Avatar Sequels no longer on Red?

[Sabovic Adis]

To be clear, when I say “code” I don’t mean actual computer code. “Cracking the code” is a figure of speech.

Sadly, ther's nothing else but code, if we talk "digital". Re-coding the code so it somehow resembles film is up to whoever does that - the film emulators with their film emulations, be it good or bad.

Secondly, as an example, look at the Venice illustration above. The most highlight latitude it can achieve is 8 stops above 18% (middle) gray at fairly high ISO values.

Film can achieve somewhere between 1-2 stops better highlight latitude than that, depending on the stock, at much lower ISO ratings.

It has always gotten the most out of its DR on the top end whereas digital video is much more sensitive and a better low light performer.

So, I’m not necessarily talking about adding more DR to a sensor so much as I’m talking about a different distribution that allows for significantly better highlights than what we can currently achieve at ISO ratings between 500-1000.

However and whatever that technology looks like is up to the engineers.

The thing with RED sensors is that they have no native ISO or ASA rating, as per G. N., so there is no overexposing, underexposing, better highlight latitude or worse. Not in "classical" sense. There's only 16 stops of overall DR (Monstro - 17) and how one chooses to use them. :smile5:

 

[Hrvoje Simic]

SEMI SPOILER ALERT

The "danger" lights. Hundreds of deeply saturated red holiday lights had to look red in order to accentuate the point and visually make sense. White, overexposed blobs with pink halos, which is what all current digital sensors in a dusk/sundown situation would yield, would not have been nearly as effective.

It was effective for me within first few shots.

"Sensor saturation code"?! "True film-level overexposure latitude"?!

And I always thought that there's DR, and how one maps/remaps linaer to 709, 2020 etc...:smile5:

Nope. There's much more than that and Brian is right.

Nothing beats film in that department. So far.

So, I’m not necessarily talking about adding more DR to a sensor so much as I’m talking about a different distribution that allows for significantly better highlights than what we can currently achieve at ISO ratings between 500-1000.

Nope, that needs upper range. Otherwise it still clips.

Different distribution with this range was already tried in previous colour sciences and while it is possible to get more saturation in upper range, holding onto peak saturated highlights too high sacrifices smooth transition to overexposure because the roof is still the same height. Remember exploding coloured lights ?

 

[Brian Boyer]

Nope, that needs upper range.

You cannot re-distribute what is not there.

I mean a different distribution as far as the sensor's light gathering capabilities are concerned. So, the photo-chemical process of film may only have 15 stops but favors highlight latitude over low light sensitivity.

If a newer sensor technology, like say, Panasonic's organic layer CMOS sensor or something else, is capable of gathering light in a similar manner, with greater saturation at the expense of some low light performance, you wouldn't need more DR than we have available now to represent those brighter values.

Or would you, because of digital's linearity? I'll admit, it gets confusing because I know how I want it to work but I'm not sure what would need to happen in order to get there.

Does greater photosite saturation capability (which, the way I understand it, would yield better highlight latitude) automatically require more DR than we currently have or can it be achieved within 15 or 16 stops?

 

[Sabovic Adis]

(...)
Nope. There's much more than that and Brian is right.

Nothing beats film in that department. So far.
(...)

What department? DR?!
Film completely fits into digital. With headroom. Including grain.
How? Scan it!
But not vice versa - printing digitally captured images to film will undoubtedly make you miss some colors. And will skew contrast.
Ergo, digital>film.


Imitating film, on the other hand

Sadly, there's nothing else but code, if we talk "digital". Re-coding the code so it somehow resembles film is up to whoever does that - the film emulators with their film emulations, be it good or bad.

is completely different mater. :smile5:

 

[Hrvoje Simic]

I mean a different distribution as far as the sensor's light gathering capabilities are concerned.

That's the light gathering saturation capacity, i.e. upper part of DR.

If a newer sensor technology, like say, Panasonic's organic layer CMOS sensor or something else, is capable of gathering light in a similar manner, with greater saturation at the expense of some low light performance, you wouldn't need more DR than we have available now to represent those brighter values.
Or would you, because of digital's linearity?

You would need more DR. In the upper range. People often oversimplify so misinterpret DR, perceiving capture + quantization part and data processing part as one, and DR as something you can move as you wish, especially due to basic DR/ISO charts like these. You cannot.

Captured signal distribution and distributing light capture are two different things.

Digital is linear. Sensor is analog.
DR is predetermined at the analog stage.

Does greater photosite saturation capability (which, the way I understand it, would yield better highlight latitude) automatically require more DR than we currently have or can it be achieved within 15 or 16 stops?

It requires more DR above.
DR is fixed to upper floor (saturation capacity) and lower floor (noise floor).

Stops and their distribution are predetermined by sensor performance. It clips at a certain light intensity and sees only noise below a certain light intensity.

It is not something which you can re-imagine afterwards in the signal chain.

 

[Karim D. Ghantous]

I still don't understand why digital sensors are linear. After all, a device which accumulates a charge can be non-linear, such as a capacitor. But I am very ill-informed about engineering etc.

 

[Brian Boyer]

Hrvoje, I don't know what it is about our exchanges but it always ends up taking several back-and-forths to get my meaning across. Based on your answers, I don't think I mean what I think you think I mean. I was trying to avoid it (for a change) but I'll take a longer and more drawn out approach.

That's the light gathering saturation capacity, i.e. upper part of DR.

That's what I thought.

You would need more DR. In the upper range.

That's what I initially stated, so we're still on the same page in that respect. I think the disconnect is where I say "more DR than we have available now." I don't think you quite understood what I was getting at. I'll explain what I mean by that below.

People often oversimplify so misinterpret DR, perceiving capture + quantization part and data processing part as one, and DR as something you can move as you wish, especially due to basic DR/ISO charts like these. You cannot.

I know they are and I know you can't, respectively. The range is the range. Those over/under exposure diagrams always add up to the same total DR range claimed by the manufacturer for each ISO.

Captured signal distribution and distributing light capture are two different things.

Captured signal distribution is what I was referring to when I mentioned "distribution."

Digital is linear. Sensor is analog.
DR is predetermined at the analog stage.

It is not something which you can re-imagine afterwards in the signal chain.

I wasn't suggesting otherwise. Everything I'm asking is based on the sensor at capture, not after.

It requires more DR above.
DR is fixed to upper floor (saturation capacity) and lower floor (noise floor).

Stops and their distribution are predetermined by sensor performance. It clips at a certain light intensity and sees only noise below a certain light intensity.

Again, I mentioned "more DR than we currently have" but you didn't address it so I don't think you caught my meaning.

I was asking is there mathematically a minimum total DR a sensor has to have in order to accommodate a light intensity of, say, 9 stops of exposure above middle gray before clipping? Or...

Can a sensor accommodate that same light intensity before clipping in as few as 13 stops, suggesting it has a high saturation point but also a relatively high noise floor as well?

Let's say we have a sensor that has 9 stops of DR total, with enough saturation capacity to allow, at best, 4 stops of overexposure and a subsequent noise floor that allows 5 stops of underexposure, in keeping with the known sensitivities of the various ASA/ISO ratings.

Increasing the saturation capacity of that sensor's next gen replacement by the light intensity equivalent of 4 stops while not improving its noise specs yields a sensor with more DR on top and more total DR, but still only 13 stops total.

As you know, manufacturers are increasing overall DR but they're doing it mainly by lowering the noise floor, not significantly increasing saturation capacity.

The "DR we currently have" falls between 14-16 stops for the top rated cinema cameras. Do we need more than 15 total stops to get substantially higher saturation capacity (highlight latitude) or is that number sufficient?

That's what I was asking. I hope it's clearer now. From your answers, I gather higher saturation capable sensors can conceivably exist within the 15 or 16 stops of DR we currently have available. What say you?

BTW, I'm not advocating for a noisier sensor, I'm just willing to accept a little more noise if it were the trade-off for gaining substantially better highlight latitude in the same number of total stops that are achievable today.

 

[Sabovic Adis]

I still don't understand why digital sensors are linear. After all, a device which accumulates a charge can be non-linear, such as a capacitor. But I am very ill-informed about engineering etc.

Problem is: they're not.

When talking DR, part of the range is fairly linear, namely highs, mids and good chunk of the lows, but the part near the noise floor has a toe, meaning - compressed. Non-linear. The other problem is that the color balance (R vs G vs B) varies depending on the part of the range you expose at (ASA), meaning - it's not linear either. It gets greeny, pinky, thisy, thaty...
That's why one always has to colorcorrect (and by that I don't mean "Tungsten to Daylight" compensation, I mean "color science" correction, or tonal ironing, which sometimes introduces another set of problems :smilewinkgrin, and is, among other things, the main reason Brian is asking for 9 stops overexposure latitude. He, actually, is asking for bigger linear part which then can be shaped to "look" like film, without being clipped in the highs, without being tonaly disbalanced and without being noisy in the lows. :smile5:

Edit:
Oh, and sensors are not digital...:rofl:

:thumbsup:

 

[Hrvoje Simic]

I still don't understand why digital sensors are linear. After all, a device which accumulates a charge can be non-linear, such as a capacitor. But I am very ill-informed about engineering etc.

Sensors are analog, linear part happens after A/D conversion.

 

[Hrvoje Simic]

That's what I was asking. I hope it's clearer now.

Yup.

From your answers, I gather higher saturation capable sensors can conceivably exist within the 15 or 16 stops of DR we currently have available. What say you?

I'd say 15-16 are mathematical stops and we have usable 12-14 stop cameras these days. Were there actual 16 practical stops available without severely molesting the material, digging into shadows and NR, we wouldn't have this conversation right now.

You are using stops as a reference to theorize about possible sensor performance.
It should be the other way around.

If instead of "saturation performance" you mean "better highlight performance with the available DR", yes that is possible. And Arri has been doing that for 10 years by mixing two parallel readouts.

I was asking is there mathematically a minimum total DR a sensor has to have in order to accommodate a light intensity of, say, 9 stops of exposure above middle gray before clipping? Or...

I don't have an answer to that question which wouldn't use guesswork as ingredients.

Can a sensor accommodate that same light intensity before clipping in as few as 13 stops, suggesting it has a high saturation point but also a relatively high noise floor as well?

Highly unlikely.

Let's say we have a sensor that has 9 stops of DR total, with enough saturation capacity to allow, at best, 4 stops of overexposure and a subsequent noise floor that allows 5 stops of underexposure, in keeping with the known sensitivities of the various ASA/ISO ratings.

Increasing the saturation capacity of that sensor's next gen replacement by the light intensity equivalent of 4 stops while not improving its noise specs yields a sensor with more DR on top and more total DR, but still only 13 stops total.

Theoretically yes. Realistically...that necessitates handling 16 times higher light intensity.
So not really up for picking in a feature store.

As you know, manufacturers are increasing overall DR but they're doing it mainly by
lowering the noise floor, not significantly increasing saturation capacity.

Yes, and you can thank the pixel count race for that.

Do we need more than 15 total stops to get substantially higher saturation capacity

You need substantially higher saturation capacity to get more stops.

BTW, I'm not advocating for a noisier sensor, I'm just willing to accept a little more noise if it were the trade-off for gaining substantially better highlight latitude in the same number of total stops that are achievable today.

You can do that by yourself. Two ways:
1) Put an ND on
2) Rate the camera higher (imagine it was more sensitive and underexpose)



I wouldn't say that matter is up for trade, many would have gladly traded it by now.

 

[Brian Boyer]

Yup.

Cool, and thanks for your insight on the matter.

I'd say 15-16 are mathematical stops and we have usable 12-14 stop cameras these days. Were there actual 16 practical stops available without severely molesting the material, digging into shadows and NR, we wouldn't have this conversation right now.

Fair enough. I went with manufacturer claims to avoid my point getting lost and derailed in a squabble over whose claims are true and "If X has more, why does it show less than Y?" type questions.

You are using stops as a reference to theorize about possible sensor performance.
It should be the other way around.

I thought I was doing both, actually. The main reason for talking in terms of theorizing sensor performance relative to stops is because of how people tend to discuss DR.

When I say I want substantially better highlight performance, as we've gone over already, it requires more DR up top. The question then becomes does the new sensor add stops to the total DR over the amount a previous gen sensor could already achieve? I think this is what most people assume.

They start mentally adding that upper DR gain to the 14-16 (or 12-14) stops that are already available in sensors, which leads them to say, "They'd have to design a sensor with 20 stops of DR for what you're asking and we already have 16, which is more than film ever had. Stop your whining."

So, I'm trying to frame the conversation in terms of designing a sensor that captures a shitload of light intensity, but that also fits within the range of dynamics we already have now and know to be achievable (14-16 stops), if such a thing is possible. Does that make sense?

If instead of "saturation performance" you mean "better highlight performance with the available DR", yes that is possible. And Arri has been doing that for 10 years by mixing two parallel analog gains.

I'm pretty sure I've only referred to it as "saturation capacity" so I'm not sure if that changes your statement any. The dual gain readout is ingenious and has done the trick for the last decade.

I don't have an answer to that question which wouldn't use guesswork as ingredients.

Again, fair enough.

Highly unlikely.

Too bad, really.

Theoretically yes. Realistically...that necessitates handling 16 times higher light intensity.
So not really up for picking in a feature store.

True. But sensor technologies capable of greater saturation are almost certainly in our future. Otherwise, it's as good as it's ever going to get. That seems unlikely.

You need substantially higher saturation capacity to get more stops.

Yes, up top. As I tried to explain above, I say it the other way around to present the possibility of achieving the higher saturation without necessarily needing a substantial amount more stops than what cameras currently work with.

It may keep people from thinking in terms of having to expand the total DR to like 20 stops (4 stops over what's currently claimed by RED, which still wouldn't necessarily mean an improved saturation capacity) to get the type of highlight performance I'm talking about.

You can do that by yourself. Two ways:
1) Put an ND on
2) Rate the camera higher (imagine it was more sensitive and underexpose)



I wouldn't say that matter is up for trade, many would have gladly traded it by now.

Yes, I'm aware of these techniques. Both require you to restrict the amount of light hitting the sensor in order to grab those extra stops up top and then turn up the gain. It's effective in practice but it's the kind of solution that ends up hurting RED in the long run for features and tv because almost no one tests cameras like that.

They gather up a bunch of cameras, use the same lenses and shoot the same subjects under the same conditions at the same settings. As long as tests are done this way, RED's highlight latitude results will always fall short of Alexa's and film's. And maybe now Venice's, too.

No one is going to set all the cameras up the same except for the RED and use whatever settings are going to make it look good for the situation. Don't get me wrong. RED's have a lot to offer and are chosen for numerous projects over other cameras.

I believe substantially better highlight performance will bring with it a host of perceived and actual pleasing image aspects that will help bring digital cinema closer to those qualities we love so much about film. RED could be at the forefront of that or they could keep chasing resolution and no light performance.

 

[Hrvoje Simic]

The question then becomes does the new sensor add stops to the total DR over the amount a previous gen sensor could already achieve? I think this is what most people assume.

It may or may not. Depends on the sensor.

So, I'm trying to frame the conversation in terms of designing a sensor that captures a shitload of light intensity, but that also fits within the range of dynamics we already have now and know to be achievable (14-16 stops), if such a thing is possible.

Why would it have to fit in same dynamics ? Why the approach in trading DR in the first place ?

True. But sensor technologies capable of greater saturation are almost certainly in our future

Yup.
Close future. I'd say.

It may keep people from thinking in terms of having to expand the total DR to like 20 stops (4 stops over what's currently claimed by RED, which still wouldn't necessarily mean an improved saturation capacity) to get the type of highlight performance I'm talking about.

I'd say that range should cover what you are talking about.

Yes, I'm aware of these techniques.
Both require you to restrict the amount of light hitting the sensor in order to grab those extra stops up top and then turn up the gain. It's effective in practice but it's the kind of solution that ends up hurting RED in the long run for features and tv because almost no one tests cameras like that.

That was a joke. Because same principle you talked about. Same DR, highlights peserved, underexposed shadows. No need to do anything in the camera.

They gather up a bunch of cameras, use the same lenses and shoot the same subjects under the same conditions at the same settings. As long as tests are done this way, RED's highlight latitude results will always fall short of Alexa's and film's. And maybe now Venice's, too.

If we put film on the side for later, I'm having high hopes for Gemini @ 1600.

I believe substantially better highlight performance will bring with it a host of perceived and actual pleasing image aspects that will help bring digital cinema closer to those qualities we love so much about film.

Sensitivity is taken care of. "Resolution" as well. Spatial. Sort of. With computative colors.
Motion and upper range DR are currently challenges for digital cinematography. Nothing still beats film in those parameters.

 

[Savannah Miller]

I feel like the lower range of digital is so much better that overall in nicely exposed shots, digital tends to have a softer contrast. Maybe you like it or not, but it's a different choice.

 

[Hrvoje Simic]

What do you mean by "softer contrast" ?

There is no inherent typical "contrast" in digital, it all depends on how the data is processed in camera and post.
Rich, well exposed log image, recorded in higher bit depth can be shaped in contrast in different approaches and outcomes. Depending on scene contrast crafted by the cinematographer, of course.

 

[Savannah Miller]

What do you mean by "softer contrast" ?

There is no inherent typical "contrast" in digital, it all depends on how the data is processed in camera and post.
Rich, well exposed log image, recorded in higher bit depth can be shaped in contrast in different approaches and outcomes. Depending on scene contrast crafted by the cinematographer, of course.

It just feels like there's more DR in an even distribution, giving you good info in the shadows too vs. just a smooth highlight rolloff.

 

[Brian Boyer]

Why would it have to fit in same dynamics ? Why the approach in trading DR in the first place ?

It might not have to fit in the same dynamics. I'm just saying if it did, I'm cool with it.

I'm trying to make it clear that I'm not proposing a sensor that's 20+ stops because anything less is inadequate for my superior needs. So, I say I'll take 14 or 15 and give up three stops of Gemini's low light prowess for 9 stops at the top, if it's necessary.

If they can keep the low light performance and significantly increase the highlight latitude, even better.

That was a joke. Because same principle you talked about. Same DR, highlights peserved, underexposed shadows. No need to do anything in the camera.

I see.

If we put film on the side for later, I'm having high hopes for Gemini @ 1600.

I'm wondering about Gemini @ 3200. I don't mind a little noise in my images.

 

[emil_h]

- Variable ND might be a stretch if they keep the same body size but I'm sure they would be a welcome addition.

Doesn't the Mini have variable ND's?

 

[Hrvoje Simic]

I'm wondering about Gemini @ 3200. I don't mind a little noise in my images.

Analog (actual) gain = trading highlight range.

1600 digital push would keep it.

Doesn't the Mini have variable ND's?

Nope.

"Variable ND" as in a single ND which you can tune and just dial-in perfect exposure.

 

[emil_h]

Nope.

"Variable ND" as in a single ND which you can tune and just dial-in perfect exposure.

I meant a motorized ND, which it does.

 

[Brian Boyer]

Analog (actual) gain = trading highlight range.

1600 digital push would keep it.

Um...huh? I know it's not generally your style but could you explain what you mean in more detail. Perhaps start by explaining, in detail, what you meant about having high hopes for Gemini @1600 - you know, the process. Maybe I misunderstood what you meant.

Then, maybe you could explain, in detail, what you thought I meant by Gemini @ 3200 - you know, the process. Maybe you misunderstood what I meant.