Board index FlightGear Release candidates 2.10

FG 2.10.0 prerelease for Mac won't start

This is the archive of topics about the 2.10 release candidates.

Re: FG 2.10.0 prerelease for Mac won't start

Postby fmg » Sat Feb 16, 2013 12:14 am

Thorsten wrote in Fri Feb 15, 2013 12:58 pm:What does /environment/ground-visibility-m say?

5000
Thorsten wrote in Fri Feb 15, 2013 12:58 pm:As for darkness:

direct sunlight: 32,000–130,000 lux
very dark overcast day: 100 lux

There's a factor 300-1000 between the actual amount of light in the scene. I think you seriously underestimate overcast weather (or should do something to your screen brightness, that may also contribute here...)

OK - My screen specs: 6500°k, Gamma 1,8, 100 cd/m2.
I do professional photo editing, so I use a hardware calibrated monitor and can't change this setting just for FG. 100 cd/m2 are off course on the low end today, but I have to stay compatible with other sources that are using CRT-monitors. Even I have still some of that kind on the start. LCD-monitors nowadays easily go up to 400 cd/m2 or even more. But I don't think, that it's wise to make it such bright, even if it's possible. CRT-monitors typically hardly reached 120 cd/m2 when they where new and now 100 for them may be more realistic. I sometimes wonder if brighter monitors may be a reason for some things getting darker.
You re probably right about the amount of light reaches the earth. But fortunately our eyes are capable to adapt. So if something is 1000 times darker we can tell this via measurement, but hardly by looking. A white sheet appears to be white to us if we look at it in bright sunlight, or under candlelight. If I take a camera with exposure set for sunlight, the picture under candlelight will be near to black. So IHMO perception should be more our guide for brightness than measurement.

I don't doubt that there are weather situations that appear so dark by perception, but the title of the METAR string is Marginal VFR (XXXX 012345Z 23010KT 5000 SHRA SCT012 BKN018 OVC060 15/11 Q1010). Overcast for me isn't VFR (OK - you can stay under the clouds. But I guess this might not be the intention).
I've done some new comparison. On the top is FG 2.10.0b375 without skydome scattering, next with skydome scattering. On the bottom is the same METAR string in FG 2.0. In this specific case I found the FG 2.0 version the most plausible, apart that skydome scattering does the better job in FG2.10. The second picture is the same scenery at 5000 ft. In FG 2.10 you can't see any ground. So hardly VFR. Apart from the brightness the METAR seemed to be interpreted in a quite different manner by FG. Maybe the optimum is somewhere in between. If your screen is much brighter than mine, the FG 2.0 version should look way to bright for you?
Image
Same scene at 5000 ft.
Image
Thorsten wrote in Fri Feb 15, 2013 12:58 pm:It's difficult to figure out the problem if you have one screenshot where the skydome comes out right and one where it doesn't.

I know. Unfortunately I haven't found a scheme yet, on that the effects rely. I had the hope my last post can clarify something. But unfortunately not. But is there another/better way than show and discuss it here?
Just let me say that for me the skydome scattering looks like the way to go and I like it very much and it has made a great progress. It mostly works fine, especial under sunny conditions. That the adapted light rendering makes light more interesting and realistic. Would be glad, if it could be best under all circumstances.

About the screen brightness: Even if you don't have a measurement device for calibration, you can get for example under http://www.eci.org/de/downloads monitor test-files, that allow a rough visual evaluation. In there are a black and a white field with the string eci in it. Your monitor is well, if can't read it, but guess that there is something.

In the end back to the red light again. Here a new example with a visibility manually set to 4032 m. The red lights appear much darker in the skydome scene. If this correct, at least FG is incoherent because the other lights stays at the same brightness. In reality red light reaches much further than blue or green.
Image
Skydome scattering in the upper one.
Last edited by fmg on Sun Jun 09, 2013 10:11 pm, edited 1 time in total.
User avatar
fmg
 
Posts: 530
Joined: Tue Jun 29, 2010 5:13 pm
Location: EDDI
Callsign: fotomas
Version: 2
OS: Mac OS X 10.6.8

Re: FG 2.10.0 prerelease for Mac won't start

Postby Thorsten » Sat Feb 16, 2013 7:04 am

What does /environment/ground-visibility-m say?

5000


Then there is no arguing - up to the uncertainty at what amount of fogging the weatherman says that he can no longer see an object (to my surprise, when I investigated how visibility is actually defined, it turns out that this is basically up to the taste of the local meteorologist - you have to 'recognize' an object at the visibility you report) - if you feed the shader 5000 m visibility, it will evaluate something lik ~exp(-dist/5000) to get the amount of fog. You can't cheat the math going in there. 8)

You re probably right about the amount of light reaches the earth. But fortunately our eyes are capable to adapt. So if something is 1000 times darker we can tell this via measurement, but hardly by looking. A white sheet appears to be white to us if we look at it in bright sunlight, or under candlelight. If I take a camera with exposure set for sunlight, the picture under candlelight will be near to black. So IHMO perception should be more our guide for brightness than measurement.


Which is why the computed intensity gets a logarithmic reweighting for perception assuming that the Weber-Fechner law describes intensity perception. Which gives me (as tested on the three different LCD screens I have available) plausible lighting all the way from noon to moonlight. What is not accounted for is of course that your eyes also adapt to the ambient light in the room - if I use FG in a dark room, moonlight even set to a small value of 0.1 gives me decent illumination of the scene. If I try the same on a sunny day, I can't see a thing. But it's impossible simulation-side to guess the amount of ambient light in your room, which is why I was asking for screen brightness.

I don't doubt that there are weather situations that appear so dark by perception, but the title of the METAR string is Marginal VFR (XXXX 012345Z 23010KT 5000 SHRA SCT012 BKN018 OVC060 15/11 Q1010). Overcast for me isn't VFR (OK - you can stay under the clouds. But I guess this might not be the intention).


I did not define this weather scenario, and I won't go into an argument whether OVC is VFR condition or not. From my end - the string says OVC, the weather model correctly goes OVC, the lighting model computes the light attenuation both by the cloud layers and by the fog, the result is logarithmically reweighted, this is what you get. The code does what it's supposed to do.

Apart from the brightness the METAR seemed to be interpreted in a quite different manner by FG. Maybe the optimum is somewhere in between. If your screen is much brighter than mine, the FG 2.0 version should look way to bright for you?


The 2.0 picture isn't an overcast scene, the lighting is plausible for the 3/8 cloud cover visible, but that's not what the string says.

What you are arguing that Marginal VFR should use a different METAR string, on which I have no opinion, but it's silly to set up the weather system to ignore what the METAR string says to that a scene comes out as you would expect it.

Likewise, if you insert 5000 m into the fog function, this is what you get, period - I'm not arguing how to evaluate an exponential function, there is only one way to do that and I'm doing it right. If you can't see any ground from 5000 ft, then you need to stay lower. Again, I don't care if the scene is really suitable for VFR or not or where the string comes from, I care if the weather and shader system generates a scene corresponding to what the string says. Which it does.

Apart from the brightness the METAR seemed to be interpreted in a quite different manner by FG.


If the METAR string really was the same, then it was interpreted in the wrong way before and is correct now. You're not going to convince me that what I see in 2.0 is a credible 8/8 (or even more than 3/8) cloud cover. And I read clearly that the report says a broken layer at 1800 ft and an overcast layer at 6000 ft . I do not accept that what you're used to seeing is different than what you see now as an argument if the past state was demonstrably wrong.

In the end back to the red light again. Here a new example with a visibility manually set to 4032 m. The red lights appear much darker in the skydome scene. If this correct, at least FG is incoherent because the other lights stays at the same brightness. In reality red light reaches much further than blue or green.


There may be two issues here:

1) exponential fogging ~exp(-dist/ visibility) is physically correct, quadratic exponential fogging ~exp(-(dist/visibility)^2) as used in the default rendering scheme is physically incorrect and just a device to get more unfogged pixels for given visibility. As a result, there are things which are less fogged in the default scheme, but I'm not 'correcting' a physically sound scheme because people argue based on the physically incorrect scheme they were used to.

2) Any light model placed into the scene which declares its own effect and does not inherit from model-default.eff or similar brings in essence its own fog model independent of what the rest of the scene does. It's up to the model/effect maintainer to make sure it brings a consistent fog model or inherits from the right effects, I can't seriously be expected to check a few thousand airports all over the world. The shader is set up in such a way that if you do it by the book and inherit from the defaults (model-default and terrain-default) then you get the desired results, if you don't code by the book and throw out the inheritance structure, then you can get any mismatch you like.
Thorsten
 
Posts: 10587
Joined: Mon Nov 02, 2009 8:33 am

Re: FG 2.10.0 prerelease for Mac won't start

Postby someguy » Mon Feb 18, 2013 11:56 pm

Just to follow up on an RC2 problem, the 2.10 release version's airport dialog takes only a couple of seconds to display. I really like this dialog, it's crammed a ton of useful info into a single, easy to use point of access. Well done!
User avatar
someguy
 
Posts: 1653
Joined: Tue Nov 25, 2008 5:54 am
Location: USA
Version: 2016.2.1
OS: Mac OS X 10.11

Re: FG 2.10.0 prerelease for Mac won't start

Postby Thorsten » Mon Feb 25, 2013 7:33 am

I think I'm makeing some progress in finding out what goes on with the lights - it would appear that they are not processed by the correct rendering scheme. I'll keep you posted.
Thorsten
 
Posts: 10587
Joined: Mon Nov 02, 2009 8:33 am

Previous

Return to 2.10

Who is online

Users browsing this forum: No registered users and 1 guest