Scarlet - Native ISO

[Simon Shasha]

Hey guys, I've been trying to figure out what the native ISO for the Scarlet is - I was over on this page but couldn't seem to find it: http://www.red.com/products/scarlet#product-tech-specs

Does anyone know? Thank you :smiley:

[Nick Wernham]

I believe it is 800, just like the Epic.

[Philipe Ratton]

The sweet spot is 800.
Edited, thanks Graeme

[Graeme Nattress]

800 is the answer.

"Native" doesn't really have a meaning.

Graeme

[Stephen Wheeler]

The Raw records 320 Iso, but the sweet spot is 800 Iso. The MX sensor is the MX sensor whether its the RED one, the Epic, or the Scarlet.

[J Davis]

800 is the answer.

"Native" doesn't really have a meaning.

Graeme

what would be the correct term to use ?

[Noah Yuan-Vogel]

ISO doesn't mean much in a system where the imager has a fixed, linear light response since there are many ways to build a gamma-corrected image from that data.

RED usually suggests their ISO800 setting's curve to be the best balance between highlight latitude and noise, which is usually what one wants out of a corrected image. Many people are happy with that default setting. As I understand there are also many ways to build your own look.

[Stephen Wheeler]

The idea is that the camera* is recording at 320. When you dial in 800 ISO you are boasting the image after it is recorded. So what is happening is you are exposing a 320 ISO sensor like it was 800 then bringing it up in post, so that you have equal protection in your highlights and your shadows. This has been discussed like a gibillon and a half time on red user.

* Edit Note: I had said sensor but meant camera.

[Graeme Nattress]

Noah's answer is good. Stephen, the sensor doesn't have a "native iso". If you go look up the definition of "ISO" there is no definition for a sensor as-is, and in terms of RAW recording the answer is "go ask the manufacturer what they recommend". We recommend starting at 800 and taking it from there.

Graeme

[Noah Yuan-Vogel]

what would be the correct term to use ?

It's not the word that doesn't fit, it is the concept.

I think it would be interesting if manufacturers published basic sensor response data/graphs so we could compare their various imagers in a more meaningful way. Perhaps a table/ graph of 16bit values at the ADC vs luminance with noise range for each color channel? It always seemed weird to me that this was never standardized, we are stuck looking at differently corrected images from different cameras with many non-standard correction options available. This might give many people a much better idea of what a particular imager is capable of and how well various manufacturer's correction curves (or "ISO" settings) perform. Not to mention it would be an amazing starting point for matching cameras with custom LUTs.

[Stephen Wheeler]

Graeme I love you. You're one of the few people on this site who I listen to. This being said I both disagree with you and agree with you.

I agree that 800 is the ideal ISO for the MX sensor. I came to this conclusion after reading a debate over 320 vs 800 and decided to shoot some tests. I also agree that with RED the idea of a "Native ISO" is irrelevant because the raw is recording at 320 regardless of your ISO settings. Another conclusion I came to while shooting tests.

What I disagree with is the idea that Native ISO doesn't exist. While it is not well defined, the popular definition is the ISO setting of least gain. When changing ISO on a digital camera you are not making the sensor more or less sensitive, but you are boasting or deadining the signal from the sensor. ISO on a digital camera is just a film like representation of gain.

If you've come from shooting hpx's or 5 and 7Ds then asking about Native ISO is natural because you are used to boasting a signal and then recording the boasted signal.

[Graeme Nattress]

Indeed with a DSLR there's analogue gain going on, which means that there is a "native" as in unity gain setting that you can test your ISO at.

I don't agree though that the RAW is at 320. It's not like that's a setting where there's least digital gain going on, as the amount of gain applied is arbitrary, and the "effective gain" is also utterly dependent upon the gamma curve. I can tell you though that REDLogFilm, the EXR floating point export and REDGamma2 are all designed around ISO800 being the neutral starting point for their math.

Graeme

[Stephen Wheeler]

Ok that makes sense. still though when I compare a grey card at 320iso in clr and the same grey card at raw they read at the same point on my waveform. I by no means am saying that 320 is the native iso of MX sensor based on this though.

[Graeme Nattress]

What do you mean by in "clr" and "raw" for your comparison? Just trying to get to the bottom of what you're seeing.

Graeme

[Noah Yuan-Vogel]

he means vmCLR and vmRAW, just view mode indicators on the top of the screen when switching between monitoring modes on R1MX by toggling the "view RAW" button. I also wondered what it is we are actually seeing when we "view RAW". I mean that's not linear light is it? I recall playing with it and it clearly wasnt linear. Is it just some somewhat arbitrary representation of the RAW data with some gamma correction and equal red and blue channel gain?

My thought would be that I don't think the fact that a grey card seems to match in its raw view and gamma corrected values is any useful indication of "native" ISO. That strikes me more as a coincidence that the gamma curve and whatever curve is used to display raw happen to intersect.

[Graeme Nattress]

Gotcha, but they're a ISO320 based representation of RAW to allow you to see the highlights better. It doesn't meant that's native ISO, just that it was chosen to help you see if you're clipping or not. No, it's not linear light as that would be absurdly dark.

Graeme

[Noah Yuan-Vogel]

No, it's not linear light as that would be absurdly dark.

Yes it would. I shot some linear light RAW footage once and it really reminds you not to underexpose digital capture, you can really see the pixel precision bits disappearing as you stop down :)

So I guess the RED view raw correction is something of a vestige of old R1-M's programming? While not at all necessary, itd be interesting to have the option for different raw view modes, it would certainly remind people to consider what RAW actually means. I can only imagine how many people don't really understand that the raw view they are seeing is still corrected to a somewhat arbitrary degree and probably dont even know that their sensor actually responds linearly. Maybe then more people will actually believe me when I tell them there is no ISO let alone "native ISO" inherent to the sensor :)

[J Davis]

When I shoot still photography and bring in RAW i can't change the iso

iso as metadata seems to be unique to red

I guess what I'm trying to say is the concept of iso with the world of RAW has to be thought of differently. Its almost
like you choose your max noise floor then expose to that.

[David Rasberry]

When I shoot still photography and bring in RAW i can't change the iso

iso as metadata seems to be unique to red

I guess what I'm trying to say is the concept of iso with the world of RAW has to be thought of differently. Its almost
like you choose your max noise floor then expose to that.

This is one of the unique characteristics of Red's approach that I particularly like. The only settings that directly affect the raw data capture are the same ones that affect exposure of film: light, aperture, and shutter speed.

You need to shoot your own tests to determine what the optimum ISO for determining exposure is, but the consensus says 800 is the general sweet spot for MX .

[Martin Stevens]

So does this mean that at ISO 800 there are an equal amount of
stops above and below middle gray?

:)

[Ryan De Franco]

Gunleik's tests revealed that the MX sensor clips highlights faster when rated lower (say, at 320 instead of 800), and the noise floor increases when you rate above 800. The sensor's actual ISO be damned, 800 is the spot for maximum dynamic range.

That being said, I've found myself rating clips shot at 800 at 640 or 500 in RedCine recently, for the right scene only ISO seems to deliver the "dim" tonality I'm looking for, not gain gamma or lift... that being said, I have no idea what I'm doing as a colorist :)

[Noah Yuan-Vogel]

Gunleik's tests revealed that the MX sensor clips highlights faster when rated lower (say, at 320 instead of 800), and the noise floor increases when you rate above 800. The sensor's actual ISO be damned, 800 is the spot for maximum dynamic range.


well yeah, but the latitude is still the same. Of course low iso clips sooner relative to grey point, raw white clip is the same for all modes, and raw noise floor is the same as well, but generally Redcolor2 and Redlogfilm are designed to simply move mid point and redistribute shadow/highlight latitude accordingly. Don't forget that higher isos also clip later and lower isos also have lower noise.

Personally I'd consider shooting at a higher iso since I love extra highlight latitude and dont care as much about shadow latitude and I'm not afraid of a little noise :)

[Ryan De Franco]

good point noah, should've said maximum dynamic range in both under- and over exposure.

[Stephen Wheeler]

Thank you everyone. This has been one of the best and most informative threads I've encountered on this site. I hope Simon got the info he was looking for.

[Graeme Nattress]

Just to be clear - Dynamic Range is constant. It doesn't matter what ISO you shoot, the total DR available is the same. This would not be the case in an analogue gain situation where typically a stop of DR is lost for every stop up in ISO you go.

What does change is highlight (and because DR is constant, consequently shadow) latitude. For each stop in ISO you go up, you gain a stop of highlight latitude and loose a stop in the shadows. Another way of thinking of "loosing" the stop in the shadows is that the shadows will get more noise in them.

Similarly if you go down a stop in ISO, you loose highlight lattitude and gain it in the shadows. This is fine until the point where you loose too much in the highlights and start to get clipping. That is why we only recommend you go below ISO320 if you're very careful and watch those highlights!

Graeme

[Martin Stevens]

Just to be clear - Dynamic Range is constant. It doesn't matter what ISO you shoot, the total DR available is the same. This would not be the case in an analogue gain situation where typically a stop of DR is lost for every stop up in ISO you go.

What does change is highlight (and because DR is constant, consequently shadow) latitude. For each stop in ISO you go up, you gain a stop of highlight latitude and loose a stop in the shadows. Another way of thinking of "loosing" the stop in the shadows is that the shadows will get more noise in them.

Similarly if you go down a stop in ISO, you loose highlight lattitude and gain it in the shadows. This is fine until the point where you loose too much in the highlights and start to get clipping. That is why we only recommend you go below ISO320 if you're very careful and watch those highlights!

Graeme

So......does this mean that at ISO 800 there are an equal amount of
stops above and below middle gray? Can you answer this question directly?

:)

[Graeme Nattress]

No, at ISO800 there's still more stops below than there are above. I think there's 4.5 above at ISO800, but the stops below are personal noise tolerance defined, so we picked ISO800 as "good safe starting point".

Graeme

[Martin Stevens]

No, at ISO800 there's still more stops below than there are above. I think there's 4.5 above at ISO800, but the stops below are personal noise tolerance defined, so we picked ISO800 as "good safe starting point".

Graeme

THANK YOU!

So at what ISO do we need to rate Epic MX so that the amount of highlight latitude
is equal to that of modern 35mm Negative?

I'm not including HDRx in this.

[Graeme Nattress]

Martin, I did measure 5219 while working on REDLogFilm, to help me understand how film "works" so to speak - I'd have to go back through my notes and see. I remember the "overall" dynamic range being very similar though between RED One M-X and the 5219.

Graeme

[Martin Stevens]

Martin, I did measure 5219 while working on REDLogFilm, to help me understand how film "works" so to speak - I'd have to go back through my notes and see. I remember the "overall" dynamic range being very similar though between RED One M-X and the 5219.

Graeme

In my mind 35mm Neg always seems to have its ISO set so that say out of the 14 stops of DR that it has the highlight latitude is about 8 stops
and the shadow latitude is about 6 stops.

So to get the MX to match this a very high ISO would be needed, maybe ISO 6400. Am I crazy?

Thank you a million times for HDRx! Love your work!

[Graeme Nattress]

Definitely more in the highlight side of things than shadow - which is severely grain-limited. It's also tricky because of the smooth highlight roll-off and how to compare that to the linear nature of a digital sensor which gives increasing information in highlights until it stops, rather than less and less and less and less until it's all gone. They're different beasts...

Graeme

[Martin Stevens]

Definitely more in the highlight side of things than shadow - which is severely grain-limited. It's also tricky because of the smooth highlight roll-off and how to compare that to the linear nature of a digital sensor which gives increasing information in highlights until it stops, rather than less and less and less and less until it's all gone. They're different beasts...

Graeme

Yes, very different and difficult to truly compare.

Film fails beautifully in the highlights. Digital can hard clip and be not so pleasant.

But then again, with your Alchemy and your FLUT curves we have beauty again.

Oh, wow....when you unleash the Dragon sensor....whoa!

:)