LC butterworth filter

Look up "gimmick" ;-) ...Jim Thompson

It looks to me that you should be doing this with stripline construction. The .12nH is only about .15 mm of 20ga wire.

John S

Those values don't make sense. Either a design approach or a filter topology not suitable for the application.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

Choose a filter topology that'll actually work?

Those values don't look right for a low-pass filter -- I gather that you're trying to design a band-pass filter using the lowpass-to-bandpass transform where you set s_bp = s / (s^2 + w_0). Doing that works if your relative bandwidth is high, but hands you extreme values of components (as you're finding out) when the relative bandwidth is low.

You should probably be looking at a coupled-resonator filter.

yeh and 0.038pf is about the same as two plates the size of a 0603 footprint separated by 1.6mm fr4

-Lasse

That's a 10th order filter! Do you want a lowpass or a bandpass?

Those values aren't practical.

What's the frequency and impedance?

John

We are trying to make narrow band band-pass filters at different frequencies. Perhaps I should try two back to back filters with overlapping regions at target frequency.

Overlapping back-to-back filters? One LP and one HP?

A resistive attenuator would have about the same effect, for less money and board space.

Coupled-resonator filters are _the_ solution to your problem, here. Why aren't you considering them?

Yes, that will work.

OK, will look into it.

That's prime turf for top-coupled coaxial ceramic resonators.

John

Technically, don't coupled resonators have repetitive stop and pass bands. I thought they need out of band filtering too. One of those things not always discussed in the design articles, much like stub filters.

Stripline makes sense to me here.

I've been hacking around in L-band and have a few filter links handy.

There are also SAW filters in that range, though hard to buy in unit quantities. You can pull them out of cable modems. SAW filters can't take much power.

do

pF

Why

Yes, we probably need wider bandpass before hitting the SAW. Unfortunately, SAW goes up to 1.6GHz only.

ualband_trx.htm

Everything is hard to buy in small qty. I brought 100 22nH smd coils from mouser before. But now, everything (any value) is 2000 minimum. I guess i can only design in linear combinations of 22nH.

This search in Mouser shows lots of 22nH smd incuctors available for pennies in quantities of one.

This search shows the same for 100nH.

John S

w do

.5pF

rim

ey

=A0Why

/dualband_trx.htm

tp://

I hope I'm not stating the obvious here, but there is the philosophical question of "how big is an inductor." Since most inductors have leakage, you need to play games to keep them from coupling. Sometimes it is as simple as orienting them in different directions to reduce coupling. There there is the issue of parasitic elements, i.e. non-ideal components. By the time you reach L-band, you really should be doing stripline filtering. I'm a lumped components person myself, so stripline is a bit foreign to me, but from an engineering standpoint, it makes the most sense. For one thing, the component values are not set to fixed values. Out of all the schemes in that band, stripline seems to be the least amount of work.

Having done some controlled impedance DUT boards, the space from the trace to ground plane doesn't have that great of a tolerance. Most PCB manufacturers are just doing logic boards. Since space wasn't at a premium in my cases, I would use the fattest boards I could find. The tolerance error was less on a thick board.

how do

trim

th

oney

=A0Why

RX/dualband_trx.htm

For lower frequencies, i like wiring them with thick wire (for example, .5mm dia) to get high Q (several hundreds). They take up vertical air spaces, but not much more PCB spaces. However, it might be difficult beyond 1G.

According to the link someone posted, striplines are around 4cm to 5cm and probably restricted from other components as well. They might not have very high Q either.

Not if the resonators are lumped-constant, but at your frequencies that may be hard to achieve.

Stripline may be the way to go, if you can get the repeatability that you need at your bandwidth, and if you have the room.

Mr. Larkin's suggestion of top-coupled coax resonators may be worth looking into, if you can find a source for the resonators.