21-Pole LP Chebychev Filter Project

[jim_griff]

Hi all.

This is about the 4th or 5th passive filter I've designed and built up to now. I have been attempting many different methods of building filters, from in-line coils with capacitors tapped to ground at strategic points, to the standard PCB style. Most failed miserably due to parasitic elements I didn't take into consideration, but this one seems to perform a lot better.

It is a 21-pole low-pass Chebychev design, with all inductors and capacitors the exact same value except for the input and output ones, which are slightly lower value. Its designed for use in the FM band (76MHz to 108MHz) so I designed the cutoff frequency a little higher (125MHz.) Funnily enough, the parasitics on the board did not reduce the -3dB frequency by any amount. It measures exactly 125MHz!

I was going to stitch both ground planes (top and bottom) together through all the holes, but got bored and only stitched enough so the performance is good up to about 500MHz. Some of those pesky microwave frequencies can obviously sneak inbetween the gaps, but my radio transmitter doesn't produce any horrible harmonics in the 1GHz+ range, so no problems there.

Enjoy the pics.

EDIT: P.S. Any advice on proper via stitching would be greatly appreciated!

[Leo Bodnar]

Good stuff!  What is the IL value?

I have been scratching my head while building 150MHz BP filter in a similar fashion and I can't figure out what would be the best way of positioning the nearby coils.  Placing them the way you did does not minimise cross-coupling. Is there science behind it?

I am surprised attenuation is around 70dB without shielding and compartmentalisation.

Why did you put capacitors alternating between the top and bottom GND strips?  If you visualise the current flowing through each LCC tank then you will see that it has to go through two caps on the opposite sides of  the strip, through the GND stitching and the bottom GND plane.  If caps are grounded to one side it would make current flow shorter.  I wonder if it would make much difference at VHF frequencies .  But at 1GHz+ it might.

Leo