there are negative and positive beads, (positives that trap your antibody negatives don't) you put one drop of each in a tube then and add your single antibody (one tube for each) except for the negative control.
then you can run your compensation on it
it is not the perfect control but it does the job.
When cell numbers are limiting, you can select only the antibodies that give complicated gating results and run FMOs only for those. You also do not need to put as many cells into an FMO tube as you would for your normal fully stained sample. It's possible to run without FMO controls, as long as your data are properly compensated and your gating is clear. For example, when I run mouse t cell panels, the CD45, CD4 and CD8 antibodies give very clear positive and negative populations, making FMOs for these colors unnecessary. Things like CD3, CD44, and CD69 may not be quite so clear and would benefit from the FMO.
FMOs generally control for negative populations that shift positive due to the assay fluorophore makeup and properties of the cell-type.
As Nikki said, you only need FMOs where this impacts on your assay and only use enough cells to see all of the populations clearly. However, there is no compensation strategy which will remove the necessity of a FMO.
Also, I would not use beads for FMO controls. Part of a FMO's job is to account for the cells impact on the assay (autofluorescence).