Design of Experiments

Klaus,
to replicate means to repeat the entire design. A 2**3 full factorial DOE contains 8 cornerpoints (=combinations). If you do 2 replicates you have to do 16 runs, so you repeat every treatment combination once. This allows, like mentioned in an earlier answer, the estimation of your experimental error.
Repeat is the repetiton of a single treatment combination. If your process does not allow fast changes (continues process like tube mill) replicates are very difficult so its adequate to take multiple sample at the same teatment combination.

The importance in knowing the difference between replication and repitition lies in the degrees of freedom allowed into the experiment for determination of significant effects.  In DOE, if you design an eight treatment combinations in your experiment and run a repeat of each treatment back to back, you have 8 degrees o freedom in the experiment.  If you randomize the 16 runs (8 treatment combinations repliacted somewhere in the experiment) and few if any are run consecutively, you have 16 degrees of freedom.  Subtle, but important.

About the differences between REPLICATION and REPETITION, the answers that have been showed in this discussion still not enough. So, I would like know if replication is better than repetition, because this design technique can bring us aditional information, such as the effects of the some condition or machine or operator or raw material.