Control　systems　which　contain　pure　time-delay　are　considered　to　be　very　difficult　in　control　process.　The　difficulty　will　grow　with　the　proportion　of　delay　time　to　dynamic　time　of　the　process.　The　traditional　Smith　predictor([1])　makes　improvements　on　the　basis　of　conventional　controller.　Even　though　the　predictor　can　solve　the　problem　of　control　scheme　with　forward　path　time-delay,　but　not　for　the　systems　contain　pure　time-delay　in　feedback　path.　This　article　puts　an　improved　Smith　predictor　scheme　which　could　overcome　negative　influence　made　by　pure　time-delay　in　feedback　path.　In　the　system　which　contains　time-delay　in　both　forward　path　and　feedback　path,　the　results　of　simulation　show　that　this　scheme　has　remarkable　positive　effect　on　control　processes.