5
\$\begingroup\$

I have a 1.8 V, 350 MHz, 100 Ω LVDS signal on an FFC on the bottom plane. Below the FFC is an aluminum metal layer; the aluminum is pretty much unconnected from the FFC. The air gap/solder mask gap is currently a few thousands of an inch, maybe 1-3, but I could put some kapton tape underneath to ensure it's not a problem and widen that gap to whatever the thinkness of the tape is 10-20 mil.

Will I have an issue with this changing the parameters of my transmission line more than say 5% or 10%?

This is what the 'stakup' looks like. The aluminum has enough impedance from the diff pair that it could be considered as mostly disconnected at 350 MHz as it's not very capacitively coupled.

enter image description here

I think the answer is no, because I assume this is like a edge coupled internal asymmetric differential pair.

In the diagram below I have this upside down from the previous diagram where H1 is the coupled distance in the cable and H2 is the aluminum. The material says FR-4 but I have the Er plugged in for Kapton FFC, but I'm not sure because H2 is disconnected.

enter image description here

\$\endgroup\$
3
  • 1
    \$\begingroup\$ My suspicion is that it's extremely relevant because the nearby conductive surface effectively shortens the distance between the two conductors (the E-field is zero within it). So it'll behave as if there was just 2x the solder mask thickness between the traces, rather than your nice and wide controlled gap. I'm not quite sure about it though, hence the comment rather than an answer. Maybe someone else knows for sure. \$\endgroup\$ Feb 9 at 18:30
  • \$\begingroup\$ maybe I'll run some experiments with an oscope if I have time. My board doesn't come until after Lunar New Year :( \$\endgroup\$
    – Voltage Spike
    Feb 9 at 18:43
  • \$\begingroup\$ In your picture from the Saturn tool, the traces will couple to both planes. So both H1 and H2, and the relative dielectrics, will determine your Zo, and not so much S. \$\endgroup\$
    – SteveSh
    Feb 9 at 21:45

3 Answers 3

3
\$\begingroup\$

The air gap/solder mask gap is currently a few thousands of an inch, maybe 1-3, but I could put some kapton tape underneath to ensure it's not a problem and widen that gap to whatever the thinkness of the tape is 10-20 mil.

The aluminum has enough impedance from the diff pair that it could be considered as mostly disconnected at 350 MHz as it's not very capacitively coupled.

We're talking distances of 1 mil at minimum to 20 mil at the maximum. It's coupled all right, I'm afraid, and coupling through a solid dielectric will only make it worse.

In any case, stick a network analyzer on the pair, and measure at various distances and dielectrics. That's the only way to be sure. Of course if you want to model it with a field solver first, that'll help, but you'll have to measure the prototypes, no doubt about that.

\$\endgroup\$
1
  • \$\begingroup\$ I'll have to plug it in and see with the VNA when I get it. \$\endgroup\$
    – Voltage Spike
    Feb 9 at 23:34
3
\$\begingroup\$

Yes. It will be a differential stripline geometry.

Note that the dielectrics may differ as well. You show FR-4 in the calculation, but FFC is made of polyimide. It's not clear to me what your stackup is, but it may contain both FR-4 and polyimide (and air for that matter), and so a hybrid approach is necessary.

The airgap must also be tightly controlled.

\$\endgroup\$
7
  • \$\begingroup\$ I can't control the air gap, it is mechanical. I can only make it wider/higher so it isn't relevant. \$\endgroup\$
    – Voltage Spike
    Feb 9 at 18:38
  • 1
    \$\begingroup\$ The answer is the same in any case: you'll likely need a field solver to find the actual impedance, or minimum height for a lack of change. \$\endgroup\$ Feb 9 at 18:40
  • \$\begingroup\$ So even if the plane isn't coupled and have currents on it? \$\endgroup\$
    – Voltage Spike
    Feb 9 at 18:44
  • 2
    \$\begingroup\$ @VoltageSpike But the plane is coupled, it is in near the signals so it is a reference plane for the signals. It does not need to be grounded to be the reference plane. \$\endgroup\$
    – Justme
    Feb 9 at 18:47
  • 2
    \$\begingroup\$ The plane essentially shorts the two traces together via a pair of distributed capacitors in series: one capacitor from D+ to the metal plane, and another one from the metal back to D-. The absolute potential of the plane is irrelevant - due to the high speed of the signal, localized voltage swings will develop in the plane via transmission line effects. \$\endgroup\$ Feb 9 at 18:49
0
\$\begingroup\$

Does there have to be an air gap? You could put a thin thermal pad in there and vacuum bond the assemblies together. This should at least give you a more controlled height (H1) and dielectric constant so that you could adjust your line widths (W) and spacings (S) to give you the differential impedance you desire.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.