引用17楼详细内容：

      Hvar:=rCycle /( 2.0 * rTmLag)=H;

      Hvar:=(INV - RueckAlt) * H / (1.0 + H);

      OutvNew:=Hvar+RueckAlt:=(INV - RueckAlt) * H / (1.0 + H)+RueckAlt:=INV*(H/(1.0+H))+RueckAlt*(1.0/(1.0+H));

//这个类比:OutvNew=Yn;INV=Xn;RueckAlt=Yn-1;(H/(1.0+H))==a;(1.0/(1.0+H))==1.0-a;

      系数是H / (1.0 + H)还是2H / (1.0 + H)除了保证系数a位于0.0-1.0之间，应该没太多约束因为(INV - RueckAlt) * 2*H / (1.0 + H)+RueckAlt==INV*(2H/(1.0+H))+RueckAlt*((1.0-H)/(1.0+H));系数之和==(2H/(1.0+H)+(1.0-H)/(1.0+H)仍然保持==1.0

      证明:假设K为任意位于闭区间0.0-1.0之间的正实数则:(INV - RueckAlt) * K+RueckAlt==INV*(K)+RueckAlt*(1.0-K);两项的系数之和==K+1.0-K==1.0,且两个系数也均位于闭区间0.0-1.0之内

      但是选择系数a==H/(1.0+H）可能是为了严格保证其位于0.0-1.0之间

----------------------------------------------------------------------------------------------------------

      2H / (1.0 + H)约等于==OUTV(t)/INV0;

//INV0类比于Xn-Xn-1;OUTV(t)类比于Yn-Yn-1;

      RueckDiff:=2.0 * Hvar + sRest

//这个类比于微分差值+残差补偿;RueckDiff:=(INV-RueckAlt)*2*(rCycle /( 2.0 * rTmLag))+ sRest== (INV - RueckAlt)*((1-e-t/TM_LAG))+ sRest== (INV - RueckAlt)*(OUTV(t)/INV0)+ sRest==(INV - RueckAlt)*((Yn-Yn-1)/(Xn-Xn-1))+ sRest==类比于斜截式直线方程Y=Kx+b,斜率是微分值,截距是残余误差

----------------------------------------------------------------------------------------------------------

       以上整体滤波模型关系,可类比于对输入做带残差补偿及一阶惯性斜坡的微分模型,残差的具体补偿数值分析关系,以及滤波系数关系,要查资料.

       以上所有不一定对.