At least you should collect 3 X 6 samples at a time for good variance of analysis. A positive control a negative control and one test sample and each should have 6 replicates for finding SD mode statistically significant. 

I agree with Ravi comment but for more confidence it would be better to collect 3 x 10 samples. Of course, it is up to the materials and space.

Hi, Dr. Ghasem and Dr. Ravi thanks for answers,

But it must be regarded that genetic materials is half-sib progenies and i want to know that how many number of plants from each family can be estimate expected mean of their family which contained all combinations of polycross nursery?

In the best random pollination situation in which any parental genotype get pollen from all of others, at least n (parent number in polycross nursery) plants for each half-sib must be presented, but it is not clear for me that 2 x n, 3 x n or ... is better?

In any case, four replicates will be enough; then standard error of means is preferable over standard deviation for the statistics.

In any case, four replicates will be enough; then standard error of means is preferable over standard deviation for the statistics.