Let's talk sensor...

[Häakon]

I have been so busy shooting that I haven't had a proper chance to reply to this thread as I had wanted to several weeks back. I would, however, like to thank Brook for taking the time to make the comparison drawings.

A lot of people seem to think that a 4:3 sensor would make a superior "Epic," though I have trouble understanding why as the majority of motion pictures are neither shot anamorphically or in a 4:3 ratio and you will never get more resolution from a chip than shooting in its native space. Brook's renders really help to illustrate something that I have attempted to bring to light several times but haven't really been able to convey properly: that - as shown in Brooks' final render (where even he declares a 4:3 Epic sensor is how things "should be") - when shooting a movie in a 2.40:1 ratio, a 4:3 Epic sensor (at 24.89 x 18.66mm recording area) gains you next to NOTHING in image area compared to a RED ONE shooting 2.40 in the upcoming 4.5K mode. This is SO important, because if you are shooting with spherical lenses the only gain you would receive from such a seemingly mammoth sensor increase is negligibly higher resolution (and a ton of vertical crop room). In fact, the wasted sensor area is so massive that you would likely have to sacrifice higher framerates or other technical improvements - things that clearly many users have made a demand for - just to record a 4:3 image.

I understand why those who appreciate an anamorphic image are trying to champion a 4:3 sensor in future cameras, but what they don't seem to understand is there is not a limit on physical sensor size in which we need to cram pixels as there is with 35mm film. The obvious solution for shooting high resolution widescreen movies is a high resolution widescreen sensor; you can still shoot anamorphically on such a chip, you would just sacrifice the extra area on the sides of the target area rather than the top and the bottom as you would have to do to shoot spherically on a 4:3 sensor.

Larger, wider sensors - which, yes, just like still film will require lenses that cover a larger area - are the future for high-resolution digital acquisition. Whether or not RED will be the company which leads that revolution remains to be seen, but I believe that forcing ones' self to think inside a 35mm box is going to be a major handicap in the years to come.

 

[Bob Torrance]

- when shooting a movie in a 2.40:1 ratio, a 4:3 Epic sensor (at 24.89 x 18.66mm recording area) gains you next to NOTHING in image area compared to a RED ONE shooting 2.40 in the upcoming 4.5K mode.

I don't follow. Wouldn't a 4:3 sensor (full frame 4 perf) just include more top and bottom area that isn't planned for anyway? And it would give you a lot more verticle resolution with anamorphics (2:1). If you wanted to shoot straight 2.40 you would just use what is already there (5K).

I'm I wrong?

bob

 

[Poi Boy]

Thanks for the great thread Brook ! I'm with Haakon on this,
I think going 4.3 would be a mistake. Give me larger and wider please.
Aloha
-A

 

[Häakon]

I don't follow. Wouldn't a 4:3 sensor (full frame 4 perf) just include more top and bottom area that isn't planned for anyway?

Yes, and that's exactly the problem - it's completely wasted space that most of the time you are likely never going to use. The only gain you'd get from buying a 4:3 Epic, then (if you aren't shooting anamorphically), is an increase from 4.5K to 5K. Is that worth a $23,000 difference to you? It sure isn't to me.

Here's a graphical example of the differences:

Note that these aren't to scale, it's just a visual representation of what's being discussed here. The first image is the RED ONE sensor - in a 4.5K 2.40 mode (which Jarred has indicated is forthcoming). We get to dip into the overscan area and use the maximum width of the sensor to record a great widescreen image.

The second (yellow) image shows Brook's proposal of a 4:3 sensor which has nearly the same recordable width as the RED ONE's. When shooting with spherical lenses, we gain almost nothing with this larger format sensor except a ton of room above and below the shot which will just be thrown away. Again, sure, a sensor like this would net us much more resolution if shooting with anamorphic lenses - but only in that case.

The third (blue) option is to develop a sensor that is physically larger than the 35mm standard we've been forced into for the last century. Since nearly all films today are shot in a widescreen aspect ratio (be it 1.85:1 or 2.40:1), it makes zero sense to have a sensor that is 4:3. Look at the size increase between the RED ONE and the 2.40 Epic - it's massive. Compared to the proposed 4:3 Epic, there's just no comparison. One could still use the middle 4:3 portion of the sensor to shoot anamorphically - and you'd get identical vertical resolution to the spherical widescreen image. The best of both worlds. Plus, larger sensors of the same resolution mean larger pixels - which translates into better sensitivity, less noise, and I would guess more room to improve dynamic range.

The still world already has large sensor (medium/large format) cameras which take full advantage of these properties. In fact, even the 35mm still "full-frame" size is significantly larger than the current RED offering. Also, for those who believe that 4K is the "most resolution we need" right now, keep in mind that oversampling is king. Even if you never end up finishing for anything larger than a 4K print in your lifetime, it is always better to have a larger canvas to work with and then downsample in post. You smooth out aliasing, reduce noise, and get an apparently sharper image as a final result. It's the same reason that you'd want to shoot with an 8MP dSLR rather than a 1MP one - even if your final output is only going to be 1MP. The larger image always looks better oversampled than the native one.

I realize that pushing the envelope in this way requires new thinking, new lenses, and a different approach than what is currently "standard." But we will never truly get an "epic" digital cinema camera until we break through those perceived limits and make an imager that goes beyond conventional "rules." There will be a limit to how many pixels one can cram into a 35mm sized box (one could argue we're already there), and thus enlarging the sensor size itself is the logical next step of evolution. RED is a company who in my mind prides themselves on not being constrained by what is deemed a physical limit but who instead smashes those perceptions and delivers above and beyond what is expected or even considered possible. They've given us 4K on a compact flash card for $17.5K - even Jim himself didn't imagine such a scenario when he first dreamed up his plans for the RED ONE two years ago. Maybe this is where "Monstro" needs to go... I'm not sure what they have cooking. But to me anyway, a 4:3 sensor of the same width that we've already got and a negligible resolution increase isn't worth double the price and it's not something I'd call "Epic." RED ONE was absolutely revolutionary in my mind, and I believe the next revolution will come when we shatter the 35mm barrier.

 

[Joel Kaye]

but I believe that forcing ones' self to think inside a 35mm box is going to be a major handicap in the years to come.

That's right. RED's biggest opportunity is leaving film as it is in the dust by going with bigger sensors and using full frame lenses. They'd instantly obsolete all the pl mount lenses out there and create demand for their own lenses which could be based on full frame SLR lenses. That's what I'd do - forget film even exists and ask "what's the best way to make a great moving picture".

 

[Joseph Ward]

If not 4:3 what about 2.76:1 or 4.00:1 :weight_lift: lol

 

[Häakon]

If not 4:3 what about 2.76:1 or 4.00:1 :weight_lift: lol

It's actually a good question - what is the best sensor ratio? Because there are a multitude of aspects in which movies are shot (and in the digital age, the choices are truly limitless), clearly there isn't a true "one size fits all" option. Because the majority of motion pictures are shot in either 1.85:1 or 2.35:1, however, and because broadcast HDTV is by definition 16:9 (1.77:1), it makes the most sense to make the sensor similar in aspect to those three approaches as you retain the highest possible resolution for your production. Keeping it 16:9 makes the most sense to me personally as it covers all three bases and allows for slight vertical reframing in post for 2:1 or 2.35:1 productions.

 

[I Bloom]

Yes, and that's exactly the problem - it's completely wasted space that most of the time you are likely never going to use. The only gain you'd get from buying a 4:3 Epic, then (if you aren't shooting anamorphically), is an increase from 4.5K to 5K. Is that worth a $23,000 difference to you? It sure isn't to me.

Well, is it 4.5K vs 5K or 4.5K vs 5.5K, I guess we'll see.
Have we gotten so spoiled that one thousand extra pixels is something we scoff at. I hope not, whether it's worth 23K USD is another question. My guess is you'll get your money's worth from Monstro in other ways. Like another digital cinema company (now defunct I think) had a slogan I liked "It's not about counting pixels it's about making pixels count."

Here's a graphical example of the differences:

Note that these aren't to scale, it's just a visual representation of what's being discussed here. The first image is the RED ONE sensor - in a 4.5K 2.40 mode (which Jarred has indicated is forthcoming). We get to dip into the overscan area and use the maximum width of the sensor to record a great widescreen image.

The second (yellow) image shows Brook's proposal of a 4:3 sensor which has nearly the same recordable width as the RED ONE's. When shooting with spherical lenses, we gain almost nothing with this larger format sensor except a ton of room above and below the shot which will just be thrown away. Again, sure, a sensor like this would net us much more resolution if shooting with anamorphic lenses - but only in that case.

The third (blue) option is to develop a sensor that is physically larger than the 35mm standard we've been forced into for the last century. Since nearly all films today are shot in a widescreen aspect ratio (be it 1.85:1 or 2.40:1), it makes zero sense to have a sensor that is 4:3. Look at the size increase between the RED ONE and the 2.40 Epic - it's massive. Compared to the proposed 4:3 Epic, there's just no comparison. One could still use the middle 4:3 portion of the sensor to shoot anamorphically - and you'd get identical vertical resolution to the spherical widescreen image. The best of both worlds. Plus, larger sensors of the same resolution mean larger pixels - which translates into better sensitivity, less noise, and I would guess more room to improve dynamic range.

Aren't you just magnifying the same image over a wider area? Which translates to less light per photosite, which balances out the additional noise of a compact sensor?
And also doesn't it depend in part on the coverage area of the lenses. If you have lenses that can cover this new format you are proposing then alright. Otherwise that sounds like suicide for Epic. Part of the genius of Red One was that they dipped into the existing pool of 35mm lenses. A 4:3 Epic would have the same pool of lenses. Likewise depth of field is an issue on larger formats that make them undesirable for practical cinematography?

The still world already has large sensor (medium/large format) cameras which take full advantage of these properties. In fact, even the 35mm still "full-frame" size is significantly larger than the current RED offering. Also, for those who believe that 4K is the "most resolution we need" right now, keep in mind that oversampling is king. Even if you never end up finishing for anything larger than a 4K print in your lifetime, it is always better to have a larger canvas to work with and then downsample in post. You smooth out aliasing, reduce noise, and get an apparently sharper image as a final result. It's the same reason that you'd want to shoot with an 8MP dSLR rather than a 1MP one - even if your final output is only going to be 1MP. The larger image always looks better oversampled than the native one.

I realize that pushing the envelope in this way requires new thinking, new lenses, and a different approach than what is currently "standard." But we will never truly get an "epic" digital cinema camera until we break through those perceived limits and make an imager that goes beyond conventional "rules." There will be a limit to how many pixels one can cram into a 35mm sized box (one could argue we're already there), and thus enlarging the sensor size itself is the logical next step of evolution. RED is a company who in my mind prides themselves on not being constrained by what is deemed a physical limit but who instead smashes those perceptions and delivers above and beyond what is expected or even considered possible. They've given us 4K on a compact flash card for $17.5K - even Jim himself didn't imagine such a scenario when he first dreamed up his plans for the RED ONE two years ago. Maybe this is where "Monstro" needs to go... I'm not sure what they have cooking. But to me anyway, a 4:3 sensor of the same width that we've already got and a negligible resolution increase isn't worth double the price and it's not something I'd call "Epic." RED ONE was absolutely revolutionary in my mind, and I believe the next revolution will come when we shatter the 35mm barrier.

I love the EPIC section. Reminds me of the good ol' days before August '07.

 

[Miguel "Macgregor" De Olaso]

Uhmm, remember that anamorphic film gate is smaller than S35mm.
Now if we use anamorphic lenses in a 4:3 S35mm sensor we would get an image that is...

 

[Ariana]

It depends in part on the coverage area of the lenses. If you have lenses that can cover this new format you are proposing then alright. Otherwise that sounds like suicide for Epic. Part of the genius of Red One was that they dipped into the existing pool of 35mm lenses. A 4:3 Epic would have the same pool of lenses. Likewise depth of field is an issue on larger formats that make them undesirable for practical cinematography?
.

I agree with that. The rental companies are not going to be happy with having to buy whole new sets of lenses to support a bigger sensor. Same with anyone else that owns PL lenses.

 

[Peter Majtan]

Häakon - the scale is not right, nor are the proportions. I have done the same exact mistake (thinking too fast) when someone mentioned first 4:3...

You need to realize that the 5K refers to horizontal resolution regardless of ratio...

So 5K Epic with 4:3 ratio will have lot more pixels then 16:9 and even more then 2.40:1:

.

What I think You (and I originally) are referring is the total number of photo-sites... Now that is a different story all together... I think You might have (like me) little more insight into how sensors are designed and made. The most critical issue and cost is related to the number of photo-sites... This would explain why we both reacted the way we did (incorrectly)...

Lets say the number of photo-sites is given:

4:3 Epic with 5K = 5120 x 3840 = 19,660,800 photo-sites

Now if we arrange the same number of photo-sites in a different ratios, we will get:

16:9 => 5911 x 3325 - and EPIC would then be 5.7K

or

2.40:1 => 6868 x 2862 - and EPIC would then be 6.7K

Conclusion:

If the 5K is locked to the horizontal resolution (regardless of ratio), I prefer the 4:3 ratio...

But if the 5K was calculated based on 4:3 ratio - I prefer 16:9, as this would give us the best overall performance and compatibility...

Just my two pesos...

Peter



EDIT:

I have also overlaid the scale resolutions known for RED ONE for comparison:

.

 

[LawrenceDinkins]

I asked an executive at BOB, why RED was not among their thosands of cameras. After all RED has 4K. The answer was: They only considered real cameras for the Olympics.
I hope the reputation of 4K RED camera improves before next Olympics.
PS: There are several non-RED 4K cameras in use at the Olympics..

 

[Stuart English]

OLympics

OLympics

I asked an executive at BOB, why RED was not among their thosands of cameras. After all RED has 4K. The answer was: They only considered real cameras for the Olympics. So, that tells something about the reputation of 4K RED camera.

No, I suggest that says a lot more about the timescale of the decisions that have to be made about equipment that is being used at the Olympics.

 

[Peter Majtan]

Keep in mind that 99% of all new digital sensors for motion pictures (both sine and video) are 16:9 ration for full compatibility with broadcast standards... This is also why Red One has a 16:9 sensor. You can assume that EPIC will follow in the same direction. In my honest opinion - it should...

So the "best" EPIC I could hope for would be the orange one on the right side...

With 5.7K it will give us 5911 x 3325 pixels of RAW resolution...
This would transfer beautifully to 4255 x 2394 of RGB resolved image...
Perfect for 4K mastering and delivery with room to spare for re-framing and/or image stabilization...

This alone would be KICK ASS!!! (pardon my French...)

Anything more then that will of course blow my mind, but I would even settle for the 5120x2880, as I never plan to use anamorphic glass...
But that is just me...


EDIT:

We can safely assume that EPIC will resolve somewhere between 78% (Red One) and 72% (Scarlet) of RGB out of the RAW files. This is because its pixel density will be higher then Red One, but lower then Scarlet...
So in the worse case scenario of EPIC only resolving 72% we need a sensor with 5688 pixels of RAW resolution to get pristine 4K (4096).
In the best case scenario of EPIC only resolving 78% we would need a sensor with 5252 pixels of RAW resolution to get pristine 4K (4096).
So 5.5K EPIC would be the "ultimate answer" to 4K pristine delivery, bet it at 4:3 or 16:9...

 

[Tom Lowe]

If the number of vertical lines (ie, the horizontal resolution) is locked at "5K," then obviously 4:3 would provide the most real estate. That is the logic many of us have been working from - since basically that is what Red One's "4K" is based on. But if we are talking about number of photosites, then yeah, an argument could be made to go with a 5.7K widescreen format, for example. I could understand people wanting more horizontal lines.

If the number of vertical lines is locked at 5K, then can we all agree that 4:3 would be the best? It allows for serious anamorphic shooting as well as possible IMAX shooting.

I do realize that no one knows for sure yet what the new IMAX 4K digital projection system will be in terms of aspect ratio, but still, if current IMAX 15/65 projects are incorporating stupid F900 footage, then surely EPIC could be useful to IMAX teams looking for a lightweight, portable camera that could be DMR'd to 15/65 cleanly.

 

[I Bloom]

If the number of vertical lines (ie, the horizontal resolution) is locked at "5K," then obviously 4:3 would provide the most real estate. That is the logic many of us have been working from - since basically that is what Red One's "4K" is based on. But if we are talking about number of photosites, then yeah, an argument could be made to go with a 5.7K widescreen format, for example. I could understand people wanting more horizontal lines.

If the number of vertical lines is locked at 5K, then can we all agree that 4:3 would be the best? It allows for serious anamorphic shooting as well as possible IMAX shooting.

I do realize that no one knows for sure yet what the new IMAX 4K digital projection system will be in terms of aspect ratio, but still, if current IMAX 15/65 projects are incorporating stupid F900 footage, then surely EPIC could be useful to IMAX teams looking for a lightweight, portable camera that could be DMR'd to 15/65 cleanly.

I agree with you, it's hard to see the point in arguing this. More vertical resolution, and a larger gate overall is better. You can always window it to 16x9 so what do you care if more sensor real estate is available.

I imagine the outcome of this discussion though really has to do with forging chips and processing throughput as well as cost. We'll just have to see what they come up with.

IBloom

 

[Tom Lowe]

I agree with you, it's hard to see the point in arguing this. More vertical resolution, and a larger gate overall is better. You can always window it to 16x9 so what do you care if more sensor real estate is available.

I imagine the outcome of this discussion though really has to do with forging chips and processing throughput as well as cost. We'll just have to see what they come up with.

IBloom

Yeah, I wish Jim would tell us if we are chatting about this uselessly, or if the aspect ratio of the chip is still under consideration.

I can imagine, though, that there might be business or competetive reasons why this 4:3 issue is being met with silence from Red Team. Who knows.

 

[Häakon]

Peter, thanks for your post; you are absolutely right in that I am thinking in terms of photosites and not a "locked" horizontal resolution. The new digital cinema "K" standards tripped me up, as I am used to thinking of sensors in terms of overall MP and not a fixed number of pixels in a sensor width... but this is just a semantics issue. The point I was trying to make was twofold; in order to really improve upon current digital offerings, we need to a) expand sensor size beyond the 35mm (motion picture) standard, and b) continue to make widescreen (and not 4:3) chips. Above and beyond that, you can "label" the resolution output whatever you like.

In looking at your new examples, the sensor I would vote for is one that is the height of the red box on the bottom and the width of the green (6,868 x 3,840). That's a 16:9 ratio that works out to about 26MP... or - surprise - very close to the size of the chip in the Canon EOS-1Ds Mark III (if it were 16:9 in shape). I guess that would make it a "6.7K" camera, though I think such nomenclature sells it short. That being said, I personally believe that you need such real estate to provide a sufficient oversample for a superior 4K final image, and while I am sure that this may seem like a huge number to a lot of people, remember that the RED ONE at 12MP seemed pretty massive to a crowd used to a maximum 2.4MP (1,920 x 1,280) sensor cap before its release. I have routinely said that the RED ONE is the best "1080" camera out there. For Blu-ray or the web, there isn't a better digital offering... I honestly believe that. The future, however, will be 4K on the big screen. The next truly "ultimate" camera should produce a 4K image similar to the way that the RED ONE produces a 1080 image today. And I dream big because RED has shown - consistently - that they can think out of the box. Perhaps that is too large an order for "Epic," but to be honest, a $20,000+ premium for a new chip seems like a pretty large order too.

 

[Joseph Ward]

Likewise depth of field is an issue on larger formats that make them undesirable for practical cinematography?

If they make a large format that the DOF is only good for a nose length(literally), in which a persons nose can only be in focus, then we have a problem.

 

[Jannard]

You are making wrong assumptions about Epic's size and possible configurations.

Jim