The NIST watt balance experiment is being completely rebuilt after its 1998
determination of the Planck constant. That measurement yielded a result wi
th an approximately 1 x 10(-7) relative standard uncertainty. Because the g
oal of the new incarnation of the experiment is a ten-fold decrease in unce
rtainty, it has been necessary to reexamine many sources of systematic erro
r. Hysteresis effects account for a substantial portion of the projected un
certainty budget. They arise from mechanical, magnetic, and thermal sources
. The new experiment incorporates several improvements in the apparatus to
address these issues, including stiffer components for transferring the mas
s standard on and off the balance, better servo control of the balance, bet
ter pivot materials, and the incorporation of erasing techniques into the m
ass transfer servo system. We have carried out a series of tests of hystere
sis sources on a separate system, and apply their results to the watt appar
atus. The studies presented here suggest that our improvements can be expec
ted to reduce hysteresis signals by at least a factor of 10-perhaps as much
as a factor of 50-over the 1998 experiment.