Is there a software that can completely handle EM/SPICE simulation?

[rodgerwxh]

I know that HFSS can work with ADS to get a complicated circuit simulation, CST support some nice GUI to design some circuits as well. However, if the circuit contains large number of non-linear devices, such as diodes and transistors or even controlled devices. Can HFSS and CST still be able to handle it for a full-wave solution? Does anyone did similar simulation before? Or is there new software that can do this?

 

[BigBoss]

No..
CST,ADS Momentum,HFSS and other EM simulators can simulate passive circuits only.
If you would test your passive and active parts together, you should transform your passive structure in ADS or MW Office and then you may use nonlinear and linear elements with passive structures together to able to simulate them.

 

[rodgerwxh]

Thanks for the reply, then I am wondering why it is difficult to directly integrate non-linear parts into the full-wave solver? The reason I ask this question is this.
Say, I have two amplifiers attached to the same structure at the same time. Then it would be impossible for me to get a passive model from the Full-wave solver side. Since on the other end is always connected to a active device. And this is nearly impossible to simulate, right? Correct me if I am missing something.

 

[volker_muehlhaus]

Thanks for the reply, then I am wondering why it is difficult to directly integrate non-linear parts into the full-wave solver?

It is not difficult, but inefficient.

From a designflow perspective, it makes more sense to separate the linear problem from the non-linear problem. The linear EM problem takes a lot of time, and we don't want to solve it again and again for different node voltages/currents. By separating the linear problem from the non-linear problem, we only need to do the time consuming EM simulation once, and then we can re-use that for all possible node voltages/currents.

Say, I have two amplifiers attached to the same structure at the same time. Then it would be impossible for me to get a passive model from the Full-wave solver side. Since on the other end is always connected to a active device. And this is nearly impossible to simulate, right?

Can you explain again?

 

[rodgerwxh]

volker, thank you for the explanation. It makes sense from a designflow perspective. However, if I understand you correct, this is assuming that different circuit components are not coupling with each other. The only connection between these non-linear circuits is the passive model extracted from EM simulation. There are two concerns as I continue the discussion:

1. Non-linear circuit itself has weak radiations by introducing voltages or currents through its output connections. Thus when circuits are put close, the mutual coupling between them might be strong, and the design method you proposed does not cover this situation, because to consider this mutual coupling you have to either modify your circuit or use full-wave simulation to capture the change in input/output voltage/current

2. To explain more clear of my example in the last post, say you have a 4-port microwave coupler, each port is connected to a non-linear circuit. My question is how can you extract a equivalent model for this 4-port coupler? is it a S-matrix?

I just want to have a knowledge of the real challenge of current circuit/EM design. Thanks,

 

[volker_muehlhaus]

However, if I understand you correct, this is assuming that different circuit components are not coupling with each other. The only connection between these non-linear circuits is the passive model extracted from EM simulation.

This is correct. It is assumed that the coupling/radiation from the non-linear elements is small. But what are those non-linear elements?

If the "non-linear components" should be included in the simulation, we can again split them into a passive linear part (layout/package/wiring of the device) and the small active part (the semiconductor itself). Some of my clients have done this for RFIC work, where they included the metal/vias of the transistor element in the EM model, and only simulated the active device (non-linear) externally through ports.

1. Non-linear circuit itself has weak radiations by introducing voltages or currents through its output connections. Thus when circuits are put close, the mutual coupling between them might be strong, and the design method you proposed does not cover this situation, because to consider this mutual coupling you have to either modify your circuit or use full-wave simulation to capture the change in input/output voltage/current

I might have answered this above? Coupling requires some finite area, which is usually in the passive linear part (layout or device package).

2. To explain more clear of my example in the last post, say you have a 4-port microwave coupler, each port is connected to a non-linear circuit. My question is how can you extract a equivalent model for this 4-port coupler? is it a S-matrix?

Yes, that's a simple 4-port S-parameter matrix. The frequency range must be chosen wide enough, so that sufficient number of harmonics are included in the S-parameters.

 

[rodgerwxh]

Thanks, volker, that's pretty much cleared out all my questions.

 

[davidryv]

The closer thing I know is APLAC, but it doesn't use SPICE, but it use its own language.

 

[RealAEL]

ADS can do EM/Circuit Co-Simulation if the layout is defined with an appropriate EM set-up.

https://www.youtube.com/watch?v=kGkJGmcsAtY

The video is for ADS 2009 and this only applies to Momentum as the EM simulator. ADS 2011 can do much more by being able to use either Finite Element or Momentum as the embedded EM solver. The circuit simulation could also be Harmonic Balance, Transient or Circuit Envelope and not just the linear simulation in the demo.

 

[volker_muehlhaus]

CST (and possibly other solvers) can also embed non-linear elements like diodes using their SPICE model. I think it is an inefficient approach that makes sense only for very carefully selected cases, for the reasons that I mentioned above.

ADS can do EM/Circuit Co-Simulation if the layout is defined with an appropriate EM set-up.

What ADS does is to solve the EM problem with ports, and then connect the circuit elements through ports. That is the better approach most many cases.

 

[rodgerwxh]

I agree that for most PCB layouts and relatively low frequency applications, ADS is sufficient. However, when mutual coupling can not be ignored, such as frequency is high, circuit elements are very close, or signal line is very sensitive to weak fields. Full-wave simulation is a must. The extracted S, Y or Z matrices extracted from a pure EM model does not contain the information of circuit mutual coupling. Where full-wave does. After all, it depends on what kind of situation of your design being applied and what kind of accuracy/cost you want to get/spend.

 

[volker_muehlhaus]

I agree that for most PCB layouts and relatively low frequency applications, ADS is sufficient. However, when mutual coupling can not be ignored, such as frequency is high, circuit elements are very close, or signal line is very sensitive to weak fields. Full-wave simulation is a must.

I think RealAEL meant ADS Momentum = EM simulation combined with circuit level simulation for the lumped elements.

The extracted S, Y or Z matrices extracted from a pure EM model does not contain the information of circuit mutual coupling. Where full-wave does.

Please elaborate what you mean by that.

I am working in RF simulation methodology for many years. I agree that unexperienced users are asking for combined EM with circuit elements, because they don't understand simulation workflows.

Above, you have linked to a "hybrid" solver . These guys connect circuit elements into the EM model. However, this will provide the same results as placing a port and doing the circuit level simulation externally. There is absolutely no advantage is terms of accuray with the Wave n ology method.

 

[xpz]

Everything has reasons for existing. ADS+HFSS exist for a long time, it has been proved that it can be used in many applications. But technology always move on, new applications appears, the old technology will not fit this situation. New ideas and new methods are required.

There are many reason that separated circuits and EM simulation cannot meet some requirements:
1. cross-talk among stripline in high freq..
2. sensitive phase control for wave port input.
3. shorter simulation time.

ADS do a good job in coupling with HFSS result. But in my several years research in GigaHz simulation and radiation problem(it include switch, amplifier and antenna array), HFSS + ADS never meet our requirements due to the de-coupled circuit and EM model (we need to obtain an accuracy higher than -50dB to obtain scattered signal). Maybe Wave n ology is a solution for this.

One situation that is not good for ADS+HFSS is wideband simulation. In order to obtain the wideband T matrix, HFSS is time consuming. For example, for mod + demod simulation, if user want to check the transient response, ADS+HFSS is time consuming and lots of additional works. But a transient circuit solver + EM solver can solve it very easy.

 

[volker_muehlhaus]

I see ... marketing buzz all around, trying to replave technical facts with funny marketing claims.

Good luck!

 

[rodgerwxh]

The fact is you are ignoring the real technical challenges that I am facing by saying you have to do it my way, b/c I am the only way,

The fact is I am designing a smart antenna, with 124 circuit elements, 45 are passive, rest is active, all these circuits are completely mounted inside the structure which is NOT 2.5D microstrips. I start to use ADS since its 2002 version. I said in previous post that I do agree that it is really powerful on most of the designs. However, for my current problem. I just can not figure out a workaround to use ADS separate the simulations of 124 circuits and 96 arms and other structures. If I extract EM models from structures, I can not see an efficiency in the "design flow". Since I probably have to buy two set of software and work on both.

Looking for something new is not a threat to something mature, it drives the evolution of technologies, and keep your mind refreshed. Many years' experience /= able to solve all kinds of new challenges. The game rule is always newbies challenge bosses and goes on and on.

 

[volker_muehlhaus]

The fact is you are ignoring the real technical challenges that I am facing

I am not ignoring them. I explained how that is solved efficiently with existing EM-circuit co-simulation. If you think that ADS Momentum planar is not appriate for everything ... agreed. Even Agilent seems to agree, and have added 3D FEM co-simulation capability to ADS, so that users can choose between different EM tools for co-simulation. Same for AWR with their circuit-EM co-simulation.

The tool that you are trying to promote here can do co-simulation, like many existing tools already offer. In both cases, the interface between EM and circuit simulation is through explicit or implicit ports, so that your promoted suggestion does not offer the superior accuracy that you claim.

by saying you have to do it my way, b/c I am the only way

I described why some methods make more sense than others.

But as I said ... good luck with your marketing efforts.