TECHNOLOGIES FOR FSGLOBAL ULTIMATE
1) Fractional Elevation Values and Optimum Base Level
Others call it "32-bit" processing, which explains as follows: altitudes in FSX are rendered 16 bit, there are
no 32 bit values. But: the number format is flexibel, how these 16 bit are used is defined by the developer. He
can define, how many bits are used before the decimal point and how many after it.Whole numbers (Integer) can be
values between -32768 and +32767. Such a large range is not neccessary for the surface of the earth. If you use
one bit after the decimal point values can be between -16384 and +16383.5 with a resolution of 0.5m. If you use
3 "Fractional Bits" resolution already is 1/8m (1/2^3) respectively 12,5cm. Altitude ranges from -4096 to +4095.875m.
You just need a small value in the negative values, but in the positve it could be some meters more already in the alps.
A BGL e.g. in the alps could cover an altitude rage of 170m to 4550m. One could optimize it by moving the zero-point
and giving the BGL a new BASE LEVEL. This could be an altitude of 2048m, which actually is a nice round value (binary).
Within the BGL the 170m are valued as -1878m and the 4550m as 2502m. Thus we get along with 13 Bits by far and can use
the remaining 3 for values after the decimal point. So what did we do? A new compiler analyses the data and sets the maximal
possible Fractional Bits (max. 7) in relation to optimalen Base level. This ensures, we ALWAYS work with the maximal vertical
resolution. Then the data is put together from the various sources, modified for the Base Level and handed over as a 32bit-value
to the Compiler of the SDK. The worst vertical resolution in this product is just 1/8m. Only three grid squares in the
Himalayas needs 14 Bits fo altitude, thus reducing vertical resolution to 0.25m. Most parts of the world can be displayed
using 7 Fractional, equal a vertical resolution of 1/2^7m = 7,8mm (!). There even are mountains of 6000m altitude, which
can be displayed using 5 Fractional Bits (3cm vertical resolution) because of their very high Base Level (Tibetan Plateau).
2) Oversampling with Smart LOD Selection
As we go for higher resolutions BGL-internally, we can draw source data much sharper. If you compare LOD 11 over LOD 13 you
will see many more subtleties in the terrain. The nearer you come, the more you see. Check the screenshots.
It is NOT the resolution in meters which is responsible for the quality of a mesh scenery, but ONLY the quality of the source
data, actually the measurement procedure. You can see it very clearly when you compare meshes of FSDreamscapes (resolution
4,75m) against FSGlobal 2010 (resolution 19m): though FSDreamscapes has a nominal higher resolution, ridges and peaks are
drawn much softer and rounder. The many points of the mesh are used to show round terrain where actually steep rugged rock
exists. Prominent peaks of the alps are drawn deformed (Matterhorn) or simply are missing (Dufourspitze). Our new compiler
analyses the terrain in terms of abrasiveness and richness of detail. It looks for the greatest gradients, the Unser neuer
Compiler analysiert das Gelände im Hinblick auf Detailreichtum und Schroffheit (Kontrast), es werden die größten Neigungen
gesucht, calculates arithmetic averages means and standard deviations, then builds a histogram. If the terrain has sharp
structures, we use a high resolution (LOD 11 instead of LOD9), which is a considerable oversampling. But we also have flat
and even areas, where as a countermove we reduce to LOD8. To draw the Greenland Ice Sheet in LOD 9 would be pure waste of
resources. LOD6 would be sufficient. Conclusion: - What used to be LOD11 in FSGlobal 2010 now is LOD13 or LOD12 (LOD12
theoretically, did not turn up). - What used to be LOD9 in FSGlobal 2010 now is LOD8, LOD9, LOD10 or LOD11.
3) Intelligent Compression
All meshdata in BGLs are compressed as a matter of principle.
Otherwise the amount of data would be exorbitant. In FSGlobal Ultimate the amount of data increases by factor 15 against
FSGlobal 2010. On the other hand this shows, that our outstanding source data has a clear potential of oversampling. To be
able to release FSGlobal Ultimate, we let our compiler evaluate terrain quality and in areas of no significant terrain
structures we allowed lossy compression, which resulted in an efficient reduction of datasize. All distinctive structures
(mainly, but not only mountainious regions) have been compressed without loss not to loose any sharpness. Guaranteed.