Differential and single-ended L and Q of an inductor

[Hamid.Kiumarsi]

Dear all

I have simulated an inductor in ADS's momentum.

For getting differential L and Q I can
1) either use one differential port
2) or use 2 single-ended ports and then get the mixed-mode S parameters.
Both above-mentioned methods of 1 and 2 for differential L and Q give same results.

However for getting single-ended L and Q I use single-mode S parameters.

There is a slight discrepancy in L and Q of the inductor if I simulate it with differential port or single-ended ports.
My question: is this common and normal?

 

[volker_muehlhaus]

Yes, you will get different results if L and Q=Im{Y11}/Re{Y11} are calculated from the properties of port 1.

 

[Hamid.Kiumarsi]

I understand that those shunt R and C networks at port1 and port2 are for pads.
Why is that in your schematic, the inductor model has another L at port1 in addition to well known combination of R and L in series with shunt capacitor.
Could you be more elaborate with some equations.

And what does this different L and Q imply? With different signaling we'll achieve different L and Q?

 

[volker_muehlhaus]

I understand that those shunt R and C networks at port1 and port2 are for pads.

No, this is the coupling of the inductor to the substrate.

If you calculate Q using Y-parameters, Q=Im{Y11}/Re{Y11}, then you get the Q factor into port 1, with port 2 shorted.

And what does this different L and Q imply? With different signaling we'll achieve different L and Q?

Yes. Different substrate parasitics for different modes, so you get different results.

 

[Hamid.Kiumarsi]

Therefore, since transformers are mostly used in differential structures, we should use differential L and Q. (correct me if I am wrong)
What if they are used only for impedance transformation?

 

[volker_muehlhaus]

Therefore, since transformers are mostly used in differential structures, we should use differential L and Q.

I tend to agree.

What I recommend to my Sonnet customers is to simulate as a two-port, so that both common and differential mode are included in the data, and then use that two-port in Cadence. This is the safest method, in my opinion. Doing only the differential case might by too optimistic if the actual circuit has some common mode component.

 

[Hamid.Kiumarsi]

I am simulating with Momentum. Since I am using patterned ground shield, I can not use ///GND/// of Momentum.
Also using ground reference ports has its own problems. Therefore I am using differential ports and I will miss common-mode data.
If you have experience in simulating an inductor (with patterned ground shield) as a two-port inductor using Momentum, please advise.

 

[volker_muehlhaus]

If you have experience in simulating an inductor (with patterned ground shield) as a two-port inductor using Momentum, please advise.

Sorry, I have no idea how to solve the Momentum port issues. I use Sonnet for all my RFIC work.

 

[jonhalice]

it is common