Drafts
Drafts and number crunching for the raytiles implementation.
By Ziv Perry, 2026-05-01
where:
- C is the circumference of the Earth (approximately 40,075 km)
- φ is the latitude
- z is the zoom level
For our fixed size tiles of 256x256 pixels, we can calculate the tile width in meters at different zoom levels (z) and latitudes (φ):
- Equator
$\phi=0$ - TLV
$\phi=\cos(32^\circ)$
| Zoom Level (z) | Latitude (φ) | Resolution (meters/pixel) | Tile Width (meters) |
|---|---|---|---|
| 15 | Equator | 4.77 | 1222 |
| 15 | TLV | 4.05 | 1036 |
| 14 | Equator | 9.54 | 2445 |
| 14 | TLV | 8.10 | 2072 |
| 13 | Equator | 19.11 | 4890 |
| 13 | TLV | 16.20 | 4144 |
| 12 | Equator | 38.22 | 9780 |
| 12 | TLV | 32.40 | 8288 |
| 11 | Equator | 76.44 | 19560 |
| 11 | TLV | 64.80 | 16576 |
| 10 | Equator | 152.88 | 39120 |
| 10 | TLV | 129.60 | 33152 |
| 9 | Equator | 305.77 | 78240 |
| 9 | TLV | 259.20 | 66304 |
From now on, all calculations will be based on
From commercial aircraft, at 43K feets (~13Km) the distance to the horizon is approximately 250 miles (~400km).
From Pythagorean theorem, we can calculate the distance to the horizon d based on the height of the observer h and
the radius of the Earth R (let say Earth is a perfect sphere):
Where:
- R is the radius of the Earth (approximately 6371 km)
- h is the height of the observer
- d is the distance to the horizon
For
We limit the GPU pixels rendeing by limit the far plane to the horizon distance.
We are always load the surrounding tiles. We render only those we can actually see, but we have to load the entire raduis around us to support aircraft manouvers that can change the camera position and orientation quickly.
The following table shows the distance thresholds for each zoom level, it can be changed in the configuration.
| Zoom Level (z) | Distance Threshold (m) |
|---|---|
| 9 | - |
| 10 | - |
| 11 | 55000 |
| 12 | 25000 |
| 13 | 10000 |
| 14 | 5000 |
| 15 | 1000 |
In zoom level 11, radius of ~10 (350km width total) is our base zoom level. You can increase the radius to cover more area, but it will increase the number of tiles we need to load and render.
Iterating from -10 to +10 in both X and Z directions, we will cover an area of 21x21 tiles around the camera position at zoom level 11, which is sufficient to cover half of the horizon distance of 400km.
While the iteration is based on the lowest zoom level, the actual rendering distance is based on the camera height and the distance to the tile.
For each tile in base zoom level, calculate the distance to the camera (XYZ distance). Check in the thresholds to decide if we need to display it or not, and if we display it, if we need to subdivide it into higher zoom level.
Update the desired_keys with the tiles that should be rendered based on the camera position and zoom level.
For any tile in desired_keys not spawned yet, spawn it.
During the creation of the desired keys list, add a check if the tile is frustum culling and mark it as so. Sort the list by frustum culling and spawn first those who are in the frustum, and then those who are not. This way we can start rendering the visible tiles faster while the others are still loading.
Each zoom level has a distance threshold. If the distance from the camera to the tile is less than the threshold, we subdivide the tile into 4 sub-tiles of the next zoom level and repeat the process for each sub-tile. If the distance is greater than the threshold, we render the tile as is.
Each tile that not in the desired keys list is a candidate for eviction. We can evict a tile only if it has a replacement or if it is not is the right distance for its zoom level.
The tiles will follow the raylib application XYZ coordinates, where X is the east-west axis, Y is the up-down axis, and Z is the north-south axis.
- Currently max zoom level is 15 (provider dependent).
- There is no limit of the min zoom level, but we will set it to 9 since zoom 8 covers an area larger than the horizon distance.
To work with float numbers required by the vectors of raylib, we have to use an anchore point for the tiles, and calculate the offset of each tile from the anchor point.
If we want to travel to far that floats are not enough to represent the position of the tiles, we can change the anchor point to be closer to the camera position, and calculate the offset.
To avoid calculate the tiles sizes, we will hold a pre-calculated size for each zoom level for a given latitude.
On construction, we will create for each supported zoom level:
- Tile size in meters.
- Min and max distance for rendering this zoom level.
- Resolution in number of vertices per tile.
- Mesh from the tile size and resolution.