GCPW-to-CPW transition

[TheUnnamedNewbie]

For a design I am working on (for my day-job) I need to use grounded CPW to interface to a flip-chip part. The reason we need to use grounded (finite width) CPW is because that is the only way we can come even close to the output impedance of the chip. We cannot get the impedance low enough using regular CPW. (we are already pushing the technology limits - 0.04mm gap).

However, I need to turn this into a regular CPW line (with higher impedance, but this is not the problem). I don't know how to do this without significant losses. And I can't seem to find any good examples in literature on how to deal with the termination of the ground plane. I can clearly see in my simulations that there is a lot of radiation at this interface between the FW-GCPW and the FW-CPW that follows, and this is by far the most significant source of loss in my design.

Does anyone have some resources or pointers to get started on this?

Here is a quick image to illustrate what I am talking about. It's flipped on its side, I'll try to take a better one later when I have my camera at hand.

[whollender]

That's an interesting problem.  You could try notching the ground plane underneath your CPW transition, so that the ground plane slowly moves away from underneath the center conductor while you transition to your new CPW geometry.

[ejeffrey]

Adding slots in the ground plane can definitely work.  Rather than ending the ground plane abruptly, you can start with a low density of very thin slots, then increase the density smoothly until the ground plane is gone and you are at your target impedance.

Here is a paper that does a very similar thing in figure 2: https://arxiv.org/pdf/1401.3799.pdf.  In this case, there is no continuous ground plane, just a gradually increasing density of grounding straps to go from high impedance to low impedance.  Note that the features here are super small: those ground straps are only a few microns wide, and the frequency is only 6 GHz.  Trying the same thing with a coarser technology might not be able to prevent radiation losses.

[TheUnnamedNewbie]

That's an interesting problem.  You could try notching the ground plane underneath your CPW transition, so that the ground plane slowly moves away from underneath the center conductor while you transition to your new CPW geometry.

I will be looking at this next. The balun I am currently building has a S21 of -2 dB, but we have been able to get similarly sized baluns below -1 dB of S21.

Adding slots in the ground plane can definitely work.  Rather than ending the ground plane abruptly, you can start with a low density of very thin slots, then increase the density smoothly until the ground plane is gone and you are at your target impedance.

Here is a paper that does a very similar thing in figure 2: https://arxiv.org/pdf/1401.3799.pdf.  In this case, there is no continuous ground plane, just a gradually increasing density of grounding straps to go from high impedance to low impedance.  Note that the features here are super small: those ground straps are only a few microns wide, and the frequency is only 6 GHz.  Trying the same thing with a coarser technology might not be able to prevent radiation losses.

I will have a look and see what these are doing. If nothing else, I can try to apply similar methods. I am worried though, since right now the technology is really limiting what I can do. The most narrow gaps I can already make are on the order of a 10th of a wavelength on the highest end of our frequency. I'll have to seem perhaps we can thin the ground planes to get narrower linewidths.

[ejeffrey]

I still think that a gradual perforating of the ground plane should be able to work, even if not by the exact method from that paper.

The key is that in order to do the transition from GCPW to CPW, you need to not only match the impedance, you need to allow the return current to transition from the ground plane up to the top layer.  Obviously you need vias for that, but I assume your problem is that a single pair of vias has too much inductance.  You need to push the current to the top plane over multiple vias.  To do that, you perforate the ground plane, which increases its inductance, forcing a larger fraction of the return current to the upper layer.  As you go past multiple vias, you increase the perforation density until nearly all the current flows in the top plane.

Also remember that CPW really needs periodic grounding straps to connect the two grounds anyway.  If you don't, bends will radiate.