Harmonic Balance (HB) Large-Signal S-Parameter (LSSP) simulation

Hi Alex,

I am afraid that it needs two ports... This is probably due to the way LSSP analysis is implemented. It launches waves from both sides because it needs to calculate S22 and S12 too. You can check Spectre User Guide for more info:

Spectre Circuit Simulator and Accelerated Parallel Simulator RF Analysis in ADE Explorer User Guide 23.1

PSP/HBSP and LSSP are used for different things... LSSP (Large-Signal S-Parameters) is true large-signal analysis. It includes nonlinear effects at different signal amplitudes (power levels). PSP/HBSP is a small-signal analysis linearized around the periodic operating point (of a large-signal) calculated by PSS/HB. It is independent of the input power. You also have SP which is also a small-signal analysis linearized around DC operating point. So, which one to use depends on what are you trying to do exactly?

BR

/Janko

Hello, Alex -

As mentioned by JankoK, the nonlinear S-Parameter measurements require two ports, but here is a solution that makes the second port a non-issue:

Since you are only using a single port for nonlinear input, Port 1 needs to be a nonlinear port ("PORT1" is the simplest).

You can use a linear port ("PORT") for the second port and you can put it almost anywhere other than the port 1 node.  After you post it, you can double-click on the value of Z for the second port (default is 50) and change it to a huge impedance that will effectively make it an open circuit and it should have no real effect on the port 1 measurements based on a 1-port circuit.

Good Luck,
Mark

Hi, JankoK.

Thank you for your kind reply. It helped me a lot.

As I read your comment, It became clear that what I want to do is the true large-signal analysis (LSSP).

In order to do LSSP simulation, Can I use second port which is not connected to DUT?

Actually, I've done it and the LSSP simulation result came out... but I'm not sure I can trust this result.

Sincerely,

Alex

Hi, Mark.

Thank you for your kind reply. It helped me a lot too.

Following your suggestion, I've used second port which is connected to DUT with huge impedance.

However, the simulation does not proceed with some errors.

Therefore, I use second port which is not connected to DUT and the LSSP simulation is well done.

But I'm not sure about this result. Do you think this result is reliable?

Sincerely,

Alex

Hi, Alex -

I experimented with an installed example  that you have in your AWR folder, "Nonlinear_Example.emp".  Here is what I did:

• I started by running the exact example at a linear input (-30dBm) and I plotted linear S11 and nonlinear LSS11.  They were identical.
• The next step was to disconnect the second port.  In order to match the results, I added a 50 Ohm resistor where the port was connected, and the results were identical (linear & nonlinear with -30dBm input).
• Next, I pushed the input power to a ridiculous value for the example, +30dBm (1 Watt input).  Now the linear and nonlinear results are very different.  The linear result doesn't change with power changes, because it is a linear measurement.  The nonlinear result changes as the power increases.  With the resistor replacing the (now unconnected) second port, the results are again identical between the connected port and the unconnected port.
• Finally, I tried to do what you did with only the input matching circuit.  In order to match the full circuit results, the input matching circuit would need to have impedances added for the items that were removed, and that is quite difficult because it changes with bias and power levels.

I don't know what kind of nonlinear circuit you are running, but I assume it is an amplifier.  The point is that in order to measure S11 (linear or nonlinear), you will always get a different result if you analyze only the input match.  The biased transistor is a key item in measuring S11, so I would recommend that you do what I did with the example.  Leave the entire circuit set up and biased - just disconnect port 2 and replace it as a resistor with the desired output impedance.

Thanks.
Mark