Physical Larger Sensors - is it progression?

[roryhinds]

With the details of RED new sensors it seems the progression of sensor design is to make physically larger sensors.

IMO this is counter productive as we already have a industry standard for lenses and the image they capture.

Sure I want more resolution and higher dynamic range with faster frame frames but I want to keep it within the industry standard physical size.

Cooke & Zeiss lenses are fantastic for film making and have been for years.
Cinema screens are a set size and post production pipelines are solid for 2k and getting up to speed with 4k.

Having to invest in a new set of lenses to capture a new image size seems counter productive to me.

Is it that RED sensors can't be made to a industry standard size with higher resolution & DR?

35mm DOF is what we all wanted with digital acquisition. The same characteristic of 35mm Film using the industry standards.

[Michael Ragen]

I think the main issue we are coming back to is the no Monstro s35 sensor. I would imagine that would just come after the Mysterium X sensor. It makes no sense for them not to make a Monstro S35 eventually.

The X sensor is going to be ready first, so for people that can get a return on that buy will do so, and upgrade later, and the rest of us will have to wait for the FF35 or possibly the Monstro S35.

I personally don't mind waiting until 2010 for my Epic. And I'm still undecided which one I will go for, although if everything pans out, I will throw down on the 645 for sure, even if a pl-mount lives on it most of the time. I don't see a problem with Red making larger sensors as well, as there is a demand, its just not as big.

[Daniel Browning]

Customers will decide if the lenses of any given format are suitable for their needs. If more FF35 lenses are built to cine standards, then that format will be even more attractive. But if not, it will still be desirable to those who are satisfied with what lenses are available.

In any case, there are definitely some very good reasons to use larger sensors. As it pertains to low light performance, RED is already very close to the maximum theoretical quantum efficiency. Using a larger sensor size, like FF35, can allow for a greater amount of light gathering ability, as long as the aperture increases. For example, going from 35mm T1.3 on S35 to 50mm T1.3 on FF35 is a large increase in aperture for the same angle of view.

All sensor sizes will benefit from read noise reductions and other technology advances.

[Justin Anderson]

Then just shoot on the FF35 sensor, cropped at Super 35 24x18mm. If you don't want to use the full sensor, you don't have to.

Also, Jim said that the REDone, which is an actual 24x18mm sensor, will be upgraded to the Mysterium-X and then the Monstoro sensor eventually, so if you want to go that route then you can.

[Chris Kenny]

I think shallower DoF is going to emerge as a major problem. Pulling focus on S35 at low stops is already very challenging, to the extent that some shots require numerous attempts even by people with years of experience, and you still occasionally see soft shots show up in major Hollywood features.

Remember, up until now formats larger than S35 have been used mostly in the still photo world, where there's autofocus that's actually useful for professional shooters, and "focus pulling" as such doesn't really exist. To the extent that larger formats are used in the motion picture world, it's typically with anamorphic lenses (which don't generally open to T1.3) and on productions of sufficiently large scope that lighting to a decent stop can be accomplished, even if you have to build a set rather than use a location to do it.

I'm not sure how practical larger formats will be for general indie-style location shooting (where you're extensively using available light and only small easily portable lighting packages), at least until someone comes up with a comprehensive automated focus pulling solution suitable for cine-style shooting. That's not a trivial problem. And even then... well, even if you can reliably keep exactly the right 1 mm deep plane in focus, when you're focusing on your subject's eyes, exactly how far out of focus are you willing to accept your subject's nose and ears being? After a point it starts to look rather strange. (One sees this effect sometimes in headshots.)

[Tom Lowe]

I wonder if adding massively to Monstro's ISO/ASA ratings might help in this respect. The new Canon 5D Mark II, for example, is said to be very usable at ISO 3200 and 6400. This would be a gain of how many stops over the Red One or regular chemical S35mm cameras?

In other words, with a higher ISO/ASA rating on Monstro, you could stop down a bit on the FF35 glass to compensate for the wider sensor, and since it starts with less DOF, you could keep the DOF in the same ballpark as S35.

[Daniel Browning]

I think shallower DoF is going to emerge as a major problem.

Short version: larger sensors can stop down to match DOF while retaining higher image quality in most circumstances.

Long version: As long as there is ample light, the larger format can simply stop down to get the same DOF as the smaller one and compensate for the exposure by changing the ND filtration. This will allow it to continue gathering more total light than the smaller format as it would if everything were the same.

However, in a low light situation, the camera with the higher QE and read noise per area for a given perspective, AOV, and DOF will yield the superior image. The reason for this is that sensors are linear, so spreading the same amount of light over a larger area (i.e. narrower f/stop on a larger sensor) will yield the same total amount of light. This technique is not possible with film if the exposure falls outside the area of linear response.

[Chris Kenny]

Consulting a focus chart and doing a bit of math, it looks like (maintaining constant angle of view by using a longer lens, of course), I think you'd need to have an ASA a bit above 2000 (if we assume the Red One is 320) to maintain the same DoF (for a given stop and angle of view) at the larger format.

[Daniel Browning]

Consulting a focus chart and doing a bit of math, it looks like (maintaining constant angle of view by using a longer lens, of course), I think you'd need to have an ASA a bit above 2000 (if we assume the Red One is 320) to maintain the same DoF (for a given stop and angle of view) at the larger format.

According to my math the larger format only needs ISO 736. The bellows factor can sometimes cause a problem, but not that big. Should we "show our work"?

If both are resampled to the same resolution, enlarged to the same size (remember the larger format is enlarged less), and have the same sensor performance per area, the final image quality will be the same (even in spite of the ISO difference).

[Chris Kenny]

According to my math the larger format only needs ISO 736. The bellows factor can sometimes cause a problem, but not that big. Should we "show our work"?


The Red One's sensor is a crop factor of about 1.5 vs. full frame, so I looked up DoF at e.g. 50 mm and then at 50/1.5 = ~33 mm, and then found a stop at the former focal length where DoF was about equal to DoF at the latter focal length. I came up with a difference of a bit less than three stops. Three stops up from (320*2^3) is 2560.

Mind you it's 4 AM here and it's possible none of this makes any sense.



If both are resampled to the same resolution, enlarged to the same size (remember the larger format is enlarged less), and have the same sensor performance per area, the final image quality will be the same (even in spite of the ISO difference).

This only works if your entire image falls within the dynamic range of the linear part of the sensor though, right? If you've got an image that exactly fills the entire dynamic range of the sensor, and you stop down two stops, suddenly you're putting that full range into the bottom 25% of sensor values. Sure, with the higher resolution sensor you see a bit more into the shadows because after downscaling to the same resolution your noise floor is effectively lower, but do you really get two full stops from this effect?

Hmm. Have to think about this.

[Daniel Browning]

The Red One's sensor is a crop factor of about 1.5 vs. full frame, so I looked up DoF at e.g. 50 mm and then at 50/1.5 = ~33 mm, and then found a stop at the former focal length where DoF was about equal to DoF at the latter focal length. I came up with a difference of a bit less than three stops. Three stops up from (320*2^3) is 2560.

Mind you it's 4 AM here and it's possible none of this makes any sense.


Here are my numbers:

36x24 sensor, 50mm lens, 39.2 degree HAOV, focus distance 5 meters, f/8.0 DOF = 6.08m using h/CoC=1200.

24.4x13.7 sensor (RED ONE with lookaround), 33.5mm lens, 39.2 degree HAOV, focus distance 5 meters, f/5.6 = 6.53m DOF using h/CoC=1200.

So f/8.0 to f/5.6 is one stop, and the 6.08m to 6.53m would be a little more on top of that (1/8 stop?). After you compensate for RED ONE's recordable area it's somewhere around 1 and 1/3 stop, which is ISO 736. QED. :) Scarlet/Epic S35 will be different due to larger recordable area.



Sure, with the higher resolution sensor you see a bit more into the shadows because after downscaling to the same resolution your noise floor is effectively lower, but do you really get two full stops from this effect?


Yes.

[mezmo]

Is It Progression?
A whole new camera system. That's what's required for FF35 and Monstro.
Bigger sensors, bigger non standard lenses, large format FOV, fatter recording media, complicated processing. It's a brave new world for RED and customers in time of global financial crisis. And what will this big stuff look like shooting and posting digital, who knows. I reckon it's gonna be big and sharp. So Good luck to the good ship RED and all those who sail in her. It may be the birth of a new age or the death of a very small but cute camera company.
Cautious people should probably not venture past a S35 Epic or MX'd RED ONE for some time to come. Seems to be no shortage of jump in quick Beta testers with plenty of cash lining up for the new RED big boxes and lenses. So break a leg guys. Cheers Mezmo.

[Tom Lowe]

Yeah, it was my understanding that FF35 was about 1 and 1/2 stops over S35/APS-C.

If Monstro has high-ISO/ASA performance like the new Nikons and Canons, ISO 700 is certainly within reason.

[Stephen Williams]

If Monstro has high-ISO/ASA performance like the new Nikons and Canons, ISO 700 is certainly within reason.

Hi Tom,

Needs loads of ND filters, hardly progress, tungsten friendly would be more useful,

Stephen

[Harrison Diamond]

Hi Tom,

Needs loads of ND filters, hardly progress, tungsten friendly would be more useful,

Stephen

But Tungsten Friendly would *sacrifice* sensitivity, hamper usefulness both outdoors and with daylight balanced lights, and is just not feasible with most current sensor technologies.

There isn't a camera out there that I know of for high end motion or still photography with a sensor that isn't natively daylight balanced.

[Stephen Williams]

There isn't a camera out there that I know of for high end motion or still photography with a sensor that isn't natively daylight balanced.

Hi,

For that very reason Film will not be slain as quickly as you think.

Stephen

[Auguste Langevin]

Hello everyone, my first post here ! (excited)

About the physical size of the sensor, 617 seems so unimaginable that no one has calculated aspect ratio of it (or maybe someone has but haven't seen it, really, I searched...)

28000:9334=3 (approximately)
186:56=3.32...

That would mean 6 micron pixel are not square?
Or have you switched 6 and 8 ?

168:56=3

and it looks more like four 645 sensor put side to side in vertical position :

42*4=168 (mm)
AND
7000*4=28000 (pixels)

Did I win anything ?

Sorry if I am mistaken, and continue your great job.

Auguste.

[Steve Sanacore]

I think the RED FF 35mm cameras are to compete in the Still Photography market. For guys like me who do both, it's perfect. For movie work, I think the S35 bodies will still be the industry standard because they can use traditional cine lenses, which the FF bodies cannot unless you shoot in cropped mode.

[David Rasberry]

Another problem with larger sensors for motion picture work is the increased size, weight and cost of glass to cover them for equal speed.
As an industrial doc shooter, the 2/3" is going to work best for me. A good fast broad range zoom is much less expensive and more compact for this format than even S35. FF35 is impractical. The larger sensors work better for still shooters.

[Jeff Kilgroe]

Auguste, your math is off Or not really off, you are just trying too hard. The 617 sensor is 3:1, that is the aspect and the pixels are square. 9334 x 3 = 28002. It's a large, panoramic format that has yet to be served by a digital solution.


I think the RED FF 35mm cameras are to compete in the Still Photography market. For guys like me who do both, it's perfect. For movie work, I think the S35 bodies will still be the industry standard because they can use traditional cine lenses, which the FF bodies cannot unless you shoot in cropped mode.

The FF35 is obviously good for still photography. However, I think most of the filmmaker crowd with the budget will opt for the FF35 models over the S35 models. Even with a PL mount installed and using S35 glass, the FF35 sensor provides several benefits, such as the higher DR and greater vertical sensor area. This is a huge benefit for shooting with anamorphic glass or using custom aspects and framing modes.

[Stuart English]

I think the RED FF 35mm cameras are to compete in the Still Photography market. For guys like me who do both, it's perfect. For movie work, I think the S35 bodies will still be the industry standard because they can use traditional cine lenses, which the FF bodies cannot unless you shoot in cropped mode.

Agreed a S35 sensor is the perfect fit, but there is no reason not to shoot a FF35 sensor in cropped mode with a PL mount and standard cinema lenses.

That gives virtually the same DOF, resolution and post production options as on the RED ONE.

[Steve Sanacore]

Yes but Jeff, does standard 35mm cine glass cover a 24x36mm sensor? I didn't think it did.

[Auguste Langevin]

The 617 sensor is 3:1, that is the aspect and the pixels are square. 9334 x 3 = 28002.

Really sorry to insist, but I am okay with what you just said (aspect in pixels)
But the size is 186*56 mm (1:3.32)
Is there some sort of anamorphic optic on that one then ?

I really don't know anything on 617 format, is it named after centimeter dimensions : 6*17 cm ? (like 645 is 6*4.5 cm)
If digital sensor is a little smaller than film, as 56 and 42 mm is smaller than 6 and 4.5 cm, I would again say 168 mm width, smaller than 17 cm.

Sorry for bothering you, but the newbie I am is completely lost here.

[Stuart English]

does standard 35mm cine glass cover a 24x36mm sensor?

No, but it will cover enough of the FF35 sensor to allow capture at 4K RAW resolution.

[Jeff Kilgroe]

Yes but Jeff, does standard 35mm cine glass cover a 24x36mm sensor? I didn't think it did.

In a horizontal sense, you are correct, but think vertically. You're leaving a lot hanging there with that 24mm height. Anyone who shoots anamorphic glass will want the FF35 as it will allow them greater vertical resolution. With anamorphics, that translates to greater overall resolution for the image.

[Antoine Baumann]

There isn't a camera out there that I know of for high end motion or still photography with a sensor that isn't natively daylight balanced.

I think you are wrong, the sony f35 seems to be native 3200k, at least that's what Art Adams is saying here:

http://provideocoalition.com/index.php/aadams/story/my_first_shoot_with_the_sony_f35/


ciao,
antoine.

[Paris Remillard]

Hi Tom,

Needs loads of ND filters, hardly progress, tungsten friendly would be more useful,

Stephen

Well, if you've got 1 1/2 extra stops, then you could throw on an 80b or 80c without penalty. Now you've got a tungsten balanced sensor with the same sensitivity of the current sensor.

[Daniel Browning]

Another problem with larger sensors for motion picture work is the increased size, weight and cost of glass to cover them for equal speed.


With a larger sensor, you don't need equal speed: you can get the same or better quality even with a proportionately narrower focal ratio (e.g. T5.6 on FF35 is the same as T1.5 on 2/3").

When light is ample, less ND with smaller apertures will yield more light in total and a higher quality image. In low light, the quality will be equal despite the smaller focal ratio because of the greater area. Of course, if you're paying extra for the larger sensor just for low light capability, then you wont get that benefit unless you also get larger glass. Also, if the larger sensor has lower QE per square micron, the focal ratio must be slightly wider to compensate.

Focal ratio (e.g. T1.3) by itself does not determine light gathering power: only the intensity of light per area. The focal length (e.g. 50mm) divided by the f-number gives the aperture (50/1.3=38.5mm), which is the true light gathering ability of the lens. Combine that with the area of the sensor, QE, and read noise to determine the low light sensitivity.

That's why, for example, you get the same amount of light with a 50mm T5.6 on FF35 as you do with a 14mm T1.5 on a 2/3". The only difference is the QE and noise of the sensor per area; the 2/3" most likely has higher technology in those areas.



As an industrial doc shooter, the 2/3" is going to work best for me. A good fast broad range zoom is much less expensive and more compact for this format than even S35. FF35 is impractical. The larger sensors work better for still shooters.

2/3" is going to work great for me too because of the price. In your situation, you're already buying the largest lenses that you want, so getting a larger sensor would, at best, only give you equal quality in low light. It would be better in ample light, but I bet no one even makes a lens that meets your needs.

[Brook Willard]

I plan on buying a 645. To me, the benefits are both creative and functional.

Functionally, I'll be able to shoot in a 24.89 x 14mm four-perforation mode. I'll also be able to do proper anamorphic sizes as well.

Creatively, it'll open me up to using all sorts of different lenses, formats and trying something new. I'll be able to slap that Nikkor 6mm on there without an adapter to get the whole circle in a shot. I'll be able to build a 65mm bayonet mount and use Imax lenses [the 645's sensor is closer to Imax size than the RED ONE's 4K recording area is to S35mm...], etc. I'm excited for both technical nerd reasons and artistic reasons... that thing is going to be great.

So is a larger sensor progress? It may not necessarily be the area that needs progress the most... but it sure isn't a step backwards.

[Stephen Williams]

Well, if you've got 1 1/2 extra stops, then you could throw on an 80b or 80c without penalty. Now you've got a tungsten balanced sensor with the same sensitivity of the current sensor.

Hi,

Adding an extra piece of glass never increased quality, needing 2 or 3 filters most of the time can't be a good thing.

Stephen

[Häakon]

A larger sensor is absolutely progress. Larger photosites mean less light-starved pixels, which in turn means better noise performance and extended dynamic range. It also means greater resolution, which has the obvious effect of superior detail rendition. Yes, it will mean that eventually you will need glass that covers a larger sensor area than traditional 35mm PL-mount optics. The sooner you can get over that, the sooner you can relish in the aforementioned advantages that a larger sensor brings. If DOF is a concern, there are a plethora of 1/3" chip cameras on the market that can assist you tremendously in this area.

[Brook Willard]

Adding an extra piece of glass never increased quality

Unless you're using the oft-misunderstood "Enhancer" filter... :clown2:

[Dominic Jones]

Well, if you've got 1 1/2 extra stops, then you could throw on an 80b or 80c without penalty. Now you've got a tungsten balanced sensor with the same sensitivity of the current sensor.
But when was the last time you had spare stops for interiors??! I know you´re assuming the sensor will be a stop and a half faster, and I hope you´re right, but even if so (and personally I think it´s unlikely we´ll see anything like that drastic an increase in sensitivity) that´s going in the wrong direction...

You´re almost always fighting for stop indoors (which is when you´re normally shooting tungsten), and you normally need to filter exterior shots to get a sensible stop anyway - so you´d be much better off with a tungsten balanced sensor and swapping your 0.9ND for an 0.6/85B combo (for instance) when shooting exteriors, where light is plentiful.

Stills cameras are daylight balanced (as are nearly all stills stocks) as flash is 5600K. That´s where the DSMC idealogy falls down slightly - drama, vfx and promo shooters will vastly prefer a 3200K sensor, stills will want 5600K...

[Daniel Browning]

Stills cameras are daylight balanced (as are nearly all stills stocks) as flash is 5600K.

The reason has nothing to do with flash, it's because the light itself has less blue in it. The sensor can't see something that isn't there.

A so-called "tungsten balanced" sensor is more blind to red and green so that they are just as weak as blue.

[Dominic Jones]

I don't know much about the technical details of sensor production, but the reason that stills stocks have pretty much always been daylight balanced absolutely is due to the use of 5600K flash lighting vs 3200/3400K continuous for moving image interiors - there's no problem with producing tungsten balanced emulsions, after all, yet virtually none are made for stills use.

My point was simply that there is no call for a sensor in the digital stills world balanced to anything other than ~5600K, whereas there is in the motion picture industry. The technical challenges of actually producing such a sensor I do not claim to have any knowledge about, and will happily leave such discussions to others...

[oldphart]

I don't know much about the technical details of sensor production, but the reason that stills stocks have pretty much always been daylight balanced absolutely is due to the use of 5600K flash lighting vs 3200/3400K continuous for moving image interiors - there's no problem with producing tungsten balanced emulsions, after all, yet virtually none are made for stills use.

My point was simply that there is no call for a sensor in the digital stills world balanced to anything other than ~5600K, whereas there is in the motion picture industry. The technical challenges of actually producing such a sensor I do not claim to have any knowledge about, and will happily leave such discussions to others...

You could change the balance of the sensor by using two blue sub-pixels to one red and one green instead of the two green of the Bayer pattern. It would be very bad in all other situations than low temperature artificial light.

Incandescent light for high-intensity uses is obsolescent. With modern light sources, you will get much better efficiency and presice control with a daylight colour balance. Cars got halogen headlights first, then it filtered down into different applications. Ten years ago, cars got xenon headlights, it has filtered into a number of applications. Now, cars have LED-based parking and direction lights, and the overhead reading lamps in buses are LED. High-output LEDs are starting to appear in specialty applications like surveillance searchlights. There are already low-power LED movie lights on the markets.

In a few years, homes and offices will be lit with a new generation of solid state lights. In places with very high electricity prices, LED-based lamps are already cost effective.

The halogen lamp fits perfectly with the "buy cheap and throw away" ideology that is now grinding to a halt. Now, sustainability is getting more attention, and a high initial price which is amortized over a very long lifespan will become more attractive. Besides, it is not obvious that LED lights will remain much more costly for very long. The abominations sold as "energy saving" mini-fluorecents were expensive only a few years ago.