An equation error algorithm is known to give a significant bias in case the
output signal is contaminated by white noise, A simple bias removal scheme
was proposed in a recent letter. Here, we provide some complementary analy
sis of that algorithm. It is shown how it can be interpreted as an instrume
ntal variable method. Further, by analytical examples, it is demonstrated t
hat the method does not always lead to convergent estimates, even for arbit
rarily small step sizes.