Hello All,
I made another try with the camouflage pattern last night.
This time, I cheated a bit:
I could not find a suitable continuous pattern for both sides, so I just used the same pattern for both sides.
In doing so, I finally figured out what was causing the texture mismatch issues.
My process is to find a continuous nice looking camouflage pattern that is about 700 pixels wide, cut it into pieces and overlay masks and then rotate the new image to be used as a texture.
(Note that the aeroplane is 29.33 feet (8.940 meters) long at a mostly constant scale of 10.24 feet per 256 pixels which makes it 733.5 pixels in length. Note also that the Spinner does not share the same texture or scale as the rest so does not need to be included.)
With just a bunch of random looking splotches and no ability to see how things overlap and fit together until it gets to the simulator, it is basically a matter of working pretty much blind until the simulator starts.
The real mismatch problem turned out not to be a Typo. It could better be described as a "ClickO".
The division of the camouflage pattern worked fine.
The conversion to BMP format worked fine.
When I did the overlay of each mask, sometimes I would forget to accept the change before continuing the edit.
Since the next step was to do a 90 degree rotation, the overlayed mask was the only thing being rotated.
The camouflage pattern underneath did not rotate and as a result did not match anything else.
With random splotches, one pattern pretty much looks like the next, but with the cheat last night, the left and right side Fuselage camouflage had to look the same. It became obvious if it did not, so I eventually saw where I was messing up.
Another item worthy of note is that although I know that my texture mapping is around 1/2 pixel or 0.24 inch away from where it should be in places, it may not always make a difference that can be seen in the model.
It is not a matter of our eyes not being able to see the difference in such a small scale; it is more a matter of the model not being generated in a way that will show such small differences:
From what I remember of SCASM coding, each vertex of a textured polygon is mapped to a pixel on the texture. Thus if the texturing scale is quite large (Greater Distance per 256 pixel bitmap), the internal conversion from AF99 to SCASM may not do well if multiple vertices are closer together than the distance between pixels. That is one problem.
Consider what this means: If your specified shift does not move a vertex enough so that it maps to a different pixel, you get exactly the same SCASM code and model as before even though your very exact specification has changed.
The second issue is that SCASM actually works on a Metric scale internally.
I believe it is 1/512 Meter resolution for the typical scale we use for these models and those are the offsets that are used from the Center of Gravity of the model to determine the location of a vertex.
So, from Aircraft Factory 99. we get to design things on a 0.01 Foot (0.12 inch) grid that then gets fitted into a 0.0769 inch grid and sometimes things that should be evenly spaced by the numbers in AF99 don't end up so evenly spaced when converted.
Note that the Wing Fillets are textured in a single colour without any pattern at all.
Note also that the Supercharger Intake still needs to be textured.
- Ivan.
Bookmarks