Page 3
Combat Flight Simulator 2 - Part II
By Jim TittleCSIM: We take it for granted that graphics will be superior to previous W.W.II Sims. Technically speaking, what kind of effort did this involve?
Tucker: Graphic improvements are never as easy as people think. First it takes a lot of very talented artists to make the objects and the textures. Then it takes a lot of equally talented people to contrive any code solutions necessary to implement them.
Corsiar in the mist
In the case of aircraft, we had to improve our ability to do animations as part of our models so that we could support folding wings, moving canopies, etc, as well as completely redesigning the way we do virtual cockpits to improve the overall look of them.
For our maximum detail models we are using texture sheets which are 4x bigger than the ones we used in FS 2000. This gives us far more detail. In addition to some technical problems that had to be resolved to get these working, the added detail put a lot more pressure on the artists, not only in terms of reaching the quality bar we wanted, but also because the added detail made visual accuracy a bigger challenge.
In the case of terrain, the FS 2000 engine had a lot of really impressive features, but there were some technical challenges we had to address. In FS 2000 the method of making shorelines involved drawing a lot of polygons, and could be a frame rate problem.
Since we knew we were going to have a lot of shorelines, we needed to make them not only look a lot nicer, but we needed to ensure that they didn't kill the frame rate. That meant completely revising the way we do shorelines. We also redid the way roads and rivers are modeled. The result is a much better look, as well as better performance.
Many of these islands aren't very big and therefore we needed to get higher detailed terrain data than we usually use. For instance, if we used a DEM (Digital Elevation Map) that had one data point per kilometre - what most of FS 2000 was modeled at - we would have had whole islands that would have been lucky to have even one data point.
We got our data at density of one point every 75 meters for much of our area of operation, which solves the accuracy problem but makes enough data to choke low-end machines, so it also necessitated rewriting a lot of the code that handles the terrain data.
