During the long shut down 3 from 2025 to 2027, the LHC will undergo the high luminosity upgrade, where the luminosity will be increased from its current value of 2 × 1034 cm−2 s−1 to about 7.5 × 1034 cm−2 s−1. To maintain high tracking efficiency under such extreme operating conditions, pixels sensors of small pitch will be used in the inner tracker. However, as the pixel pitch is reduced, charge sharing effects like diffusion start to play an important role in determining the resolution of these sensors. In this thesis, non-irradiated pixel sensors of geometries 50 × 50 μm2, 25 × 100 μm2 and 17 × 150 μm2 are tested using an electron beam of energy 5.2 GeV at the DESY test beam facility. These are n+n sensors with a thickness of 285 μm and are bump bonded to low noise read-out chip ROC4SENS. These sensors are much thicker than the currently used pixel detectors at the CMS experiment and are ideal for studying diffusion. A dedicated configuration of three sensors called the Dreimaster was used to perform the measurements. From these measurements, the spatial resolution, cluster size, and efficiency were ex- tracted and studied as a function of bias voltage and incidence angle of the beam. The best resolution measured for the 50 × 50 μm2, 25 × 100 μm2 and 17 × 150 μm2 at a bias voltage of 300 V were 3.09 ± 0.03 μm, 2.16 ± 0.03 μm and 1.42 ± 0.02 μm respectively. The effects of diffusion were studied by taking measurements at various values of bias voltage for the 25 × 100 μm2 at the vertical incidence and optimal angle. The effect of diffusion on resolution depends on the beam incidence angle. At vertical incidence, the resolution was observed to improve with decreasing bias voltage, and at the optimal angle, the resolution was observed to deteriorate with the decrease in bias voltage. The test beam measurements were used to validate the simulation of pixel sensors. The pixel sensors were simulated using PIXELAV and Synopsis TCAD. Qualitatively a good agreement between the measured and simulated data was observed. Quantitatively, significant deviations between the measured and simulated data were observed. However, a good agreement between the simulated and measured cluster size as a function of bias voltage and incidence angle was observed.