higher bound for a

float point error function a