Pushing terrain sizes one more step.

[bigben]

I'm still experimenting with just how far I can push my terrain tiling method. I finally reached a brick wall with GM4.78 and bit the bullet and upgraded to v9 last night (3x the cost of upgrading from TG0.9 to TG2 )  There are of course numerous benefits for me with the upgrade, but the main one was creating TER files from data a *LONG* way out of the UTM zone used for the data projection.

[attachimg=#1]
This is the beginning of a repeat of my very first TG0.9 animation <http://www.path.unimelb.edu.au/~bernardk/terragen/gallery/mov/gdfo.mov> (6.1Mb) made from a 513 TER of Australia.

The first thing to note is that because the path of the animation will be mainly along the east coast, I've used a UTM zone centred on the path, and then placed the base TER to the west in TG2 (X offset -1.75e+006). What that means is that I have just as must space to the east to potentially fill with data. The two smaller terrains that you can see near the bottom are 90m resolution TERs (4097x4097), with the one on the right being close enough to X=0.

If you watch the animation, there's a tall pointy peak with "snow" on it about half way through. That is the mighty Mt Koscuisko soaring to an impressive 2,200'ish metres . Here is a quick render of the same mountain from this terrain set.
[attachthumb=#2]

That's getting close to an increase in resolution of 2 orders of magnitude.

So it's definitely possible to cover a significant portion of a hemisphere with real elevation data. I've calculated that if I stick to 4097x4097 terrains I'll need around 300Mb of TER files to cover the path of the animation, but I should be able to cut that down using non-square terrains (after a bit of reprogramming of my database)

The hardest part now, is going to be keeping the camera "horizontal".

[attachthumb=#3]
This resolution test is looking NW across the Grampians which is near the right hand edge of the second smaller terrain (camera at -424193,-104995,-1.07628e+006)... So far so good... while we may not be able to cover an entire globe with terrain data, it is certainly a lot less limiting than some might think 

PS, the base terrain is 1100m resolution, over twice that of my Swiss Alps base.

[cyphyr]

Scrapes jaw from the floor.
WOW is all I can say
Richard

[bigben]

You might want to rest it there    I just found a set of 5" bathymetry data for the globe I downloaded a while ago that I couldn't open up before. Now I can add sea floor data to my terrain sets even better than my previous hack (http://picasaweb.google.com/ozbigben/TerragenRenders/photo#5096531565120271250, note small dark glitches around the coastline from manually editing ascii data). 

I'm not sure whether a water glitch with high camera positions has been resolved yet. If anyone wants a TER to test for an orbital render I'll take suggestions for a location.

[rcallicotte]

Ben, I love the way that green looks so realistic in the last picture.  The color might be a little off, but the implementation of your idea is neat.

[bigben]

Having a bit more fun with GM9. I decided to keep the bathymetry separate from the rest of the terrain for 2 reasons... 1: It makes it harder to calculate vertical adjustments (minimum altitude = 0 is easy) 2: The bathymetry data is only 1km resolution so I can create a much larger tile at the data's resolution than trying to include it with much higher resolution terrain data.  Now I just have to figure out the best way to merge the 2 in TG.

Here's a screengrab of an upcoming terrain set... All of New Zealand at full SRTM3 resolution (6 non-square TERs, 240Mb)
I had to also modify my blending masks to accomodate non-square terrains (only 3 extra nodes). While there are advantages to non-square terrains it would be really handy if you could obtain the altitude range from the TER file within TG2 since I can't adjust the altitude range for these in TG0.9