Robust control design hinges on the availability of an accurate uncertainty model. The aim of this paper is to develop an approach for accurate uncertainty modeling. The proposed method is based on H∞-norm estimation, or peak amplitude. A new approach is developed that explicitly takes into account inter-grid frequency behavior while only requiring a reduced experiment time, modeling effort, and limited user intervention. In particular, the proposed method relies on local rational models. Experimental results on an active vibration isolation system confirm that the approach is able to handle lightly-damped systems with significantly less data compared to spectral estimation and local polynomial estimation techniques.