When discussing error estimates of the point-source mechanism and the
source time function obtained by the two-step procedure by Sileny, Pan
za & Campus (1992), the authors insist that in the first step-inversio
n of seismograms (after Sipkin 1982) to get the moment tensor rate fun
ctions (MTRFs)-a homogeneous variance for all the data is needed to ke
ep the advantageous symmetry of the normal equations. We show that thi
s is too strong a requirement and can be dropped.