EMRFD Message Archive 1559

Message Date From Subject
1559 2008-04-18 07:11:48 Alan Ibbetson Filter layout to minimise unwanted coupling
I have a question about the shape of the ground current distribution in
a real world filter, so I can visualise the difference between a good
and a bad layout.

I want a 7 pole Cauer LPF for a 160M transmitter. The plot facility in
the excellent ELSIE filter design package predicts three nulls in the
stop band. When I built the filter and examined it on my old (but
inexpensive) spectrum analyser and tracking generator I could only see
the first two nulls. The response where the third higher-frequency null
should have been was pretty flat at around -60dB, with no evidence of
the predicted -100dB dip.

I have a solution, but I don't really understand how it works. When I
first built the filter I laid it out using Manhattan construction as a
physical copy of the geometry of the circuit diagram, with the grounded
end of all the shunt capacitors in a row along one of the long edges of
the board. This layout had no "third null" in its response. After seeing
the note on p3.12 of EMRFD I added a 1cm high vertical wall made from
thin copper sheet along the edge of the board where all the grounded
ends of the capacitors are. Now the response looks pretty much like the
design software says it should. I added more walls round the other edges
and also between the inductors. These improved the response a little bit
more and brought it within 1 or 2dB (measurement error) of the design

There are a few words of explanation about ground currents and mutual
inductance in the book. I can produce the two null response in SPICE by
adding a 1nH mutual inductor as in fig 3C but I have to admit I don't
have a good mental picture of the physics that is creating this
inductor. This makes my future designs rely on an empirical approach to
layout rather than something based on understanding. Not good :-) For
example, would just using a 1cm longer/wider board have the same effect
as my vertical walls? The board in the "bad" filter in fig A p3.12 seems
to have plenty of unused area around the components, although the BNC
earthing looks suspect.

Please can someone on the list explain the shape of the ground current
distribution in my good and bad filter layouts, how to predictably get
the good distribution, and what is actually happening to make the good
distribution good? Or point me at a reference if there is one I've not
found by Googling.


Alan G3XAQ
1560 2008-04-19 01:24:37 groundplane1 Re: Filter layout to minimise unwanted coupling
I have found the following reference to be useful, it may help with
some of your questions.
The hosting site usually requires a log in (I havn't had any spam
issues frim them), you may need to go through:
first and select EMC Information Centre, Keith Armstrong ortfolio.
There are a number of interesting EMC articles on this site, which
focuses on EMC and is targeted at UK electronics engineers.
(I had some difficulty acessing the site this morning, they may be
reorganising it, so links might change.


The information suggests that to be properly effective the ground
plane has to extend beyond the components.

I think 2dB corelation simulati