mm wide sensor have no real difference in depth of field characteristics? That is not logical.

Let's just be honest about the differences and find solutions to deal with it, rather than tell ourselves that it's not really an issue.

 

[Cristina S]

Besides, we're comparing the depth of field difference between the S35 sensor of the RED versus the FF35 sensor of the EPIC, so film's response to light is not a factor. That difference will be, practically speaking, about 1.5 stops once you match f-stop, field of view, and distance focused.

Since it's 2x sized, isn't just 1 stop of difference?

 

[David Mullen ASC]

Generally if you look at depth of field charts... you see that simply switching to a focal length of half or double the previous one causes a four-stop difference in depth of field.

But you also have to double or halve the Circle of Confusion figure because you're comparing images blown-up or shrunk to match each other, i.e. you're assuming that comparisons in depth of field are being made between images presented at the same size. So that four-stop difference between two lenses with half or double the focal length becomes only a two-stop difference once you halve or double the Circle of Confusion figure.

So generally you can figure that the magnification factor is also the difference in practical depth of field in f-stops. So, for example, 35mm is 2.5X the difference in size with the 2/3" format, and the depth of field difference is also 2.5 stops.

Now all of this ignores a few real-world issues like MTF, etc. that may affect perceived focus fall-off. For example, I've done some quickie unscientific tests of depth of field differences between 2/3" and 35mm digital cine cameras (using a row of receding focus charts) and they more or less fall into what the traditional math tells you about the practical depth of field, i.e. a 2.5-stop difference. In other words, when I stopped the 35mm digital cine camera down by 2.5-stops, it matched the depth of field of the 2/3" camera, once I matched distance and field of view.

But camera assistants constantly claim that 35mm digital cine cameras have less depth of field than 35mm film cameras of the same size target area. I think one factor is this higher MTF of the digital cameras, causing a more obvious or visible fall-off or difference between what is considered in focus and out of focus. Some people have made the same claims just comparing a Zeiss Master Prime to a Cooke S4, that the snappier look of the Zeiss makes it appear to have less depth of field. I'm not sure if that's true or simply that the higher contrast of the lens makes it easier to see when the focus is on or off. I found when shooting on the Genesis, for example, that it was always really obvious when the lens was slightly off-focus.

 

[Tom Lowe]

The question is then, where do we get high-MTF FF35 cine lenses?

Yep, that is the question! And that question is going to start getting asked more and more and more often, I bet, as all of us come to grips with FF35 digital cinema.

Regarding still lenses on the wider end, two lenses that interest me on the Epic/Scalet FF35 are the Nikon G 14-24 and the Contaxt N Zeiss 17-35. The latter is said to be damn near flare resistant, and the former holds crazy edge sharpness. I'm going to jump into the Nikon G 14-24 soon and start shooting 5.6K RAW timelapse with it on the 5D2. I'm guessing that the results will be very nice.

Of course, as you have pointed out many times, these still lenses are not ideal in terms of mechanics and focus throw.

 

[Geoff Yerke]

The Nikon G 14-24 is a fantastic lens. Highly recommended.

 

[Pawel Achtel]

Tom,

Yes, the Nikon G 14-24 on Canon 5D2 has the potential to work nicely for night time lapse, but remember, it is not going to be as bright as a Master Prime fully opened. I wonder if the 5D2 could be chopped up and patched with PL mount welded onto it? :w00t: Hmmm, tempting....

 

[Tom Lowe]

Tom,

Yes, the Nikon G 14-24 on Canon 5D2 has the potential to work nicely for night time lapse, but remember, it is not going to be as bright as a Master Prime fully opened. I wonder if the 5D2 could be chopped up and patched with PL mount welded onto it? :w00t: Hmmm, tempting....

Okay, you chop up yours first and try it! :construction:

BTW, the EF 24 f1.4 I seems to shoot really well with the 5D2 for night timelapse. The resolution and sharpness look very good. If the Mark II is a big improvement, then that is a serious lens for FF35, too.

 

[Sanjin Jukic]

Tom,

don't get afraid about Master Primes.

It's like a make up or face lifting.

Nothing special.

Movies are not changed because master primes just arrived!!!!!!!!!!

It was and it will be another type of low light shooting.

You are already on 5DMkII.

Rent or get 50mm f/1.2 oe 85mm f/1.2 and you'll see what you get with MP.

And last but not least, best if you would rent MP to test by yourself on R1.

 

[Pawel Achtel]

Okay, you chop up yours first and try it! :construction:

Butchering brand new and perfectly good filming equipment was always appealing to me. I can probably try with the 50D first and see what comes out. Where would I buy the camera side PL mount from?

 

[Tom Lowe]

Sanjin, a 14mm 1.3 Master Prime would be a little bit of a game-changer for me, since the closest thing on FF35 would be the 14-24 @ f/2.8. For night/star shooting, that 14mm MP would be HUGE, assuming I had a body to shoot with.

 

[Pawel Achtel]

Where would I buy the camera side PL mount from?

Oh, I have an idea.

Sanjin, how much do you want for your "obsolete" Red One PL mount? :biggrin:

 

[Daniel Browning]

My point is that the larger format, on equal footing [*], is capable of all the same things that the smaller system is, including diffraction, resolution, DOF, and light gathering power. In some circumstances (e.g. noise and dynamic range in bright daylight), the larger format can be superior "for free". But in other circumstances (e.g. low light), it can only be superior if the aperture is increased (and the DOF made thinner).


[*] Equal footing, to me, means both systems have a lens of equal physical aperture, the same sensor performance per area, the same total resolution, nonlinear capture, and totally raw (no compression). In real life, compression will favor the smaller system because the larger system must underexposed, which is harder to compress. I don't know how REDCODE will react with that, but I'm guessing the loss of 1/6th stop. Nonlinear mediums, such as film and "video", don't respond to underexposure linearly, like RAW digital, so that workaround does't work as well. Also, sensor performance (read noise, QE, FWC) per area doesn't scale perfectly with total sensor area, so that sensors 1/20th the size tend to have twice the performance per area, and smaller differences in size have smaller differences in performance per area. I'm guessing that the difference in size between S35 and FF35 will result in another 1/6th stop, leading to 1/3rd stop total difference.

OK, let's take the consideration one step further. Let's consider diffraction limit of two lens systems from two extreme examples Daniel provided:

Scarlet 2/3": 14mm f/0.9
617: 220mm f/14

They both have approximately 15mm physical aperture and similar angle of view. Is the resolution affected equally by diffraction?

Yes, as long as you have the same expectations for both then they are affected equally by diffraction. It's only when you expect *more* from the larger format that it is affected more by diffraction (in a manner of speaking). To get more resolution one would have to accept at least slightly thinner DOF, but at least you can know that any resolution+DOF combination the smaller system is capable of can be done at least as well (if not better) on the larger system.

The reason is that when both systems have the same effective resolution, the diffraction effects are the same. The effective resolution would be the same if both were projected at 2K. At higher resolutions (e.g. 20x60 prints), the 617 28K will show higher resolution than the 3K 2/3", unless the f/stop is so narrow that diffraction knocks it down to the resolution of the 3K, but at no time will the 617 28K have *less* resolution for any given DOF, which leads to my point about both being affect by diffraction in the same amount for a given physical aperture.

The optical system with the larger sensor will be more affected by diffraction...because the distance to the focal plane also increases with focal length, and so the airy disk diverges more over this greater distance.

The greater divergence of the airy disk is compensated perfectly by the much larger CoC, as when you scale the CoC with the format size (i.e. expect the same resolution from both systems). In my post I mentioned the effect of both cameras being limited to the same capture resolution (3K), where the pixels are so large on the 617 that the larger airy disk caused by diffraction has no effect. The effect is the same if they have different capture resolutions but the same output resolution, such as 2K projection.

In other words, your position is based on the use of a *smaller* h/CoC (horizontal AOV divided by CoC), whereas my point of view is based on using the *same* h/CoC, such as h/CoC=1200. So h/CoC is essentially the resolution of the system (capture through display). The typical method to calculate the scaled CoC is with the diagonal AOV, but I think horizontal AOV is more useful in cinema.

It's only when you *expect* more resolution from the larger format that equality is lost. For example, if you choose a physical aperture that is *almost* diffraction limited on the 2/3" 3K, then use the same phsycial aperture on 617 28K (a much longer focal length and much narrower f/number), you will not get anywhere near the increase in resolution expected from 3K to 28K. But if you only expect 3K as in the 2/3", you'll definitely get that.

I promise that I'm not trying to make this more complex than it is. Hopefully by explaining my point here I will learn to describe in a simpler and clearer manner in the future.

 

[Daniel Browning]

Thank you for the response, David. The clarity and succinctness of your posts are something we all strive for. I'm always too laconic or overly wordy.

A digital sensor does not break the classic rules regarding depth of field; being able to underexpose is only a workaround solution to compensate for the inherent difference in depth of field due to the size of the sensor, focal length chosen to match field of view, distance focused, f-stop, and Circle of Confusion figure used.

If the classic rule you are referring to is "larger formats are not capable of the same deep DOF as smaller formats in low light", then I think the workaround does break that classic rule. A RAW uncompressed digital sensor is completely linear, and reducing the exposure 7-fold will result in exactly 7 stops less image information per area. If you compensate for the per-area loss by increasing area 7-fold, then the total information captured is the same.

Essentially, RAW uncompressed digital is a medium where the workaround works really well. Two factors that make it imperfect, though, are compression (even REDCODE will not be perfect in this regard, I expect) and sensor performance per area (i.e. smaller sensors tend to do better per area).

Besides, we're comparing the depth of field difference between the S35 sensor of the RED versus the FF35 sensor of the EPIC, so film's response to light is not a factor.

True, I only mentioned it because the expectations of most people are based on nonlinear systems, such as film, video, and JPEG. Uncompressed RAW, as in DSLR and some digicams, is the medium where the workaround scales perfectly.

I'm not a film expert, but humor the following attempt to illustrate with an example. Imagine Super8 film with dimensions of 5.46mm X 3.07mm (1.78 aspect). A 7mm lens would have a 42 degree AOV. Shooting in relatively low light, say the best exposure is at T/2.8 (for a nice, meaty neg). On top of that, the very deep DOF provided by a 7mm T/2.8 on Super8 is critical for the shot.

To shoot the same scene with a VistaVision camera and the same film stock as in the Super8, a 46mm lens is needed (for the same 42-degree AOV and perspective). Shooting at f/2.8 would also yeild a nice, meaty negative of much higher quality than the Super8, but it would also cause much, much thinner DOF, ruining the shot (or at least making it much more challenging on the puller!). One must stop down to f/18 to get the same, deep DOF and focus pulling difficulty (or lack thereof). That's over five stops of underexposure. Even though the VistaVision has over 50 times more area, the underexposure will cause *more* than 50 times less light to be recorded on the film, because more of it falls on the non-linear part of the negative. (That's my understanding, at least. I don't know if flashing the film with 5 stops of light would get enough of the response up into the linear range to get results as good as the Super8, but it would be neat if it did.)

With RAW uncompressed digital, 50 times more area perfectly compensates for 50 times less intensity per-area.

That difference will be, practically speaking, about 1.5 stops once you match f-stop, field of view, and distance focused. You may be able to compensate by rating the larger sensor faster and thus stop down the lens, but that doesn't mean that the true depth of field difference is less than 1.5 stops, just that there may be solutions to work around that.

The point was made that FF35 has more challenging/difficult DOF than S35. My response is that it's only more difficult if you expect higher quality (such as resolution, light gathering power, etc.) in low light. For the *same* quality level, the DOF is the same. So if you shoot low light, and you don't want thinner DOF than S35, then FF35 will not give you increased quality. But when you shoot in ample light, the FF35 will let you have the increased quality *without* the penalty of thinner DOF, until diffraction starts the law of diminshing returns. In other words, FF35 can do the same, and in some circumstances, it can do more.

1/3 of a stop falls within a margin of error in exposing, so you could practically say that a 1/3 difference is ignorable, which means that a 36mm wide sensor and a 24mm wide sensor have no real difference in depth of field characteristics? That is not logical.

It will make much more sense if, instead of thinking about the sensor:

a 36mm wide sensor and a 24mm wide sensor have no real difference in depth of field

Think about the lens:

a 15mm aperture lens and a 15mm aperture lens have no real difference in depth of field

The fact that one is "49mm f/3.2 on FF35" and the other is "34mm f/2.2 on S35" are almost beside the point: they both have the same amount of glass. An imaginary 11mm f/0.7 on some super-tiny format would also have the same DOF/performance. Essentially, it doesn't matter how big the sensor is, as long as the physical aperture of the lens is the same, the DOF and light gathering power in low light will be too.

One has a very bright light focused on a tiny area, the other has a dim light focused over a large area. When you magnify them both for display at the same size, the bright light of the tiny sensor gets magnified a *lot* and spread out over a large area, where as the dim light of the large sensor gets magnified a *little* and spread out over the same area.

My central theme is that larger formats, if they are totally linear, can do anything a smaller format can do, and sometimes do it even without the penalty of thinner DOF. For many people, that "sometimes" will not be worth the extra cost and other downsides of FF35 (assuming it even had proper lenses to begin with!). But the reason I bring all this up is that for some people, such as those shooting in ample light, or those who can somehow deal with thinner DOF in low light, will definitely benefit from FF35.

 

[Pawel Achtel]

Thanks Daniel, you made some excellent points.

On a side note, I was imagining a "normal" lens for the 617.

The size of airy disk diameter is approximately:

D = 2.4 x λ x f, where λ is the wave length, f is a focal ratio

From this equation, we can estimate that the lens will start to be diffraction limited at around f/5.

Let's assume a "normal" lens, say 600mm. A subject at 3m, will have a DOF of 3mm, and this is with the lens stepped down right to diffraction limit!

Open it up to f/2.8 and somebody tell me how to follow focus with 1.5mm DOF.

Of course, if we use a modest telephoto lens for this format, say 1200mm, and get 2m from our subject, the DOF will be a width of a hair....

Any comments?

Sorry, I got side tracked away from the original subject, but it kinda flows from the above considerations.

Edit:
Not to mention the physical aperture of a 1200mm f/2.8 lens: 428mm. That's a lot of glass! If someone can make it lighter than 200kg (500lb) it would be a genius.

 

[Daniel Browning]

Let's assume a "normal" lens, say 600mm. A subject at 3m, will have a DOF of 3mm, and this is with the lens stepped down right to diffraction limit!

If by "normal", you mean the AOV of a 50mm lens on FF35 and a 32mm lens on RED ONE, that is a horizontal AOV of 39 degrees. A 39 h-AOV lens on 617 is a 220mm lens. 600mm has a 5.5 degree h-AOV on 617 format, which is like a 220mm lens on RED ONE, well into super telephoto territory.

Not to mention the physical aperture of a 1200mm f/2.8 lens: 428mm. That's a lot of glass!

Pfft. I have a 1250 f/5 in my garage that's $500 brand new. Perfectly corrected for chromatic abberation over the entire visible spectrum, though it does have some S.A. and coma. 428mm apertures sell for under $2200 now-a-days and weigh less than 70 pounds.

Happens to be a reflector, though.

 

[David Mullen ASC]

The depth of field difference between a larger and a smaller sensor assumes matching field of view, distance, and f-stop, which also implies the same sensitivity -- if you can just stop down on the larger sensor because of a higher sensitivity or a greater ability to underexpose, of course you will gain more depth of field to compensate for the inherent difference.

But this doesn't mean that the 1.5-stop effective difference between a S35 and a FF35 sensor is imaginary -- it is something to be factored in, something to be acknowledged and dealt with, just as when comparing shooting between any two formats.

A 1/3" DV camera has 2-stops more depth of field than a 2/3" camera, which in turn has 2.5-stops more depth of field than a 35mm camera -- those are all issues that have to be factored in, often by controlling the f-stop.

So yes, the depth of field difference between a S35 and a FF35 camera can be compensated for... if you stop down the FF35 camera by 1.5-stops, either by adding more light, or by underexposing, or because the FF35 camera is more sensitive than the S35 camera. But that doesn't mean the 1.5-stop difference doesn't exist, just that there may be solutions.

Just take a S35 camera and an FF35 camera, match field of view, match distance, and match f-stop and then look at the depth of field difference. You'll find that if you stop down the FF35 by 1.5-stops, the depth of field will then look more or less identical to the S35 shot. If I owned a FF35 digital camera, I'd shoot some charts and demonstrate that easily. It would also demonstrate that just stopping down by 1/3 of a stop would not be enough to compensate.

it works the other way as well.. anyone who shoots 2/3" video and 35mm digital on the RED could tell you that there is a depth of field difference of about 2.5-stops. Like I said, I did a test of that comparing the F900 and the 35mm Genesis camera over at Panavision using a receding diagonal line of focus charts, and the practical difference came out to 2.5-stops.

 

[Daniel Browning]

So yes, the depth of field difference between a S35 and a FF35 camera can be compensated for... if you stop down the FF35 camera by 1.5-stops, either by adding more light, or by underexposing, or because the FF35 camera is more sensitive than the S35 camera. But that doesn't mean the 1.5-stop difference doesn't exist, just that there may be solutions.

Agreed.

 

[Douglas Underdahl]

I agree with David - and I can tell you that from shooting a feature on 35mm at a 2-2.8 split, that using a FF35 frame digital camera is going to have some mighty tiny depth of field, unless you could rate it at at least ASA 1000. It's not a happy experience to see actors moving in and out of focus just by leaning toward the camera slightly. Really, T4 is a good stop to work with for RED or 4 perf 35mm for interiors - the actors can move and it gives your AC a "stop he can live with". I suppose if FF35 Monstro gives us 14 stops of DR and has more sensitivity than RED1, then it could be worth it, but otherwise, for most interior work, that tiny depth of field is going to make it not something that I want to deal with. I mean, look at poor Heath Ledger leaning in and out of focus in the interrogation scene in Dark Knight. What, are we going to tell him that he's only allowed to move side to side, or that he has to work out each lean in and out with the focus puller in advance? I like to let the actors have as much leeway during a performance as possible.

 

[Tom Lowe]

This is why higher ISO/ASA ratings are very important. The 5D2 has ridiculously awesome high-ISO capability, so I am assuming that RED will also make big advancements in ISO/ASA ratings by the time Monstro FF35 hits the streets.

 

[David Mullen ASC]

Well, I've shot four features in 35mm anamorphic and it's a similar challenge -- basically you try to commit to working at an f/4 inside unless there is a good reason for a wider stop, like in a candlelit scene, night exterior, etc.

If there is extra quality from shooting in FF35, just as in shooting in 35mm anamorphic, someone is going to find a way of making it work for them.

My main concern is not so much the lower depth of field, which can be compensated for to some extent, it's the lack of cine optics that cover FF35 -- if focus-pulling is going to be more challenging whenever you get into a low-light situation, you don't want your lenses to be making things even harder due to their mechanical design. At least 35mm anamorphic lenses are designed for cine use.

But it's all doable -- after all, "Dark Knight" was partly shot on IMAX, which uses adapted large-format optics, and the depth of field challenges are even greater with that format compared to FF35. The IMAX frame is 70mm wide, that's almost a 3X difference compared to S35 and a 2X difference with FF35.