With the benefit of being able to adjust to varying acoustic conditions in real-time, adaptive approaches to active noise control (ANC) show promise to attain optimal noise attenuation. A reliable estimate of the secondary path is necessary for adapting the control filter. An estimate can be obtained by using an auxiliary noise signal played back via the loudspeaker. Depending on its power, temporal and spectral form, the auxiliary noise might be perceived as unpleasant. This reduces the comfort of use. In this contribution, we introduce pre-processing to the auxiliary noise signal, such that it mostly remains under psycho-acoustic masking thresholds. It is therefore less audible and can be of higher power without creating unpleasantness. This increases the overall ANC performance, due to an improved signal-to-noise ratio for the secondary path estimation. We evaluate the ANC performance using measurements of acoustic paths and realistic disturbance signals.