Dot3 bump mapping on a GF2

I STILL have to ask some more questions, because there is NO easy understandable tutorial on dot3 bump mapping. There are great demos out there with source, but they are so hard to read and carry so much stuff with them I´m currently NOT interested in … pretty uncool sigh!

So here are my questions:

  1. How can I convert the light vector to tangent space and who can exactly explain WHY it´s needed and WHAT it is.

  2. Light vector consists of 3 components that have a range from -1.0f to 1.0f after normalisation, right?

  3. What is, if I use mip mapping for the normalmap, will the results be false, because the normals are scaled wrong? Should I only use linear filtering or NO filtering or doesn´t it matter and I can use trilinear filtering for it?

  4. If I need the light vector from every vertex, that´s involved in the bump mapping, I have to know the exact position, this is simple at the beginning, but what´s after a translation or rotation?
    How can I get the values I need to compute the light vector?

  5. Perhaps anyone knows a GOOD and easy to understand resource on the web for my questions (don´t point me to NVIDIA, I have got most of their docus here, BUT … TOOO much stuff for me, that I don´t really understand)?

Regards,
Diapolo

btw.: I´m really getting crazy, because of the bump mapping topic … am I that stupid or is it because I´m german and don´t get everything that easy?

first, diapolo, i give you the tip to download some documentation, cause there the WHY is easy described… goto nvidia.com/developer and have some fun there

1.) how is describet, too, just how to generate tangentspace not really, but i can give my code… it works great

2.) after normalization, every vector has 3 components with range [-1,1], yes, and how to compress em to [0,1] i have described above

3.) yes, they go “corrupt”, and how to generate your own mipmaps can be found in one code, and its not so difficult to use it really ( this part of code ) i dont use mipmapping at all anymore, simply GL_LINEAR for filtering looks good enough anyways, cause i like it at the moment it has some filtering problems, then you can really see the bump, else its filtered away and smoothed like that…

4.) for this you have to do some transformations from the different spaces you are working in… terrible to get it under control… so much here:

you set up your position ( glLookAt or something ), then you transform your lightsource by it ( cause it is in objectspace i think ), then you set up the transformation for your current to render mesh and transform the light by its inverse matrix, then you transform it per vertex into tangent space… funny, not?

  1. as i mentoined above, the quellcodes are not so good to understand, but the documents give you much bout theory… i try to set up a nice sourcecode for all the helpless guys out there cant wait to get it finished… think at the weekend or something it should be finished… hm tomorrow i am away, thurstday i am away, friday, too, saturday too, sunday, too… ok next week i think

btw, ich weis nich, hab die ganze scheisse auch verstanden obwohl ich deutsch spreche, und eigentli sogar schwitzerdütsch und nidemol dütsch, vo däm här no schlimmer abr sone problem isches nöd, aber so ein problem ist es wie gesagt nicht, du kannst es, you can do it just give me little time to write it for everybody…

(crazy germans :slight_smile:

anyway, I do have a real question…when you’re doing the ‘special mipmap’ thing, do you just add the 4 vectors together, then normalize?

yep, that works nice i think… ( never done it, as i said, protesting mipmaps )