Is the future FF35?

[James T Mather]

The only thing I can really say about FF35 is: "Once you go FF35, you will never go back." Or at least you won't want to go back.

Don't underestimate the boosts in ASA/ISO ratings to be able to compensate for DOF. A year ago, I was pegged out at ISO 800 on my Canon 350D. Now I shoot comfortably at ISO 3200 on my FF35 5D2. I think Drew and some others will confirm this.

I have no doubt you are right Tom about FF35 - I absolutely prefer to have more options than less - and FF35 offers more pictorial options as well as potential noise floor than other formats (I am not arguing against this - just our work requirements are different).

Wow, 3200 ASA is beyond my "normal" shooting range or usual comprehension (once shot a commercial at 12,000asa for a nasty tough edge - on film though and it was certainly "extraordinary" requirements in the story). Digitally I usually end up at 320asa or 500ASA with a gun to my head. 1000ASA for (brief) 2nd unit cutaways at night - and if I can take my name off the credits TK tests show that Red is workable to about 640 asa I thought for what I would describe as "process normal" - someone else might have a different opinion.

Here's looking forward to 3200ASA - it'll be like christmas. A 1.2k HMI will light up the desert.

 

[Tom Lowe]

One of the main reasons I got the 5D2 was for the clean high ISO frames you could get with it. If the FF35 Epics can do the same thing that would be awesome. But being able to shoot at that high of an ISO really does change the way you shoot and what you can shoot. Plus I was really surprised by FF35 on my 5D2. I could never go back to a DSLR that shoots with a cropped sensor. Hopefully soon I can say the same thing about shooting FF35 on the Epic.

"Once you go FF35, you never go back." Write it in stone.

 

[Daniel Browning]

Thanks, Mike.

What you just said above is, in fact, true, but there are other conditions that must be met.

It's true that there are conditions, but I think the conditions are all pretty reasonable. They are same ones that I think most people would assume as a matter of course: same exact lighting, level of technology, software, angle of view, subject distance, etc.

...while true in theory, is not quite as true in practical situations.

That's true. For example, theory predicts that 2/3" at f/1.6 has less noise than 1/3" at f/1.6. But if your 2/3" camera is from 1999 and your 1/3" camera is from 2009, you may find the 10-year-old camera has more noise. But that does not mean 1/3" has less noise than 2/3" in general.

As another example, theory predicts that RED ONE 4K at f/2.8 has less noise than RED ONE 2K at f/2.8. But if you apply a ton of detail-smearing noise reduction software to the 2K footage, you can make it look less noisy than the 4K. But that does not mean 2K has less noise than 4K in general. Other people would find that level of NR software to be very distasteful.

My position is based on the idea of all other things being equal (AOTBE). If your specific circumstances introduces some sort of inequality (e.g. one camera has better technology, or the other has different post processing), then your mileage will vary, but that doesn't take away from the validity of the theory.

It's necessary to separate things that are always true from the things that are only true in certain circumstances.

At the same camera to subject distance, with the same amount of light, it is not true - and this is often the deciding factor under practical shooting conditions.

My position does not depend on changing camera to subject distance. I'm talking about keeping that the same. Also, I'm not talking about changing the amount of light. I'm saying that FF35 can stop down and have the same DOF without adding more light. ISO 1600 on FF35 has the same noise level as ISO 640 on S35, all else being equal (same technology, etc.).

Physical conditions often dictate where the camera must be, and in turn, what lens is used. Practical considerations often dictate where the lights are, and how much is required.

Agreed. And that's what's neat: the larger format can still hit the same DOF with the same amount of light.

Going to a deeper stop affects the exposure in a nonlinear way,

For some formats that is correct. Film and video, for example, record information non-linearly. But raw is different. It really, truly is linear. (Compression probably introduces some non-linearity, but if so, that too should be separated from sensor size.)

So while what you say is true in the literal sense, it also involves other considerations that are sometimes practical and sometimes not. Hence why James and others (myself, Dave Mullen, and some others) have sometimes questioned your conclusions based on practical realities.

I understand that there are practical realities. For example, say the DOP is advocating shooting all the interiors at RED ONE 4K, 24p, f/5.6, 1/48, ISO 640. But the director wants to shoot RED ONE 2K, 24p, 1/48, f/2.8, ISO 160 because he thinks it will be less noise and more DOF. The truth is that they are the same noise and same DOF, but if the director gets his way, the practical reality is that it doesn't matter. That doesn't take away from the theory.

Daniel, James is an awesome guy, tons and tons of experience. So while he may not be as technically deep on these particular issues, his knowledge and life long real-world expertise is pretty intense, so it's a bummer to drive him away.

Yeah, I feel bad about that. I really think I have the right content, but my delivery must be very poor. I'm probably being too rude and prideful. I'm not trying to be a jerk, honest. (That's because I don't have to try, it just comes naturally! ) I'll try to work on being more respectful. I welcome corrections (PM is great) and specific suggestions.

You have an awesome bucket of knowledge here that is going to get buried in this thread. Have you considered making a blog entry, complete with these test photos that you're advocating? It'd be helpful for me to actually see all these bullet points instead of just reading them.

Good idea. For what it's worth, here's one (slightly flawed) visual comparison for you, showing a larger format with the same noise when used at the same DOF (slower f-number and higher ISO):

http://forums.dpreview.com/forums/read.asp?forum=1032&message=16107908

 

[M Most]

I'm saying that FF35 can stop down and have the same DOF without adding more light. ISO 1600 on FF35 has the same noise level as ISO 640 on S35, all else being equal (same technology, etc.).

It may have the same noise level (don't know how exactly you could measure that), but my contention is that if you did what you're suggesting, you would have two different images, not the same image. One component of those two images that may (or may not, I still say what you're claiming is theoretical and not something I have ever seen proven) be relatively equal is noise, but there are a lot of other components that make up an image, and these are likely not the same if you are shooting at vastly different stops with vastly different ratings. I would really like to see some visual confirmation of what you're claiming before I concede that the theory has real world validity, because experience tells me that's simply not the case - especially when you're shooting with a RAW capture, in which by definition, the sensor response is fixed regardless of the metadata settings.

 

[Michael Hastings]

It won't be predictive, it will have to be hyper-accurate and smoothly instantaneous, as you say. But that is not beyond the realms of possibility for technology.

In 5 years instead of lighting t2.8 on 500asa we will light the same and shoot an 11/16 split at 12000ASA (noise free of course) and press a "look" button in software to choose selective focus and add bokeh/flare/anamorphic/etc. In ten years it will be built into the camera.

I'm joking but it makes you think. (Canon already does lens correction for aberration and vignetting in the camera - rebel T1i $700)

 

[Daniel Browning]

It may have the same noise level (don't know how exactly you could measure that),

I suggest just eyeballing it to make a subjective judgement, but if you want to get really scientific you measure the std. dev. from a part of the image that has uniform brightness.

...but my contention is that if you did what you're suggesting, you would have two different images, not the same image.

In most cases, yes, the images will be different in some ways. The fact that you have to use a different lens (or the same lens but zoomed to a different focal length) means the possibility for different optical characteristics such as bokeh, aberrations (contrast, resolution, flare, distortion), etc. But if there is any confusion about what is caused by the lens itself and what is more general to the format, it can usually be resolved by using different (better) lenses.

EDIT:

I would really like to see some visual confirmation of what you're claiming before I concede that the theory has real world validity, because experience tells me that's simply not the case.

I would be happy to do the experiment myself and post the results (including raw files), but I don't want to do it if you have some objections to the experiment itself. Here is what I propose to do:

• Canon 5D2 in raw mode.
• Same exact raw conversion software.
• Same exact lighting.
• Same subject distance.
• Picture A: 70-200 f/4 L IS at 70mm, f/4, ISO 200.
• Picture B: 70-200 f/4 L IS at 200mm, f/11.4, ISO 1600.
• Crop Picture A to match the angle of view of Picture B.
• Downsample both pictures to the same resolution.
• Compare for noise, DOF, diffraction, etc.

Sound good?

 

[D Fuller]

It won't be predictive, it will have to be hyper-accurate and smoothly instantaneous, as you say. But that is not beyond the realms of possibility for technology.

If it's not predictive, it will often be late.

 

[Tom Lowe]

If it's not predictive, it will often be late.

In 1992, at Fort Carson, I used to be a gunner on a US Army Bradley fighting vehicle. The stabilization system was able to cope with 40mph vehicle speeds over extremely rough terrain, and still allow me to hit a human-sized target from 1,500 meters..... almost every time. Do not underestimate technology. :driving::biggrin: :gun:

 

[M Most]

I would be happy to do the experiment myself and post the results (including raw files), but I don't want to do it if you have some objections to the experiment itself.

As an exercise to prove the theory, what you're proposing is fine. But here's the rub: since we're discussing this on Red User, and since the subject under discussion is a FF35 sensor vs. a S35 sensor (or a 2/3 inch sensor, for that matter) it's a bit academic because your experiment is using the same sensor, even though it might be proving out the math and the theory. The problem here is that once again, real life considerations dictate that there are many more complications when comparing disparate devices. To assume that the theory will be true when applied to a device that hasn't been manufactured yet is a bit presumptuous. The fact is that Red has chosen to move towards larger sensors in order to improve specific performance characteristics because with the current technology, there are no other ways to achieve that. They are attempting to come a bit closer to the dynamic range and exposure flexibility of current film emulsions, and the larger sensor is seen as a way to do that. But as with all choices, it comes with a price, as there are other implications and compromises. One of the implications of going to a larger target is that it requires a physically larger lens with more coverage. And although the sensitivity of the chip is naturally going to be higher than the current size, how much higher is still a bit of an unknown because nobody has been able to test it yet (other than Red, that is, assuming they already have test silicon). So any assumptions about being able to comfortably and predictably shoot at ISO's of 1600 or higher are conjecture. I don't have a lot of doubt as to your conclusions, given the conditions you're assuming. I just don't know that those conditions will be available in these new devices. So while I'm not doubting your conclusions, I am doubting the ability of future devices to achieve the conditions required for those conclusions, as well as the ability to do it under actual production lighting and shooting conditions.

 

[David Rasberry]

Sensitivity per se does not necessarily follow sensor size. Sony's top line 2/3" cameras, at least two of which share the same sensor block are rated as follows:
F900 Cine Alta 320ISO
New F800 Cine Alta 800ISO
PDW700 ENG/EFP camera 1600ISO

 

[Tom Lowe]

Sensitivity per se does not necessarily follow sensor size. Sony's top line 2/3" cameras, at least two of which share the same sensor block are rated as follows:
F900 Cine Alta 320ISO
New F800 Cine Alta 800ISO
PDW700 ENG/EFP camera 1600ISO

It doesn't follow exactly, but bigger sensors help. :001_cool:

 

[Daniel Browning]

As an exercise to prove the theory, what you're proposing is fine. But here's the rub: since we're discussing this on Red User, and since the subject under discussion is a FF35 sensor vs. a S35 sensor (or a 2/3 inch sensor, for that matter) it's a bit academic because your experiment is using the same sensor, even though it might be proving out the math and the theory. The problem here is that once again, real life considerations dictate that there are many more complications when comparing disparate devices.

You are throwing the baby out with the bathwater. The complications do not make the theory any less useful. Quite the contrary. It informs us about exactly how size plays a role. It does not make other factors (e.g. technology level, sensor performance, etc.) unimportant, but lets us separate their effect from the effect of size.

For example, theory says that, all else being equal, an 8L engine will give you more horsepower than a 4L engine. But in real life, all else is not equal. Your 8L engine is 50 years old and not running right, whereas your 4L engine is brand new with the highest technology, so in this circumstance the 4L has more horsepower. That does not mean 4L engines are always better than 8L, or that the theory needs to be thrown out. The two can co-exist.

In the same way, all else being equal, large formats can stop down and increase ISO to get the same DOF as smaller formats, without having more noise. In real life, some smaller formats have superior technology. Both are valid and useful at the same time.

The experiment I described only proves that if the sensor performance (per area) is the same between a large format (FF35) and small format (2/3"), then you can shoot them at the same DOF and get the same noise, but it's still useful knowledge when the performance per area is different.

To assume that the theory will be true when applied to a device that hasn't been manufactured yet is a bit presumptuous.

But that's not what I'm doing. Others are stating that smaller formats have a DOF advantage compared to larger formats. I am debunking that myth by describing what happens when both formats have the same technology (same performance per area).

If I assumed the FF35 performance was going to be 2 times better, with newer, high technology, that would be presumptuous. If I assumed S35 was going to be older and worse technology, that, too, would be presumptuous. But I'm not: I'm discussing what happens when the performance and technology level is the exact same between sensor sizes. That makes it possible to talk in generalities about the format size.

Technology and sensor size are both important. If one of them is the same (e.g. same sensor size), then the only difference left will be technology. If the other is the same (e.g. same technology), then the only difference left will be size. Let me give you a metaphor of how I see the situation:

• Engine size = sensor size
• Engine RPM = f-number
• Speed (MPH) = depth of field

There are some little fuel-efficient cars with 1L engines (1/3" Bayer sensor). If you drive these poor little guys at 3000 RPM (f/2.8), the speed is comfy 25 MPH (deep DOF). Safe as a go-cart. (These are made-up numbers, obviously.)

Then there are sedans with 2L engines (2/3" Bayer sensor). If you drive these at the same 3000 RPM as the 1L go-cart (f/2.8), you get a much faster speed, we're talking 50 MPH (thinner DOF).

Then there are the mid-level sports cars. A nice BMW M3 4L (S35). If you drive it at the exact same 3000 RPM (f/2.8), you go faster still, more like 100 MPH. For many purposes this is unsafe, so it's better to go down to 2000 RPM (f/4.0). Some are a little crazy and they'll redline it at f/1.3, hitting ridiculous speeds (dangerously thin DOF).

Now we have a company announcing a new 8L sports car (FF35). Everyone is saying that it is far too dangerous, because it goes 150 MPH at just 3000 RPM (f/2.8). I am responding to this by saying that you don't have to drive them at the same RPM. Sure, the larger engine will let you drive faster if you want to. But if you scale the RPM with the engine size, you will get the same speed in the end. 8000 RPM on a 2L engine will give you the same speed as 2000 RPM on 8L, just as f/1.5 on 2/3" will give you the same DOF as f/6 on FF35.

The controversy occurs because most people think that when you drive larger engines at lower RPMs, you get more noise. But that's the surprising part: noise is the same. When all else is equal, noise scales with the MPH, not the RPM.

One of the implications of going to a larger target is that it requires a physically larger lens with more coverage.

They are sometimes physically larger (length and width), but often not heavier, as described above.

And although the sensitivity of the chip is naturally going to be higher than the current size, how much higher is still a bit of an unknown because nobody has been able to test it yet

My position is not predicated on the performance of future RED sensors, but on the idea of what it would be like if two sensors had the same technology, but different sizes.

Let's look at what happens when you vary sensor and f-number. First, we'll assume "all else being equal":

• Same low light scene
• Angle of view
• Subject distance
• Technology/sensor performance
• Same raw conversion, post, etc.

There are two remaining factors to consider, sensor size and f-number:

1. Sensor size:

• Bigger sensor = less noise and thinner DOF
• Smaller sensor = more noise and deeper DOF

2. F-number:

• Faster f-number = less noise and thinner DOF
• Slower f-number = more noise and deeper DOF

Here's what I think happens when one or both of the factors are changed in some way:

• Bigger sensor size but keep f-number the same: less noise but thinner DOF.
• Faster f-number but keep sensor size the same: less noise but thinner DOF.
• Bigger sensor and compensate with f-number: same noise and same DOF.
• Smaller sensor and compensate with f-number: same noise and same DOF.
• Smaller sensor with same f-number: more noise and deeper DOF.
• Same sensor with slower f-number: more noise and deeper DOF.

F-number scales with sensor size.

 

[Daniel Browning]

Sensitivity per se does not necessarily follow sensor size.

"Sensitivity" is a Marketing term, and every Marketing department has their own idea of what it means. The manufacturer can assign whatever ISO number they want, it's completely arbitrary (even among different models from the same camera line).

What matters is the real performance, such as SNR at a certain number of stops below clipping. If you look at that, then it does follow sensor size, as long as the performance per area is similar.

 

[david farland]

Phew.....
don't know where this thread to going anymore but suffice to say there's a little chat on sensor/lens technology and how/whether it will or should be used.
Fair to say everyone has legimitate opinions born of out of science/practice (or just a little googling) and suffice to say we all see open ends to each other arguments.
I'm sure sensors will be developed/marketed to camera companies who'll utilise them for reasons most of us will never know. But the short answer is whatever's prevalent and has better noise/larger gamuts/less artifacts and cheaper than what's around then we'll use it.
Now for whatever reason commercial/technical reasons a company decides to launch new and competing technology/format great. If they work out obstacles like cost/compatability to existing kit/practice/knowledge then bless them.

It good to hear what people see these obstacles are because it doesn't matter what a scientist says is better for you if it doesn't fulfill all the priorities of the user then it's indeed tail wagging dogland.

so I think we're agreed the sensor will work. There's still conjecture on cost/design of FF35 cine lens and the focus/aperature/light issue is still to be fully born out. My feeling is lens companies (& Red) know the metrics/business case around introducing new kit and Red sound like their going to introduce FF35 whether we agree or not and whether all our issues have been worked out or not.
Tip to those oldies who can....look back at what you were using 20 years ago and think do you really give a damn about the technology or what it did!

Chill.......
Dave

 

[Stephen Williams]

You're very welcome, Hans.


But you don't have to light to T4. FF35 with a dark exposure gives you the same amount of noise as S35 with a normal exposure. This fact is very easy to prove to yourself with a simple experiment using any raw camera, such as a DSLR or even RED ONE:
.

Hi Daniel,

My only reason to shoot FF35 would e for better quality.
If my lighting stays the same the noise for FF35 is identical to S35, there is no advantage in shooting FF35. Thank you for clearing that up for me.

FF35 seems to be a soloution looking for a problem.

Best

Stephen

 

[M Most]

What matters is the real performance, such as SNR at a certain number of stops below clipping. If you look at that, then it does follow sensor size, as long as the performance per area is similar.

I don't disagree. But a camera - even a digital camera - is not just a sensor. It is a complete optical system. All lenses have "sweet spots." The reason that most DP's like to shoot in controlled situations at about an f2.8 to f4.0 is because for most lenses, that is the sweet spot of performance. Stopping down from there allows greater DOF, but you pay a price. Opening up yields shallower DOF, but you also pay a price. Changing the shooting parameters by changing F-stops and ISO ratings doesn't necessarily yield the kind of gains your theories imply. That's part of what I meant by "other considerations."

Stephen Williams posted:

FF35 seems to be a soloution looking for a problem.

I agree, to a point. I see it, as I previously mentioned, as an attempt to wring better performance out of relatively current technology - an admirable quest, but not without other compromises and issues. One could be cynical and look at is as part of a marketing numbers game (bigger is better, so "FF35" is "better" than "S35," and "6K" is "better" then "4K"), and for some that might work, but I don't see it that way.

 

[Gavin Greenwalt]

I don't disagree. But a camera - even a digital camera - is not just a sensor. It is a complete optical system. All lenses have "sweet spots." The reason that most DP's like to shoot in controlled situations at about an f2.8 to f4.0 is because for most lenses, that is the sweet spot of performance. Stopping down from there allows greater DOF, but you pay a price. Opening up yields shallower DOF, but you also pay a price. Changing the shooting parameters by changing F-stops and ISO ratings doesn't necessarily yield the kind of gains your theories imply. That's part of what I meant by "other considerations."

But isn't that also the classic unobtonium fallacy?

"If we replace this electric car's steel frame with unobtonium it'll be 10 more efficient than a normal gas car!" The falacy being that if you had unobtonium existing products also benefit.

The "sweet spots" are often the product of diffraction though aren't they? If we're replacing a s35 sensor with a FF35 sensor then the diffraction characteristics of the FF35 lens will favor a lens stopped down more. Or at least that's been my experience, maybe Daniel covered that in a chart somewhere a few pages back. In which case a FF35 stopped down an extra stop to match a S35 for DOF could very well still be within its sweet spot.

Stephen does the ability to produce the same outcome make it pointless? Sometimes yes. But that also means it's more flexible. If you are shooting your nice big panoramic exteriors you could shoot FF35 or if you're shooting inside on tungsten then you could shoot S35. For me that's less expensive than buying two cameras but I get the benefits of both.

 

[Daniel Browning]

My only reason to shoot FF35 would be for better quality.
If my lighting stays the same the noise for FF35 is identical to S35, there is no advantage in shooting FF35.

That position makes sense.

The only significant advantage FF35 would have in that situation is if FF35 cine lenses had fewer aberrations (more contrast, less flare, smoother bokeh, less distortion, etc.), but they don't exist yet.

There will be others that use the camera in places where there is ample light to stop down "for free" (e.g. daylight exteriors). They'll get the noise, dynamic range, color, and other advantages of the FF35 sensor while keeping DOF the same as S35, but when they go back to the studio it goes back to being only the same as S35.

I don't disagree. But a camera - even a digital camera - is not just a sensor. It is a complete optical system. All lenses have "sweet spots." The reason that most DP's like to shoot in controlled situations at about an f2.8 to f4.0 is because for most lenses, that is the sweet spot of performance. Stopping down from there allows greater DOF, but you pay a price. Opening up yields shallower DOF, but you also pay a price.

That's true, but the sweet spot also scales with sensor size. On a large format 8x10 camera, the sweet spot on many lenses is f/64. Medium Formats have a sweet spot around f/22. FF35 sweet spot is f/8. And so on.

The entire reason for a "sweet spot" is one thing: diffraction. And diffraction scales in perfect proportion with DOF. f/11 on S35 causes a noticeable loss of contrast due to diffraction. f/11 on FF35 has less diffraction (higher contrast), but only because the DOF is thinner. If you stop down to f/18 to get the same DOF, then your diffraction goes up to the same level as well.

Changing the shooting parameters by changing F-stops and ISO ratings doesn't necessarily yield the kind of gains your theories imply. That's part of what I meant by "other considerations."

There is a certain elegance to the way everything scales with format size. It takes care of so many of the considerations: DOF, light, noise, diffraction, sweet spot, lens weight, file size. One way to see it is to forget about the format size and look only at the apparent aperture diameter. If you have a lens with a 5mm diameter opening, it doesn't matter whether it's a 10mm f/2 on small format, 25mm f/5 on S35, or 50mm f/10 on a larger format: the DOF, light, noise, diffraction, etc. will be the same.

The only things that don't scale are the ones affected by humans, such as the level of technology, type of lens designs available, prices due to economies of scale, etc. The higher cost of larger formats have generally tended to give smaller formats the advantage of economies of scale, and therefore better technology and wider variety of lens designs, etc. (S35 has more cine lenses than FF35, FF35 has more still lenses than MFDB, etc.) But it is not intrinsic to the nature of the larger format.

One of the considerations that no one has brought up is camera size. Current technology tends to cause FF35 cameras to be larger than S35. But the actual sensor itself is not really that big. In the future, I think FF35 and S35 cameras will be about the same size, as Leica did with their diminutive FF35 still camera. Even 645 will be as small as current DSLR cameras in the future.

The most practical consideration is that FF35 cine lenses just don't exist in any appreciable quantity.

Are there any others?

 

[Stephen Williams]

That position makes sense.

The only significant advantage FF35 would have in that situation is if FF35 cine lenses had fewer aberrations (more contrast, less flare, smoother bokeh, less distortion, etc.), but they don't exist yet.

One of the considerations that no one has brought up is camera size. Current technology tends to cause FF35 cameras to be larger than S35. But the actual sensor itself is not really that big. In the future, I think FF35 and S35 cameras will be about the same size, as Leica did with their diminutive FF35 still camera. Even 645 will be as small as current DSLR cameras in the future.

Hi Daniel,

The weakest link going FF35 is lack of cine lenses. If there is a real possibility is there will be no improvement in quality, sounds like a format that wont go very far IMO.

FF35 Film cameras don't need to be an bigger than S35, it's just hardly any were made in the last 50 years, they were used for specialist applications so size was not important. Panavision have a 28 pound 65mm camera from the 1960's (?) so size was never really the issue.

Best,

Stephen

 

[Stephen Pruitt]

And that's the point exactly. There are about 1.21 gigalenses out there right now for the S35 world. Except for companies wanting to sell more lenses, I have a hard time believing that FF35 is a real positive for anyone.

Again. . . if we say our prayers, perhaps RED will hear our pleas and give us a S35-sized 5K Monstro.

That would be my camera. No iffs. . . no ands. . . no buts.

Stephen