"EMS" stands for "Expected Mean Square".
An EMS table is a method for determining the appropriate numerators and denominators for the F-tests performed during an ANOVA study. The form of the EMS table depends on several things, to include:
the form of the model equation,
the type of effects (fixed, random, or mixed; see link) and associated hypotheses, and
any restrictions on randomization, to including blocking, nesting, etc.
Both Hick's and Montgomery's texts provide reasonably easy to understand instructions for creating an EMS table for a study. (see attached .pdf files for excerpts)
For an in-depth study of the theory behind fixed, random, and mixed effects models, see the book "Variance Components" by Searle, Casella, and McCulloch.