If the "gravitational centers" of the two proof masses are not identical, a radial offset error will look like an EP violation. For 10^-18 g , the centers must be known to 3 x 10^-12 meters. The method of AC cancellation is an attempt to measure the difference in the gravitational centers by using a spinning gravity-gradient field to cause an oscillation in the difference output of the two proof masses proportional to the separation error.

The paper referenced in this link describes the report at the 8th Marcel Grossmann Meeting in Jerusalem in 1997 of a satellite free-fall EP experiment using AC Excitation/Cancellation (PDF). The method of AC Excitation does not work very well, and at best could probably only measure down to 10^-17 to 10^-18  g.

It is, however, possible to combine this method with DC Cancellation in a technique which will be called AC Cancellation. This has the advantage providing a second measurement in addition to a result based on the separation of the two proof masses as a function of time.

DC Cancellation changes the orbit dynamics of a Drag-Free Satellite to eliminate the degeneracy between a semimajor-axis error and an EP violation and to cause a response which varies as t² not simply t. DC Cancellation is mechanized by placing two large fixed masses on either side of the proof masses to change the gravity gradient felt by the proof masses.

AC Cancellation achieves the same value of the gravity gradient as DC Cancellation by using a spinning toroidal mass, and in addition the spinning toroidal mass provides the AC gravity-gradient field necessary to measure the separation of the two proof masses. AC Cancellation is also discussed in the MG8 AC Paper (PDF).