In this paper, we analyze and model the impact of floating dummy fill on the signal capacitance considering various parameters including signal dimensions, dummy shape and dimensions. Intra-layer dummy has its greatest impact on coupling capacitance while inter-layer dummy has larger impact on the ground capacitance component. Based on this analysis, we propose simple capacitance models (Cc for intra-layer dummy and Cg for inter-layer dummy). To consider realistic cases with both signals and metal fill in adjacent layers, we apply a weighting function approach to the Cg model. We verify this model using benchmark circuits and find that total net capacitance with floating fill can be extracted within 1% of field solver results on average with total extraction runtime reductions of up to 40%. When evaluating the incremental capacitance due to fill alone, average error of the models range from 2%-15% across benchmarks and fill-related runtime overhead is reduced by 60%-88%.