Re: [SI-LIST] : Proposal: Rs correlation/collaboration for W-Elements

Larry Smith (ldsmith@lisboa.eng.sun.com)
Tue, 3 Aug 1999 12:24:23 -0700 (PDT)

Check this out. Should I send it? Any glaring mistakes?

> From: "Dr. Edward P. Sayre" <esayre@nesa.com>
>
> If we are going to check the accuracy of the .W model, I wish to make the
> suggestion that we use measurements as well for verification. Most field
> solvers do not account for losses in copper or dielectric to as correct an
> extent that I would deem acceptable. Does anybody have a set of test board
> that they would volunteer?
>
> ed sayre

Ok, we will bite. It turns out that we are having trouble getting
model to hardware correlation for the W element. At the end of this
note, I am attaching measured data. (Many thanks to Tanmoy Roy for
providing this data!) The data is for 5 inch and 10 inch transmission
lines, both microstrip and stripline. The data is self consistent.
The losses on a 10 inch line are twice as much as the losses on a 5
inch line. The losses on the microstrip are almost twice the losses on
the strip line.

I believe this illustrates an important point. The current on the
microstrip line travels mostly on the bottom surface, as close as it
can get to the return current. The proximity effect causes this to
happen. The stripline looks up to see a return plane above, and down
to see a return plane below. Current therefore travels on both the top
and bottom surfaces of the stripline. For this reason, I believe it
makes sense for the microstrip to have almost twice the loss of
stripline.

The industry has taken it for granted that skin effect causes current
to travel on the skin of a conductor and for the resistance to have a
sqrt(f) dependency. The W-element makes this assumption. It has two
resistive parameters, one that is frequency independent (R0) and one
that goes as sqrt(f).

Unfortunately, the data below does not indicate a sqrt(f) dependence
but rather loss that is proportional to frequency. Normally, we would
associate this with dielectric (G) loss, but I have a hard time
believing that FR4 is as lossy as these measurements indicate, below 1
GHz.

Could it be that proximity effect is giving us something other than
sqrt(f) losses? I'll let the data speak for itself. We would be
interested in any thoughts that the SI-list has on the mechanism for
the loss and EM extractors that predict resistance values that will
account for these losses. The measurements are very repeatable...

The microstrip is 6 mil wide, 1.5 oz copper separated from the
reference plane by 2.8 mils of FR4. The stripline is 4 mil wide, 1 oz
copper. It is sandwiched between reference planes that are 3 mils
above and 6.5 mils below the trace. We have assumed eR=4.0 for the
dielectric. The measurements were taken with a network analyser, S21.
All losses are in dB.

layer: uStrip uStrip stripl stripline
length: 5" 10" 5" 10"
freq(MHz)
50 -0.117 -0.334 -0.171 -0.379
100 -0.245 -0.554 -0.304 -0.586
200 -0.418 -0.903 -0.419 -0.894
300 -0.608 -1.244 -0.547 -1.145
400 -0.797 -1.621 -0.665 -1.364
500 -0.990 -1.981 -0.774 -1.583
600 -1.198 -2.310 -0.893 -1.767
700 -1.397 -2.699 -0.968 -2.021
800 -1.592 -3.059 -1.047 -2.274
900 -1.838 -3.594 -1.256 -2.483
1000 -2.188 -4.106 -1.465 -2.691
1200 -2.474 -4.676 -1.557 -2.937
1500 -3.026 -5.762 -1.788 -3.522

regards,
Larry Smith
Sun Microsystems