S-param accuracy issue during Spectre transient simulation of low coupling inductor structure.

[Aman101]

Hi,
I am simulating a 4 port S-param file of two inductors with weak coupling between them (about -110dB).
First inductor: ports 1-2
Second inductor: ports 3-4
Nport settings: interp=linear, imptrunc=1.0e-6
Frequency: 0 : 100MHz : 300GHz

Simulation is comprised of a voltage source across port 1-2 while measuring voltage across ports 3-4.
Simulation is performed in AC and Transient (setting a pure 1V sinusoidal signal at 14.5GHz).

AC response is very smooth without any strange unexpected artifacts.
Tran response is converged without any blow-up artifacts.
Unfortunately, Tran response is 4 times the AC @ 14.5GHz!!!

I suspect it is originated from Spectre Impulse Response calculation.

Therefore I set the following Nport settings:
imptrunc=1.0e-12
imptrunc=0
datatrunc=0

All did not alter the results much.

When using S-param of stronger coupling between the inductors (-80dB), Tran and AC are identical!
using bbspice only makes things worse...

What are the Spectre settings for most accurate Impulse response?

Thank You.

p.s. I have read the article "7 Habits of Highly Successful S-Parameters" from 2019.

 

[BigBoss]

Transient response is not essential for those frequencies.. I don't know why you're using for ??
s-parameters are discrete type of parameters therefore continuity will always be a problem while doing continues time simulations.
Spectre converts s-parameters into continues form by interpolation and sometimes extrapolation but if there are some steep discontinuity between two values, spectre may fail..

 

[Aman101]

Transient response is not essential for those frequencies.. I don't know why you're using for ??
s-parameters are discrete type of parameters therefore continuity will always be a problem while doing continues time simulations.
Spectre converts s-parameters into continues form by interpolation and sometimes extrapolation but if there are some steep discontinuity between two values, spectre may fail..

Thank You for your reply,
This test bench is used to emphasize the problem...
Actual test bench is much more complicated and includes multiple coupled inductors of two DCO's operating at two frequencies separated by few ppms.
Narrowing the problem to a single frequency was done in order to simplify the test bench even more.

 

[vivekroy]

I have faced this issue a lot. The only thing that kind of solved this issue for me was to go for a low-frequency step (1MHz) in the EM simulation. If I remember correctly, the issue for me was the interpolation between DC and the first step.
Also, try replacing your input voltage source with a port with some non-zero real impedance and see if that is of any help.

@BigBoss, sometimes I did the transient on such s-parameter files as a final check on the system.

 

[timof]

You can ask Ken Kundert at https://designers-guide.org/forum/

 

[Aman101]

I have posted on Designers Guide, Thank You