I just got 4.21 and am back at this.
Can’t get my head around it as usual.
At 125000 lux and sunny 16 settings why is my desert so dark?
Bump. whether I use manual settings or putting EV100 on 16, I’m getting same dark results. Really need some help with that.
I’m not even sure if I’m doing it wrong or it just doesn’t work properly, like before.
1/125, ƒ16, and ISO 100 is closer to EV15. With a pure white card, fully rough facing up, I increased the sky brightness until the card read about 25,000 Lux. I increase the sun to about 135000 Lux, bringing the final card reading to 125,000 Lux. With these settings, the previous values work well(also pretty close to perfect exposure for 18% gray), but EV16 is pretty dark, but if we follow typical EV measurements, EV16 is mainly for bright subjects on snow or sand, which explains why it’s much darker than we expect. EV15, for 1/125, ƒ16, at ISO100 should look correct provided the intensities are all accurate and your materials are in the proper range.
Try raising your sun angle to be more in line with a mid-day angle and increase the intensity. Also try measuring a white card with the Pixel Inspector. My sun at 135000 is only about 99,000 Lux, so that’s probably why it looks so dark to you.
Are you using a custom skydome ? If so, you also need to crank up the brightness multiplayer of your HDRI.
I have several questions but, how do you measure the surface illuminance? I don’t see such option in pixel inspector nor HDR visualizer.
I do, I just had hidden it in that picture. Wondering about how to measure surface illuminance though.
.sch documented a good method a few pages back, which is similar to what I do at work as well. Essentially, you can do it from a sphere(easy to sample from any direction) or use a plane oriented towards any particular direction you want to measure. The material on either mesh needs to be pure white(1.0, not PBR white) and fully rough. If you lay the plane flat, horizontally, you’re recreating a horizontally placed light meter, which is usually the case since IRL you don’t really want to burn out the meter by pointing it directly at the sun.
With the Pixel Inspector enabled, and you’re in game mode(nothing will happen if you’re not), you can sample an area on the mesh. For the sphere, it’s most likely the direction of the dominant light source or the top pole. For the plane, it’s anywhere on the top face. It’s a good idea to disable bloom, fog, DFAO/SSAO, and any other post-effect that might interfere with the scene color. The HDR input in the Pixel Inspector measures everything in luminance or cd/m² or nits, which is why you want to use a pure white material since luminance is based on color. Taking that HDR value and multiplying it by pi will give you the Lux value.
Lux is easier to find for any given lighting condition. Luminance is a good measurement if you’re really trying to nail a sky material or emissive surface. With most Lux measurements online, they will include the sky contribution too. I will usually measure the skylight by itself and then the sky and sun to get the ambient in range first before the sun, since the sun measurement on a Lux meter can vary drastically based on the angle of the sun and card.
Thanks for the explanation. I do still find it darker than expected though.
- I’ve adjusted the sky dome brightness until skylight at 1 intensity gives me an HDR value of around 7960 which results in 25,000 Lux after multiplying by 3.14.
- Then added directional light on top of that, with 100,000 Lux which results in HDR value of around 40,000 which results in around 125,000 Lux after multiplying by 3.14.
Here is what I get:
Here’s what I expect to get:
There’s a very large difference in both pictures and I still fail to see why. Also consider the fact that my light source is facing down, it goes darker with a 45 degree angle.
I’m far from beeing an expert on lighting but for me it looks like your dark albedo eats up most of your light. Especially since you camera setup looks just right.
Ground texture isn’t really that dark. It looks about right actually. View from buffer
Edit: Also white plane in previous post isn’t shining white either.
Yeah but it is a lot darker than the ground on your reference… Even you don’t want to use it for your actual scene i would try “for test purpose” a brighter sandy texture like on your reference to check how lighting acts or simply change the albedo color.
Albedo looks correct. Dessert sand has an albedo of .45-.35.
An update for folks here on this UDN issue: https://udn.unrealengine.com/questions/462412/skylight-intensity-should-not-use-cdm2.html
Epic will change Skylights to show as IntensityScale and not show cd/m2, but the change will go in to 4.22. This is good news, as cd/m2 is misleading on skylights and would add to confusion.
Increasing albedo brightness doesn’t work, further, it breaks PBR. Locking EV100 to ~11 gives a little nicer exposure/brighter picture, but still looks way off and that’s not even how it should be setup. At EV100 16 using 300,000 Lux is more desirable for a desert, but that’s also not how it should be setup. I don’t think Epic will even notice these issues until they try to create a handful of outdoor test scenarios.
I think you guys take these PBL units as panacea … actually units itself or light values that you set not so important… you can set anything you want and calibrate your scene to 18% grey (See rosegoldslugs post)… and you know what? it will be correct even if values that you put for lights or exposure are different from web charts for real world values for different lighting conditions… what you are actually do is just trying to bring values in Engine to that values from charts for real world to the same value… which in the end give you the same correct 18% grey value and so on … but the difference is that you can control your lights (Outdoor) but photographer not… so they search their “correct or what they think is right” exposure manually too … because conditions are so different and so many factors affect values. Sometimes even 45 angle differents in shot require different exposure values… And dont forget about shot in photographie is static and you need correct exposure for every angle view what is too not possible to create… so… what you try to do is a little no sense and only help in lookdev to calibrate your assets with consistency…
And here start to play role lighting ratios… sun and shadows for outdoor and fill and key for studio … as example for outdoor 1/4 or 5 for typical direct sun light and here you can set anything you want…5000 sky, 20000 sun or 20000 sky and 80000 sun and calibrate both to 18% grey with exposure will give you same result in the end but with different exposure value…and both will be correct
That image looks like it was shot with a different fstop than 16 though. Shadows are really bright!
If you change your setting from 16 to 14 you will be 2 stops brighter and it will look like your example image…all good here I would say 
I think it is very difficult to compare a stock photo with a scene in Unreal that is using completely different ground and sky textures.
It is better to go somewhere and capture textures along with a color chart as well as a light probe (for IBL usage) and measure the sky and sun intensities with a light meter in real life.
That way it is easier to make sure you know what you are looking at, and discern why something might look off.
So, I finally had time to update to 4.21 over the weekend and try out a few things! 
To make it quick:
It absolutely works as intended as far as I can tell and it actually feels pretty satisfying! 
On the flip side though, I would still like the streamlined system for camera and auto exposure and not two different ones^^
Also, I found a few remaining rendering issues, I believe, and I will share them further down as well
Full version:
My idea was to take some photos I shot and use the same exact camera settings while trying to make stuff in Unreal look pretty much the same value wise, while using kinda known values for my settings in Unreal.
I quickly threw together something in Unreal and pretty much just put in values and called it a day^^
For now I went with a sunny daylight shot taken at 11:23 or something and a dusk shot from during July this year. Both images were shot with a Sony Alpha III and a Zeiss 55m 1.8 lens.
I tried to get close with overall values and feel in both sets, however…Unreals tonemapper is of course different from Sonys curves, so the images do have certain differences.
Overall though, I am super happy the way I was able to incredibly quickly achieve results that hold up
I might try some interior comparisons as well at some point^^ This is actually quite fun!
First one shows some nice colors and warm skin tones. You can see the values I shot at to the top right
Now lets look at what I quickly threw together…remember, the camera settings in Unreal are exactly the same! I used 128 000 lux for the sun as it was full sun that day and I went with 10 000 candela/m2 for the white in the clouds directly to the opposite of the sun. Skylight alwas at 1 in all these tests.
Here are some more shots of the same environment whit these settings (its slightly overexposed as it was tweaked towards a character slightly in shadow )
Also, I wish I would have chosen a sky that pops a bit more^^
Now lets look at dusk. This one is slightly tricky as the lights we see here are extremely weak lights and have only about 1 candela at max. I turned off the sun and tweaked the sky brightness until I visually had something locked in that matched the reference photo as I wasnt sure about a candela reference.
Now the Unreal scene! I went with lights that are around 1.5 candela inside sub surface spheres, so the brightness of those spheres is tied to the intensity of the light source. The big light by the rocks is a 1000 lumen light and I believe the brightness and spread makes a lot of sense
Again, this was just some quick testing and I havent looked at auto exposure, yet…But as far as the values go…this makes sense!
Cheers!
PS: I totally forgot to mention the bugs I encountered!
- eye shader seems to be broken as eyes render black when using physical values (I have pre-exposure turned on)
- skin doesnt show spec anymore
- the hair of her is set to black, for some reason it renders light brown when using physical values
- black dots on water reflection on the daylight shot
I have not tried 4.21 yet but it does look promising! Nice work on the comparisons Daedalus51
How is the auto exposure working now? Do you have the option to set a spot meter size and also get an average luminosity value reading from it?
That is a great way to compare real world vs in game data.
Cheers buddy, highly appreciated!
Nah…sadly auto exposure is still the same crap it has always been 
Also the whole metering is pita via the pixel inspector. I really wish there was something like a spot meter!
Hi my 3rd “vate” Tilmann
(in time order koola 1st and Rabello 2nd :D)
is like a week i’m tring to focus on your posts and youtube channel (btw amazing class i just give to you a little contribute
i have a question about lighting and building:
i alsways used **skylights **specific cubemap and ( mostly with the multibounces) i get many colours from HDRI (for exemple in the interiors i have colours from the windows came from hdri) now with the **captured **scene i have no this colour shades
is it normal?