HFSS Simulation - Return path

Hi all,

When we simulate an on-chip spiral inductor we typically use lumped ports to interface with the inductor ports (assume you have an inductor with input and output connections). The lumped ports are connected between two conducting materials so one connection is the inductor itself and the other is the PEC material. Between the lumped ports we connect a ground ring for the return current path. I have several questions about this:
1. What if we do not connect the ground ring around the inductor? How severe will the impact be?
2. What if we connect a MOM cap? Do we still need to have the ground ring? In general, the DC current isn’t flowing through the cap so do we need to have the current return path to Port1 from the reference of Port2?
3. Can we use a waveport instead of lumped port to simulate the inductance?
4. I haven't seen any "return path" for the current when using waveport. What am I missing here? Is it because the nature difference between lumped port and waveport so we can think of the lumped port as a Voltage/Current source?

I am attaching the picture below for demonstration of the "problem".
Regards,

Jim cage said:

1. What if we do not connect the ground ring around the inductor? How severe will the impact be?

Click to expand...

I assume this is a series inductor. The need for a "ground" conductor arises from transmission-line theory, which assumes an equal current on two conductors. If you only have one conductor, you won't support a TEM mode. This may or may not be a problem for you. In your case, I would say it depends on how electrically large the inductor is. You can probably get away without it if the inductor is small.

Jim cage said:

2. What if we connect a MOM cap? Do we still need to have the ground ring? In general, the DC current isn’t flowing through the cap so do we need to have the current return path to Port1 from the reference of Port2?

Click to expand...

DC current is not the issue -- again, it's supporting the right mode. I'm not sure how well the ring does this, so it may not make much of a difference without it.

Jim cage said:

3. Can we use a waveport instead of lumped port to simulate the inductance?

Click to expand...

Yes.

Jim cage said:

4. I haven't seen any "return path" for the current when using waveport. What am I missing here? Is it because the nature difference between lumped port and waveport so we can think of the lumped port as a Voltage/Current source?

Click to expand...

Once again, it depends what mode you are exciting. Waveports can excite non-TEM modes which only need one conductor. If you only see one current being excited by the port, then it is not a TEM mode.

Jim cage said:

2. What if we connect a MOM cap? Do we still need to have the ground ring? In general, the DC current isn’t flowing through the cap so do we need to have the current return path to Port1 from the reference of Port2?

Click to expand...

RF current also needs a return path - in simulation and also in real world hardware.

Jim cage said:

3. Can we use a waveport instead of lumped port to simulate the inductance?
4. I haven't seen any "return path" for the current when using waveport. What am I missing here?

Click to expand...

In those cases where you use a wave port, the wave port area covers both signal and return conductors, so that there IS a well defined return path. This is not fundamentally different from using lumped ports.

The problem that you have/see is certainly worth a discussion - what is my return path in simulation and in hardware? How to model an N.5 turn inductor with terminals on opposite sides if we don't know the return path? Good questions, but there are no trivial answers, and wave ports is not an answer at all. What comes closest to your intended method are these somewhat unphysical ports in MoM solvers (Momentum etc.) with port reference at infinity. But much better is to find out and include the actual return path as implemented in the hardware.

Thank you for your response.

I am not sure If my answers were properly answered so I'll ask it again:
1. Connecting the MOM cap as it is without ground ring. Is that acceptable or not?
2. Can we assume that the lumped port is a "voltage source" where the reference signal is the minus terminal while the signal is the plus terminal?
3. If you have to design MOM cap in HFSS. what will be your test structure? with or without ground ring for the return current path?
4. Do we need to connect the reference terminals of the lumped ports (as depicted in the figure)?
5. Can someone share an HFSS example with the lumped ports?

Regards,

Jim cage said:

Thank you for your response.
I am not sure If my answers were properly answered so I'll ask it again:
1. Connecting the MOM cap as it is without ground ring. Is that acceptable or not?

Click to expand...

I'm not sure if you got my point. You need a ground ring or another way of connecting the port ground references, otherwise you will see nosnense result. And the way how you create that ground connection will change results, because that path will be part of the results.

Jim cage said:

2. Can we assume that the lumped port is a "voltage source" where the reference signal is the minus terminal while the signal is the plus terminal?

Click to expand...

Simply speaking, yes.

Jim cage said:

3. If you have to design MOM cap in HFSS. what will be your test structure? with or without ground ring for the return current path?

Click to expand...

Most EM experts will stay away from HFSS for this task, because planar MoM solvers are more appropriate. But to answer your question: If you use HFSS, you need the ground ring or another defined ground return.

Jim cage said:

4. Do we need to connect the reference terminals of the lumped ports (as depicted in the figure)?

Click to expand...

Absolutely. Otherwise you get nonsense results (open circuit).

Jim cage said:

Thank you for your response. I am not sure If my answers were properly answered so I'll ask it again:

Click to expand...

Hi Jim,

What Volker said is correct -- so long as you're only looking for a solution which only involves TEM modes. This isn't always the case -- It could depend on your operating frequency.

Are you going to try to model this system with a high-accuracy equivalent circuit? If so, then yes, you want to ensure there is a return path.

Is the inductor electrically large? You may need to consider the effect of the distance/orientation of the "ground ring", as the transmission properties will depend on the TEM mode supported between them.

Do you only have one conductor with which to connect the inductor (on either end)? You could excite a non-TEM mode (although this *could* be detrimental to the operation of your inductor). You can do this with a waveport.

I have simulations with lumped ports, but they involve complicated setups. There is a great tutorial included with HFSS that uses lumped ports: "Getting Started with HFSS: A Silicon Spiral Inductor". The example uses a ground ring.

Good Luck

@Planar: These on-chip components are designed and used as lumped elements (more or less), with dimensions that are small compared to the wavelength.

- - - Updated - - -

Some thoughts on port ground reference in RFIC EM simulations:
http://muehlhaus.com/support/ads-application-notes/inductor-em-ports