Gravitational classification chambers (GCCs) were generally never used to classify fine powder because of its poor accuracy. However, in this study, a high quality closed-loop wind tunnel was designed and manufactured to improve this GCC drawback. A high quality closed-loop wind tunnel was designed to provide a low wall interference and low turbulence level in the test section. A wide-span feeder fed the powder smoothly as a curtain with a uniform aerodynamic interaction by the cross wind. The classification performance of GCCs has therefore been improved, and it could now classify lead-tin (Pb-Sn) powder with particle sizes smaller than 200 pm by controlling feeding rate and cross flow speed. Our experimental results show that GCCs have good classification performance from the view of the index of classification size ratio (CSR) by controlling cross flow speed and feeding rate. For GCCs the best sharpness index is about 0.4, the CSR is 1.09, and the cut size is 15.8 mu m. The results demonstrate that GCCs had a fine powder classification capability and its cut-point was accurate enough for use in a classification operation. Furthermore, GCCs could be used to reduce ultra fine powder from raw powder. The volume percentage under 10 mu m of raw material decreased from 7.61% to 0.64% through three repeatable classifications. This function could be considered as a good pre-process for specific classifiers (sieves or centrifugal classifiers). The product efficiency of this specific classifier could be increased by coupling it with GCCs. (C) 2009 Published by Elsevier B.V. on behalf of The Society of Powder Technology Japan.