Singular Value Decomposition

to all, I'm sorry my english was bad, but I hope you understand about this question
So i want to ask about svd
does the svd can calculate the observation data that contains zero value ??
i have try it with R software,
in trial with eigen() function, i got some negative eigenvalue
in svd() function, it has a same result, i got the positive eigenvalue still in same value
but the negative eigenvalue that i got from trial with eigen() function was not negative anymore, but it changed with positive value and the number is different too.
because in a diagonal matrix in SVD, we need the square roots of the non-zero/positive eigenvalue, right??