siwave port
wlcsp shalom,
I'll start with the easiest question. HFSS extrapolates to DC, it can give a better guess when used with the DC thickness option, but it's not a DC calculate per-se.
SIwave has a DC solver which gives accurate results.
When you say the simulation failed, I suppose that you mean that the spectre simulation failed. If I were you I would do two things. First try running Spectre with the original SIwave S2P file (there should not be any difference between the S2P file and the FWS file since you have only 2 ports). Generally speaking Spectre should be able to transient an S2P file directly. If you run a more complex SIwave simulation (and export a DDR2 64bit bus 128 S-Parameter file) then you should use Ansoft Nexxim (which is Spectre compatible but has a special S-Parameters transient algorithm). Second , please connect the FWS (or S-Parameter) file properly in the Spectre schematics:
If the N port S-Parameter (N=2 in your case) is represented with a symbol that has 2*N terminals then connect all of the N negative terminals to the circuit zero ground node. If the N port S-Parameter represented with a symbol that has N+1 terminals then connect the last terminal to the circuit zero ground node. Connect the remaining N positive terminals to their respective power or signal voltage (there is no terminal for the ground net, more on this below).
In some circuimstances you might want to circuit simulate a Full Wave extracted results (such as SIwave or HFSS) together with a Quasi Static extracted results (such as Q3D , TPA or IBIS pkg RLC models). In this case (only) the thumb rules above break, and you need a better understanding of what you're doing.
So here is the explanation. Let's imagine that you have an IC on each side of the Transmission line. Each one of them has their own local ground terminal and their own local signal terminal (right?). In SIwave or HFSS we do not assume that the ground voltages accross all of the model is zero. The ports are referenced to a specific place in the ground plane very close to the signal. The only question that matters is this: How much power (or voltage) does the circuit simulator excite between the local signal and the local ground. SIwave couldn't care less if the Signal is 1V and the ground is 0V or if the signal is 101V and the ground is 100V, the voltage difference on the port is still 1Volts! For the sake of simplicity let's say that 0.1V is lost in the line, so the output of the S-Parameter would be a difference of 0.9Volts (again SIwave does not care if the signal is 0.9V and the ground is 0V or if the signal is 100.9V and the ground is 100V ). In other words, the ground is always local ! Any inducatance or capacitance between the trace and the ground are part of the S-Parameters and you cannot separate the trace from it's return path. Back in the circuit simulator each port is represented by 2 terminals (in the 2*N symbol case). The voltage difference between these two nodes is all that matters for the S-Parameter block. So the first IC excites the Transmission line with 1V and the other IC sees 0.9V. You cannot tell from the results if the signal is 0.9V and the ground is 0V or if the signal is 1V and the ground is 0.1V. For the IC it doesn't matter either, since it uses its LOCAL ground node as reference. So even if you assume that the ground in the second IC is zero, it might not be the case. The buty of all this is that even in reality (in the lab) the measurement is always done with a local ground reference and not the "global" ground of the power supply. By definition high speed (and RF) signals always travel with its return path, and the measurement is done between the signal and the local ground reference.
There's much more and I encourage you to contact your local Ansoft support.
Regards,
Itai