The separation characteristics of the PM2,5 aerosol size selectors used in
speciation samplers developed for the U.S. EPA National PM2.5 Chemical Spec
iation Trends Network were evaluated under clean renditions. Measurement of
particle penetration versus aerodynamic diameter was conducted using an AP
S 3320 in conjunction with a polydisperse test dust, The resulting penetrat
ion curves were integrated with assumed ambient particle size distributions
(40 CFR Part 53, Subpart F) to obtain an estimate of measured mass concent
ration and to predict bias relative to the PM2.5 reference separator. The c
utpoint of two sharp cut cyclones, from the family of cyclones developed by
Kenny and Gussman (1997), compares favorably with the WINS, although posse
ssing a slight tail that extends into the coarse particle mode. A second cy
clone used by the Andersen Corp., AN 3,68 demonstrated the sharpest cut cha
racteristics of the devices tested; however, it possesses a D-50 cut size o
f 2.7 mum at its design flow rate. The separation characteristics of the Sp
iral separator were observed to be the shallowest and were >2.5 mum when gr
eased or ungreased. Estimated mass concentration bias relative to the FRM w
as within +/-5% for the idealized fine and typical assumed ambient distribu
tions for all separators. For the idealized coarse distribution, estimated
bias ranged between +4% and +8%. A comparison of these results with actual
field measurements made by the samplers that employ these separators demons
trates that a laboratory evaluation under clean conditions, while useful fo
r design purposes, is not sufficient to predict separator behavior in the "
real world".