I want to convert a recursive real formula to rational number representation, but I get the wrong response. For the real formula: k = 1.9903694533443939 u1 = -12.485780609032208 u2 = -6.273096981091879 u3 = k * u2 -u1 /// 1st iteration u3 = -1.7763568394002505E-15 // approx. zero /// 2nd iteration u3 = 6.273096981091875 /// 3rd iteration u3 = 12.485780609032204 which are the correct values For the rationalized formula we have kn=39270881, kd=19730448 u1n=-234750193, u1d = 18801403 u2n=-181086902, u2d=28867225 u3n = u1d * u2n * kn - u1n * u2d * kd; u3d = u1d * u2d * kd; /// 1st iteration u3n = -21418193, u3d = -4484756162545334248 u3n/u3d = -3.552713678800501E-15 // approx. 0, ok /// 2nd iteration u3n = -417378632937067183, u3d = -909312338189447552 u3n/u3d = 6.273096981101384 // ok /// 3rd iteration u3n = -366571621992223341, u3d = 216113855164233728 u3n/u3d = 0.913588924930172 // wrong! Can anyone figure out what's wrong? Thank you in advance.