Over a scale of 10 to 100 km, rainfall is almost never uniformly distributed. Understanding the characteristics of the horizontal variability has many implications to, for example, hydrology and satellite remote sensing. For satellite passive microwave measurements of precipitation, it is important to understand this variability to undertake the corrective measures for so called beam-filling problem. The main thrust of this paper is to understand the variability of the precipitation at typical passive microwave radiometric pixel scale (~25 – 50 km). The two important attributes of the variability – the fractional rain cover and the probability distribution of the precipitation within the pixel-scale are studied. The study uses radar data of TRMM validation missions from Gunnpt-Darwin, Guam, Houston, Kwajalein, and Melbourne. The study also uses data from the two other radars, namely, TOGA-COARE and AMeDAS, Japan. In case of TRMM validation radars, the convective and stratiform cases were separated and studies separately. Out of these 7 radars, 4 existed in tropical region and the other three in sub-tropical areas. The study reveals a double lognormal distribution of precipitation in the small scales (~25 - 50 km) in the tropical and sub-tropical region. A comparative study of the distribution from different radars is carried out, and discussed in terms of their geolocations and precipitation forms.