Exposing + Native ISO

[Terry VerHaar]

<snip>

So, rating 1000 ISO outdoors is forcing you to stop down (close the aperture and let less light hit the sensor) to protect the highlight range, and rating at 320 indoors is forcing you to open up the aperture let more light in from the shadow areas of the scene.

<snip>

Battistella

But, David, isn't this where we get back into the territory where we are assuming what is seen in the monitor is effecting the choice of other exposure parameters (aperture and shutter). As much as humanly possible, I avoid this trap as I am so used to shooting based on stop lights and goal posts, with support from the histogram, that i don't let the look in the monitor creep into my decision.

 

[Scott Crawley]

Again, it depends upon the exposure method being employed. Yours (like mine) is to make shutter and aperture choices based upon RAW sensor data. That way our aperture settings are not driven by ISO settings as in David's example. I am of the opinion that this is the only way to do it right.

But, David, isn't this where we get back into the territory where we are assuming what is seen in the monitor is effecting the choice of other exposure parameters (aperture and shutter). As much as humanly possible, I avoid this trap as I am so used to shooting based on stop lights and goal posts, with support from the histogram, that i don't let the look in the monitor creep into my decision.

 

[Terry VerHaar]

And just to be clear on this skin tone thing. Do we agree that, as long as the skin falls completely within the DR of the exposure - i.e. skin is completely captured, we will not lose color fidelity, i.e. proper skin tones (or at least the potential for them, in the case that WB or tints were off during shooting)? I understand that all bets are off if skin itself is over or under exposed, thereby losing some of the "bits" that represent it.

 

[Adrian Jebef]

This is the other major thing, I think many people are using ISO as gain instead of as a basis for measuring light.

In video terms +3, +6, +9, +12 db, how much noise can you tolerate? Vs. how much picture you want to see?

So some of what is creeping in now is the merging of video and film. Traditional Film guys work in one way, video guys work in "boost and gain" to make a brighter picture. I think that this creates some real problems when ISO is viewed as gain.

Battistella

On target. BTW: if you change your ISO setting in camera on the EPIC or DRAGON you are essentially adding or subtracting VIDEO GAIN to your output image. It is the same as if you were working with an old Betacam. The difference is that the underlying R3D doesn't change. The more deviation you make "away from the center" the more your shot may be perceptually over or underexposed.

WYSIWYG is an essential working method in digital cinematography. Early in the process older cameras (Betacam, DVX100, Red One) had a much smaller dynamic range. The only way in which to capture your scene exposure correctly was to "light" it into your camera's DR. You'd basically have to squeeze a 12 Stop scene into a 7 Stop DR container.With modern cameras (DRAGON, ALEXA, F65) we've got so much more DR to use. Sometimes our scene contrast at 12 Stops doesn't matter because the DRAGON can easily record 16 Stops of contrast. This is why folks love the term ETTR. If you have more than enough available DR range in your camera it makes sense to open up and expose the lower DR range of the scene towards the clipping point of your camera sensor's limit. In essence you're still capturing the entire DR of the scene but you're now capturing it with less noise.

Terry, this is exactly why it "feels right" to bump up your ISO when shooting high-key or high contrast scenes vs. lowering your ISO when shooting low contrast scenes. In a high contrast scene it's going to be your highlights that clip first. The best way to guard against that would be to close down the lens (exposure) and then add fill light to your mid and shadow tones in the scene. It's very much akin to the desert example I posted earlier. When you bump up your camera's ISO the WYSIWYG image gets too bright. So you close down by 2 Stops. Now the highlights look OK but your mid-tones and shadows are too dark. So you light them up a bit with added fill. The same is true in the opposite direction. For a low-con night interior it's better to expose your mid and shadow tones as far up the camera's DR scale that you can because it leads directly to less noise in Post. Again ETTR. So you take your ISO down to 320 which makes the image on the monitor look a bit too dark and pushes you to open up the stop (exposure) a bit.

There are a lot of shooters that have gotten accustomed to working this way in respect to ISO and WYSIWYG. Over time and with experience most can find a working methodology that gives them consistent results. But I think it's important to actually understand what is going on when one makes exposure decisions. For me personally ISO, or ASA, or EI is just a base reference for even exposure calibrated to 18% Grey. Beyond that it's fairly useless. I don't expose at that Base ISO every time because it doesn't make sense for every scene. What is actually far more important is a camera's dynamic range. If you know that you have a total of 10 Stops of range you will have to work much harder than if you know you have 16 Stops of DR. Recognizing your camera's DR range and them intentionally using it to expose your scene the way you want it to look when finished is what photography is all about. Exposing 18% Grey at stated ISO is the technical conclusion but deciding where your range of DR falls above and below your "subject" is the art.

 

[Terry VerHaar]

Things seems to be going full circle here, and heading back into confusion...

To explicit, bumping ISO does NOT add electronic gain before the recorded path. So, it does NOT change the recorded image.

If, like me, you DO NOT let the WYSIWYG image effect your exposure choices, you will not be in danger of over/under exposing unless you ignore your goal posts and stop lights.

I don't agree WYSIWYG is a part of shooting RED.

 

[Scott Crawley]

Yes, but you are not confused.

Things seems to be going full circle here, and heading back into confusion...

To explicit, bumping ISO does NOT add electronic gain before the recorded path. So, it does NOT change the recorded image.

If, like me, you DO NOT let the WYSIWYG image effect your exposure choices, you will not be in danger of over/under exposing unless you ignore your goal posts and stop lights.

I don't agree WYSIWYG is a part of shooting RED.

 

[David Battistella]

Terry

Iso setting has zero effect on what hits the sensor and is recorded. ISO rating will have an effect.

Rating it at 800 mean you are metering at lighting it around the ISO 800 value.

If you rated an Alexa and an epic at 1000 ISO you'd have the flexibility with red to raise or lower the value later whereas with Alexa that is baked into the file.

Battistella

 

[David Battistella]

Fwiw.

In camera.
I pretty much much leave the camera in raw view. I use the histogram an goal posts and I always shoot a color checker on scene and lighting changes. I just know what I'm going to get so I don't let the meta distract me much. I'd be happy if it was monochrome too. I will use exposure to determine if there is a clip or excessively dark shadows.

In post
i set everything to rc3 rlf
white balance
set curves
maybe affect ISO up to about 1000 max
saturation tweaks.

Thats the the file I edit with.

Then I grade from rc3 rlf 320iso. This is best case scenario.


Battistella

 

[Hrvoje Simic]

The more dense and accurate color sample means that you can push values in post further before the calculation goes wrong

"Wrong" as in "failing to represent perceivable accuracy".

This part is correct.

delivering an incorrect color value to a pixel that is interpreted by the viewer as noise.

Right?

Noise is not the only incorrect value which may emerge.

Noise if captured - is there.

Some types of picture manipulations make it more visible, some less, but noise if captured is there. Post does not create it, it only enhances its visibility.

Not to be confused with other post induced artifacts. Noise is not the only issue when the underexposed footage goes "wrong". If the data is missing due to underexposure, "wrong" pixel does not have to have noise but still looks "wrong", as in "poor color", or image is perceived with "wrong tonalities", or just subliminally "less natural".

Imagine the range of color tonalities of a face. If properly exposed, those tonalities are fully represented in the best possible way camera can capture them. Well is full, gradations look "realistic" or "natural" and generally appealing.

If underexposed, those tonalities are lacking full range and further pushing of the material further reduces that range. If face is underexposed 3 stops A/D had 1/8th of the light to convert to digital values. It may be noise free and one doesn't see those little things jumping all around, but the picture is much "thinner". Viewer can perceive that as "poor color" or "not natural" or "plastic looking" and colorist can additionally feel it while manipulating the material because there is a noticeable lack of available finesses.

Even if it is 16 bit A/D, underexposed > pushed portion of the image will not look as good as 12 bit A/D which had properly exposed tonalities to convert.
So yes, there is a more usable range with 16 bits, but no - it should not be relied upon for these salvaging purposes.

The reason this is brought to attention is the increasing tendency to rely on technological advancements to compensate for inadequacies,
instead of expanding through mutual growth.

Prior is a crutch, latter is a ladder.
Prior stagnates, latter evolves.

 

[Gunleik Groven]

Even if it is 16 bit A/D, underexposed > pushed portion of the image will not look as good as 12 bit A/D which had properly exposed tonalities to convert.
So yes, there is a more usable range with 16 bits, but no - it should not be relied upon.

The reason this is brought to attention is the increasing tendency to rely on technological advancements to compensate for inadequacies,
instead of expanding through mutual growth.

THIS pretty much represents what I think I see is often forgotten.

Better tools does not mean that you should forget photographic skills. It should be an inspiration to use them even better...

 

[David Battistella]

Salient posts herjove.


Battistella

 

[Mark L. Pederson]

THIS pretty much represents what I think I see is often forgotten.

Better tools does not mean that you should forget photographic skills. It should be an inspiration to use them even better...

Amen.

And again WYSIWYG is NOT the way to learn the fundamentals that will serve you better even if you decide to use WYSIWYG as an exposure method. And it's the LAST exposure method I would personally choose.

When you shoot RAW - YGMTWYS (You Get More Than What You See).

 

[Scott Crawley]

Noise is not the only incorrect value which may emerge.

Noise if captured - is there.

Yes yes, but captured noise is simply an incorrectly sampled pixel. It is the result of an insufficient sampling technogy and/or insufficient exposure.

Edit:

Oh, and we can absolutely create noise in post, in addition to the other things you describe. We do not just amplify pre-existing noise.

 

[David Battistella]

This thread really comes back to the importance of lighting and exposure.

Cranking the ISO and shooting in the dark is a disaster waiting to happen with any camera.

Whats worse is folks who think a miracle camera should solve these common problems.


Battistella

 

[Gunleik Groven]

Yes yes, but captured noise is simply an incorrectly sampled pixel. It is the result of an insufficient sampling technogy and/or insufficient exposure.

I would not say it is incorrect as it is a predictable and inevitable part of the sample. YOU decide to which degree the relative amount of noise is present, through your exposure decisions.

 

[Adrian Jebef]

And just to be clear on this skin tone thing. Do we agree that, as long as the skin falls completely within the DR of the exposure - i.e. skin is completely captured, we will not lose color fidelity, i.e. proper skin tones (or at least the potential for them, in the case that WB or tints were off during shooting)? I understand that all bets are off if skin itself is over or under exposed, thereby losing some of the "bits" that represent it.

This is a tricky thing and it gets better or worse depending on your camera's overall DR. The more DR range you have to play with the further down the underexposure scale you can place skin tones without too much compromise in color/exposure fidelity. Digital cameras are weird. They "see" and record linearly. Something that is the complete opposite to how we "see". Because of this linearity a digital camera's DR range gets filled up with a disproportionate amount of mid-highs and highlights taking up about 2/3rds of that total DR range. Just the bottom 1/3 is actually used to record or capture all those mid and shadow tones. The bummer is that those mid and shadow tones are usually the most important things in our scenes!

So if you have a scene like this one and you want it to "look" like this when finished you have two choices:

You can close down your stop until it "looks" right by eye on your calibrated 25" OLED monitor. It's sillhoutte-y and "feel" like dusk. You've still got detail in the sun and your actor is mostly in shadow with just a hint of detail. Because your camera is rated at a Base ISO of 800 and you haven't deviated from that setting you feel you're going to be fine in Post. You'll have enough stops above and below middle-grey just in case you want to make tweaks later. The problem is that 18% Grey in this scene is not your actor's shadowed face. It's probably closer to the sky at the top right of the frame. Your actor's face could be a good 5 Stops under key (mid-grey). If you decided later to build a power window over your actor's face and lift the exposure there so you could see his eyes you may be disappointed. Now you've got dirty, noisy skin tones. And there's really not a lot you can do about it at this point.

Or you can go the opposite route and decide to ETTR or fill up the sensor's DR range as fully as possible. You'll get a brighter looking picture on your 25" OLED but you already know that you'll pull it all down later in Post Processing. Depending on your needs and how much contrast range (stops) there is between the sun and your actor's face you'll need to decide just how much you'll "overexpose" the scene to get your "fat negative". Too much and you'll end up with a clipped sun. Too little and you may not get enough exposure or color separation in your actor's face.

So how do you know what the "right" amount of over-exposure is? If you've read Ansel Adams' books and own a meter you might be fine. If you've been shooting with the same camera for 3 years you might also be fine. But the only real way to be sure is to have a consistent RGB output from your camera that shows you that camera's entire DR range at all times. This is the case for a proper Log output. For EPIC at ISO 320 and RedLogFilm we get very, very close. Setup this way you basically have access to all of the raw sensor data (it's just converted into an RGB image so we can see it!). Is the sun going to clip? Just look at the monitor and check the waveform. Oops! It is clipping, but just a little. Close down a 1/3 Stop. Sweet! Perfect. How's the actor's face? Hmmm. It's a little low. Only 17% on the waveform. Will it work? Maybe? It's hard to tell when you're looking at a Log image. The camera has such DR that everything just looks lo-con and flat. It looks like crap actually. What to do?

3D LUT !!!

Boom! All of a sudden you can set a "look" over your Log image (entire RGB video interpretation of raw DR) and shift it to your desired final output image. You've got access to both the raw sensor DR and your intended scene "look" or "exposure". Is a 3D LUT the same as the metadata tools found in the RED cams and in RCX? No. Because REDs tools only work for RED products. A proper 3D LUT should be platform agnostic.

As Mark has already mentioned the importance of getting proper 3D LUT and CDL support via RED and as he's said it's kinda a big deal (which it is) and as he's suggested maybe it would be better discussed in it's own thread... all I'll say for now is that when a Log/LUT workflow works, it's pretty frickin' awesome.

 

[Scott Crawley]

It is only inevitable until sampling technology becomes so good that we are unable to fuck the sample up.

I would not say it is incorrect as it is a predictable and inevitable part of the sample. YOU to which degree the relative amount of noise is present, through your exposure decisions.

Edit:
In my haste I overlooked the greater point you were making. Sure, right now we can choose to embrace noise as one of our creative choices... and we do. I happen to believe that one day in the future, when clean images are the norm and film is long behind us, that people will no longer look upon noise favorably. It is likely to become a quaint nuisance. That's just my opinion though.

 

[Mark L. Pederson]

What to do?

3D LUT !!!

Boom! All of a sudden you can set a "look" over your Log image (entire RGB video interpretation of raw DR) and shift it to your desired final output image. You've got access to both the raw sensor DR and your intended scene "look" or "exposure". Is a 3D LUT the same as the metadata tools found in the RED cams and in RCX? No. Because REDs tools only work for RED products. A proper 3D LUT should be platform agnostic.

As Mark has already mentioned the importance of getting proper 3D LUT and CDL support via RED and as he's said it's kinda a big deal (which it is) and as he's suggested maybe it would be better discussed in it's own thread... all I'll say for now is that when a Log/LUT workflow works, it's pretty frickin' awesome.

When the Dragon goes down this road ... shit's gonna get crazy

 

[Gunleik Groven]

When the Dragon goes down this road ... shit's gonna get crazy

I cannot but agree....

 

[Adrian Jebef]

Terry

Iso setting has zero effect on what hits the sensor and is recorded. ISO rating will have an effect.

Rating it at 800 mean you are metering at lighting it around the ISO 800 value.

If you rated an Alexa and an epic at 1000 ISO you'd have the flexibility with red to raise or lower the value later whereas with Alexa that is baked into the file.

Battistella

Just a side note here but we should stop referring to the ALEXA as a camera that "bakes in it's ISO". It shoots raw just like the RED cameras. We should stop assuming that it can only shoot HD ProRes. BTW: If you actually test ISO ratings on the ALEXA between ProRes Log-C and Arriraw you'll find the differences surprisingly small. That camera has a very smooth response curve.