On waring's problem : some refinements