The double central composite design with 29 points rcpresents an extension of the central composite design and was developed for the study of three factors in more than three levels. Essentially it is composed of two 2³ factorials (levels 1 and B), of two stars (levels alpha and 2alpha) and of one central point. In the present paper the origin of the design, the double central composite completely randomized not orthogonal in 5 levels (-2, -1, 0, +1 and +2), the completely randomized design, orthogonal, in nine levels (-3.02; -2.00; -1.51; -1.00; 0.00; + 1.00; +1.51; +2.00; +3.02) and the orthogonal design divided in two blocks (levels -7.262; -4.391; -3.631; -1.000; 0.000; +1.000; +3.631; +4.391 and +7.262) are shown. One example of the late design is presented with its correspondent analysis. According with Box and Wilson criterion, this last design is more efficient than the 3 x 3 x 3 factorial. They can be used in fertilizer programs aiming the most efficient economical recommendation of dosages and in other areas of scientific research in which we want to evaluate the surface response equations and its extremum properties looking at the maximum response.