What condition must have the weights in PLSR?

Well, this can be an easy question: "they are orthogonal between them", anyway I just want to check it, and the simple way is to combine all the weights in a matrix, and to multiply:

(Weight.matrix transposed)  * Weight.matrix

and the result must be the Identity Matrix. So let´s check it:
Xodd_pls3_w1
Xodd_pls3_w2
Xodd_pls3_w3
Xodd_pls3_w4
w.matrix<-cbind(Xodd_pls3_w1,Xodd_pls3_w2,
                Xodd_pls3_w3,Xodd_pls3_w4)
round(t(w.matrix) %*% w.matrix,4)

> round(t(w.matrix) %*% w.matrix,4)

            Xodd_pls3_w1 Xodd_pls3_w2 Xodd_pls3_w3 Xodd_pls3_w4 Xodd_pls3_w1   1 0     0 0 Xodd_pls3_w2   0 1     0 0 Xodd_pls3_w3   0 0     1 0 Xodd_pls3_w4   0 0     0          1

 
So the condition of orthogonality  is fine.