Orbiter horizon curvature and other stuff

[cjp]

[Moderator note] This thread continues with posts split off from the 'MS Flightsim staff canned' topic.

Uh. I doubt, very doubt it. Orbiter needs to change significantly and we'd need some professional designers and coders who work on it many hours each day for years. Orbiter would need a real atmospheric model, including wind, clouds, rain, fog and so on (also, the atmosphere gets way too dark too early if you climb up, and the curvature of the horizon is visible too early too). Orbiter also needs a new or significantly updated physics model for atmospheric flight. At the moment it's still far away from being a realistical atmospheric flight simulator. It's rather static.

I don't think the curvature of the horizon can be visible too early in Orbiter. It uses realistic dimensions for everything. Maybe the only unrealistic thing is the opening angle of the camera, but you can change that to match your monitor at its distance from your eyes.

For the rest:

• Weather and atmospheric effects
• Landscape rendering and collision detection
• Detailed maps, airports, well-known landmark buildings & so on
• More realistic aircraft & helicopter simulation
• air traffic control & so on

Some of these require adjustments in the Orbiter 'kernel', or at least the preferred way to implement them would.

I don't say Orbiter can't become what MSFS is today. But without a lot of money and rather capable coders and designers, forget about it immediately

And without a lot of patience. You either need lots of capable coders, or lots of patience, or both.

 

[Moonwalker]

I don't think the curvature of the horizon can be visible too early in Orbiter. It uses realistic dimensions for everything.

You can see it already in about 6 to 7 km altitude regardless of the FOV (unless its 10° which is anything but realistically). But maybe that's because the visibility is almost endlessly in Orbiter and the earth is just a plain ball. There is no terrain, no real haze and so on. In MSFS and X-Plane this is totally different. But even when you fly in reality at 10 km altitude you can't even see the earth's curvature through the cockpit windows. A few pilots claim they can but whenever I tried that I did not notice it at all. But from the Concorde it was definitely possible to see it at about 17 km altitude.

 

[Erupter]

I dunno...
Orbiter can surely be made so that dimensions are correct, but everything seems so small and short. But that may be because of lack in details, and also because anything in orbiter flies much faster then anything (or almost) in MSFS.
Altough i remember one fictional mach 3+ craft and i did not get the same sensation i have in Orbiter.
It just seems too small... you take off and you're 20km in the atmosphere in notime. I don't know... sure i have no experience in space flight, but i have some in real atmo flight and both MSFS and Xplane give a very good representation of it.
Alwayse depending on the accuray of the implementation of the model.
Some MSFS payware product do give a very realistic feeling, some freeware too.
And some payware and freeware feel like crap.
Usual mixed bag.

 

[cjp]

It just seems too small... you take off and you're 20km in the atmosphere in notime.

That's because you usually fly powerful rocket propelled crafts in Orbiter. Try flying the [ame="http://www.orbithangar.com/searchid.php?ID=1897"]White Knight[/ame] instead.

On the horizon curvature issue: I tried to measure the curvature in a picture I made @ 10km altitude, but I don't have any conclusions yet (with 2288 px = 49 degrees, it goes 6px away from a straight line). Wikipedia says:

At an altitude of 10 km (33,000 ft, the typical cruising altitude of an airliner) the mathematical curvature of the horizon is about 0.056, the same curvature of the rim of circle with a radius of 10 metres that is viewed from 56 centimetres. However, the apparent curvature is less than that due to refraction of light in the atmosphere and because the horizon is often masked by high cloud layers that reduce the altitude above the visual surface.

Edit:
I have more information now: if I'm correct, the curvature of the earth should give a distance(*) of 14 pixels in my image, while I only measure 6 pixels. Probably caused by the clouds. BTW, the curvature is hardly visible:

I chose this picture because the horizon is extremely sharp here. In most daylight pictures there is a lot of haze.

(*) distance of the center point of the horizon from the straight line between the lefmost and rightmost points on the horizon.

 

[Scrooge McDuck]

I dunno...
Orbiter can surely be made so that dimensions are correct, but everything seems so small and short.

The environment in Orbiter already has the right dimensions. The fact that everything feels way too fast or small, is 100% due to the lack of atmospheric haze and details. Also, you are right that when we do fly atmospheric in Orbiter, we fly almost exclusively (rocket-powered) craft capable of insane speeds, like DG. That's the reason why everything seems so small, and why you can get everywhere so fast in Orbiter.

You can see it already in about 6 to 7 km altitude regardless of the FOV (unless its 10° which is anything but realistically). But maybe that's because the visibility is almost endlessly in Orbiter and the earth is just a plain ball. There is no terrain, no real haze and so on.

True, I think the main reason why the curvature appears so clear in Orbiter, is because of the unlimited visibility. In the real atmosphere, there is never such a thing as unlimited visibility. Also details like cloud bands and ground elevation play a role indeed.
If there is anything a new Orbiter OVP graphics client need, I would vote for haze/limited visibility on lower atmpsphere. (not sure if OGLA already has this?).

regards,
mcduck

[edit] ah we posted at the same time cjp, posts content are overlapping a bit

 

[Urwumpe]

If there is anything a new Orbiter OVP graphics client need, I would vote for haze/limited visibility on lower atmpsphere. (not sure if OGLA already has this?).

Seconded.

Maybe also volumetric cloud layers.

 

[Hielor]

Orbiter horizon curvature can be visible sooner than it should because the visible area is noticeable not an actual sphere. That is, the textures are mapped to a polygonal object which is circumscribed by the actual spheroid shape of the planet.

You can notice this at places where there are no hi-res textures. The visible ground is far below the place where you land (and the place where your shadow shows up). In some places it rises to meet you, and then falls back quickly due to the shape.

It's just not noticeable because the textures on planets have very little detail at low altitude.

 

[Scrooge McDuck]

Orbiter horizon curvature can be visible sooner than it should because the visible area is noticeable not an actual sphere. That is, the textures are mapped to a polygonal object which is circumscribed by the actual spheroid shape of the planet.

No, I don't think that this is the biggest cause. For planets in Orbiter, you can set the sphere detail (segments). Even if you set this to a high level, you will still see the curvature at altitudes where you wouldn't directly expect it at first. I still think it would be the same in reality when there was unlimited visibility.
The other effect you mentioned, only appears for non-spherical objects in Orbiter, not earth. Also consider that looking at the world in Orbiter would be the same as looking to the real world through a wide angle lens (depending on your FOV), which affects the curvature of the horizon depending on your attitude.

I tried to measure the curvature in a picture I made

But let me add, that horizon curvature in a photograph (or in a simulation) also heavily depends on lens barrel distortion. In your case, also the aircraft window (as lens) might affect it too.
Therefore, it may be quite hard to know if a line in a photograph was straight or curved in reality. Unless you can completely and accurately compensate barrel distortion, I think this measurement is not accurate.

 

[T.Neo]

There was Artlav's Nebo for clouds, but the clouds were generated as smoke and they were lit up at night.

Maybe cloud vessels, that display cloud meshes?
As for weather, I think there is a weather mfd on OH. How difficult would it be to model weather in Orbiter?

 

[cjp]

Also consider that looking at the world in Orbiter would be the same as looking to the real world through a wide angle lens (depending on your FOV), which affects the curvature of the horizon depending on your attitude.

If Orbiter is only a little bit similar to other 3D applications (and I think so, because it would be hard, and not really useful, to make it different), its perspective projection follows a perfect pinhole camera model. This means that straight lines in 3D project to straight lines on the screen. So, whenever something is not straight on the screen, it can't be straight in 3D.

But let me add, that horizon curvature in a photograph (or in a simulation) also heavily depends on lens barrel distortion. In your case, also the aircraft window (as lens) might affect it too.
Therefore, it may be quite hard to know if a line in a photograph was straight or curved in reality. Unless you can completely and accurately compensate barrel distortion, I think this measurement is not accurate.

I already measured this effect on the same camera for a different purpose, and I estimated it to be less than a pixel. I think the camera already corrects for its own lens distortions in a pre-processing step.

That leaves the aircraft window of course. I have more pictures from aircraft windows (using the same camera), so the next step would be to measure lines of which we know they are straight in 3D.

 

[FordPrefect]

Well, in my point of view it still looks phantastic. At least there is nothing which looks better if we talk about watching the earth from space, not FSX, not X-Plane and not even Space Shuttle Mission 2007

While it does not simulate spaceflight in the Newtonian meaning, it does look a lot better in Celestia.

Here's an unedited Screenshot of Earth (with my customized Earth texture) in the current Celestia beta release.

Also the stars look fantastic. And the moon displays a nice zero-phase glare, I don't know how it is done, but it's a great effect (just like in the real world).
Celestia's graphic engine for deep space visuals is pretty nifty I must say. Admittedly, I got a bit spoiled by playing around with Celestia.

 

[Scrooge McDuck]

If Orbiter is only a little bit similar to other 3D applications (and I think so, because it would be hard, and not really useful, to make it different), its perspective projection follows a perfect pinhole camera model. This means that straight lines in 3D project to straight lines on the screen. So, whenever something is not straight on the screen, it can't be straight in 3D.

I meant this: the first image (attached), shows Orbiter using extreme fisheye-wideangle FOV, looking at the horizon.
The second image, is exactly the same wide-angle FOV camera, at exactly the same position, but now looking just slightly more downward. Notice the difference in curvature?

I already measured this effect on the same camera for a different purpose, and I estimated it to be less than a pixel. I think the camera already corrects for its own lens distortions in a pre-processing step.

What camera do you use? Most cameras don't do this. The effect I describe, depending on the lens and orientation, can be hundreds of pixels...
Again: I think the accuracy needed for this photograph (we're indeed talking about a few pixels to measure), is way larger than the error generated by lens (and cabin window) effects, making this measurement useless in this case.

 

[Zachstar]

I think it is rather useless to think about adding many of the things that make atmospheric flight more realistic. Look at how long it took flightgear to implement many things.

If Martin decided to open source it one day (This is not a call to do so just an obvservation) The code might be able to be used to do so but that would be ALOT of work still.

And yes Orbiter knocks the pants off of FSX when it comes to high speed, high altitude flight.

As for Orbiter being bad looking. I do not think anyone can effectively deny that. It is a DX7 game without modern shaders. And only a single layer of texture mapping as well. With the ability to use shaders and other effects Orbiter craft can look just as good as FSX.

 

[Urwumpe]

Just start at adding bump mapping... Orbiter's rendering is not comparable to other programs, so it is hard drawing comparisons.

A shader which looks good inside the atmosphere of Earth, looks bad in space and even worse on Mars. Getting space looking good (for example, the extreme contrasts) is more important.

Environment mapping (for example the blue glow of Earth, but also the reflection of the lunar surface on Apollo spacecraft) would next be something, FSX does not need to show, but what Orbiter would require as advanced visualization.

I recommend everybody visiting http://www.nasaimages.org and look at some spaceflight photographs. it is visually a strange place, which the current DirectX7 renderer already does better than FSX.

For example, look at this: http://hubblesite.org/gallery/spacecraft/10/lg_web.jpg

 

[cjp]

I meant this: [..] Orbiter using extreme fisheye-wideangle FOV [..] Notice the difference in curvature?

Yes, I see the difference. But still, if the pictures were taken from a height of 1 meter (or less), the horizon would be a straight line in the first picture, and not visible at all in the second picture. It would be a straight line in all pictures where it's visible.
OTOH, the answer to whether the curvature is visible may depend on the camera orientation, with such an extreme fish-eye camera.

What camera do you use? Most cameras don't do this. The effect I describe, depending on the lens and orientation, can be hundreds of pixels...
Again: I think the accuracy needed for this photograph (we're indeed talking about a few pixels to measure), is way larger than the error generated by lens (and cabin window) effects, making this measurement useless in this case.

I use an Olympus C-480 (see the EXIF metadata in the image file). Not quite the most expensive camera. Still, when I make a picture of a chessboard pattern, and fit lines through corner points that should be on the same line, the corner points are never more than a single pixel away from the fitted line.

'hundreds of pixels' would be way too much, even for a cheap camera like this. At a resolution of 4Mpx, that would be extremely visible, and unacceptable to most consumers.

In fact, for my master thesis, I worked with 'cell-phone' type cameras (the OV9653) which had this built-in correction feature. I think this feature is very common, especially for low-end cameras, where it's cheaper to correct the image in software than to equip the thing with better hardware.

BTW, I think we're drifting off-topic here.

 

[Scrooge McDuck]

Yes, I see the difference. But still, if the pictures were taken from a height of 1 meter (or less), the horizon would be a straight line in the first picture, and not visible at all in the second picture. It would be a straight line in all pictures where it's visible.

Just tested in orbiter at low altitude, and same effect visible: horizon gets pretty curved.
In fact, you can even reproduce something like the 2nd image I posted in realworld photography, while standing on the ground (using extreme wideangle or stiching).

'hundreds of pixels' would be way too much, even for a cheap camera like this. At a resolution of 4Mpx, that would be extremely visible, and unacceptable to most consumers.

Not at all, this is actually a very common effect. (maybe we're confusing different effects?) Even DSLR cameras clearly show this whenever you make a wide angle shot. See this page for example (for example, look at the horizon in the bottom 2 images on that page). Here's a page in Dutch about wide angle distortion.
When I take a wide angle shot with my Nikon digital SLR camera, with the horizon on the bottom, it appears curved the other way around. When the horizon is at the top of time image, it appears curved like in space. But there are some (extremely) expensive lenses that somehow manage to compensate this without post processing.

Anyway, you're right, we are drifting off-topic here.
I found some good links on the original MS Flightsim staff topic, will post them later.

 

[cjp]

Just tested in orbiter at low altitude, and same effect visible: horizon gets pretty curved.

Weird. I'll try it myself.

In fact, you can even reproduce something like the 2nd image I posted in realworld photography, while standing on the ground (using extreme wideangle or stiching).

I think that is different. In real-world photography you have lens distortion which causes a similarly looking effect. Especially fish-eye lenses: a 'pinhole camera' without distortions (like in 3D rendering) can never have a FOV >= 180 degrees, so any lens with a greater FOV must have distortions. Stitching also creates its own distortions.

Not at all, this is actually a very common effect. (maybe we're confusing different effects?) Even DSLR cameras clearly show this whenever you make a wide angle shot. See this page for example (for example, look at the horizon in the bottom 2 images on that page). Here's a page in Dutch about wide angle distortion.

I tried the example images in 'this page'. The difference between 'before' and 'after' is hardly visible, but it's more than I expected. In fact, calculated back to the resolution of my camera, it is 35 pixels at some places.
However, these are 16mm pictures: very wide-angle. My camera only goes to 38mm (which I used in my horizon picture), with a horizontal FOV of 50 degrees.
Also, the lens on my camera can not be replaced with a different one. That makes it easier for the camera electronics to 'know' the distortions caused by the lens.

Anyway, you're right, we are drifting off-topic here.

Maybe we can continue using PMs?

 

[Scrooge McDuck]

Maybe we can continue using PMs?

I applied some 'secret powers' to create a new thread...
Now we're not off-topic anymore

regards,
mcduck

 

[T.Neo]

Orbiter horizon curvature can be visible sooner than it should because the visible area is noticeable not an actual sphere. That is, the textures are mapped to a polygonal object which is circumscribed by the actual spheroid shape of the planet.

You can notice this at places where there are no hi-res textures. The visible ground is far below the place where you land (and the place where your shadow shows up). In some places it rises to meet you, and then falls back quickly due to the shape.

Could this be fixed
A:Increase the number of polygons
B:change collision detecting to match to polygons?

 

[Erupter]

My 2c about barrel distortion.
A fish-eye (or super-wide) lens causes extreme distortion below or above the central horizontal axis.
Giving in to such an approximation is the necessary trade off to cramp more things in a small area.
The tighter the viewing angle, the less distortion: super-tele give results similar to classical 2d drawings without perspective.

But such an effect becomes visible in the range of 25mm equivalent optics and below.
So there is not much point in showing real world fish-eye photos.
Instead try to find high-altitude shots at normal focal lengths.