An improved Short-Open-Load-Thru (SOLT) on-wafer vector network calibration method for broad-band accuracy is proposed. Accurate measurement of on-wafer devices over a wide range of frequency, from DC to high frequencies with a minimum number of space conservative standards has always been desirable. Therefore, the work is aimed at improving the existing calibration methods and suggesting a best "practice" strategy that could be adopted to obtain greater accuracy with a simplified procedure and calibration set. Quantitative and qualitative comparisons are made to the existing calibration techniques. The advantages and drawbacks of each calibration are analyzed. Prior work done at the University of South Florida by an improved SOLT calibration is briefed. The presented work is a culmination and refinement of the prior USF work that suggested that SOLT calibration improves with more complex definitions for the calibration standards. Modeling of the load and thru standards is shown to improve accuracy as the frequency variation of the two standards can be significant. The load is modeled with modified equivalent circuit to include the high frequency parasitics. The model is physically verified on different substrates. The relation of load impedance with DC resistance is verified and its significance in SOLT calibrations is illustrated. The thru equation accounts for the losses in a transmission line reflections and phase shift including dielectric and conductor losses. The equations used are important for cases where a non-zero length of thru is assumed for the calibration. The complex definitions of the calibration standards are included in the calibration algorithm with LabView and tested on two different VNA's -- Wiltron 360B and Anritsu Lightning. The importance of including the forward and reverse switch terms error correction in the algorithm is analyzed and measurements that verify the improvement are shown. The concept using same foot size calibration standards to simplify the calibration process is highlighted with results to verify the same. The proposed technique thus provides for calibration strategy that can overcome the low frequency problems of TRL, retain TRL accuracy at high frequencies while enabling the use of a compact common footprint calibration set.