500MHz wideband filter design
 

I was curious if anybody had ideas as to how to deal with the fractional pF
capacitors that come out of 'first pass' filter design programs such as
Elsie for filters up in the 70cm (and slightly shorter) range (50 ohm
terminations on both ends, LC style)? I'm attempting to build a bandstop
filter up there (20dB attenuation, 10MHz bandwidth), and I'm aware of the
usage of microstrip traces as equivalent capacitors (and inductors), but the
filter needs to be (approximately) flat all the way back down to the ~30MHz
and using, e.g., lamba/8 microstrip sections would create repetition in the
filter's response every quadrupling of the center frequency.

So... other than attempting to be clever and transforming the filter such
that capacitance is increased... or scaling the system impedance (to... 20
ohms? Then I'm still stuck with 2-3pF values)... are there any good methods
of implementing such a wideband filter design at these frequencies?
Pointers to books, magazine articles, etc. would be appreciated.

Thanks,
---Joel Kolstad

"Joel Kolstad" wrote in message
...
I was curious if anybody had ideas as to how to deal with the fractional
pF
capacitors that come out of 'first pass' filter design programs such as
Elsie for filters up in the 70cm (and slightly shorter) range (50 ohm
terminations on both ends, LC style)? I'm attempting to build a bandstop
filter up there (20dB attenuation, 10MHz bandwidth), and I'm aware of the
usage of microstrip traces as equivalent capacitors (and inductors), but
the
filter needs to be (approximately) flat all the way back down to the
~30MHz
and using, e.g., lamba/8 microstrip sections would create repetition in
the
filter's response every quadrupling of the center frequency.

So... other than attempting to be clever and transforming the filter such
that capacitance is increased... or scaling the system impedance (to... 20
ohms? Then I'm still stuck with 2-3pF values)... are there any good
methods
of implementing such a wideband filter design at these frequencies?
Pointers to books, magazine articles, etc. would be appreciated.

Thanks,
---Joel Kolstad

Hi Joel,
Perhaps you can give us some additional number? What sort of attenuation
level do you require? Where are the -3dB corners? What is the tolerbale
passband return loss?
You mention the filter needing to be good down to 30MHz, but no mention of
characteristics above 70cM. A 10MHz wide bandstop filter with decent
attenuation at 70cM is going to be complex. Would perhaps an elliptic LPF
also work? Do you need the full 10 MHz of stop frequency?
I use the Genesys simulator/synthesizer package which is highly flexible,
allowing real time filter tuning- but ALWAYS try to avoid a wide spread in
component values and very small capacities.

73,

Dale W4OP

What order filter (how many capacitors)? What sort of power level?
Guess I'd look at trimming the capacitances. I have a bunch of
0.5-5pF trimmer caps that I might use for such a filter. Or, I'd get
a big pile of 1pF, 1.2pF, 1.5pF, ...etc and select caps that measured
correctly. Since 10MHz really isn't a very big percentage bandwidth,
I'd probably go for trimmer caps, and tune the whole filter using a
network analyzer to monitor its response. Does your filter design
program give you the resonant frequencies of each resonator? -- Does
your design not have some really small inductances, too??

Cheers,
Tom


"Joel Kolstad" wrote in message ...
I was curious if anybody had ideas as to how to deal with the fractional pF
capacitors that come out of 'first pass' filter design programs such as
Elsie for filters up in the 70cm (and slightly shorter) range (50 ohm
terminations on both ends, LC style)? I'm attempting to build a bandstop
filter up there (20dB attenuation, 10MHz bandwidth), and I'm aware of the
usage of microstrip traces as equivalent capacitors (and inductors), but the
filter needs to be (approximately) flat all the way back down to the ~30MHz
and using, e.g., lamba/8 microstrip sections would create repetition in the
filter's response every quadrupling of the center frequency.

So... other than attempting to be clever and transforming the filter such
that capacitance is increased... or scaling the system impedance (to... 20
ohms? Then I'm still stuck with 2-3pF values)... are there any good methods
of implementing such a wideband filter design at these frequencies?
Pointers to books, magazine articles, etc. would be appreciated.

Thanks,
---Joel Kolstad

Hi Tom,

"Tom Bruhns" wrote in message
m...
What order filter (how many capacitors)?

Looks like I can get away with a third order; just a pi or tee.

What sort of power level?

1W.

Does your filter design
program give you the resonant frequencies of each resonator?

Yes, or at least I can calculate it readily enough.

Does
your design not have some really small inductances, too??

They're in the 'hundreds of nanohenries' range. With them I've been more
worried about obtaining decent Q's than the actual values, in that
parasitics would seem to be much more comparable to the size of the caps
than the inductors.

Thanks for your advice,
---Joel

Another issue you'll have to deal with if you don't do some scaling of
some sort is that the inductors that go with the fractional-pF
capacitors will have significant parasitic capacitance...you're likely
going to get self-resonant frequencies not far above the filter
stopband frequency (or worse). It would help a lot if you could make
the filter stopband considerably wider. Seems like with a 10MHz
stopband, you're asking for filter performance not too unlike a
duplexer.

You mentioned the need for the filter to be flat down to low
frequencies, but what about higher frequencies? Are the repetitions
which are common in distributed (transmission-line) filters really a
problem? Or do you need to pass all the higher frequencies too?

Cheers,
Tom


"Joel Kolstad" wrote in message ...
I was curious if anybody had ideas as to how to deal with the fractional pF
capacitors that come out of 'first pass' filter design programs such as
Elsie for filters up in the 70cm (and slightly shorter) range (50 ohm
terminations on both ends, LC style)? I'm attempting to build a bandstop
filter up there (20dB attenuation, 10MHz bandwidth), and I'm aware of the
usage of microstrip traces as equivalent capacitors (and inductors), but the
filter needs to be (approximately) flat all the way back down to the ~30MHz
and using, e.g., lamba/8 microstrip sections would create repetition in the
filter's response every quadrupling of the center frequency.

So... other than attempting to be clever and transforming the filter such
that capacitance is increased... or scaling the system impedance (to... 20
ohms? Then I'm still stuck with 2-3pF values)... are there any good methods
of implementing such a wideband filter design at these frequencies?
Pointers to books, magazine articles, etc. would be appreciated.

Thanks,
---Joel Kolstad

"Tom Bruhns" wrote in message
m...
Another issue you'll have to deal with if you don't do some scaling of
some sort is that the inductors that go with the fractional-pF
capacitors will have significant parasitic capacitance...you're likely
going to get self-resonant frequencies not far above the filter
stopband frequency (or worse).

Yes, I can definitely appreciate that.

It would help a lot if you could make
the filter stopband considerably wider. Seems like with a 10MHz
stopband, you're asking for filter performance not too unlike a
duplexer.

Well, it's supposed to be an adjustable notch filter, from 30-500MHz, with a
2% stopband width... and it's at the high end where it's looking pretty
challenging.

You mentioned the need for the filter to be flat down to low
frequencies, but what about higher frequencies? Are the repetitions
which are common in distributed (transmission-line) filters really a
problem?

At the higher frequencies, no. At lower frequency, yes! This might suggest
breaking it up into two filter, since I could probably get 300-500MHz out of
distributed filter. The problem there is how to set the center frequency...
although simply using PIN diodes or similar to ground the transmission line
somewhere along its length sounds promising.

---Joel