Re: [SI-LIST] : Guard banding

Ray Anderson (raymonda@radium.eng.sun.com)
Tue, 11 Nov 1997 09:26:34 -0800 (PST)

Mark-

Thanks for your comments. I can completely agree
with your experiences regarding your use of grounded
shield traces in a low-level RF environment. I was
with GTE Spacenet for about 12 years doing circuit
level design of earth station hardware from DC thru
14 GHz and have encountered similar situations.
Sometimes a well grounded piece of metallization
appropriately placed can provide those critical
few dB's of isolation that can transform a system
from parasitic oscillation to stability.

Since becoming involved in SI issues over the past
several years it has become apparent that the
digital world is just recognizing certain phenomena
that the RF and Microwave guys have been dealing
with for ages.

In _most_ high-speed digital cases it seems that allowing
more space (about the same space that a grounded trace would
take) has the effect of markedly reducing crosstalk. If you
have the luxury of being able to put the metal down on your
board in the form of a grounded trace things will get all
the better providing it is truly effectively GROUNDED at the
frequency of interest. Insufficient grounding on the ground
trace can make things worse as the trace can re-radiate or be
a mechanism of unintended coupling.

These effects can be effectively analyzed and simulated to
predict what is going to happen. I'm sure someone, somewhere
has published a study or paper on it.

As far as low frequency and DC applications go, I believe that the
scheme you speak of (driving the guard trace from a lo-Z source)
is common practice in a lot of instrumentation circuitry. At
DC sometimes people will create a grounded "ring" of metal around
a really critical hi-Z node to prevent picoAmp level parasitic
current flow across the surface of a board the may be ever
so slightly conductive due to contamination and moisture from
humidity.

You are right, there is a fundamental difference between guarding
and shielding. I tend to think of "guarding" as primarily a DC sort
of thing, while shielding is a AC/RF concept.


Best Regards,

Ray


/\
\\ \
\ \\ /
/ \/ / / Raymond E. Anderson WB6TPU
/ / \//\ Signal Integrity Engineer
\//\ / / Sun Microsystems
/ / /\ / 901 San Antonio Road MS MPK15-103
/ \\ \ Palo Alto, CA 94303
\ \\
\/ (650) 786-3272
(650) 786-6457 fax
raymond.anderson@Sun.Com

>
> But there are places where a WELL "GROUNDED" "guard trace" can
> act as a shield between adjacent signal paths. The application
> I'm thinking of was for a 1.5GHz part that had a VCO next to
> an LNA. Getting THAT combination to settle down so the VCO was
> oscillating on frequency and the LNA wasn't required these
> "extreme" measures. Luckily, even though the pitch of the IC
> was fairly fine, we had several grounds next to each other for
> low inductance grounding. It also supplied sufficient room for
> the shield trace and included vias. We didn't make any isolation
> measurements other than to note that the unwanted coupling appeared
> to be reduced sufficiently to eliminate the oscillations. There
> wasn't easy access to both traces and the problem was solved.
>
> It did what a shield is supposed to do, intercept and redirect
> currents to where you want them to go. I was surprised that the
> vias didn't change the impedance of the 50 ohm line next to it
> very much.
>
> As I recall aren't guard traces driven by a low impedance sources
> to about the same potential as the signal traces, thereby reducing
> current leakage from the signal trace? I've heard of it used for
> very low current "DC" signals. (BTW, I'm asking if this is correct)
> There seems to be a fundamental difference between "shielding"
> and "guarding".
>
> --
> ---------------------------------------------------------------
> Mark Randol, RF Measurements Engineer | Motorola SPS, Inc.
> (602)413-8052 Voice | M/S EL379
> (602)413-4150 FAX | 2100 E. Elliot Road
> ryvw50@email.sps.mot.com | Tempe, AZ 85284
> ---------------------------------------------------------------