I would like to know if any one knows the texture size per unit size that the quake 3 engine uses. Meaning, I am using 1 opengl unit to equal 1 foot, how many texture pixel would be good to use per foot. Also, i heard that lightmaps use much much fewer pixels per foot than the base image. What is a good pixel per foot on those also? And if i used such a small light map, what is the best interpolation to use??

hello
OpenGL units have nothing to do with size of geometry you see on the screen. It may look the same with OpenGL unit 1 and 10000.
Question about how many texels (edit: pixels) per foot would do best doesn’t have good answer either, just make it look detailful enough.
Lightmaps are always in lower resolution because they’re unique textures, but fortunately they usually look quite good in low resolutions along with highres mesh textures.

[This message has been edited by MickeyMouse (edited 02-05-2003).]

I think 1 texel/foot is ok for lightmaps. The other textures are typically repeated over a wall, so they can be more accurate. Just find the minimum resolution that gives you the result you want.

Q3 uses mainly 256*256 sized images. Those are used to cover a distance of around 6 foot. This is only a vague estimation, i didn´t measure it.
I don´t know Q3´s lightmap resolution.

In quake3, one unit is one inch. It really is more convenient to have units in inches than feet for most thing, since (for example) all integer numbers of inches can be exactly represented. In fact, many things turn out nicer if you can snap your geometry to an integer grid.

In Q1 and Q2, there was one texel per unit. In Q3, there are 2 texels per unit, so each texel is 1/2 inch square.

Lightmap density is controlled with -samplesize q3map command line option. The default value is 16, which means 16 units between samples. This is 1 lightmap sample for every 32x32 = 1024 texture samples.

I’m not sure what you mean about your interpolation question.

Oh yeah, the interpolation. Linear is about the best you can do, but you can improve that a lot by using well antialiased lightmaps, especially for the shadows. A little blur there also helps a lot.

Why would locking things to a 1 inch grid, instead of a foot grid change things that much? If i want a half a footi simply use .5 units. Or is floating point calculations realy that much of a problem?

Floating point represents all numbers as a rational number, where the denominator is a power of 2 that you get from the exponent. This means that you cannot precisely represent any rational number with the denominator not a power of 2. This means you can only represent 0, 3, 6, and 9 inches exactly; all others are approximated. If you do a lot of processing of vertices, you could get subtle problems that show up because of this approximation.

In other words, epsilon issues are a little easier to handle if you have a power of 2 grid, even if that power of 2 is negative (eg, 1/8 or 1/32).

As to metric units… they are better in many ways, but most artists are more comfortable in inches.