Balanced approximation can be optimal

It is shown by example that there exist systems whose balanced approximations are optimal. Specifically, for various norms including the Hankel norm and the L2 norm, it is shown that the approximation error can achieve the minimum possible.