Since Federal Communications commission (FCC) of the USA released the microwave frequencies from 3.1 GHz to 10.6 GHz for commercial radio systems in February, 2002, the ultra-wideband (UWB) radio has been paid attention as one of the future high speed wireless communication system. A UWB bandpass filter is one of the key components for the UWB system to control the spectrum of the UWB signals. In the development of an UWB bandpass filter, the main difficulties one may face come from the ultra-wideband (about 110 percent fractional bandwidth as FCC’s mask) and the hard requirements for out-band performance. In this thesis, a compact ultra-wideband (UWB) bandpass filter composed of two parallel coupled stepped impedance resonators (SIRs) is proposed. The SIR resonators are tapped with open stubs to create two transmission zeros at both edges of the rejection bands providing a very sharp transition band. The genetic algorithm is used to determine the physical dimensions of the filter. Using Ansoft HFSS EM simulator, the fitness values are calculated during the algorithm. The interaction between MATLAB and Ansof HFSS is done through functions that create visual basic script. The optimized bandpass filter shows a quasi-elliptic function response and a fractional bandwidth of about 110% (from 3.1GHz to 10.6GHz). Simulated results of the filter indicate low insertion loss and a flat group delay within the paand and a very sharp attenuation at the stopband edges (better than 120dB/GHz). Finally the resultant UWB bandpass filter is fabricated and measured to provide an experimental demonstration of its actual UWB paand and skirt performances. Good agreements between the measurement and simulation results are achieved and demonstrated. Keywords: bandpass filter, Ultra-wideband, resonator, insertion loss, microstrip.