In this study, we conducted WRF simulations in two summers in 2006 and 2007 at three spatial resolutions (36 km, 12 km and 4km), and found the 4 km simulations overall resulted in best performance in terms of the spatial distribution and diurnal cycle of precipitation in US. Two cumulus schemes, Kain-Fritsch (K-F) and the scale aware scheme Grell-Freitas (G-F), were compared, and simulations with K-F scheme in general over predicted the precipitation in the southeastern U.S., New Mexico and Colorado, and this overestimation is largely reduced when G-F scheme was applied. At a spatial resolution of 4 km, G-F scheme is comparable to that without cumulus scheme. In addition, this study comprehensively evaluated the diurnal cycle at the three spatial resolutions (36 km, 12 km and 4km) with two cumulus schemes (K-F and G-F). Compared to observations, 4 km convection permitting simulations reproduced well the diurnal cycle over both Great Plain and North American Monsoon region and captured the eastward propogation of precipitation. With spatial resolutions of 36 km and 12 km, the G-F scheme captures the diurnal cycle of precipitation reasonably well whereas the K-F schemes produces diurnal peak a few hours earlier. Further analysis shows that the ratio of large scale to total precipitation is not sensitive to resolution for K-F scheme, but it increases substantially from 36km to 12km and 4km with the scale aware G-F scheme. This study highlights the importance of the application of scale aware scheme and convection permitting scale in simulating the diurnal cycle of precipitation.