Jump to content

Your DREAM SPECS for a Digital Cinema Camera?


Tom Lowe

Recommended Posts

There has been some interesting discussion lately in the Epic forum at Reduser about possible specs for a next-gen digital cinema camera. Some of the debate revolves around increasing the size of the sensor from the "legacy" 35mm cine-size negative, which is roughly an APS-C "cropped" sensor, to a "full-frame" sensor.

 

My dream specs:

 

- "Full frame" chemical still or APS-size sensor (roughly 36mmx25mm), which is also roughly the size of the legendary Vista Vision format. Who hasn't dreamed of shooting on Vista Vision?

 

- 6K resolution, which would easily downsample to pristine 4K for finishing

 

- RAW

 

- 60 to 120fps

 

- Non-linear recording on HD or solid state

 

A lot of the criticism about this larger format involves trying to pull focus on a sensor that size. But hey, the Vista Vision guys did it, right?

 

What are your dream specs?

Link to comment
Share on other sites

  • Replies 93
  • Created
  • Last Reply

Top Posters In This Topic

  • Premium Member

65mm size sensor. If you've ever compared the look of a 65mm lens and its 35mm equivalent, you'll know what I mean. The longer lenses one uses in 65mm make everything just look nice, without distortion.

 

Highlights that blow out gracefully like film, not that ugly clipping that the Red and any other digital camera show.

 

Random sampling, to avoid aliasing, like film

 

Better and more natural color rendition especially skintones, like film.

 

But I'm not holding my breath, film offers me all those things already.

Link to comment
Share on other sites

Six chips of 65mm size

a complete avoidance of codecs by storing all data as single-frame TIFF's (32-bits per color)

A solution to the color curve and blowout issue (solved by the 6 chips, which are 3 RGB + 3 RGB highlight)

A physical mirrored shutter, no electronic junk in the way

Storage directly to a RAID HD array

No audio stored on the unit

Link to comment
Share on other sites

OK, dreaming is free.

 

So, if it's a DREAM camera, then it should definitely have:

 

- 65mm sensor which should be:

- 8K [if CMOS Bayer], so that we could have full-luma+full-chroma mindblowing 4K out of it.

- or a 4K [future FoveoNx3 type of cinema sensor]

- RAW. There's no going back from RAW, that's for sure.

- 120 fps at 8K.

[OMG, uncompressed RAW from 8K would be... 1,3 GB\sec for 24 fps and... what? - ~ 6GB\sec for 120 fps... OK, that's futuristic :) ]

- Random sampling.

- Record to SSD drives.

- The camera should be called: "RED Lawrence" :)

- under 30k $... Ok, under 50k $ :)..... O-oh, ok, under 100k $ :(.

 

...

 

But for the moment... A Scarlet would do :).

Link to comment
Share on other sites

  • Premium Member
Highlights that blow out gracefully like film, not that ugly clipping that the Red and any other digital camera show.

For that to happen, either:

 

  • Manufacturers would have to find a practical way to give silicon sensors a non-linear response, like film. They are currently hampered by the fact that nobody has the slightest idea how that could be accomplished! (Well, they understand what would need to be done, but nobody has any idea how this could be turned into a manufacturable product. The problem is that all the trouble happens in the analog part of the chip, and very little R&D is being done in that field, because commercial applications are somewhat limited).
  • Manufacturers would need to find a way to drastically reduce the noise floor of the sensor, which would allow the iris to be stopped down and so allow the sensor to linearly capture more highlights, which could then be "massaged" to produce a film-like response curve. However current generation silicon sensors appear to to be as sensitive as they are ever going to get. Once the incident photon level drops below a certain point, the noise from vibration of the silicon atoms starts to become intrusive. The chip would become exquisitely sensitive to other forms of electrical interference as well.

Random sampling, to avoid aliasing, like film

 

That (or at least the desired outcome) is actually do-able, by simply increasing the pixel count.

 

 

Better and more natural color rendition especially skintones, like film.

Unfortunately, manufacturers have to live with the physical properties of materials that are available to them. The dye filters used on a single-chip colour filter not only have to produce the right spectral response, they have to be screen-printable and UV-hardenable, and stand up to years of exposure to light without fading or discolouring. Silicon also has a far-from-ideal spectral response and the necessary IR filtering further upsets this.

 

Film dyes are dissolved in a water-based gel which sets on the film base at room temperature, there is a much greater range of chemicals to choose from, and each individual frame of the negative usually only sees less than a second's exposure to light in its entire operational life.

 

Other problems with skin tones are that our eyes are exquisitely attuned to recognizing faces and detecting subtle discolorations that may indicate disease, so we notice colour errors far more on human skin than anything else. Also real faces are not at all like the face of a painted mannequin; skin is actually a 3-D translucent gel with an enormously variable reflectivity range. (Which is why TV studios routinely apply half an inch of makeup on some people :lol:)

 

 

But I'm not holding my breath, film offers me all those things already.

Which is why there is no great incentive for video camera manufacturers to try to solve these problems. On the global scale, the market demanding film capture is very small, and there already is a workable technology: film!

 

However, various manufacturers' attempts to provide "film-like" performance from video cameras over the years has certainly given us much better (and cheaper!) video cameras, so even if you have never shot a frame of film in your entire career, you are still benefitting from it :lol:

Link to comment
Share on other sites

  • Premium Member
A solution to the color curve and blowout issue (solved by the 6 chips, which are 3 RGB + 3 RGB highlight)

There has been a lot of discussion about that on Reduser. Can I just cut to the chase and point out that that simply will not work. :lol:

Everybody seems to under the impression that you can simply discard the information from the overloaded pixels on the "full aperature" sensor, and substitute the equivalent pixels from the "ND'd" sensor. This is a very old idea. The trouble is overloaded sensors do not just overload, they inject stray current carriers into the non-overloaded sensors and degards them as well. This is the effect known as "blooming".

While you can theoretically get a small improvement in dynamic range this way, this is usually nullified (or more than nullified) by light losses and/or degradation of the signal to noise ratio by extra circuitry needed on the sensor.

 

While the Arriscan uses a variant of this technique in its "double-exposure" system, it does not scan film in real time, and remember, a couple of stops improvement in DR from scanned film equals a lot more stops in the actual shooting conditions, and so is well worthwhile.

Link to comment
Share on other sites

  • Premium Member
What are your dream specs?

I think a lot of people would get very excited about something like a Genesis, (ie with all its "plug and play" industry workflow compatibility) but with a RED-type front end and a RED-type price.

 

That's probably all the industry needs at the moment.

Link to comment
Share on other sites

  • Premium Member

If you're still busy dreaming about a camera that's superior to the choices that already exist, then that camera won't be good enough either once it becomes real. You have tools available to you now that people could only dream of a short time ago.

Link to comment
Share on other sites

There has been a lot of discussion about that on Reduser. Can I just cut to the chase and point out that that simply will not work. :lol:

Everybody seems to under the impression that you can simply discard the information from the overloaded pixels on the "full aperature" sensor, and substitute the equivalent pixels from the "ND'd" sensor. This is a very old idea. The trouble is overloaded sensors do not just overload, they inject stray current carriers into the non-overloaded sensors and degards them as well. This is the effect known as "blooming".

While you can theoretically get a small improvement in dynamic range this way, this is usually nullified (or more than nullified) by light losses and/or degradation of the signal to noise ratio by extra circuitry needed on the sensor.

 

While the Arriscan uses a variant of this technique in its "double-exposure" system, it does not scan film in real time, and remember, a couple of stops improvement in DR from scanned film equals a lot more stops in the actual shooting conditions, and so is well worthwhile.

If you'd notice, my design would also not be in real-time. It would be storing the frames, for computers to work with later on. I am quite aware that you cannot simply discard, but I so know that with enough time and computing power, it can be done. I've even done it myself, but yes, it takes a lot of CPU time.

Link to comment
Share on other sites

Nate, I guess we can forget portability with your system?

No reason why you would, as there's a lot less parts within it. The RAID array would be the largest component, and it would be roughly the size of a 200' 35mm magazine. It's size should be roughly the same as the Aaton. By removing as much as possible, and instead focusing on just raw recording,

Link to comment
Share on other sites

  • Premium Member

Would it be wrong to ask for a digital camera that gets amazing battery life!

 

Seriously, though, my "dream camera," would be something high-rez with interchangeable lenses, which is affordable enough to employ in "dangerous," situations, with a small form factor.

I don't want a camera to "kill film," rather an HD content producing machine akin to the current crop of JVC/HVXs etc, that I can slap a PL lens on or a Nikon still w/o a wonky adapter and get just a bit more control of my image (something with a 2/3' chip, or hell, a 35mm chip. )

Of coure, I know that's a pip dream for now.

 

 

And/or a camera which can make me coffee. Not that it would be useful at all most of the time. .. but for those long days a built in peculator would be a dream come true :lol:

Link to comment
Share on other sites

Manufacturers would have to find a practical way to give silicon sensors a non-linear response, like film. They are currently hampered by the fact that nobody has the slightest idea how that could be accomplished!

 

This issue has been explored. It is an active research area. Please search online for nonlinear CMOS sensors. It is not terribly difficult to add an analog nonlinearity to sensors.

Edited by DJ Joofa
Link to comment
Share on other sites

  • Premium Member

> Six chips of 65mm size

 

I think you only need two, if they're bayered and sufficiently high res, and 65mm is just wannabe chasing of an expensive format. 2/3" video is perfectly usable for dramatic work and full ap 35 is actually a pain in the neck depending on sensitivity; going any larger than that is just masturbatory "I want to shoot 65" without actually having given any thought to the grief you'll cause yourself.

 

> a complete avoidance of codecs by storing all data as single-frame TIFF's (32-bits per color)

 

Craziness. All this drivelling on about high bit depth is daft - few modern video cameras have more than four or five bits of real data. Absolutely none of them have more than eight, and yet people blast on about ten bit like it's more important than oxygen. For grief's sake just record it eight and diffusion dither it up to whatever you want - at least the noise will be gaussian that way!

 

> A physical mirrored shutter, no electronic junk in the way

 

I'd rather have a viewfinder that showed me what the image actually looked like, but ideally a hybrid design with information optionally superimposed over the image using a transreflective display would be ideal. Flickery mirror shutter cameras give me a headache. Personal preference I understand, but all this banging on about a colour viewfinder is gibberish when it previews neither colour, exposure or sharpness with any accuracy.

 

> Storage directly to a RAID HD array

 

Surely you want flash.

 

> No audio stored on the unit

 

This is just curmudgeonliness. Single system sound takes expensive make-work drudgery out of postproduction and should be used unless there's a reason why not - and the only reason why not is if the radio link broke down on the use take.

 

This is a lovely list of the requirements of the dedicated film luddite who's come up with a spec list based on nothing more than the fact that it's difficult to meet. It's not very practical.

 

P

Link to comment
Share on other sites

Here here to Keiths post!

 

 

Fuji tried a "ND filter" (my description) on neigbouring pixels last year in still cameras and a DPreview exhaust test indicated it wasn't a success.

 

 

Obviously those shooting docs will have a different wish list than those shooting drama.

But both genres would like as much control of depth of field as possible.

 

By stipulating a 65mm sensor one is limiting control of depth of field on one direction whereas 2/3 inch limits it in the other.

 

Ultimate control of depth of field (in theory) is most likely to come from a camera with an interchangeable or adjustable sensor, say from 2/3 inch to 65mm. the challange is to do this and still keep a common texture resolution noise signature ect of the image.

 

Here is a relevant question. What percentage of 35mm dramas regularly shoot T1.5? im not saying that T1.5 on 35mm doesn't have a place it does!

 

I believe that a 1 1/3 inch sensor with T 1.5 lenses is more practical and would satisy both day to day narrow depth of field requirements (lets say equivelant in 35mm shooting T2.8) but also offer greater scope in achieving deeper depth of field required for docs.

 

For many applications 2/3 has too much and 35mm too little, what about 1 1/3?

 

 

The brave, so called, revolutionary world of new digital cinema cameras is still being tied to the legacy of a sensor size based on achieving acceptable film grain.

 

I'd love to see a digtial cinema camera design that starts from scratch that is not dependant on integration with existing film lenses.

 

RED 3ks format is interesting in its size is around 1 1/3 (?) it just needs more res and a set of dedicated lenses :)

 

 

 

 

Mike Brennan

Link to comment
Share on other sites

  • Premium Member
This issue has been explored. It is an active research area. Please search online for nonlinear CMOS sensors. It is not terribly difficult to add an analog nonlinearity to sensors.

Please do not quote out of context, which includes truncating quotes. What I said was:

 

Manufacturers would have to find a practical way to give silicon sensors a non-linear response, like film. They are currently hampered by the fact that nobody has the slightest idea how that could be accomplished! (Well, they understand what would need to be done, but nobody has any idea how this could be turned into a manufacturable product. The problem is that all the trouble happens in the analog part of the chip, and very little R&D is being done in that field, because commercial applications are somewhat limited).

 

It's one thing to make small numbers of non-linear CMOS sensors with matched characteristics, quite another to make 12 million of them! Or even 2 million or 640 x 480.

 

Can you answer the obvious question: "If it can be done, then why isn't anybody doing it?"

Link to comment
Share on other sites

  • Premium Member

There's a lot to be said for a camera system that can use currently available optics and not need a dedicated line...

 

As for non-linear response, surely it would be better and probably ultimately simpler perhaps to someday increase the linear range recorded so you can apply whatever gamma curve you want. Color negative film holds about 14 stops of dynamic range so in theory if you could record more than that, let's say 16 stops of range, as a linear signal, you could emulate the 14-stop curved gamma look of film.

 

One argument I've discovered for an optical viewfinder is when you've got a 90-second boot-up time with the RED, it can be frustrating to grab a powered-down RED camera and quickly set-up a shot, only to find you can't see anything until it's done booting. Film cameras do have the advantage of only drawing power (other than for some accessories) when actually triggered, and don't have to be powered to look through while setting up a shot.

 

But personally I don't have a problem with the concept of electronic viewfinders, since it's nice to see the signal that will be recorded... just that the reality of electronic viewfinders on the market leave a lot to be desired. Ideally there would be a hybrid approach as Phil mentioned, an optical image and an electronic image, either side by side, or overlaid in some manner, whatever. Of course, you could have an optical viewfinder and a decent onboard HD monitor.

Link to comment
Share on other sites

  • Premium Member
As for non-linear response, surely it would be better and probably ultimately simpler perhaps to someday increase the linear range recorded so you can apply whatever gamma curve you want. Color negative film holds about 14 stops of dynamic range so in theory if you could record more than that, let's say 16 stops of range, as a linear signal, you could emulate the 14-stop curved gamma look of film.

 

Which is I what I just said 14 posts back.

 

Plenty of peolle know what needs to be done, but nobody has any practical means of actually doing it as yet.

If you are talking about the point where no further useful information can be obtained from colour negative by any means, film will do considerably than 14 stops. However, that is about the point where it becomes impractical to build lenses that can provide enough contrast to take advantage of it. You are talking about a situation where the ratio of brightness between the darkest and brighest pixels on the sensor is about 16,000 to one! It is simply not possible to make the inside of the lens barrel and the camera dark enough.

 

The record for the least reflective substance known was recently developed by researchers at Rice University. Based on carbon nanotubes, it only reflects only about 1/2,000th of the light falling on it, which is only 11 stops!

Link to comment
Share on other sites

Can you answer the obvious question: "If it can be done, then why isn't anybody doing it?"

 

Trust me, people are actively working on it. If you have not seen them mass produced then that does not mean that there are insurmountable obstacles to it. Several experimental non-linear sensors have been fabricated.

 

Please refer to the following publication as an example where the authors claim that they implemented a 120 db DR !!! sensor. (Using the standard notion of 6db per bit, that would turn out to be 20 bits, or 20 stop, well into HDR range!)

 

Hara, K. Kubo, H. Kimura, M. Murao, F. Komori, S., "A linear-logarithmic CMOS sensor with offset calibration using an injected charge signal", Solid-State Circuits Conference, 2005. Digest of Technical Papers. ISSCC. 2005 IEEE.

 

Enough work is being done actively in this area. This technology is just maturing and sooner or later you will see the results.

Edited by DJ Joofa
Link to comment
Share on other sites

I'd actually be quite happy if my D3 or something like it could do 24 fps sustained for even short Bolex-like bursts, 4K (3.2 whatever) Bolex or Digital Eyemo.

 

Yes I know the shutter would not survive long at that rate and would certainly mean "mirror up" (I've thought for a long time even film SLR's could benefit from spinning mirror vibration and sound wise - drive em like a CD walkman)

 

And even 24 as a limit would work for me (36 would be nice) I do weird things with Shake....

 

Obviously this is not a "production camera" but the as thread says dream specs I'm being conservative !

 

The form factor is beneficial to me and I frankly love shooting motion imaging sequences on a 24 x 36 area; I'll save medium format motion digital for the upgrade dream B)

 

Yes I want more DR, highlight control (sensitivity I'm happy now: EI 800 without a second thought, 1600-3200 as a "push")

 

Now do I get dream software & computer to go along ?

 

-Sam

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Forum Sponsors

Metropolis Post

New Pro Video - New and Used Equipment

Gamma Ray Digital Inc

Broadcast Solutions Inc

Visual Products

Film Gears

CINELEASE

BOKEH RENTALS

CineLab

Cinematography Books and Gear



×
×
  • Create New...