Transient simulation of nport (s-par) in spectre cadence

[fathi]

spectre transient simulation time s-parameter

I have s-parameter data exported into Cadence, and used in nport component. Depending on whether I choose windowing=yes or windowing=no, or set the maximum frequency limit for the n-port, or using equivalent y-parameter or z-parameter instead of s-parameter, I either get error in transient simulation (Signal V(...) = 1.3 GV exceeds the blowup limit for the quantity `V' which is (1 GV). It is likely that the circuit is unstable. If you really want signals this large, set the `blowup' parameter of this quantity to a larger value.) or get different result which are not consistent with each other. The component whose s-parameters I'm using is not passive. Any suggestion for the setting or overcoming this issue?

Thanks.

 

[Alex_IC]

z-parameter data file spectre

If it is an active device S-parameters, it is not possible to simulate transient with S-par.
The S-parameters measured for particular bias point, where the circuit is assumed linear with small signal variations.
I've heard about P2D S-par, may be you can use it...

 

[fathi]

cadence error blowup limit

I'm sorry, my mistake; the architecture is passive. It's not active.

 

[Alex_IC]

blowup limit spectre

So use Spice model generator tools in ADS to get lumped elements spice model for transient simulations.

 

[fathi]

cadence blowup parameter

Thanks for your reply.

1- I tried fitting a Pi-Lumped Network Model (which would physically be able to describe the behavior of my block across a narrow frequency range) in a narrow frequency range using ADS Spice Model Generator, but the generated/predicted S-parameters of this lumped model are very different from the original S-parameters, and as a result, the simulation result of the circuit is not right.

2- I simulated the same circuit using S-parameters (SNP compoenets) in ADS, and got consistent result for transient simulation there. So, basically, there should not be any problem in time-domain simulation using S-parameters. I am not probably using the right settings in Cadence.

 

[Alex_IC]

s-parameter passivity check cadence

You have to adjust the Pi model after generation with optimization tools in order to match the result with S-par and Pi model.

 

[milkdragon]

s-parameter spectre dc

For transient simulation, you have to set your n-port in your CDF to spline in the interpolation method. It is better if you can carefully modified your .s2p file such that there is a zero freqeuncy data

 

[fathi]

spectre s model view cadence

How would you set it up in ADS to find the components?

I couldn't find any element in ADS for matching the S-parameters. The way that I set it up is that I put the S-parameter data as a n-port device (SNP) and I put the schematic of the lumped model to fit with n additional ports in the same schematic. So, I have a schematic with total of 2*n ports. Then, I set the goals to be magnitude(Sij)=magnitude(S(i+n)(j+n)) and phase(S(ij))=phase(S(i+n)(j+n)). I tried satisfying these goals by setting the RELATIVE error be smaller than 1%, but that causes a problem around zero phase, not mentioning the 2pi discontinuity points. Finally, I couldn't make it converge to a good result. Do you know any standard and robust way of fitting the model to a given s-parameter data in ADS?

Thank you very much.

 

[Alex_IC]

spectre s-parameter transient

I do it by hand with parametric simulation.

 

[fathi]

s-parameter file in spectre dc block

I don't get your method.
If the number of ports are large and/or the model has many compoenents, it's not possible to find the values by hand.

 

[Alex_IC]

s parameter simulation spectre cadence

Yes, you are right. I did it for spiral only...

Added after 13 minutes:

Have you tryied the method milkdragon suggested? If it works inform me pls. I'm intrested in this problem also.

 

[fathi]

s2p spectre transient

Based on the nport application note from Cadence, linear interpolation usually works better than spline. By zero frequency, you mean resonance frequency? I even increased the frequency range to include the resonance frequency; didn't help. Actually, since the variation around resonance frequency is large, it's better not to go above that and cut the data before it; still didn't help....

 

[fathi]

fitting s parameter to lumped model

My .snp file already has the DC (zero frequency) data.

 

[fathi]

cadence spectre nport

Using rational fitting with no DC point in the s-parameter file works.