PCB (FEM) PI and SI simulation questionaier. What do you use,...

[pointhi]

I posted those questions in https://www.eevblog.com/forum/eda/pcbpcad51pcbpcad720-review/msg3495214/#msg3495214, and as noted it would make sense to add a new thread for it.

As a bit of background. I'm currently doing initial research about FEM integration for KiCad. So, hopefully we get the first analysis modes into KiCad 7. My current plan is to start with current density / voltage drop analysis for PI because it seems to be fairly straightforward to achive and there are not much other free tools out there who allow this. To get a better sense of what potential users might need, I would like to ask a few questions about your current usage habbits as well as ideas for features. Note that I'm talking about PCB simulation, and not some abritrary structures like an MRT coil:

* Can you rank the analysis modes you use by importance, and for what (High-Power, RF, High-Speed-Signals,...)
* How regularly do you use each analysis mode?
* How long does the analysis require approximatly for your boards (seconds, minutes, hours)
* What Program do you use and why (What do you really like/dislike about it or the competitors)
* What would you like to see in features which no one implemented yet, but look not that complicated, and would help you?

Thanks for your responses.

[asmi]

* Can you rank the analysis modes you use by importance, and for what (High-Power, RF, High-Speed-Signals,...)

I don't have any power nor RF sims, so that leaves the high-speed stuff - impedance, signal integrity and crosstalk.
If I were to have them all, I'd prioritize high-speed stuff over PI since the latter is easier to just "wing" by over-speccing your PDS. I don't do RF stuff (yet), so not sure where that would be, but I expect it would be ahead of PI as well.
That said, I don't design boards for chips which consume 100's of Amps of current (like modern desktop CPUs), but I imagine that PI analysis would be much more important if I did.

* How regularly do you use each analysis mode?

I use impedance and SI analysis all the time - whenever situation warrants. If PI would be available to me, I'd use it as well, if anything as a sanity check.

* How long does the analysis require approximatly for your boards (seconds, minutes, hours)

If you mean how does a simulation run take - then it's up to few minutes (stuff like impedance analysis). Typical SI simulation run is a couple of seconds.
But I don't think run time is that important - I would absolutely wait however long it takes, as long as at the end of it I get results I need.

* What Program do you use and why (What do you really like/dislike about it or the competitors)

Orcad Pro - SigXplorer for SI simulations (both pre- and post-route analysis), impedance and crosstalk "visions" for board-level analysis.
I tried using high-speed tools in Altium, but either I am too stupid, or tools are crap, but I didn't get anywhere and abandoned my attempts after a couple of days of trying to get it to do what I want.

* What would you like to see in features which no one implemented yet, but look not that complicated, and would help you?

Do you mean in simulations, or in layout as a whole?

Thanks for your responses.

You are welcome, and thanks for the great job to you and other KiCAD developers!

[zeqL]

I posted those questions in https://www.eevblog.com/forum/eda/pcbpcad51pcbpcad720-review/msg3495214/#msg3495214, and as noted it would make sense to add a new thread for it.

As a bit of background. I'm currently doing initial research about FEM integration for KiCad. So, hopefully we get the first analysis modes into KiCad 7. My current plan is to start with current density / voltage drop analysis for PI because it seems to be fairly straightforward to achive and there are not much other free tools out there who allow this. To get a better sense of what potential users might need, I would like to ask a few questions about your current usage habbits as well as ideas for features. Note that I'm talking about PCB simulation, and not some abritrary structures like an MRT coil:

* Can you rank the analysis modes you use by importance, and for what (High-Power, RF, High-Speed-Signals,...)
* How regularly do you use each analysis mode?
* How long does the analysis require approximatly for your boards (seconds, minutes, hours)
* What Program do you use and why (What do you really like/dislike about it or the competitors)
* What would you like to see in features which no one implemented yet, but look not that complicated, and would help you?

Thanks for your responses.

I'm doing mostly SI simulation and also PI simulation. I'm not a specialized SI engineer doing simulation all day, so my use of simulation is from time to time. I've been able to get my hands on professional suites (Sigrity and Hyperlynx) as well trying open source and free solutions.
As I'm looking at doing boards with 28G and 56G serdes, I'm looking for IBIS-AMI simulation, which is out of the scope of FEM. For a small budget I decided to give a shot to Simbeor THz and use PyBERT.

For FEM and related simulation here are few usage that are usually quick to setup for the user and could be really useful for simulation :
- S-parameters / RLGC extraction of a trace/differential pair with or without coupling (not proposed by any free program)
- PI simulation : current flow and density map along with via current repartion

After that you can reproduce features like directly in Kicad layout:
- Impedance "heatmap" over pcb trace (Orcad like) to visually see impedance change
- Coupling "heatmap" over pcb trace (Orcad like) to visually see coupling factor

All of these need a good stackup impedance calculator/setup first.

[Pitrsek]

 S-parameters / RLGC extraction  - as long as you are happy with 2d cross section, mmtl can do that http://mmtl.sourceforge.net/ Can also calculate cross talk over length of trace.
For DC drop/current density you can go with https://www.femm.info/wiki/HomePage
For more general geometry, FEMM can calculate Ls and Cs. Ie you could slice the PCB perpendicular to the trace and feed crossection to FEMM. Not a full 3d field solver, but might be more than useful.

[zeqL]

S-parameters / RLGC extraction  - as long as you are happy with 2d cross section, mmtl can do that http://mmtl.sourceforge.net/ Can also calculate cross talk over length of trace.
For DC drop/current density you can go with https://www.femm.info/wiki/HomePage
For more general geometry, FEMM can calculate Ls and Cs. Ie you could slice the PCB perpendicular to the trace and feed crossection to FEMM. Not a full 3d field solver, but might be more than useful.

I tested MMTL, yes it's accurate. But when you have dozens of diff pairs (to perform IBIS-AMI simulation), recreating all the structures in MMTL is very time-consuming. So being able to have a single S-parameter file for a whole pcb trace is more than welcome.
The goal is not to look at the trace parameters but to use the S-parameter file in an IBIS-AMI simulation (with PyBERT for instance) where you have a TX and RX simulation model an then you put the link modelization.
With scikit-rf (python) I'm able to aggregate several S-parameters files into a single one (for instance a 2 PCB connected by a connector).

FEMM is way too "general", physics oriented and I clearly don't have the background. I tried openEMS with Kicad, as it is explained on kicad.info, but this also very time consuming. I know it's free and I can do some effort but if I spend 5 min with a paid software while I need a day with a free software to do the same thing...
Also if I remember correctly at some point you need to go through FreeCAD for the layout export, but unfortunately the user-script developped doesn't take into PCB trace thickness on external layers (FreeCAD limitation).

What I see is like in Hyperlynx, you select your net or trace -> Tools -> Extract S-parameters -> Option window -> "Extract" -> coffee -> Extraction done 

[Pseudobyte]

I use Ansys SIwave and HFSS. Keysight ADS is much easier to setup/use though.