[RESOLVED] EV vs. Corona EV

[chris_widdowson]

@Juraj Talcik

It’s been long time since that event but I would definitely remember all the people from beer afterwards :- )

That default intensity has some really funny name that only Ondra can write :- ).

But it works like this: You put simple white color into environment (255/255/255), create some object ( Teapot with 128/128/128 Grey color ), hit render, and the teapot at EV=0 will be exactly 128/128/128.
So it’s more like ‘no’-exposure giving you exactly what someone without photographical background might expect. I.e exact intensity of the HDRi how he sees it in material editor, or material to look exactly as well.

But I don’t really want to defend it, not sure if this is truly practical in some kind of scenario…I would be perfectly happy with 100perc. photographic values. Currently the only difference is the real EV and Corona EV are shifted, though I never bothered to find out by how much.

This is a quote from a 2017 discussion regarding exposure settings.

For a camera at default settings (film width : 36mm, ISO: 100, F-stop: 16, Shutter speed: 1/50), what is the corresponding EV in the VFB?

Seems to be ~4.5?

What are the maths behind this calculation?
How many times has this already been answered on the forum?

[Juraj_Talcik]

I had bit more complex post about this later somewhere but any such math would be pretty useless practically because the value is inverted anyway.

But “absolute” EV values are not used in photography either…for at least two decades, since inception of digital cameras.
EV is primarily used as “offset” ( i.e “+2 EV” meaning two stops more light ), not as absolute table value (i.e “2 EV” being exact number resulting from combination of shutter and focal length normalized per ISO).

Why would you even need this? The full settings (Sensor size, ISO, Shutter, Focal Length) are identical between Corona and cameras. None of my cameras (Sony Alpha 7r series) have option to input absolute EV value, it’s long abandoned concept.

I am of opinion that Corona should go full way and allow EV modification even in full settings, so that is it absolutely identical to how digital cameras work. You set your settings (f-stop, shutter, iso..) and still have option to use EV as offset.
I’ll make that a request when I’ll remember.

[chris_widdowson]

@Juraj I recorded a short video in response to your question, but also to better illustrate why I posed this question in the first place (watch x1.25-1.5 speed).

EDIT: really sorry if I mispronounced your name(?)

EV vs. Corona EV - Tonemapping Settings

[romullus]

If you want simple exposure to match photographic exposure by default, then set EV to -4.5 (or whatever you find appropriate) and save render settings as defaults. As Juraj already explained, there’s very good reason why simple exposure is set as it is and i really don’t see a reason why it should be changed. In fact i would be pretty upset if it would.

[chris_widdowson]

@Romullus Right, and your suggestion is a good one, although I am still curious what the conversion formula might be. I think it’s clear enough that -4.5 approaches a visually similar result, but if we follow the typical exposure formula:

EV = log2 ( 100 * aperture² / ( ISO * time ) )

we arrive at 13.6; given default camera settings.

13.6/3 is approximately 4.5 (4.53)(?)

[Fluss01]

@Juraj Talcik

I am of opinion that Corona should go full way and allow EV modification even in full settings, so that is it absolutely identical to how digital cameras work. You set your settings (f-stop, shutter, iso..) and still have option to use EV as offset.
I’ll make that a request when I’ll remember.

Definitely need this. Has it already been requested?

[Juraj_Talcik]

@Fluss

@Juraj Talcik

I am of opinion that Corona should go full way and allow EV modification even in full settings, so that is it absolutely identical to how digital cameras work. You set your settings (f-stop, shutter, iso..) and still have option to use EV as offset.
I’ll make that a request when I’ll remember.

Definitely need this. Has it already been requested?

I think I or someone else might have asked for it long time ago, but I didn’t make any request yet :- ). Go ahead!

[Fluss01]

@Juraj Talcik

@Fluss

@Juraj Talcik

I am of opinion that Corona should go full way and allow EV modification even in full settings, so that is it absolutely identical to how digital cameras work. You set your settings (f-stop, shutter, iso..) and still have option to use EV as offset.
I’ll make that a request when I’ll remember.

Definitely need this. Has it already been requested?

I think I or someone else might have asked for it long time ago, but I didn’t make any request yet :- ). Go ahead!

here : https://forum.corona-renderer.com/ Enable VFB EV spinner while in photographic exposure mode

[chris_widdowson]

Just realized it’s pronounced ‘yoo-rye’ - sorry man

[maru]

I understand cjwidd’s request and the idea behind it. I will contact our devs to find out if there is some simple way to swap between EV and photographic values. My concern right now is that the photographic exposure is controlled by 3 parameters, so it appears to me that there will be actually an infinite number of possible combinations giving exactly the same EV equivalent. I guess providing a few examples, or locking two specific values at their defaults (e.g. shutter speed 1/50s and F 16) and providing the third one (ISO) would be enough here.

(Internal ID=503222427)

[maru]

I don’t have any exact info yet, but here is how I started. Maybe someone has time and courage to take it further:

• Created a scene with 0 black background
• Placed a Corona Camera
• Placed a plane
• Painted the plane with Corona Color set to pure white plugged into self-illumination slot, self-illumination intensity set to 1
• Used a secret spell to remove gamma 2.2
• Rendered

This way we can read RGB values from the plane. Simple EV set to 0 gives Tonemapped sRGB 1, meaning that the surface has 1 Watt per steradian per square meter. Great.
So I tried with various simple EV values vs RGB and with various ISO settings vs RGB. I left shutter speed and f at 1 to simplify things.
I stored the results in Excel and drew a graph with extrapolation and showing the function which is used to draw the graphs.
Here is what I got:
EV = 1.4427LN(RGB)
ISO = 0.1757RGB

It is still guesswork, but it more or less works.

Now, with this, you can enter any RGB value and it will print the EV and ISO values for you. We would like to have EV<>ISO conversion, but I am simply too stupid for this.

I am sharing my XLS file “as is”, without any instructions.
The table you are interested in is the lowest one titled “ev and iso based on rgb”.

Scratch that, unless you want to have a for fun exercise. See the post below.
evs2.xlsx (21.6 KB)

[maru]

@cjwidd - Here is the formula used by Corona which defines the dependence between photographic settings and simple exposure:

a = 78 * fStop^2 / (0.65 * shutterSpeed * ISO);
exposure = log2(683/a)

For some reason, I did not get it to work yet, but I will try again. Feel free to share your findings!

[maru]

Here is an XLS where you can enter the fstop, shutter speed, and ISO values, and it will print the corresponding Corona EV. I doesn’t work the other way around, though.
Photographic-to-CoronaEV.xlsx (11.9 KB)

[chris_widdowson]

this is amazing

[chris_widdowson]

It actually was -4.491148827

This is so cool, thank you so much for sharing this Maru - thank you!

[Fluss01]

Thx Maru saved, this could be useful.

[jpjapers_1]

That spreadsheet is cool thanks Maru!
Juraj posted a thing a while back (which i think he referenced above) that i found pretty interesting and is now basically how i set up every scene because its so formulaic and works every time.

@Juraj Talcik

I also do not get the obsession with following F-Stops/Shutter for CGI work to set exposure. In physical realm, F-Stop and Shutter speed end up modulating exposure, but primarily affect depth of field and motion blur.
And those are almost always used creatively in photography.
Photographers in fact go to great lengths to decouple exposure from those effects, for example by using strong (up to 15 EV !!) ND filters so they can achieve strong motion blur independent of exposure.
Or inventing crazy techniques like ‘Brenizer’ stacking to achieve shallow depth of field in wide angle setup to get past the limitations of physical shutter size.

Remember, intensities of lights are mostly absolute, we only have one sun, but exposure is completely relative value, the less light penetrates the sensor, the higher you would compensate it.
There is no such thing as “F8, 1/16, ISO 100” for midday interior. Maybe it’s cloudy day and the windows is 20cm2 and floor is black. There goes your setup.

Don’t set your F-Stop, set your exposure, depth of field and motion blur. You can’t set those effects unrealistically and the fact that you can achieve them independently, is amazing freedom of CGI, use it. Locking them in exact ratio might be helpful if you’re integrating into exact filmed footage, but otherwise, you can liberate your approach.
ISO literally has no meaning in CGI if you aren’t matching footage, it’s the same as simple EV compensation. So might as well just stick with it, making life simpler with one less value and broadening creative potential as well as simplifying the whole setup.

Here is how I set my camera in CGI:

0. I start with some base lights, mostly it’s Sun.
1. I set my exposure in “simple” EVs. I go up until the desired brightness is achieved. Simple.
2. I set depth of field until I achieve desired creative effect in my scene. It ends up being obviously realistic value: F8-F16 for various broad shots where I want just tiny natural blur at the end of my scene to soften the CGI look, and F4 for something like close-up vignette. But setting that value only affects Depth of Field.
3. If it’s shot with motion blur (car trail, breezy tree) I set my motion blur with shutter.

Physically correct workflow can be creative and user-friendly. Not puzzle.

[chris_widdowson]

This is exactly the type of thinking I was trying to address in the ‘Plz Halp’ thread. The idea of artistic freedom implies an unconstrained question; if you are a beginner, you do not know what the reasonable limitations are and that can lead to confusion.

As an interpretation, it might follow like this:

Step 0: add sun / sky
Step 1: adjust to ‘desired’ brightness
…

At that instant the formula is derailed because, as a beginner, what (visually) constitutes desired brightness?

At the same time, an inversion of that logic (Step 2), “It ends up being obviously realistic value: F8-F16” points to a range of conceivable values that help to constrain the question.

I’m not trying to castigate the very helpful suggestions that were being put forward in the quoted example, but I am vaguely asserting that there is a (sort of) quantitative approach to implementing photorealism in CG. When we use subjective terms to describe photorealism, we lose track of the question.

Inb4: ‘artistic freedom is what make great images’ → I’m not talking about artistic choices that make an image emotional or worthwhile. I’m talking about the qualities of an image that make it indistinguishable, or nearly indistinguishable, from a photograph.

[jpjapers_1]

@cjwidd

This is exactly the type of thinking I was trying to address in the ‘Plz Halp’ thread. The idea of artistic freedom implies an unconstrained question; if you are a beginner, you do not know what the reasonable limitations are and that can lead to confusion.

As an interpretation, it might follow like this:

Step 0: add sun / sky
Step 1: adjust to ‘desired’ brightness
…

At that instant the formula is derailed because, as a beginner, what (visually) constitutes desired brightness?

At the same time, an inversion of that logic (Step 2), “It ends up being obviously realistic value: F8-F16” points to a range of conceivable values that help to constrain the question.

I’m not trying to castigate the very helpful suggestions that were being put forward in the quoted example, but I am vaguely asserting that there is a (sort of) quantitative approach to implementing photorealism in CG. When we use subjective terms to describe photorealism, we lose track of the question.

Inb4: ‘artistic freedom is what make great images’ → I’m not talking about artistic choices that make an image emotional or worthwhile. I’m talking about the qualities of an image that make it indistinguishable, or nearly indistinguishable, from a photograph.

Photorealism surely has to be subjective because its not measurable in itself. There isnt a metric for photorealism. But there are metrics for camera settings and physically based workflows that are ‘known good’ and produce consistent results. Outside of that, is where the sujective questions start like ‘how bright should it be’ or ‘are my shadows too dark’ etc. I think the best we can do is to build a physically plausible foundation in a scene and then do the art within those restrictions.

I don’t think there is a way to talk about photorealism without it being subjective.

[chris_widdowson]

Probably not - at least not in some constructive or meaningful way. That being said, what Ludvik points out in the ‘Time to ditch sRGB…’ - about how it would be better if the viewport camera defaulted to photographic emulation - would probably help alleviate some of the confusion.