Sorry, another Dragon optimal ISO thread...

[Marcus Friedlander]

Does anyone else rate their ISO's based on intended grain structure?

There's so much confusion about the whole "Native ISO" question, so I just ignored it and found, what I think is, a better method.

Instead of focusing on trying to see exactly what the sensor sees (which is impossible). I always focus on what my intended grain structure is that I'm trying to deliver, and then, I'll figure out the light levels I'll be shooting in, which will determine the OLPF.

After extensively testing the three main OLPF's, Ive found:
STH @ 250&320 - Clean as glass and great for that high budget commercial look.
STH @ 400&500 - A beautifully textured image that is amazing for narrative projects.

STD @ 400&500 - Same as top line of STH
STD @ 640&800 - Same as bottom line of STH

LLO @ 640&800 - Same as top line of STH
LLO @ 1000&1280 - Same as bottom line of STH

I should note however that the data above has been tested for a 6k capture with an acceptable REDCODE ratio of 5:1-10:1, and may not work as a blanket strategy for all capture settings. NOTHING WILL REPLACE TESTING FOR YOURSELF!!!!!!

Also, the delivery resolution and screen size is a key factor as well. I listed two numbers above for each category. The lower ISO is for if I'm delivering a 4k master or anything theatrical/larger format exhibition, and the bigger ISO is for a 2k master or anything tv/web.

Hope this helps!

EDIT: When it comes to colors, specifically skintones, STH>STD>LLO. So, if possible, I'll always try and shoot on the STH, but often I won't have the appropriate light levels, so I have to settle for the STD or LLO.

 

[Terry VerHaar]

Yep. Can't help but think that Jordan left the captured images at the ISO at which they were captured; so 800 for one, 500 for the other. The real test would be whether the images captured at 800 (assuming all else being equal) looked noisier when the ISO was dropped in post compared with ones actually captured at 500. (Or, a vice versa comparison with adjustments in post.) I have to admit, I remain unconvinced that lowering ISO from 800 to 500 - ALL ELSE BEING EXACTLY EQUAL - would result in a "cleaner" image.

I think the last sentence in my previous post left this statement open to an interpretation that was not what I meant. So, to explain more clearly...

I still remain unconvinced (for all the reasons better explained by Bjorn) that shooting a given image at ISO 500 vs. 800 will yield a cleaner image than simply changing the metadata in post from 800 to 500. This of course assumes, as I did clearly state before, that both (presumably identical) images were lit the same and were properly and equally exposed - i.e. all else being exactly equal. To suggest that one image is cleaner than the other simply because one piece of metadata was different really does make no sense.

 

[Mike Nagel]

I don't disagree with you. But when you say RAW histogram, you are talking about , in camera yes? IN this case that histogram is displaying based on an ISO selection in camera.
In RAW view that histogram is based on DC/RLF 800 ISO 5600K. So in that case one are basing the exposure on that.

Which is a bit different than what Mark was suggesting earlier with ETTR at 250ISO. And the DOF is effected if you are using aperture and not shutter to control the amount of light hitting the sensor.

David

raw view in cam is the raw data in log space - same data as linear light, simply in log space, no gamma applied - same data, just different way looking at it

a lot of peeps grade digital from log, I don't know anybody that grades directly linear light ;-) even if they start linear, they convert to log and/or optionally to linear video afterwards...

color space and color temp etc don't matter - as I said before - because u can't make any critical color decision on a non-calibrated crappy small LCD and most importantly all of this can and will be changed in post.

ISO is (at best) a brightness setting for the LCD, it means nothing. If you - changing the ISO - trick yourself into exposing your scene differently, then well that's one way of working..... ;-)

simply control how much light hits the sensor, and for that (at least) I want to see what the sensor is actually getting

simple way of seeing the difference in workflows is looking at the histogram and switch from standard view with RG4 applied to raw view... what's the difference here ? nothing except that a gamma curve has been applied which obviously (in a post-film, digital capture, digital finish era) is a decision to be made in post and NOT an on-set decision...

the intended DOF changes the amount of light hitting the sensor... shallow DOF, wide aperture may now flood more light onto the sensor and vice versa... again, look at the raw data and either cut light or add light to the scene... if u can't change aperture (u want that DOF and/or concern about diffraction) and can't change shutter (u want that motion blur) and u can't cut/add light to the scene (meaning: u are operating on a zero budget, as light is the base of all photography), well then u gotta live with whatever it is u shoot... ;-) (and RED files will still always be very forgiving)


I feel that some peeps may confuse some stuff a little bit by comparing shooting DSLR and/or the old film days vs. shooting digital raw on RED...

all of this is rather easy, the real trick is to how to distribute the light mass over the stimuli range to get the most out of the sensor... and M, MX and Dragon all have different characteristics...

 

[jordan levy]

Changing iso in camera or changing it in post makes no difference. Giving the sensor more light, makes a difference / gives a cleaner image. fiddling with iso while shooting does not affect the amount of light hitting the sensor. ISO is nothing but a meta data setting that says what amount of digital gain should be applied to the image. So what you write above does not make sense, or it´s something I misunderstood.

A simple test, shoot a gray wall in 2k. Then while you are recording go trough all the iso numbers available without stopping the camera. Then in red cine apply a 800ISO setting for the whole clip. I can assure you the noise pattern will be the same all trough the clip.

that was exactly what we did, and then with a person with a black shirt. i cant explain it, i dont know the science behind it, but the image shot at 500 was cleaner then the 800, even if i matched the iso in post. this was a t 2kws actually. when we did the same test at 4kws i we saw no difference. again, i cant explain it, but i know what we saw and i know how i will handle the shots i need to do. so regardless, if it doesnt mean anything to you, then of course ignore your iso setting.

 

[Björn Benckert]

that was exactly what we did, and then with a person with a black shirt. i cant explain it, i dont know the science behind it, but the image shot at 500 was cleaner then the 800, even if i matched the iso in post. this was a t 2kws actually. when we did the same test at 4kws i we saw no difference. again, i cant explain it, but i know what we saw and i know how i will handle the shots i need to do. so regardless, if it doesnt mean anything to you, then of course ignore your iso setting.

Camera going from cold to working temperature can explain it...

As mike writes: "iso is (at best) a brightness setting for the LCD, it means nothing. If you - changing the ISO - trick yourself into exposing your scene differently, then well that's one way of working..... ;-)"

 

[jordan levy]

i mean i guess i was compensating the exposure, so maybe over exposing the shadows a bit helps the entire grain structure.
whatever it is, it's definitely tricking me and works, so ill go with that method when needed.

 

[Alberto Guglielmi]

Hello Alberto,

you are right I sounded very rude and i feel sorry for that. It had never been my intention though - I´m a friendly person, please apologize.

May be I even misinterpreted what you said, it makes way more sense for me in your latest post. I see that ISO thing pop up every now and then and I simply wanted to be clear that ISO setting does not effect anything regarding the recorded data. With a high ISO setting it is even possible to have a clipped histogram at the right side and be in fact heavily underexposing with no clipping in the recorded RAW data. When you say you expose the LL OLPF like 800ASA we´re exactly in the same range. I also agree with David there can be a reason to "rate" STH OLPF Dragon at 1000ASA if this results in the final image you´re after. The only thing that might be misunderstood easily and that is what I tried to point out is that ISO "setting" is just one variable of interpreting the same recorded data as it is perfectly shown in Davids example above. I completely agree with Patrick, RED provides excellent education on their 101 site and it is well worth in dive into.

Hey Ulf, no offense taken. We chatted on other posts. I know you are friendly and I always find your point of view very informative.

 

[Ulf Krentz]

I would love clarification and/or confirmation as I don't think this is entirely correct. It is my understanding (because this does come up every couple of years) that the Histogram is showing ISO-interpreted data, but the Goal Posts show raw data in all settings. (Scurrying to the RED site to find the answer myself... lol)


EDIT: From the RED site regarding goalposts:"Unlike the histogram though, these are not affected by the ISO speed or look setting, and instead represent raw image data."

Thanks for correcting that, it slipped my attention and my wording was wrong, I was actually talking about the histogram and its limiting borders. You are again spot on regarding the traffic lights and goal posts.

 

[Ulf Krentz]

Hey Ulf, no offense taken. We chatted on other posts. I know you are friendly and I always find your point of view very informative.

Thank you!

 

[Ulf Krentz]

Does anyone else rate their ISO's based on intended grain structure?

Marcus, I very much like this approach. I did not test the Standard OLPF yet, how do the skin tones come out, do you feel the STH OLPF is indeed still better and necessary at all?

 

[Björn Benckert]

i mean i guess i was compensating the exposure, so maybe over exposing the shadows a bit helps the entire grain structure.
whatever it is, it's definitely tricking me and works, so ill go with that method when needed.

Yes, more exposure in the shadows brings noise down. Compensating with exposure / giving sensor more light brings noise down. Not adding little instead of a lot of digital gain (iso) keeps noise down. but it does not matter where it´s done in camera or post. That and color temperature is the whole idea with raw.

 

[Marcus Friedlander]

Marcus, I very much like this approach. I did not test the Standard OLPF yet, how do the skin tones come out, do you feel the STH OLPF is indeed still better and necessary at all?

Hey Ulf,

I probably should have made a note of this in my original post. EDIT: it has now been amended.

IMO the STH has BY FAR the best colors, which includes the skintones. If I have the ability to shoot between 250-500, I will always choose the STH. However, more often than not, I find myself not having enough light to shoot at those levels so I usually end up with the STD. And if worse comes to worse, and I don't have nearly enough light, I'll shoot with LLO. But that's always a last ditch measure for me.

But I absolutely believe, despite the fact that the STD OLPF is gorgeous, that the STH should be a mandatory purchase.

Glad you liked this method! If you end up using it, I would love to hear about your results!

 

[Mark Toia]

Great write up, Toia
Just to clear things up a bit, because one could misunderstand your info about the traffic lights.
The term traffic light is just being used because it looks like one, not because it works like one.

So, the exposure level is not "green" = safe / "red"=dead but it states which channel is clipping!
So it's : "Red"= red channel clips highlights, "green"= green channel clips highlights...

Yes correct.

 

[David Battistella]

Thanks for the lesson on exposure Mike.

If ISO is so useless and so unnessasry, why would RED even call it ISO. Although it is true that their ISO does not work in a Linear way, as Bjorn has pointed out here in some of his tests as well.

The fact is, it is there, it's labeled as ISO not "monitor brightness setting" and it can definitely act as a basis and factor into your exposure and lighting decisions.

As an example, why would Phil, in his extensive testing show the different ISO values spread throughout his examples as a basis for mapping middle grey. http://www.reduser.net/forum/showth...ude-Examined&p=1561590&viewfull=1#post1561590

I get that ISO is metadata, I get that it is a METADATA parameter that can be changed, but that does not discount the fact that it can be used to determine exposure strategies in different situations.

Please don't act like I don't understand this stuff. I really so understand it far better than you might realise.

David

ISO is (at best) a brightness setting for the LCD, it means nothing. If you - changing the ISO - trick yourself into exposing your scene differently, then well that's one way of working..... ;-)

simply control how much light hits the sensor, and for that (at least) I want to see what the sensor is actually getting

simple way of seeing the difference in workflows is looking at the histogram and switch from standard view with RG4 applied to raw view... what's the difference here ? nothing except that a gamma curve has been applied which obviously (in a post-film, digital capture, digital finish era) is a decision to be made in post and NOT an on-set decision...

the intended DOF changes the amount of light hitting the sensor... shallow DOF, wide aperture may now flood more light onto the sensor and vice versa... again, look at the raw data and either cut light or add light to the scene... if u can't change aperture (u want that DOF and/or concern about diffraction) and can't change shutter (u want that motion blur) and u can't cut/add light to the scene (meaning: u are operating on a zero budget, as light is the base of all photography), well then u gotta live with whatever it is u shoot... ;-) (and RED files will still always be very forgiving)

 

[Mike Nagel]

Thanks for the lesson on exposure Mike.

If ISO is no useless and so unnessasry, why would RED even call it ISO. Although it is true that their ISO does not work in a Linear way, as Bjorn has pointed out here as well.

The fact is, it is there, it's labeled as ISO not "monitor brightness setting" and it can definitely act as a basis and factor into your exposure decisions.

As an Example, why would Phil, in his extensive testing show the different ISO values spread throughout his examples as a basis for mapping middle grey.

I get that ISO is metadata, I get that it is a METADATA parameter that can be changed, but that does not discount the fact that it can be used to determine exposure strategies in different situations.

Please don't act like I don't understand this stuff. I really so understand it far better than you might realise.

David

I know that u are a great shooter and I'm just explaining things from my perspective - if we can further improve your beautiful pics of Italy and especially Florence it's a win-win for all... ;-)

ISO and all other metadata are there so that on-set (a) u have a preview of the scene and (b) the DOP can set his mark of intent, needed for some, especially for clients etc - all of these things can be changed in post

ISO as a "brt setting for the monitor" is a naturally just a (pretty good) metaphor... like I said, u are simply tricking urself into protecting the highlights more... u can do this w/o raising ISO... ;-)

we can now rip apart the "exposure strategies" u mentioned but IMO, only if u want to add noise texture on purpose u'd opt shooting on purpose a higher ISO, which means only one thing: pulling light from the sensor... and why would u ever do that ?

as I said, I can add noise/grain over a clean image fully controlled in post - way more effective and obviously way more controlled and flexible...


my posts were my opinion on best/cleanest image acquisition on a given sensor, a digital sensor that has specific characteristics and knows nothing about your ISO setting - not about "creative on-set decisions" (as in: starving a light hungry beast).... that will (usually) be regretted later ;-)))

 

[David Battistella]

It's all good Mike,

You are describing it exactly correctly. When you know what is going on under the hood, using an ISO rating higher or lower (maybe more texture or sharper, if you want to put it that way) you gradually learn to leave the ISO where it is and fill the sensor where it need to be for that shot. That comes with experience.

Rating higher, is indeed a way of pulling light from the sensor because you can then decide to stop down based on ISO (and people should only do this after their own testing and determining what is acceptable to them)

I agree with you. For a noob who wants the cleanest images Marcus post above is an excellent guideline. For those who want a bit of grainy texture, there is a way to accomplish that as well. Or you can opt to stay with the cleanest images and do the rest in post. everyone can figure it out based on need, budget, etc.

Sometimes, sensor underexposure is "intended" underexposure based on the rating, but everyone needs to understand that. AND test test test for themselves.

David

I know that u are a great shooter and I'm just explaining things from my perspective - if we can further improve your beautiful pics of Italy and especially Florence it's a win-win for all... ;-)

ISO and all other metadata are there so that on-set (a) u have a preview of the scene and (b) the DOP can set his mark of intent, needed for some, especially for clients etc - all of these things can be changed in post

ISO as a "brt setting for the monitor" is a naturally just a (pretty good) metaphor... like I said, u are simply tricking urself into protecting the highlights more... u can do this w/o raising ISO... ;-)

we can now rip apart the "exposure strategies" u mentioned but IMO, only if u want to add noise texture on purpose u'd opt shooting on purpose a higher ISO, which means only one thing: pulling light from the sensor... and why would u ever do that ?

as I said, I can add noise/grain over a clean image fully controlled in post - way more effective and obviously way more controlled and flexible...


my posts were my opinion on best/cleanest image acquisition on a given sensor, a digital sensor that has specific characteristics and knows nothing about your ISO setting - not about "creative on-set decisions" (as in: starving a light hungry beast).... that will (usually) be regretted later ;-)))

 

[Hrvoje Simic]

Mmm, in my opinion ETTR is strictly to combat noise and nothing else. With weapon noise isn't an issue with the standard. In your example it sounds like you are protecting highlights at the expense of the human subject, which would be the opposite of ETTR

Unless I'm misunderstanding your reference

ETTR combined with the digital imaging's knee jerk heritage of protecting the highs at all cost leads - and many times has been shown on this forum - that skin tones or other important elements of the shot can suffer by underexposure. Noise is less of an issue today but underexposed shot is still underexposed shot and pushing it may compensate the luminance but not density and reduced tonal distribution.

 

[Stacey Spears]

Does anyone else rate their ISO's based on intended grain structure?

My approach mirrors yours. I figured out the ISO, for a given OLPF, that produces the "clean as glass" results that I watned and then I set it and forget it. In my case, I decided to shoot everything on the Standard OLPF @ ISO 400 for now.

Isn't this the basis of the ETTR philosophy?

ETTR works pretty well for still photography. Its more difficult to use in motion imaging when you need to match multiple shots. If you ETTR each shot, and your primary subject is lit differently in each shot, then it can be a PITA to match them later.