This is an algebra question. My problem is following: I am reading a list of coordinates: line#0: rx ry rz line#1: rx ry rz ... line#n: rx ry rz All variables (floats) represent a rotation arround each axis in radians. Which means: float rx = rotation on x axis [-2pi..+2pi] float ry = rotation on y axis [-2pi..+2pi] float rz = rotation on z axis [-2pi..+2pi] My problem is, I am usisng DirectX D3DX functions. When i say: D3DXMATRIX mx, my, mz, matrix; D3DXMatrixRotationX(&mx, &rx); D3DXMatrixRotationY(&my, &ry); D3DXMatrixRotationZ(&mz, &rz); matrix = rx * ry * rz; Everything works fine. Now i tried this: D3DXMatrixRotationYawPitchRoll(&matrix, ry, rx, rz); This one does not work! It is not a gimbal lock problem, because i tried it also with quaternions and get the same results. On the other way, when I just try one component: D3DXMatrixRotationYawPitchRoll(&test, 0.0f, rx, 0.0f); test == D3DXMatrixRotationX(&mx, &rx); So each compmonent is the same as above, but when using YPR they get corrupted when they are multiplied all together. Please do not point me to GIMBAL LOCK, because I am sure it is not a Gimbal lock problem. Thank you in advance!

First let's have a look at your input. You said:

float rx = rotation on x axis [-2pi..+2pi]float ry = rotation on y axis [-2pi..+2pi]float rz = rotation on z axis [-2pi..+2pi]

This seems a bit strange; a rotation expressed in radians has a range [0,2Pi]. Your rotations have a range [-2Pi,+2Pi]; basically it defines that any object can have twice as many (unique) rotations which is not the case. I'm not sure how DX handles this; maybe they expect the angles in the normal [0,2Pi] or [-Pi,Pi] range.

Second I see you are very sure it is not gimbal lock. Please not that just constructing a quaternion out of thee Euler rotation angles is not magically going to remove gimbal lock. But since it is not gimbal lock...

Greetz,

Illco

Quote:Original post by Samurai Jack
D3DXMATRIX mx, my, mz, matrix;
D3DXMatrixRotationX(&mx, &rx);
D3DXMatrixRotationY(&my, &ry);
D3DXMatrixRotationZ(&mz, &rz);
matrix = rx * ry * rz;

Everything works fine. Now i tried this:
D3DXMatrixRotationYawPitchRoll(&matrix, ry, rx, rz);

try this:
D3DXMatrixRotationYawPitchRoll(&matrix, rx, ry, rz);

note that I just changed the yaw axis from y to x, as in your first code sample

That is not correct.

D3DXMatrixRotationYawPitchRoll( Y, X, Z ) is the correct order.

Yaw = Y rotation
Pitch = X rotation
Roll = Z rotation

I checked if seperatly:
D3DXMatrixRotationX(&matrix, value) == D3DXMatrixRotationYawPitchRoll(0, value, 0)
also
D3DXMatrixRotationY(&matrix, value) == D3DXMatrixRotationYawPitchRoll(value, 0, 0)
also
D3DXMatrixRotationZ(&matrix, value) == D3DXMatrixRotationYawPitchRoll(0, 0, value)

I double checked that for sure.

Quote:Original post by Samurai Jack
D3DXMATRIX mx, my, mz, matrix;
D3DXMatrixRotationX(&mx, &rx);
D3DXMatrixRotationY(&my, &ry);
D3DXMatrixRotationZ(&mz, &rz);
matrix = rx * ry * rz;

What you are doing here is:

1. rotate rx radians around the x axle
2. rotate ry radians around the y axle
3. rotate rz radians around the z axle

Quote:Original post by Samurai Jack
D3DXMatrixRotationYawPitchRoll(&matrix, ry, rx, rz);

Now the difference here is that D3DXMatrixRotationYawPitchRoll does the rotations in a different order:

1. rotate ry radians around the y axle
2. rotate rx radians around the x axle
3. rotate rz radians around the z axle

and that will not give the same result.