I've been reading Wes Hayward's articles on coupled resonator filters, and
I've simulated some of the results and they're what I'd expect. However,
his examples are all bandpass filters... and I need a bandstop filter.
Chanigng the parallel LC shunt resonators to series LC series resonators
doesn't really work at all... although there's a perfect notch at the
desired center frequency, the passband response is very low (tens of dB
insertion loss) and recovers very, very slowly. It seemed as though I just
needed a different size coupling capacitor, but playing around some doesn't
improve the response significantly (for instance, if you take the example on
page 85 of Introduction to Radio Frequency Design and change the parallel
resonators to series resonators, even if you change the 7.2pF top coupling
capacitor to something astronomical such as 1nF the response is still poor).

I've ordered a copy of Zverev's book (and Zverev w/Mathei), where the theory
of all this comes from, but in the interim... can anyone suggest what the
fundamental problem might be? Interesting, from looking at a few pages of
Zverev on Amazon.Com he does have schematics of what I think I'm after (top
capacitor coupled series resonators to create bandstop filters).

Thanks,
---Joel Kolstad

It's been a while since I've done this. But one way to realize a band
stop filter is to begin with a lowpass prototype of the desired response
(Butterworth, Chebyshev, etc.). Then add a capacitance in parallel with
each inductance of the lowpass, of value 1/(w0^2 * L), and an inductance
in series with each capacitance of the network, of value 1/(w0^2 * C).
Beginning with a common series L - shunt C lowpass, this results in
series-connected parallel resonant circuits with shunt series resonant
circuits in between. Intuitively, the topology is correct in that you'll
have zeros due to both the series and shunt networks, and the response
will be unity at DC and infinite frequency. Another topology which would
work is alternating shunt series resonant circuits and series parallel
resonant circuits. The topology you describe can't possibly work since
the coupling capacitor prevents unity response at DC.

This methodology doesn't account for inductor loss, as Wes' analysis
does. You'll find the necessary information about this in Zverev, and
making use of it will give you a real appreciation for the programs now
available for the purpose.

Roy Lewallen, W7EL

Joel Kolstad wrote:

I've been reading Wes Hayward's articles on coupled resonator filters, and
I've simulated some of the results and they're what I'd expect. However,
his examples are all bandpass filters... and I need a bandstop filter.
Chanigng the parallel LC shunt resonators to series LC series resonators
doesn't really work at all... although there's a perfect notch at the
desired center frequency, the passband response is very low (tens of dB
insertion loss) and recovers very, very slowly. It seemed as though I just
needed a different size coupling capacitor, but playing around some doesn't
improve the response significantly (for instance, if you take the example on
page 85 of Introduction to Radio Frequency Design and change the parallel
resonators to series resonators, even if you change the 7.2pF top coupling
capacitor to something astronomical such as 1nF the response is still poor).

I've ordered a copy of Zverev's book (and Zverev w/Mathei), where the theory
of all this comes from, but in the interim... can anyone suggest what the
fundamental problem might be? Interesting, from looking at a few pages of
Zverev on Amazon.Com he does have schematics of what I think I'm after (top
capacitor coupled series resonators to create bandstop filters).

Thanks,
---Joel Kolstad

"Joel Kolstad" wrote in message
...
I've been reading Wes Hayward's articles on coupled resonator filters, and
I've simulated some of the results and they're what I'd expect. However,
his examples are all bandpass filters... and I need a bandstop filter.
Chanigng the parallel LC shunt resonators to series LC series resonators
doesn't really work at all... although there's a perfect notch at the
desired center frequency, the passband response is very low (tens of dB
insertion loss) and recovers very, very slowly. It seemed as though I
just needed a different size coupling capacitor, but playing around some
doesn't improve the response significantly (for instance, if you take the
example on page 85 of Introduction to Radio Frequency Design and change
the parallel resonators to series resonators, even if you change the 7.2pF
top coupling capacitor to something astronomical such as 1nF the response
is still poor).

I've ordered a copy of Zverev's book (and Zverev w/Mathei), where the
theory of all this comes from, but in the interim... can anyone suggest
what the fundamental problem might be? Interesting, from looking at a few
pages of Zverev on Amazon.Com he does have schematics of what I think I'm
after (top capacitor coupled series resonators to create bandstop
filters).

The Elsie filter synthesis program has a bandstop filter design option. I
just tried it and it works OK. None of the designs are top-coupled
resonators, though.

Leon
--
Leon Heller, G1HSM
http://www.geocities.com/leon_heller

Correction:

Roy Lewallen wrote:
. . .Another topology which would
work is alternating shunt series resonant circuits and series parallel
resonant circuits. . .

That's the same topology I described just above.

Roy Lewallen, W7EL

"Joel Kolstad" wrote in message
...
I've been reading Wes Hayward's articles on coupled resonator filters, and
I've simulated some of the results and they're what I'd expect. However,
his examples are all bandpass filters... and I need a bandstop filter.
Chanigng the parallel LC shunt resonators to series LC series resonators
doesn't really work at all... although there's a perfect notch at the
desired center frequency, the passband response is very low (tens of dB
insertion loss) and recovers very, very slowly. It seemed as though I
just
needed a different size coupling capacitor, but playing around some
doesn't
improve the response significantly (for instance, if you take the example
on
page 85 of Introduction to Radio Frequency Design and change the parallel
resonators to series resonators, even if you change the 7.2pF top coupling
capacitor to something astronomical such as 1nF the response is still
poor).

I've ordered a copy of Zverev's book (and Zverev w/Mathei), where the
theory
of all this comes from, but in the interim... can anyone suggest what the
fundamental problem might be? Interesting, from looking at a few pages of
Zverev on Amazon.Com he does have schematics of what I think I'm after
(top
capacitor coupled series resonators to create bandstop filters).

Thanks,
---Joel Kolstad



The side arms are OK as series resonators but your middle C needs changing
(transforming) into a C+L parallel resonator.

The way to go is firstly design a prototype low pass filter. Then transform
it to its high pass equivalent. Then transform to the band stop design.

regards
john

"Leon Heller" wrote in message
...
The Elsie filter synthesis program has a bandstop filter design option. I
just tried it and it works OK. None of the designs are top-coupled
resonators, though.

Right... the reason I started down the 'coupled resonator' path was due to
the fact that -- for bandpass filters -- the component values at high
frequencies tend to be much more realizable than with the 'tradition' L-LC,
C-CL structure.

In any case, I've found a paper that addresses the 'unreasonably small
component value' problem -- it interleaves a low-pass filter with the
band-stop filter, and then applies some transformations to get back to a
'reasonable' looking topology. The extra degree of freedom provides a lot
of flexibility in obtaining 'big' component values (at the expense of having
more components involved, of course).

Thanks for the help. I have played with Elsie, and it's a pretty nice
program (even if it does insist on properly scaling its displays only when
its window is maximized! :-) ).

---Joel

"john jardine" wrote in message
...
The way to go is firstly design a prototype low pass filter. Then
transform
it to its high pass equivalent. Then transform to the band stop design.

This (alone) doesn't work if you want a 2% bandwidth filter at 500MHz built
using lumped L's and C's with finite Q's. Try it. :-)

---Joel

"Joel Kolstad" wrote in message
...
"john jardine" wrote in message
...
The way to go is firstly design a prototype low pass filter. Then
transform
it to its high pass equivalent. Then transform to the band stop design.

This (alone) doesn't work if you want a 2% bandwidth filter at 500MHz
built
using lumped L's and C's with finite Q's. Try it. :-)

---Joel


Most certainly not up at that frequency. Just too many parasitics. :-).
(Even 150MHz is pushing it).
regards
john