New visibility model

So, I've been tinkering with some angular diameter calculations... which I knew nothing about until minutes ago lol So, here's what I believe I've determined, assuming I'm using the calculator correctly.

 

 

A 20 meter object, at 20,000 meters (20km), is a only about 0.06 degrees in apparent diameter, which according to what I'm reading, if you hold your pinky finger up at arm's length it covers approximately one arc degree, so we're talking a milimeter or so? Likely visible as a tiny speck. A 20 meter object at 50,000 meters would logically be several orders smaller, or effectively invisible. Based on that, you shouldn't be seeing a damn thing of a fighter sized target beyond 15 miles or so. Larger bomber size objects would likely be visible out to 20 miles or so, if you knew where to look and were particularly eagle-eyed. Atmospheric conditions would heavily influence this, as any kind of distortion, haze, cloud, etc, would massively reduce these numbers. NOTE : See addendum at end

 

 

We've kind of wandered through this thread from close range to long range, what's visible, what isn't, and what should be. The numbers I provided above would appear to be absolute maximum, under ideal conditions view distances, and that you knew where to look. This also seems fairly consistent with what I've read about aerial visibility in the past.

 

 

 

 

 

 

So, at close ranges, some of the stuff people have described are 3d model issues (vanishing parts, flat gray textures, etc). This is something each developer will have to examine within their product and correct. Some other things about long distances, I doubt it's a 3d model, but an artificial rendering of generic "contacts" at distance, this is likely to apply equally to all aircraft regardless of designer and would fall in ED's court.

 

 

Some of what Philstyle there noticed about dots appearing and disappearing with FoV, that's probably it trying to render a single pixel but as you zoom out, the pixels get condensed and it disappears. Any SSAA or MSAA is highly likely to influence things at this stage also, ESPECIALLY if someone is using an external shader like ReShade or SweetFx, etc. I've definitely run into that in other games as well.

 

 

 

 

 

 

So what I'm getting at, various devs need to double check their LoDs, preferably using a consistent scale amongst themselves, and ED should take a look at their long range rendering methods to insure a degree of consistency not only across contacts, but across distances so you don't have things bizarrely becoming more difficult or popping in and out as you close on them.

 

 

-edit

A flaw in my calculations, this assumes an object equally large in all dimensions (ie a box) whereas a plane is long and skinny. So those numbers presented above are a bit generous. You could logically assume that a plane in a head-on orientation would be much less visible, despite technically being so many arc-degrees, etc. Example, a 747 may be 160ft wingspan but most of that would not be visible at a distance (being effectively a slice), only the round fuselage would even factor in. A very important fact to keep in mind when thinking what's visible or not. Likewise, it may be 160 ft long, when viewed from the side, but it's a fraction that in height. Again, coloration and atmospheric conditions would greatly influence apparent visibility.

 

 

TLDR ; you shouldn't be seeing anything at extreme ranges. That's what radar is for. If you could see stuff clearly at 30+ miles, we wouldn't need it.

Edited June 2, 2018 by zhukov032186

I've done testing on varying hardware setups and there is massive spread of visual ranges at which the same contact on the same mission can first beocme visible (from 55km down to 10km).

Good test there. Although it appears as though 1440p is actually functioning more or less correctly but that 1080p is not. You shouldn’t be able to see a fighter size aircraft at 50km.

The biggest concern for me is consistency, everybody should see things at the same distance the same way, no matter what that distance is.

The biggest concern for me is consistency, everybody should see things at the same distance the same way, no matter what that distance is.

 

You would need to know the pixel density of each and every screen and adapt the settings to that dynamically, Not sure if that's even possible?

 

I posted this before and seems very believable for distances, IF you have 20/20 vision or corrected 20/20with glasses.

 

Predicting Visibility of Aircraft

 

 

Quote

 

Howell carried out a field study in which pilots attempted to detect another aircraft (DC-3) approaching on a collision course.

 

(DC-3) collision course Over various conditions (detection distance)

5.5 to 8.7 km

3.4 to 5.4 Miles

 

Knew exactly the approach angle of the target aircraft

17.3 to 23 km

10.7 to 14.9 Miles

 

About three times larger than the detection distance. (Knew exactly the approach)

 

 

I would wait till ED locks down this graphic engine some more, it's much better in the beta I believe, needs tweaking still. ED and 3rd party aircraft models LOD's could also now need tweaking now because of more changes?:cry: They must get frustrated sometimes....

Edited June 2, 2018 by David OC

The most worrying thing I've seen there is the fact that you want to make planes visible from 56 km away in 1440p too, instead of making them invisible at 1080p. You guys have any idea of what 56 km is?

 

 

Who said they wanted to make all contacts visible at 56km also?

The video simply points out that there is massive difference when viewing in 1440p v 1080.

At no point does it pass judgement about which is more realistic.

Good test there. Although it appears as though 1440p is actually functioning more or less correctly but that 1080p is not. You shouldn’t be able to see a fighter size aircraft at 50km.

 

IMO you are right, that the 1440p version is closer to reality.

First I'd like ED to get consistency on this, then move to more realistic distances for initial spotting. But so long as there is such a wild disparity between resolutions, that would remain my primary issue.

The biggest concern for me is consistency, everybody should see things at the same distance the same way, no matter what that distance is.

A part of the problem is, that not everybody has the same monitor and screen size, does not use the same resolution, hasn't the same gamma, contrast and brightness settings and finally not the same eyesight and maybe is sitting at a different distance from their screen.

 

If we let out the eyesight, the sitting distance and assume nobody would alter the FOV what are the parameters you can identify from inside the game code?

The resolution, as you set it up in game? Check.

The rest? Not really.

 

The screen size (24-27" monitor vs. 38-58" TV) isn't known to DCS.

Neither is the contrast set on your monitor, the gamma value or brightness setting available to do adjustments.

 

Technically it boils down to adjust the "imposter" (the pixel marking the plane, when the 3D- model's size is smaller than the angular size represented by a pixel) size and appearance distance, to the resolution set in DCS.

You can now even take into account the zoom functions and changes to the FOV.

 

From what I've seen and what I have tested myself, the current system shows contacts at pretty long (unrealistic) distances.

 

I would suggest to dynamically adjust the alpha value of the "imposter" from 0% (total transparency) at something like 30 NM to 95% at the point the 3D-Model is visible and blend it back to 0% between 6-4 NM.

 

That would enable eagle-eyed pilots to spot a faint tiny dot at maybe 12-10 NM (that may need tweaking).

When somebody adjusts the FOV to zoom out we may need to adjust (lower) the alpha value, again? This is particularly hard to solve, as we have a lot of people using multi-monitor or widescreen monitor/projector setups that require(!) adjustments of the field of view.

 

I am not absolutely sure this whole thing can solve this, it is just an idea.

As well, I don't know if there are blocking points like performance impact, or other limitations I am not aware of and I can't tell what isn't already implemented, or I misinterpreted my tests...

Edited June 2, 2018 by shagrat

A part of the problem is, that not everybody has the same monitor and screen size, does not use the same resolution, hasn't the same gamma, contrast and brightness settings and finally not the same eyesight and maybe is sitting at a different distance from their screen.

 

If we let out the eyesight, the sitting distance and assume nobody would alter the FOV what are the parameters you can identify from inside the game code?

The resolution, as you set it up in game? Check.

The rest? Not really.

 

The screen size (24-27" monitor vs. 38-58" TV) isn't known to DCS.

Neither is the contrast set on your monitor, the gamma value or brightness setting available to do adjustments.

 

Technically it boils down to adjust the "imposter" (the pixel marking the plane, when the 3D- model's size is smaller than the angular size represented by a pixel) size and appearance distance, to the resolution set in DCS.

You can now even take into account the zoom functions and changes to the FOV.

 

From what I've seen and what I have tested myself, the current system shows contacts at pretty long (unrealistic) distances.

 

I would suggest to dynamically adjust the alpha value of the "imposter" from 0% (total transparency) at something like 30 NM to 95% at the point the 3D-Model is visible and blend it back to 0% between 6-4 NM.

 

That would enable eagle-eyed pilots to spot a faint tiny dot at maybe 12-10 NM (that may need tweaking).

When somebody adjusts the FOV to zoom out we may need to adjust (lower) the alpha value, again? This is particularly hard to solve, as we have a lot of people using multi-monitor or widescreen monitor/projector setups that require(!) adjustments of the field of view.

 

I am not absolutely sure this whole thing can solve this, it is just an idea.

As well, I don't know if there are blocking points like performance impact, or other limitations I am not aware of and I can't tell what isn't already implemented, or I misinterpreted my tests...

 

 

Your idea would work better than how things are currently. I know it's impossible for ED to adjust for all setups but there has to be a better way than it is now.

 

 

 

Sometimes even things like contrails don't show up until I zoom in then they pop in (but that is a different thing entirely I assume).

Real fighter pilots said, that if you know where the contact is, you can see at 4-5 miles.

 

- DCS, 2 miles, impossible to see.

 

- DCS 1 miles, open formation line abreast, super difficult to see, you have to zoom a lot (not real, you should see super easily)

 

- Dogfight, you i´m following the tail till get some distance (no more than 2 miles) and dissappear...

 

In this cases, doesn´t matter what monitor you have, it should be easy for everybody to see contact.

 

 

 

 

 

This is a very specific scenario, forget medium distances, or trying to see contacts at 5-10 miles, SA...

Real fighter pilots said, that if you know where the contact is, you can see at 4-5 miles.

IF you know where to look exactly and after a considerable amount of training you were undergoing to become a real fighter pilot.

 

- DCS, 2 miles, impossible to see.

From my tests easily at 5-6 NM that is IF you know where to look exactly! With scanning the area and verify suspect "spots" with zoom (as substitute for unimpaired vision and the ability of the brain to focus on points in 3D space opposed to a 2D screen), you can manage even if you don't know exactly where to look.

- DCS 1 miles, open formation line abreast, super difficult to see, you have to zoom a lot (not real, you should see super easily)

That, like in real life, depends on contrast to the background, angle on target and the position of the sun for example.

 

As for the it should be "easy" part: that is what labels are for! I would rather have less intrusive labels (as in more like marked dots) than adapting the 3D models to a point where spotting planes is "super easy".

 

We still need a bit of optimization on a few things, but if you deem it "impossible" to spot a plane beyond 2 miles (I guess you meant 2 Nautical Miles?) than no halfway realistic tweak can help you. That is where you need labels or red arrows/boxes with flashing "Enemy here!" labels.

A part of the problem is, that not everybody has the same monitor and screen size, does not use the same resolution, hasn't the same gamma, contrast and brightness settings and finally not the same eyesight and maybe is sitting at a different distance from their screen.

DCS does know the resolution you have set in the graphics menu. So the sprite size could adjust based upon that.

1080p = 1 pixel

1440p = 2 pixels

2160p = 4 pixels.

Something like that would be fair. The sprite should fade out beyond about 10 miles as well.

Shagrat

 

 

I only want DCS improve, i would not be here if my interest is zero in DCS, i repeat i love this sim, but actually other people has this big problem.

 

 

I guess, how BMS do this? because in BMS i don´t have this problem.

 

 

Can DCS do it in the same way BMS?

 

 

I think this is a big issue that need to be solved.

 

 

We (virtual pilots) are not in equal conditions.

Edited June 2, 2018 by Rikus

Can DCS do it in the same way BMS?

No, according to Wags DCS cannot just scale up the 3D models, it affects their radar cross section and something related to “object fusing” I’m not sure what that means.

https://forums.eagle.ru/showpost.php?p=3497229&postcount=40

Edited June 2, 2018 by SharpeXB

I thought spotting was solved in the late 1.5 build. You could see a dark dot very far away. Now I keep thinking the stars and moon are contacts in Vr.

From what I've seen and what I have tested myself, the current system shows contacts at pretty long (unrealistic) distances.

 

I would suggest to dynamically adjust the alpha value of the "imposter" from 0% (total transparency) at something like 30 NM to 95% at the point the 3D-Model is visible and blend it back to 0% between 6-4 NM.

 

That would enable eagle-eyed pilots to spot a faint tiny dot at maybe 12-10 NM (that may need tweaking).

When somebody adjusts the FOV to zoom out we may need to adjust (lower) the alpha value, again? This is particularly hard to solve, as we have a lot of people using multi-monitor or widescreen monitor/projector setups that require(!) adjustments of the field of view.

 

I am not absolutely sure this whole thing can solve this, it is just an idea.

As well, I don't know if there are blocking points like performance impact, or other limitations I am not aware of and I can't tell what isn't already implemented, or I misinterpreted my tests...

 

Your thought is similar to mine. However, the way I see this:

 

What is taken in consideration is the size the aircraft takes up on the screen: when it's far enough (or zoomed out enough) to be the size of a pixel, it's drawn as a pixel. Both distance and zoom change size, but in the end it's not looking at distance or zoom, it's looking at the end result: the size the aircraft will occupy on the screen, and this will define how many pixels will be drawn. If it's just one pixel, draw just one pixel. If it's smaller than one pixel, blend it with the background to make it less visible, blend more and more as it becomes smaller. So I don't think you sould be worried with FOV as you mentioned.

 

BUT I have to say that a simpler and more effective solution would be to just not draw this damn pixel if aircraft is smaller than one pixel. Why go through all the trouble of micro calculating the blending of one freaking pixel that's at a distance that's not supposed to be seen anyway, when there's much worse visibility problems with aircraft that's close to you? Priorities.

 

This whole 'pixel imposer' thing that DCS is using apparently was designed with 4k or 2k screens in mind, where a person apparently can't see a single pixel (I don't have 4k or 2k screens, so please correct me if I'm wrong). If you can't see the pixel, then don't render it and it will at the same prevent players at 1080p like me from seeing it. Because those pixels stand-out here like a black whale lying on a white sand background.

Real fighter pilots said, that if you know where the contact is, you can see at 4-5 miles.

 

- DCS, 2 miles, impossible to see.

 

- DCS 1 miles, open formation line abreast, super difficult to see, you have to zoom a lot (not real, you should see super easily)

 

 

Except for, you know, that being utter BS :)

 

 

@Shagrat

Actually, the bit about shading transparency on the fly, starting at a set range, and gradually lessening, that sounds pretty reasonable, and in theory at least, should provide consistent results. Even if a resolution had a theoretical advantage, the contact being translucent would nullify that, putting everyone on roughly equal footing.

 

 

@DavidOC

I realise you've posted that before already here, and I've read similar, but it's interesting those numbers align with the calculations I did the other day independently. I figured "math" would be pretty unbiased, though who the hell knows if I even used the calculator correctly :music_whistling: Them numbers and big words, you guys

Can DCS do it in the same way BMS?

 

Smart scaling is just one of possible solutions and, although I have defended it, what I want is SOME soltion, not necessarily smart scaling.

 

My impression (I could be wrong) is that ED is trying a different solution: increased contrast between the aircraft and the world. After a good time not doing any close range dogfights, I recently installed 2.5.2 and let the gamma at the default value of 2.2, instead of lowering to 1.6 - 2.0 as I was used to. I also don't use any AA (in-game or external) and use a 1080p monitor. I'm now seeing aircraft get dark and contrast not only against the sky, but against the ground too, which was impossible some time ago. Also, it's happening not only with "dot" aircraft, it's happening with medium and close range aircraft too. I started the mission "F-15C Agressors", where I always had difficulty seeing the enemy (F-5 plane with desert camo over the Nevada desert) and now I can maintain visual contact on them for about 80% of the duration of the dogfight, while before I couldn't see them for about 80% of the duration of the dogfight.

 

I don't know if it's the effect of gamma or if something changed with 2.5.2, but visibility is definitely different. Also, not using AA causes the aircraft to shimmer, making them more visible.

No, according to Wags DCS cannot just scale up the 3D models, it affects their radar cross section and something related to “object fusing” I’m not sure what that means.

 

Not 100% sure, but "Object fusing" could be complex objects with moving parts.

Where actually two (or more) "objects" will be rendered with a relation to one another in 3D-space.

Now if you scale the objects, "grow" into each other...

 

Well, the radar cross section calculation alone, is pretty much a no-go for scaling distant objects... I dare not imagine the shitstorm if jet-jockeys find they can track and lock targets at 300-400 NM. ;)

DCS does know the resolution you have set in the graphics menu. So the sprite size could adjust based upon that.

1080p = 1 pixel

1440p = 2 pixels

2160p = 4 pixels.

Something like that would be fair. The sprite should fade out beyond about 10 miles as well.

 

 

This could be a good solution.

 

 

For example, imagine that for a F15 Ed asigns 10 pixels.

 

 

 

10 pixels in each monitor will look bigger or smaller plane, so:

 

20 pixels for 1080p

25 pixels for 1440p

30 pixels for 4K

 

 

Or also use percentage:

 

X pixels for 1080p

Increase X 25% for 1440p

Increase X 40% for 4K