PCA generated initial matrix in gaussian and ellipse?

I have to do PCA in Matlab for object recognition.

For the input matrix / data, it should be fulfill the condition (to be distributed in Normal Gaussian) and in elipse shape (when I plot it).

For now, I generate matrix randomly :

   [a,InputMatrix] = sort(rand(100,20)); %Rows=100 Columns=20

But if I perform PCA in that Random Matrix (InputMatrix), the shape of eigen vectoor e1 and e2 that I get for the PCA result will be wrong (when I plot).

I think, I have to do rotation, scalling and other things to do it..

But I dont understand..

Could someone please help me to generated random matrix with Normal Gaussian and in ellipse shape?? Please.. help me T_T