Vector Rotation Around Axis

I didn't proof your code carefully, but it looks like you're applying a rotation about the world up axis, correct?

Is this for a typical first-person-style controller? If so, there shouldn't be any need to rotate vectors manually as you're doing. (Actually, it's not really necessary to do that for any type of camera, although it makes more sense for some types than others.)

If it is a first-person controller, I'd recommend representing the orientation using yaw and pitch angles, and building a matrix from them as needed. You can then extract the direction vectors from the matrix (if you need them for, say, linear motion).

For this method I simplified the rotation around an arbitrary axis matrix with one that simply rotated around the Y axis; believing I would get a different result. However, the result of the wavey camera remained. I am still unsure where to go from here except to take a completely different direction, either using quaternions or storing the yaw/pitch/roll individually.

Can you show [color=#660066]ResetViewMatrix()[color=#666600]?
[color=#666600]

[color=#666600][color=#000000][font=arial, verdana, tahoma, sans-serif]

I suspect the problem is in building the view matrix. Like the concatenation order is off. Could be you need to reverse the order and values of the translation and rotation. I think of building a camera matrix as the opposite of building a world object matrix - reverse the order of concatenation and negate the values. Like instead of translating(10,10) and then rotating by 10.0 degrees you would first Rotate by -10 and then translate by (-10,-10). Another way to do this is to build the matrix as you would a normal world object usual and then calc the inverse of it to use as the camera matrix.[/font]



int CD3D9Camera::ResetViewMatrix()
{
CVector Z = (m_Forward - m_Position);
Z.Normalise();

CVector X = m_Up.CrossProduct(Z);
X.Normalise();

CVector Y = Z.CrossProduct(X);

m_View.m_f11 = X.m_fX; m_View.m_f12 = Y.m_fX; m_View.m_f13 = Z.m_fX; m_View.m_f14 = 0.0f;
m_View.m_f21 = X.m_fY; m_View.m_f22 = Y.m_fY; m_View.m_f23 = Z.m_fY; m_View.m_f24 = 0.0f;
m_View.m_f31 = X.m_fZ; m_View.m_f32 = Y.m_fZ; m_View.m_f33 = Z.m_fZ; m_View.m_f34 = 0.0f;
m_View.m_f41 = -X.DotProduct(m_Position); m_View.m_f42 = -Y.DotProduct(m_Position); m_View.m_f43 = -Z.DotProduct(m_Position); m_View.m_f44 = 1.0f;

return 0;
}

This method appears to have been working fine for a couple weeks, I'm not sure how it could be causing problems. I've been working on this all this evening too and I'm still getting nowhere. I'm not far off making a quaternion camera just to overcome this.



...



This method appears to have been working fine for a couple weeks, I'm not sure how it could be causing problems. I've been working on this all this evening too and I'm still getting nowhere. I'm not far off making a quaternion camera just to overcome this.

Assuming m_Forward is a forward vector, then i don't believe you want to calculate Z the way you are, this may have worked if your camera was at <0,0,0> (or any where on the -Z axis), but if you put pen to paper with any other position, you should see that this is wrong... If you are already storing forward, up, and right vectors in world space, you can use them for ZYX respectively. The assignments to m_View look accurate otherwise.

Solved thanks, the view matrix was incorrect. Thanks!