Abstract: Soil gas diffusion coefficient is an important parameter for describing soil gas diffusion process. Its value varies with soil texture, water content, bulk density, and thus is difficult to predict. In this study, we designed a testing apparatus of soil gas diffusion coefficient based Currie's one chamber method and investigated its relationship with air filled porosity. The apparatus was composed of 2 components: a polyvinyl chloride soil chamber on the top with cutting ring and O ring, and gas chamber below. The soil chamber was connected with air and gas chamber on both sides. The gas chamber was made of transparent resin as an observation window. A stainless steel plate was located between soil chamber and gas chamber for gas diffusion control. In addition, KE-25 O2 sensor from Figaro Inc., Japan was used to transmit voltage to computer screen. From prepared calibration curve of gas concentration and voltage, the gas concentration and the gas diffusion coefficient could be calculated. A preliminary experiment showed that the measurement error caused by gas leakage was 0.000006, smaller than measurement results, indicating that the gas leakage could be ignored and the measurements were relatively accurate. In the laboratory experiment, quartz sand of 3 particle sizes were prepared to investigate the changes of relative gas diffusion coefficient with different total porosity and undisturbed and disturbed sandy loams were used to study the changes of relative gas diffusion coefficient with air-filled porosities. Meanwhile, the relative gas diffusion coefficient estimation based on measurements was compared with 3 models (Buckingham method, Millington and Quirk method, structure-dependent water-induced linear reduction model SWLR model). The results showed that relative diffusion coefficient of quartz sand increased with increasing total porosity regardless of particle sizes and also increased with particle sizes regardless of total porosity. The relationship between relative diffusion coefficient and air-filled porosity followed power function. For the sandy loam, disturbed and undisturbed soils yielded similar results with values smaller than 0.006. Among the three models, the relative diffusion coefficients estimated only from the Buckingham method was closer to that from measurement-based results with bias of -0.219×10-3 and root mean square error of 0.799×10-3, indicating caution should be paid when estimators from empirical models were used and the value couldn't replace the measurements. The study would provide valuble information for relative gas diffusion coefficient measurements and calculation.