Measurements and production at ever decreasing length scales give rise to increasing demands on active vibration isolation. Where conventional passive vibration isolation can typically isolate frequencies above 100 Hz sufficiently, high-end applications like gravitational-wave detection or semiconductor manufacturing require active vibration mitigation schemes. This drives the development of inertial sensors with good low-frequency (< 100 Hz) sensitivity. In these sensors, conventional passive readout is replaced by an active (force-feedback) topology, with interferometric readout. Although it is generally well established that an active readout topology is capable of achieving lower self-noise levels, no literature was found on a quantified sensitivity comparison of active and passive readout. Moreover, no literature was found quantifying the influence of control performance on the inertial sensor output signal robustness. This thesis aims to clarify what the influence of control performance is on the sensitivity and readout robustness of an active topology interferometrically read out inertial sensor. A comparison is made between active and passive readout, as well as multiple control stiffness levels in active readout. To do so, an experimental setup is designed and constructed. It is capable of digitally switching between passive and active readout, without the need for hardware changes. The digital implementation of the controller allows to measure the sensitivity for different control bandwidth in active readout. An incremental reduction in 1/f sloped readout noise spectrum is observed for incrementally increased control bandwidth. Compared to the highest noise spectrum associated with passive readout, a 100 fold noise reduction is measured for the 60 - 1000 Hz frequency band by applying the maximum performance controller in active readout. At 150 Hz this is a 1000 fold reduction in noise. Further improvements of the 1/f sloped readout noise are not measurable, as the spectrum is dominated by physical disturbances, ground motion and flat spectrum readout noise. In passive readout, the inertial sensor output signal relies on accurate estimation of the stiffness, damping and mass parameter of the mechanical design. Active readout is not dependent on the stiffness and damping parameter, but within the control bandwidth does require characterization of higher order parasitic modes of the mechanism. These modes are susceptible to frequency drift and can cause a change in the readout magnitude. Above the control bandwidth the dependency is reduced, similar to passive readout. The conflicting requirement on the desired control bandwidth, for maximum sensitivity and readout robustness, requires a trade-off to be made in controller design. The incremental reduction of the readout noise spectrum provides a quantitative relation between control performance and readout noise. It confirms a position and velocity dependent magnitude and demonstrates that benefits of reduced readout noise in the active topology already exist for the lowest control performance. These results strengthen our understanding of readout noise and output signal robustness in active topology interferometrically read out inertial sensors.