On 22 Lug, 15:12, "Frank" wrote:
I never was an expert on Smith Chart calculations, so I would adopt a
somewhat less precise (and definitely more suck-it-and-see) approach.

(1) Make 'A' a quarterwave at the centre of the FM band. This can be done
by temporarily connecting it as a simple shunt stub, and snipping for a
notch centred at around 98MHz.

(2) Make up stub 'B1' by temporarily connecting it as a simple shunt stub,
and snipping for a notch centred at around 93MHz.

Now, at 70MHz, the frequency response will be rolling off into the 93MHz
notch, and the RLR will be getting rapidly worse. It will be a capacitive
mismatch, so it can be corrected by adding shunt inductance in parallel
with the stub. This can be an actual inductor, or a parallel stub with its
end short circuited.

[The advantage of a stub is that you don't need additional screening. It
can be tuned by pushing a shorting pin through the coax. The short can
later be made permanent.]

So,

(3) Tune the shunt inductor/stub for best match and lowest through loss at
70MHz.

(4) Make up stub 'B2' by repeating (2) and (3), but for (say) 103MHz.

(5) Connect up the complete filter, with the 'matched' stubs separated by
the quarterwave 'A'.

You should now have a filter with minimal loss and a good match at 70MHz,
but with two deep notches at 93 and 103MHz. If you are happy with the
results, finalize any temporary short circuits etc. If you are not happy,
you can still make a few tweaks to the tuning. If all else fails, it costs
virtually nothing to replace a bit of coax which is too short.
--
Ian

Interesting Ian, but of course you cheated by adding two extra components.
Just the same, it works exactly as you describe.

In my model I used inductors with a Q of 50. Step 3, above, works out
to 47.6 nH, and for stub "b" the shunt inductor is 62.5 nH. S21 from
88 to 108 MHz is -20 db. S11 from 65 to 75 MHz is - 20 dB.
The insertion loss is a nominal 0.4 db at 70 MHz. I used open ended stubs,
not that it makes any difference to the end result. The series stub is not
very
critical, but response symmetry (S21 65 - 75, and S21 rejection 88 - 108)
appears better. Note b1 = 94.5 degrees, and b2 85.7 degrees for my
open ended stub model. Also the network is perfectly symmetrical; i.e.
S11(f) = S22(f),
and S21(f) = S12(f).

If anybody is interested I can provide JPEG copies of the response,
and schematic, including a Smith chart plot.


I would be interested of course, I tried to come up with an "all coax"
filter and with minimal parts because the thing must live under the
antenna and pass all the power on TX which can be a few hundred watts
one day, that's why I was using 1/2 inch cellflex (have also 7/8 inch
but no connectors for it).
I still don't get why a perfectly reasonable filter on the smith chart
turns out to be trash when realized, I'd really like to learn
something.

73
Francesco IZ8DWF