A brief background. If more data or information is needed, let me know.
I have 2 boards. Different stackups (as described below). Both have
differential lines designed with the same target impedance (102 ohms).
Both have (finished) 4.5 mil line widths (as measured by the venders
after cross-sectioning) and are using 1/2 oz copper (that I believe is
actually 0.6 mils thick after a roughening process that both venders use.
Board 1 has differential trace lengths of roughly 20 inchs (actually 2 10"
traces that I've jumpered together). The traces start with a set of vias
that get the signal pair to the internal layer, and they make a straight
run to the 10" via pair. The second 10" run is identical (i.e. the lines
are straight with no vias in between. The nearest lines are 50 mils away.
The D.C. resistance of a single 20 inch trace is roughly 7.2 ohms. Using an
11801 with 6150 probes and a kludged miniture semi-rigid coax tips to make
contact with the differential pairs, the differential impedance is 109 ohms.
The impedance (within an ohm) is 109 ohms from the start of the line to the
end (except for the discontinuity at the probe tips and at the jumper where
the 2 10" traces are connected).
I connected the 2 10 " traces together to get the lines longer than board 2.
But the measurements for the 10" are identical (except the D.C. resistance
is about 3.5 ohms for the single 10 inch trace)
Board 2 has differential trace lengths of roughly 16 inchs (single run, same
layer from connector via to connector via). The nearest traces are 20 mils
away and run in parallel for the full 16 inchs (this is a bus). The D.C.
resistance is 5.6 ohms. Using the same measurement technique, the differential
impedance is 111 ohms at the beginning of the line, BUT, the impedance
increases from 111 to 119 from the start of the line to the end in a very
linear fashion. When I go to the other end of the trace and repeat the
measurement, I get the same reading (111 at the start and increasing to 119
and the other end).
So the question is:
Why does board 2 exhibit this increasing impedance effect? Is it possible
that the materials are so different (terrible) that I'm seeing significant
losses or other effects?
The following describes the stackups and line spacing:
Board 1
--------------------------- gnd
4 mils dielectric core
-- -- signal layer (4.5 line, 8.5 mil space, 4.5 lin)
8 mils dielectric prepreg
-- another signal layer, 50 mils away
4 mils dielectric core
--------------------------- gnd
Board 2 (I do not know yet which is core and which is prepreg if this is
significant)
--------------------------- gnd
6 mils dielectric
-- -- signal layer (4.5 line, 8.5 mil space, 4.5 line
8 mils dielectric
--------------------------- gnd