Designing LC Ladders

[Rob Gaddi]

So I need to do some wideband low RF filtering (flat to 0.5 dB from
40-100 MHz, fairly sharp cut off on the high end, low end can cut off or
not as convenient), and have found that my lossless LC ladders are in
fact anything but. Who knew that several ohms of resistance in my 0603
inductors would add up, other of course than anyone who had stopped to
think about it.

I've managed to massage the filters back into shape through trial and
error in RFSim99, but in the process have beaten my nice little
Butterworth and Chebyshev filters into unrecognizable scattered heaps of
very sore poles, making irrelevant decades of filter theory. It strikes
me that there must have been an easier way of doing it.

So, for future reference, how should one go about doing wideband RF
filtering? I've got books here with plenty of suggestions if I felt
like blowing ten thousand dollars designing an analog ASIC, but I'd like
a solution with a few less zeros, not to mention not having to learn
analog ASIC design.

-- Rob

 

[Max]

Rob Gaddi wrote:

... So, for future reference, how should one go about doing
wideband RF filtering?

I think you were on the right track. Karl Willy Wagner dealt with some
of the same issues in his very classic paper "Kettenleiter und
Wellensiebe" in 1919 (_Arkiv für Elektroteknik_ was the journal that I
recall, though I should verify). There are also some good solid-state
monolithic filters today with some flexibility.

 

[Nico Coesel]

Rob Gaddi <rgaddi@bcm.YUMMYSPAMtmc.edu> wrote:

So I need to do some wideband low RF filtering (flat to 0.5 dB from
40-100 MHz, fairly sharp cut off on the high end, low end can cut off or
not as convenient), and have found that my lossless LC ladders are in
fact anything but. Who knew that several ohms of resistance in my 0603
inductors would add up, other of course than anyone who had stopped to
think about it.

Besides that, component variations will kill your filter as wel. I
recently designed a low pass filter in about the same frequency region
using an active sallen-key second order filter. To make it roll off
more sharply I added an LC series snubber.

The topic is more or less covered on page 3 of this document:
http://www.edn.com/contents/images/93004di.pdf

--
Reply to nico@nctdevpuntnl (punt=.)
Bedrijven en winkels vindt U op www.adresboekje.nl

 

[Max]

My previous reply on this will teach me not to rely on memory.
Wagner's 1919 paper is "Spulen und Kondensatorleitungen." _Arkiv für
Elektrotechnik,_ Band 8, Heft 2 u. 3 (July 1919). His paper
"Kettenleiter und Wellensiebe" appeared in _Elektrische
Nachrichten-Technik,_ Band 5, Heft 1 (January 1928).

 

[Joerg]

Hi Rob,

So I need to do some wideband low RF filtering (flat to 0.5 dB from
40-100 MHz, fairly sharp cut off on the high end, low end can cut off
or not as convenient), and have found that my lossless LC ladders are
in fact anything but. Who knew that several ohms of resistance in my
0603 inductors would add up, other of course than anyone who had
stopped to think about it.

It depends on what the words "fairly sharp" mean. If you have to be down
to -30dB or more at 110MHz it'll be quite a challenge. Then those tiny
0603 inductors won't cut it. I'd start with 'real' inductors.

Also, depending on band rejection outside your range you may have to
make compartments or at least spread it out a bit. Another question
would be whether this is a one-off or something for mass production and
what the cost and other constraints are.

One trick that may seem a bit obscure but has worked for me in the past
is to up-convert the whole chebang to above a GHz where you then can
design a regular filter, abandoning the lowpass-highpass combo and using
resonant filtering instead. Then just convert back down with the same
LO. Make sure to suppress the unwanted sideband. Who knows, maybe you
could even convert to some range where you can buy commercial filters
that happen to fit a 60MHz bandwidth. If you want to scope out this
route check Murata's 'Gigafil' series of dielectric filters. I believe
there was one that has 60MHz bandwidth. In case you go for it be
prepared for some really suspicious looks from the other engineers in
the design review.

With respect to mixer chips for such a task that has become much easier
these days, thanks to the bargain pricing going on in the cell phone
industry. Same for filters.

Regards, Joerg

http://www.analogconsultants.com

 

[Rob Gaddi]

It depends on what the words "fairly sharp" mean. If you have to be
down
to -30dB or more at 110MHz it'll be quite a challenge. Then those tiny
0603 inductors won't cut it. I'd start with 'real' inductors.

Did some poking around on Coilcraft's site. Yep, lo and behold a little
less stinginess on the footprints could have doubled or better my sorry
Q of 40. Yeah, fairly sharp in that I'm trying to keep that whole band
while putting at least 20dB of hurt on the 3rd order harmonics that
start at 120 MHz. Hence I was using 6 pole filters, and the parasitics
added up.

One trick that may seem a bit obscure but has worked for me in the past
is to up-convert the whole chebang to above a GHz where you then can
design a regular filter, abandoning the lowpass-highpass combo and using
resonant filtering instead. Then just convert back down with the same
LO.

Ideas like that make me wish I had budget enough to design with
honest-to-god impedance controlled boards.

 

[Tony]

Probably no good to you, but AFAICR the early 10baseT magnetics
included a 7th order Butterworth LPF, which might be a cheap easy way
to filter a lesser DDS.

On Wed, 05 Jan 2005 15:18:07 -0600, Rob Gaddi
<rgaddi@bcm.YUMMYSPAMtmc.edu> wrote:

So I need to do some wideband low RF filtering (flat to 0.5 dB from
40-100 MHz, fairly sharp cut off on the high end, low end can cut off or
not as convenient), and have found that my lossless LC ladders are in
fact anything but. Who knew that several ohms of resistance in my 0603
inductors would add up, other of course than anyone who had stopped to
think about it.

I've managed to massage the filters back into shape through trial and
error in RFSim99, but in the process have beaten my nice little
Butterworth and Chebyshev filters into unrecognizable scattered heaps of
very sore poles, making irrelevant decades of filter theory. It strikes
me that there must have been an easier way of doing it.

So, for future reference, how should one go about doing wideband RF
filtering? I've got books here with plenty of suggestions if I felt
like blowing ten thousand dollars designing an analog ASIC, but I'd like
a solution with a few less zeros, not to mention not having to learn
analog ASIC design.

-- Rob

Tony (remove the "_" to reply by email)