Night lighting issues still present - The community solutions

[CptLucky8]

@Grinde81 thank you once more for sharing these great comparison shots.

I’m really wondering whether Asobo developers are really testing VR at night with a G2. What I mean is that if they are trying with a Vive, or an old “low res” HMD, sure they’ll blame the HMD panels. But using an Index, a G2, or a Q2, there is no way during QA testing, because lights are creating so much a soup of fuzzy blobs, no one can feel the urge to remove the headset and wear glasses.

Otherwise, maybe there is another explanation: if someone is young and has never had the experience of having bad vision to the point of needing corrections, maybe they can’t relate the bad visual experience to what you see when you need glasses and they find this normal. But this would be IMHO not showing much of common sense and therefore shouldn’t be the case.

Nevertheless, it seems to me, like I’ve written somewhere in this topic, they are using the semi transparent gradual halo as a mean to reduce the perceived light bulb size. In lowering drawn light intensity, this is lowering halo size (more pixels discarded at the periphery) and this is reducing perceived light size.

To me this is plain wrong, because no matter the size, light bulbs are always fuzzy and no matter the intensity, the core won’t change below a minimum size which is the size of the opaque disk in the middle.

Instead, I believe a better approach is to effectively reduce light bulb size by distance, and when closer to a certain distance let appearing some more of the halo around to emulate the perceived intensity diffracting/hallowing around because of the moist in the atmosphere (I don’t have the proper English words to describe this).

The former method (actual) is simpler to code in the pixel shader because you’re just varying an alpha blending value per light. The latter might be slightly more complex because you’re varying both size (always) and alpha (up close) but I doubt it is that much complex in practice anyhow, especially given XP11 is doing this and displaying thousands of lights, including every single star in the sky, and every single moving car head and tail lights, at 60fps in 4K…

[Grinde81]

@CristiNeagu

It’s most apparent in the light cast by streetlights. In the pre release, the area under street lights is very bright, whilst now it’s not as bright. If the light is the same, that would mean more dynamic range. And of course the sim doesn’t have more dynamic range, it’s just that the light is dimmer now, but it creates an effect of more dynamic range.

Excellent points, hadn’t thought about it that much myself, but paying attention to that now while comparing the photos, I definately agree. the “lightsplash” under the early alpha shots are indeed too localized and too bright (in my taste), while streetlights usually have a wider “throw” and being slightly dimmer.

[CristiNeagu]

@Grinde81

the “lightsplash” under the early alpha shots are indeed too localized and too bright

That’s exactly what you would see in a photo that is exposed for the sky and the plane: overexposed streetlights. Now some may prefer this look, because people will mostly use reference photos to determine if the lighting looks realistic. But i agree that the goal here should be replicating what we see with the naked eye.

[GrilledChippie]

I think they use the blurred, fuzzy, and fading light orbs in the distance to try simulating the effect of city’s warm glow in the distance.

IRL the warm glow are due to atmospheric effect, lights covered behind buildings, and reflected lights from building surfaces etc… But in game, no vegetation / building after a certain distance, so they use the blurred / fuzzy lights to create a similar effect.

[Grinde81]

@GrilledChippie

I think they use the blurred, fuzzy, and fading light orbs in the distance to try simulating the effect of city’s warm glow in the distance.

That might very well be the case, I can’t come up with any good guess for the bloated blob-lights other than watching too many out of focus night-cityscape photos.
The game does support volumetric light, so I don’t really understand why this isn’t used more for “city glow”, right now it’s mostly visible with ridiculously high aerosol settings.

[GrilledChippie]

Yeah there are lots of things that we don’t understand why they’ve decided to go that way…
From early alpha night lights turned to become sepia at release… then update 5 which all lights becoming monotone and too bright, and they’ve toned it down a bit with the USA update plus some randomisation of light placements.

If devs want to remove sepia (baked texture) as per the Q&A, then why introduced sepia at the first place? Not only sepia, but a weird version of it as shown on several screen shots above!

Also, you can see the jarring difference between early alpha night lights to our current night lights, is the long distance lights quality! Early alpha is way more realistic! The lights jumbled together, randomised, several tones of light warmness, no bloated blurry lights, no “string of pearls” effect etc…
I want to see what they come up with in SU6 to address “long distance light fix” now the community is more aware of what was in early alpha. They better not disappoint!

[SeaW0lfBis]

The overexposed effect of the highways in the release is just a small detail in a much larger picture. Even questionable if they would tweak it, because from what I see, the overexposed effect was localized in some roads, meaning that it wasn’t intentional and would require just some tweaking. And please don’t bring high altitude prints because I never flew the big birds. I’m saying from the GA point of view.

The overall effect was much more realistic from a general aviation point of view than what we have now, with just light dots all over in a TRON grid. Now if the widebodies people messed the lighting just because cities were not like they are used to see from high above, that is messed up as well. Because now we have neither of them.

Low flying at the release was tenfold better than now and it was the right direction. They could have tweaked it, not gone to a whole different approach. Because now we would never get it back. It is always going to be a black canvas with some light dots in it - just like in X-Plane and all the old sims that came before that.

Anyways,

[GrilledChippie]

I can only hope that devs do spend time and read most of the comments here. There are good points and suggestions raised by us that can really help them in getting this right.
It would be more encouraging if one or two of them do interact with us here so we know they’re listening and taking onboard our suggestions etc.

[Grinde81]

Regarding night lighting from the latest dev Q&A:
Long distance, the team is working on it right now, it’s going to affect different areas,
there’s also night lighting in the world map, so it’s both systems that we need to improve so…
we’ve done a lot of work on night lighting close up, it’s really just the long range fix that is missing, Sim Update 6

NO! There’s ALOT left to do with the lighting close up!
Even though Asobo is listening to the forums and their customers, I feel there’s a very big gap between the community and the guys adjusting the night lighting.
Although “a lot of work has been done”, we’re still worse off than what we had in the Early Alphas, more than one year ago in terms of lamps.
What we have now is inferior on many levels explained earlier, so how in the world do they consider close range lighting “done”, just the long range missing…?

Timecode for night lighting: 00:47:41

[MAX84848]

I am hopeful about update 6. Do you think Directx 12 will provide an improvement in visuality and lighting?

[GrilledChippie]

Sigghh… They don’t really inspire confidence in the community do they?
not only a big gap between us and the guys adjusting the night lighting, but also a disconnect.

[CristiNeagu]

Hearing the Q&A as well as some comments from some pretty knowledgeable people on the forum makes me think that Asobo are holding back some things waiting for DX12 so they don’t have to implement them twice. I can imagine that the people who would be tasked with improving the lights would have to be pulled off the DX12 team. That makes sense to me, so after we get DX12 we should start seeing improvements.

[sehnsedahamses]

One major fail is that all the lights are still omnidirectional.

This is completetly wrong!

The absolute majority of lamps are shaded upwards to prevent light pollution.

So what we should see are primarely the reflections from the ground instead of oversized glowing balloons. Not to mention all the floating light’s. There is still a lot of work to do!

[hobanagerik]

Would having directionality to the light sources lead to a cost in performance. The sim can do it, for example with airport approach lighting, and taxi lights etc. But having hundreds, if not thousands of light sources with some added property to them to accomplish this might be another source of pain for those already suffering from poor performance.

[CristiNeagu]

Well… it would create issues. All you’d need to know in order to hide a light or show it is how far away it is and the height difference between you and the light, so it’s not extraordinarily costly. But then you run into issues. It will possibly lead to visual artefacts (such as weird patterns in how lights appear and disappear due to floating point rounding errors) and it will definitely get shot down by the same people who have been complaining about long range lights not being visible. And if you want to fix that and say “well, some lights are shaded, but others aren’t” you now need to discriminate between the two. And you need to have a way of specifying which is which. And now it starts getting more difficult to process.

My opinion is that the cost associated with this would not justify the result.

[sehnsedahamses]

Is’nt it enough to cover the lights with an simple pyramid polygon? 4 points, 4 triangles? Even if multiplied by thousands that should not have an impact. We need no raytracing here (by now ), just simple shading.

[hobanagerik]

Those polygons have a cost, too.

Look what bad PG can do.

[CristiNeagu]

That is probably the worst thing you can do, performance wise.

[sehnsedahamses]

Can you explain why this is so?

I am not an dev but from my restricted point of view I think that in the night terrain and building shading options can be reduced. So why not safe some polygons on buildings, trees or terrain and other so or so then invisible objects and use them for the lights?

[CristiNeagu]

If you’re going to use pyramids to cover the lights, that would add 4 polygons for each and every light. That can quickly get into the thousands. And even if you’re not showing them, you still have to calculate whether or not the light is behind or in front of them. And calculating if the line between you and the light source intersects the pyramid is pretty intensive, and it basically is raytracing in a way.

There are simpler ways of doing it. I’m no expert either but the easiest way that comes to mind is by using shaders to compute the angle between you and the light and determine if the light should be visible or not. Shaders are run on the GPU in parallel so they’re very efficient. And doing simple maths with maybe a lookup table instead of a trigonometric function makes it even more efficient.

Either way, like i said, i don’t think the result will look anything like what you’d expect it to look. The result will most likely be that a lot of lights would simply disappear, especially near to your viewpoint.