Night lighting issues still present - The community solutions

A classification method for light intensity would be interesting to see.(Who am I kidding, even the early builds randomised lighting would make me happy) Of course everything depends on the quality/resolution of the data and some places could be rendered falsely completely dark if the resolution is too low. Some other methods could be used on top of the light classes, like having lights around buildings - business districts or using any probabilistic equation with land classes. If they used this data(combined with other methods I referred to) to classify their light placement for the whole world, we could eventually have an extremely realistic representation of VFR nighttime conditions for the whole world.

You can check this out too, if you haven’t already. Light pollution map. VIIRS 2019 night light data. I am using this site for a few years now to find dark areas, following the newly released VIIRS global maps. They have a meaningful classification of light intensity(Radiance in10^9 W/cm^2 * sr )which essentially can be used to measure light pollution over specific areas. The map has some data glitches here and there, but it is very accurate and hi res in most places.(polar regions and seas contained most of the glitched areas,sometimes ship lights etc) , but realistically using these data for light classification in the sim , wouldn’t yield visible artifacts) . The other problem , would be the “spread out glow” measurements of medium light intensity. There is a glow around large metropolitan areas and heavily lighted industrial districts, which can be seen also in the Black Marble database. An AI algorithm can eliminate most of the excessive “glow” regions, by finding the right light clusters(given the right analysis sensitivity and value thresholds) ,which essentially are light sources and not atmospheric glow.

Finally another thing(or two) I’d like to pinpoint.

  1. It seems like they use the Black Marble map for their sepia masking and the light pollution simulation (cloud illumination,atmospheric glow) as well. They could get rid of the artificial(unrealistic) ambient light you see when going over large cities(suddenly clouds in the horizon are very bright if it’s not overcast) , if we really wanted the most realistic nighttime conditions. The sim already has a basic exposure adaptation mechanism in place. The atmospheric light pollution could and should stay constant and not changing unrealistically depending on where you are. If you watch some clouds in the distance over a city for example and you were some kilometres away from any village or a place with light pollution(a dark place), those clouds above the city would appear pretty bright, reflecting any light from the city beneath. If you went closer and closer to the city, the cloud brightness wouldn’t change, but by the time you enter the first avenue or highway with hundreds of buildings and lamps around, the clouds can even become nearly black. This is how the eyes adapt in different light conditions. Of course the sim hasn’t got any extremely adaptive and intelligent way of adjusting screen(pixel brightness percentage) exposure , but right now the adaptive ambient light looks totally off. The only exception , is when it’s totally overcast. Its only then that the ambient light is not so visible. In the photos below you can see what I mean. Testing location , London without a moon in the sky , ultra settings. Moving from further out, towards the city centre. Some may like this ambient light feature, others don’t.



  1. Today I also located a weird bug(no clue if it is new or not) and I hope I am not the only one. The clouds around me(a square shaped area ) were darker than the clouds far in the distance. To replicate this bug , moon has to be high in the horizon. The problem may be connected to moonlight.

5 Likes

Aspen and Chicago is with the release lighting, while the Sandpoint and San Francisco (best looking) are Early-alpha lighting.

Note that in the release(sepia) version using 4K and/or increasing rendelscale would significantly increase lamp visibility range due to the wrongfully scaling of lamp visibility/transparency (which was fixed in Release 5)

Early alpha before reduced lamp visibility and sepia mask got introduced

Release version many is saying was “perfect”,
The the severely reduced lamp visibility is painfully obvious.

7 Likes

omg san francisco actually looked so nice !
what happened to the sim :frowning:

5 Likes

Brightness and exposure aside, the eye or tiny cellphone camera sensor would still pick up the same photons. Light concentrated to the roads with non illuminated areas dark/featureless.
Reducing the exposure of that Chicago photo, the human eye would see something more like this:

This is what MSFS got wrong up until now, which instead did illuminate the whole city uniformly like a huge brown lamp, making non illuminated surfaces glow in the dark as much as illuminated roads, instead of representing the grids of bright roads you would see in real life.
Photos does an excellent job of showing WHAT would be illuminated IRL, no matter the exposure. No eye in the world would see deserts and non illuminated areas glow in the dark, period.
We see the same lamps, same sources of light as the cameras(excluding the really dim ones due to the cameras ability to collect photons for an extended period of time)
But the general grid like features in the picture would definitely be visible to the eye in the same layout, and NOT like a brown evenly illuminated daytime texture (MSFS sepia mask)

1 Like

I agree with you. The night lighting already looks like photos in the sim though, far brighter than real life visually but most people are probably expecting something more along those lines. Laminar tried the “natural vision” approach and many complained it was too dark. Its going to be hard to strike a balance whatever they do.

lol. Are you kidding me?
Floating bulbs 100m above groud, super even, on every small countryroad that never has actual light in real life?
Right direction? Yeah sure, if they want us to completely abandon MSFS2020.

7 Likes

It was WAY better before. It’s pretty simple. Now every countryroad looks like a highway, we have double lights 100m above ground everywhere, and the lights are way to evenly distributed.

6 Likes

Yeah, seeing the last one of San Fran , it is clearly visible that the draw distance isn’t there yet and the light sprites in the current version are looking huge, uniform in the distance, without distinct halos around them. They look , i don’t know… Fake?
“Neighbourhood”, smaller road lights and between buildings , could also have their size of halos and the size of their ground splashes randomized.

*Buildings also have no lights from that distance? Or is it just a different time of the day issue?

Thank you for sharing this, this is definitely a bug of some sort which will be easily corrected in the next update. I also believe the main point in the discussion is not only these floating bulbs in particular, but the algorithm responsible for generating credible night lighting visuals in general.

The later seems to be the one for which the matter is not just asking bug corrections, instead, it is about debatable ideas and suggestions, so that we can help Asobo differentiating what it is the community is wanting in terms of lighting, based on as accurate as possible comparisons and analysis (per their asking in the last Q&A).

4 Likes

I’ve noticed the same indeed. The buildings are not showing as much ‘window’ lighting as pre-alpha.

Amazingly I’ve just realized I’ve left FS2020 running in the background at the very same spot all paused, so now it is the night in the simulator and I’ve posted a new screenshot (with the weather panel showing time of day).

I find update #5 lighting in SFO is not too bad though in the distance. It clearly shows the NASA Black Marble texture further away than #4 as promised (can be seen better in the night shot), and the overall “density” of lights are quite realistic around KOAK and northerly from there.

However up close, in the SFO city, it looks like OSM synthetic streets painted with a repetitive pattern of poles, with the main road to the bridge highly lit and all others almost not. Furthermore, pre-release is showing much present building window lights than update #5 and this makes it as-if there is a power outage in the city as a whole!

This is so disappointing, pity we’ll never know why they changed it. What was the point of alpha testing if they were just going to change it all anyway?

1 Like

Additional material to compare my SFO screenshots with

This website has very nice pictures of SFO from the air, some at sunset/night:
https://tobyharriman.com/aerial/

What is nice is the photographer is also commenting about his aperture, speed and iso which helps getting an idea how close these photos are to what you’d see IRL.

#1 I find this one quite close*:

#2 This one as well:

#3 This one is a little bit over exposed I believe but it gives an overall view of what lighting looks like in this city:

It is clear to me the very problem of night lighting is not really how far they are drawing what, rather it is what they are drawing.

When comparing just these shots it seems night lighting is a 2 factor problem mostly when seen from above:

  • pole lights are illuminating the ground.
  • car lights are illuminating the eye.

This means the approach I’d like to see, for lack of experimenting, is combining these:

  1. Use the OSM data as they are doing and as @Grinde81 has been showcasing in order to generate self-illuminated ground texture.

It is the equivalent of the “sepia mask” (which is in fact the NASA Black Marble texture) but restricted to the street and roads. Self-illumined is to be read as a rendering engine material (a texture which intensity does not depend on the sun to keep it simple. Color and width would be based on actual OSM metadata and heuristics if data is missing (near/far city centre, dense/light population area, highway/road etc…)

You can clearly see in photo #2 the “sepia mask” very close representing the highway and parking lot ground, but there is no discernable orange ground at a distance.

  1. Don’t use the “sepia mask” texture at a distance because there is not enough resolution.

You don’t need too much resolution even up close but at a distance, if there is not enough resolution, such texture won’t suffice to discern roads or street which are close. In effect, there could be not enough pixels to represent it all (1 pixel orange, 1 pixel black, 1 pixel orange is the strict minimum under which orange pixels blend together)

This is clearly how it looks like in photo #3 and photo #2 (see my comments above).

  1. Overlay all these roads and streets with actual lights emitted from actual cars.

This is what gives the most wow effect in X-Plane in my opinion and this is depicting perfectly red/yellow lights when seen in the correct direction, unlike the baked-red-dots seen in FS2020.

Photo #1 shows clearly most of the dots are cars, not poles (except on the pier). Photo #2 as well shows this clearly on the highway and in the most luminous streets. Photo #3 on the bridge up close shows the car dots are less visible than the orange ground.

  1. At a distance, because there is no longer the “sepia mask” texture, artificially increase either the car light dot density or the car light dot texture size in order to compensate.

Here is a gross overview of this idea which I certainly can’t illustrate nor validate right now at all, but which makes some sense to me on paper (and this could prove being a bad idea in the end).

*photos and credits from this page: https://tobyharriman.com/an-evening-above-sf/

5 Likes

Agreed, i thought i was in p3d

I’m posting this one to illustrate the use of OSM data as a basis for street light intensity/density.

It is not meant to be representing correct exposure but I find it clearly shows light density features which are matching road classification found in OSM, as well as light density depicted in my FS2020 SFO shots above.

Here is a photo:

Compare with this OSM “Transport” layer view:


https://www.openstreetmap.org/#map=13/37.7934/-122.4483&layers=T

*credits and source: https://www.gizmodo.com.au/2015/03/these-incredible-aerial-views-of-san-francisco-are-just-jaw-dropping/

@CptLucky8

Here are some images generated from OSM data :




9 Likes

Very impressive!
That would be a great foundation to build upon instead of the Sepia Mask brown lamp illuminating the whole city’s daytime-texture.
A texture overlay like this could also reduce the lamp-rendering distance required, saving even more FPS, being faded/replaced by “real” sim-lamps and their effect on the ground.

What software(s) did you use?

Thanks you ! I will publish more images of other places to come very soon on my twitter account (MSFS forum is not the place for this, I posted the images here just to show the potential of OSM data). I can give my twitter by mp is you are interested, just ask me :wink:

This is generated by a C++ software I developed that is still work in progress. The objective is to cover the entire globe since the OSM data is very accurate and complete enough to do it.

In MSFS, these kind of baked OSM images would certainly do the job for high altitudes (in place of the sepia mask), but there is still a difficulty in the transition between low altitude levels (where you can see light bulbs everywhere), and high altitudes.

Night lighting is a very complex subject. The other day I generated some night cities pictures in a totally dark room and was satisfied with the luminosity. When I looked at them the day after, with the shutters opened, I could barely see something on the images because the ambiant luminosity ruined it completely. :thinking:

3 Likes

I have read the exact opposite …if you turn it off…

Early build on alpha there was no colour variation, big bulb lamp on ground larger then cars\windows on city, too much bright, a lot of complain not realistic that why they changed. Light are very hard to do, you improve things you degrade another things on others area, reducing, increasing will introduce another things questionable, dealing with graphics engine and code with procedural is not like photoshop.

Until there is no XML or CFG tweak for night lights, no one will be happy, this was always the case on tweaking for taste with sim, and this will never change, you have the camp of IFR or VFR, you have the camp from camera view or in flight view etc…how to fix for all = XML or CFG files tweaking.

These tango dance with Asobo will cease until it’s more open to users for tweaking, include water, reflection and others texture, and they could move to others area without distraction of vote anymore, at the moment it’s all locked and compressed on big files not editable.

I offered some suggestion previously like this to avoid all issue from users taste or more work for Asobo to move on from night light for once: Option is to add xml files or CFG, users could edit to their own taste, this is the only solution to meet all users taste. Editing is know for popular mod.

NightLight XML or CFG file.
Street (house\condo)
R,G,B; 0.0,0
Lamp size; 0.0
Brightness; 0.0
Contrast;0.0
Ground reflection; 0.0

Street (Manufacture\store)
R,G,B; 0.0,0
Lamp size; 0.0
Brightness; 0.0
Contrast;0.0
Ground reflection; 0.0

Boulevard
R,G,B; 0.0,0
Lamp size; 0.0
Brightness; 0.0
Contrast; 0.0
Ground reflection; 0.0

Highways
R,G,B; 0.0,0
Lamp size; 0.0
Brightness; 0.0
Contrast;0.0
Ground reflection;0.0

Others etc…
R,G,B; 0.0,0
Lamp size; 0.0
Brightness; 0.0
Contrast;0.0
Ground reflection;0.0

4 Likes

I tried with off…no change for me.

1 Like