Post subject: filter design Posted: Mon Mar 05, 2007 2:52
Joined: Thu Mar 01, 2007 3:45 pm
When designing an interdigital
or a combline bandpass filter in stripline, I frequently have a
S21 dip, about 0.5 to 0.8dB in magnitude, near either low side or
high side of the passband. For example, if the passband is 900MHz
to 1.3GHz, this dip may occur at around 990MHz or 1.22GHz. There
is a corresponding S11 spike at the same frequencies. What causes
this? Can anyone enlighten me on this? Thanks.
Post subject: Posted: Tue Mar 06, 2007 6:28
Joined: Fri Feb 17, 2006 12:07 pm
Location: London UK
This will always
happen when using an ellyptic function or a Chebychev design, and
it is due to internal resonant paths within the structure.
Constant Phase types avoid this, but have a slower roll-off for
the same number of elements.
The attenuation variation (S21)
across the band for a Chebychev should never exceed the design ripple
factor (eg. 0.25dB) by any more than the small margin for dissipation
due to finite Q values of the elements. If you are looking for maximum
stop band roll-off rate, choose ripple factor at say 1.0dB, and
that will be the variation limit across the band.
that the -3dB bandwidth and the equal ripple bandwidth are different
from each other, which is warning you that between the ripple factor
frequency and the -3dB frequency the S21 (and S11) will vary wildly.
There is a formula relating them, but I will have to look for it.