FSXA Transparency problems Z-order?

[Vitus]

Hi,

I was fed up with the Electra for the last days so I started something new - the Bleriot XI. Unfortunately I ran into some trouble and I hope you can help me sorting things out.

I setup a couple of semi-transparent material sets. One for the tire, some of the steel structures to mount engine and fuel tank and other stuff, one set for the cloth of the tail (stabilizer, elevator and rudder) which is just a little bit transparent so that the wooden framework beneath is visible.

But it gets better: the Bleriot XI didn't have ailerons, instead it used "wing warping" and I am planning to use skinned mesh animation for that. To reduce the amount of materials and textures, I want to use one material for the wooden framework and the cloth that covers it (not yet finished). Thus this material needs to be (semi-) transparent as well as "skinned".

As an example, here are the material settings for the wings:

Now the problem!
It looks fine in Gmax, and it also looks fine at a first glance in FSX but if you pay attention you will see the problem. On the first picture I want to show you the problems with the wing - it looks distorted, like the normals are flipped (its NOT a mirror issue, btw!).

With the second picture I want to show you some really strange behavior of the transparent parts. When a transparent part is in front of a non-transparent part it looks alright - see for instance the middle of the rudder where the wooden framework is barely visible through the cloth. But as soon as a transparent part is in front of another transparent part, the part underneath is shown completely (see the marked spots and keep in mind that the wings' material is in fact a transparent one!).

The same is going on with the steel elements that holds the engine and fuel tank. Here is everything ok:

But when you tilt the camera that one strap appears in front of another, it is messed up.

I hope you have some useful advise!

Regards,
Vitus

 

[hcornea]

Your Z-test is set at "Never" ... perhaps try "GreaterEqual" .. or even "Greater"?
May also need to check "Z-write"

The slider seems to be set at around 30 (even though the numeric is 0.0) .. this may need to be higher.

 

[Vitus]

Hi hcornea,

thanks for the advise! It worked!

but now I am curios: what did I do?

 

[n4gix]

In a nutshell, you're telling FS to render the objects using this specific material last.

IOW, it will always "be on top" of the render stack.

 

[Heretic]

Hi hcornea,

thanks for the advise! It worked!

I told you the same thing via ICQ...but nooo, Mr "Stone Age Technology is awesome!" didn't want to listen...

 

[Vitus]

nononononono! You said I should waste my time by fiddling around and by chance will get good results.

 

[Vitus]

Transparency problem #2

Unfortunately, the issue is not solved entirely. The first picture shows the prop-prob:

And if you have a closer look at the wheels you will notice that in the just-a-little-bit-transparent area, instead of the wing, the sky blends in, thus resulting in a blueish corona along the spokes:

The third thing I noticed is this:

One should be able to see the root of the wing through the side, but it is blended out and you can see parts of the landscape instead.

How do I have to manipulate the z-order parameters to get rid of these effects?!

Cheers,
Vitus

 

[Vitus]

Hey guys,

unfortunately, this issue still persists and it gets more annoying the more parts I add to the Bleriot. Therefore any advice is appreciated.

Regards,
Vitus

 

[lionheart]

I have problems also with this via multiple materials that are transparent, overlapping each other.

In the Kodiak, the visors are messed up when viewing from the outside in. They appear gone from outside, and inside are fine. (They are a semi-transparent dark plastic).

 

[Kai503]

Hi Vitus,

Oh well, I have fighted weeks on the problem and didn't figure it out. It was a hangar with semi transparent windows. At last I made those as seperat objects.

But yesterday evening I figured something out, what MIGHT help. Maybe somebody of the gurus can say more about it:
I was working on a house with a balcony. The handrail as every fence; a plane, cloned, normal flipped, both attached, DTX1 with alpha. However, one of the three handrails I didn't flip, but rotated it 180 degrees (yeah, it was quite late.. ). Z-writing in the material checked. In FSX all sides were fine, with one exception: the rotated one. The flipped ones worked fine.

Can it be that the poly which you want to see through an other one, must be flipped? I was running through hell with that hangar, inside was cut out with the boolen operation. At a test box I flipped the normals and it worked fine.

Just my

Regards,
Kai

Btw. May I ask how it comes that a guy from New Zealand is using a German version of FSX?

 

[hcornea]

How do I have to manipulate the z-order parameters to get rid of these effects

What may help (and took ages to get my head around)

FSX doesn't think about drawing in terms of physical space (juxtaposition of one bit of geometry in front of another)

It thinks of them in terms of Z "value" ... which you are specifying in the alpha.

DXT5 Alpha will give you more resolution.

By Z-testing, you are giving the material a weighting (independent of its position, and colour) but relative to the other materials that it overlaps (line of site)

The material that gets drawn is the one for which the Z-test is True. What is rendered depends on the Framebufferblend settings ... and you may get bleed-though if the SrcAlpha doesn't match well with the Z-test value used.

Beyond this ... ie if all the Zweighting / value is equal (or if it is not used) ... then the material "drawing order" comes into play.

The way the material drawing order is determined for a given model depends on the hierarchy used within the Gmax model - and in particular ... is different if the model has LODs.

TIP: Load your compiled MDL into Arno's ModelConverterX .. the materials are listed in compilation order ... which correlates with drawing order .... so you can see what is going wrong.

EDIT: For semitrans materials "Greater" rather than "GreaterEqual" may work better

Oh ... and good luck

 

[hcornea]

Can it be that the poly which you want to see through an other one, must be flipped?....

No ... I don't think so, the direction of the normals is enough. I have rotated planes many times to create.

Incorrect Z-test parameter or Test type, and possibly the poly is not parallel?

Don't know ... but it is possible to have it work.

 

[Vitus]

Hi Kai,

thanks for your thoughts, but the normals are ok. I changed the material to be non-transparent and the part looked alright.

Btw. May I ask how it comes that a guy from New Zealand is using a German version of FSX?

Gosh, you have a good eye! I am living in NZ for two years now but I am (and my FSX is) German.

Hello hcornea,

yeaaah! I was hoping for an explanation like yours, that tells me what I am really doing by moving these levers! But maybe you can give me an example for that? For instance how should I set the material properties for my canvas in the example above? The wing (canvas & wooden parts) using the same material as the cockpit covering (the canvas left and right to the pilot). I am using DXT5 with the alpha of the canvas being ~230 and the wooden parts of course 255. What Z-Test value would be appropriate?

Regards,
Vitus

 

[Kai503]

I am living in NZ for two years now but I am (and my FSX is) German.

Sorry for going shortly off-topic..

Dachte ich mir doch.. Same here, even I made it only 1600 km to the North.

Kai

 

[Vitus]

FYI: I couldn't solve the problem yet, so I decided to build the parts that caused most problems completely in 3D.
The all-new landing gear (one wheel 992 faces):

U-beam engine mounting (one beam ~1000 faces):

 

[lionheart]

FYI: I couldn't solve the problem yet, so I decided to build the parts that caused most problems completely in 3D.
The all-new landing gear (one wheel 992 faces):

U-beam engine mounting (one beam ~1000 faces):

My goodness man, that is some brilliant detail work....

Thanks for keeping us up with the progress...



Bill

 

[Blogga]

I would say don't be too afraid of higher polygon numbers. It is less importaint these days as even the casual flight sim user tends to run multi cored systems with more than enough ram to make poly processing just as smooth as texture processing.

I learned the hard way that the current breed of flight simmer want's the absolute maximum detail possible. Far beyond even the latest Mainstreen game levels. Software developers fuel hardware development because nearly every new release that comes for the pc pushes the boundries of curent hardware.

This is looking great by the way mate. Those woodgrain textures are stuning.

 

[Vitus]

Thanks for your comments! I am quite happy with the results so far and I think the decision to use geometry instead of semi-transparent texture sets was the right one.
With an aircraft like this, you can push the boundaries and really go into detail with the modeling, because there is almost no CPU/GPU power needed for the instrumentation. There simply are no instruments

The overall number of faces is now almost reaching 200,000 and there are still some parts missing. On the other hand this model barely needs texture maps, as all the parts involved are rather small. I can't sense any significant frame loss with this aircraft.

Cheers,
Vitus