Use of narrower frequency bands: the possibility of concentrating the analysis
in narrower frequency bands provides important advantages:
Perfect fit more easily obtained. Better sensitivity.
The effects of the distributed delays introduced by the transmission
lines can be correctly approximated by a transfer function of reasonable
order.
Frequency responses that are very complex with too many resonances can
be cut in sub bands that are analyzed separately.
Once the critical band is known, there can be a significant gain in
simulation time when performing parametric analysis.
it is strongly recommended (very strongly indeed) focusing on critical bands:
To re-confirm that unstable poles are obtained at a particular
frequency, always repeat the identification centering the analysis on
the frequency of the unstable poles.
Anytime you doubt about your results (as when detecting whether a
right-half plane quasi-cancellation is physical or just appears
mathematically because of an excessively high order) center the analysis
on the critical bands.
Repeat the identification increasing the phase tolerance. A value of the phase
tolerance too strict (small) may give raise to numerical
quasi-cancellations.
Repeat the identification changing the identification order manually.
From the SISO method, the order used by
the Automatic algorithm can be used as a starting point. If reducing the Order
of the identification, there is a good fitting in magnitude and phase and the
quasi-cancellation/low sensitivity poles have disappeared, they were numerical
poles.
Perform a parametric analysis. If varying a circuit parameter, the
quasi-cancellation/low sensitivity poles move to the left-hand side (LHS) of the
plane, they are physical poles.
