Reduced-scale refrigeration compressors supported on gas-lubricated bearings have been identified as a key technology for domestic heat pump applications, in order to improve both the system efficiency and the reliability. Unfortunately, reduced-scale machines suffer from increased aerodynamic losses compared to large-scale industrial machines, which are caused by the small feature size and manufacturing tolerances. Due to the small feature size, the machine Reynolds number is low and higher frictional losses occur. The manufacturing tolerances as well as form and positioning tolerances for assembly require larger relative clearances in reduced-scale machines compared to large-scale ones, leading to increased losses caused by tip leakage. Centrifugal compressor design and performance prediction strongly depends on empirical guidelines and correlations, which are deduced experimentally and numerically from test data of large-scale machines. Hence, the question arises whether these empirical design implications and correlations are applicable for reduced-scale compressors. This PhD-thesis presents results on the investigation of the impact of low Reynolds number flow and large relative tip clearances on the compressor performance and flow patterns. The investigation is performed both experimentally and numerically (CFD and one-dimensional compressor model) on two reduced-scale centrifugal compressor designs. An experimental test facility of an up-scaled refrigeration compressor has been realized and tests in terms of impact of clearance ratio alteration as well as performance improvements by impeller design have been performed. Large relative clearance ratios of up to 20 % have been investigated. Empirical correlations accounting for the impact of geometrical scaling as well as altering the tip clearance gap are assessed and improved. These correlations are tested and validated at design and off-design conditions. Design implications on how to design a centrifugal compressor with respect to large clearance ratios have been postulated. The hub and shroud end-wall distribution as well as the shroud blade angle distribution have been identified as main design parameters for mitigating the negative effect of large relative clearance ratios. Furthermore, it has been shown that compressor optimization in terms of tip leakage induced phenomena depend on tip leakage and even more on the guidance of main and splitter blade tip leakage vortices. A mid-loaded shroud blade loading distribution offers the best tradeoff between both. Superior efficiency of such a loading configuration has been verified both numerically and experimentally. The loading distribution is suggested to have a minor impact on compressor efficiency at low relative tip clearance ratios, however, at large relative clearance an efficiency difference of up to 2 % occurs between the different loading distributions.