The experimental design of mixtures for multivariate calibration is introduced. The idea of this design is based on uniform distribution of experimental points in a concentration hypercube and does not employ the concept of fixed concentration levels.