Up　to　now　the　approaches　of　fault　diagnosis　at　subnetwork-level　for　analog　circuits　using　self-testing　and　mutual-testing　are　confronted　with　an　essential　core　problem:　Is　there　an　error　area　when　the　constrain　that　all　torn　nodes　are　accessible　has　been　broken　through?　What　rule　must　be　followed　for　the　distribution　of　the　accessible　nodes?　This　paper　presents　a　theorem　to　answer　it,　and　proves　that　if　all　torn　nodes　are　accessible　it　is　a　special　situation　of　this　research　for　no　error　area　in　existence.