Turbulent flows laden with inertial particles present multiple open questions and are a subject of great interest in current research. Due to their higher density compared to the carrier fluid, inertial particles tend to form high concentration regions, i.e. clusters, and low concentration regions, i.e. voids, due to the interaction with the turbulence. In this work, we present an experimental investigation of the clustering phenomenon of heavy sub-Kolmogorov particles in homogeneous isotropic turbulent flows. Three control parameters have been varied over significant ranges: $Re_{lambda} in [170 - 450]$, $Stin [0.1 - 5]$ and volume fraction $phi_vin [2times 10^{-6} - 2times 10^{-5}]$. The scaling of clustering characteristics, such as the distribution of Voronoi areas and the dimensions of cluster and void regions, with the three parameters are discussed. In particular, for the polydispersed size distributions considered here, clustering is found to be enhanced strongly (quasi-linearly) by $Re_{lambda}$ and noticeably (with a square-root dependency) with $phi_v$, while the cluster and void sizes, scaled with the Kolmogorov lengthscale $eta$, are driven primarily by $Re_{lambda}$. Cluster length $sqrt{langle A_c rangle}$ scales up to $approx 100 {eta}$, measured at the highest $Re_{lambda}$, while void length $sqrt{langle A_v rangle}$ scaled also with $eta$ is typically two times larger ($approx 200 {eta}$). The lack of sensitivity of the above characteristics to the Stokes number lends support to the sweep-stick particle accumulation scenario. The non-negligible influence of the volume fraction, however, is not considered by that model and can be connected with collective effects.