Subject: Time: 7:20 AM
OFFICE MEMO re>impedance across ground plane Date: 2/16/96
Joe, you sent an inquiry "We are searching for information on the impedance
across the ground plane on a pcb. We have some harmonics near 3 Ghz, and would
like to see if we can reduce the impedance to earth ground through our pcbs,
and backplanes."
Welcome to the club of "serious" printed circuit design. Your problem is most
probably due to "cutouts" in the power.ground planes. As you to indicate most
of you concerns are near 3 GHz, you want to look at near "quarter-wavelength"
conductor patterns. For example, the wave length in air for 3 GHz is about
100 mm, if you are using epoxy-glass with a dielectric constant of 4.3 @ 3
GHz, the propagation velocity is about 50% of air for about a 50 mm
wavelength. A quarter-wave is then about 12.5 mm [0.5 inch]. In theory, any
conductor that "electrically-looks-like" a quarter-wavelength stub
electrically looks like an infinite impedance. Any conductors that are
shorter than a quarter wavelength have a finite impedance.
G 1/4 L Z G 1/4
L Z 1/4 L G
signal conductor (sc) scscscscsc scscscscscscscscscsc
via short (v) v v
v
signal ground (sg) sgsgsgsgsgsgsgsgsgsgsgsgsgsgsgsgsgsgsgsg
Note 1 -- In the above, G represents a "good" electrical ground (short), 1/4 L
represents a quarter-wavelength (L is used for lambda), and Z represents a
high impedance (in theory infinite).
Note 2 -- The "signal ground" is the signal's "reference or return" conductor
path and does not have to be the low-frequency (dc) grounded (in NEC parlance
grounding) conductor.
The above, (hopefully) illustrates two conditions, the left half a
quarter-wave "fork" or "stub" condition; and the right half a half-wave
condition where the conductor is grounded at each end and is "floating" in the
middle. The voltage at "G" should be "0" (zero) volts, and the voltage at
"Z" could be at least the applied voltage (or more).
As a rule-of-thumb, a "good" electrical ground(ing) conductor should be less
than about 1/20'th of a wavelength. If the product is high-speed
analog/digital instrumentation or signal conditioning, or high-sensitivity
radio receivers, then you may need grounds of 1/50 L or less.
Your problem may be related to non-imaged conductive patterns transmission
lines, an illustration follows:
scscscscscscscscscscscscscscscscscscscscsc
sgsgsgsgsgsg sgsgsgsgsgsgsgsg
s g
g s
sgsgsgsgsgsgsgsg
In this situation, the signal will see an increase in voltage due to the
increase in transmission line (or signal path) impedance where the conductor
spacing is increased.
Your solution will be to eliminate all cut-outs and if at all possible to use
as "solid" as possible signal "ground" or reference/return plane(s).
A couple of other thoughts/info.:
* The impedance of a 35 micrometer thick copper foil at 3 GHz is
approximately 20 mohm per square.
* The impedance of a 1 mm printed board (ciruit) conductor at 3 GHz
is in the range of 13-15 ohms per square.
Thus it becomes obvious if you require good "low-impedance" power, ground, and
signal grounds/references/returns you don't want to route them around the
printed board design as traditional conductors. Nor do you want to use
power/ground planes that resemble "swiss cheese".
References:
* I remember a good article in PCDesign magazine that presented some good
stuff on high frequency designs. It also addresses "following the current"
for the effect of "ground loops" at higher frequencies.
* Your concern is an EMC (electromagnetic compatability) issue. Most of my
references and experiences have been based based on courses and text's from
Interference Control Technologies, which I found out has been renamed the
"emf-emi control" or "the EEC Press" in Gainsville VA.
If you need some further information, you can contact me in one or more of the
following methods.
Ralph Hersey
Lawrence Livermore National Laboratory
P.O. Box 808, L-225
Livermore, CA 94551
Phn. 510.422.7430
FAX. 510.424.6886
e-mail at work: [log in to unmask]
e-mail at home: [log in to unmask]
|
