Join us on Wednesday, Sep 16th, 2026 for our next LIVE Ask Me Anything Session on "Trusting Simulation Clarity at 50 GHz and Beyond" and discover how to achieve accurate, reliable results for next-generation high-speed designs.

If my simulation result looks okay, how do I know they accurately reflect behaviour at 50 GHz and beyond? 

For getting started, great question. 

For getting started, great question. 

For someone new to simulation , any best practices that you can share? For results at 50Gbz

1. Never trust the first run. Always perform at least one mesh refinement study.
2. Check current flow and return paths, not just S-parameters.
3. Use realistic material and roughness data.
4. Simulate beyond the target frequency (e.g., 110 GHz for a 50 GHz design).
5. Ask "Why do I see this behavior?" for every resonance, notch, or loss peak. If you cannot explain it, investigate further before signing off the design.

Any example that you can share , practical one where convergence criteria failed  and mat  be simulation appeared accurately 

Yes, this is actually a common situation in high-frequency EM simulations.

Consider a differential via structure that initially appears to perform very well up to 50 GHz. The S-parameters are smooth, insertion loss looks realistic, and there are no obvious signs of a problem. Based on the initial results, one might conclude that the design is acceptable.

However, after performing mesh refinement, a previously hidden resonance may shift by several GHz, leading to a different engineering conclusion. In this case, the critical electromagnetic behavior was concentrated around the via anti-pad region, where strong fringing fields existed.

The original mesh was too coarse in key areas such as:

1. Via barrels
2. Anti-pad edges
3. Plane transitions

As a result, the solver technically converged and produced a numerically stable solution, but the local field behavior was not sufficiently resolved. The simulation looked accurate because the S-parameters were smooth and physically reasonable, yet the solution had not fully converged with respect to the mesh.

This example highlights an important point: a simulation can appear correct and even meet convergence criteria, while still missing important physical effects due to insufficient local mesh resolution.

Therefore, confidence in a result requires both:

1. Numerical convergence of the solver, and
2. Physically accurate modeling and sufficient mesh resolution in regions where electromagnetic fields are concentrated.

Thank you everyone for the great questions and engagement today. Feel free to post your questions and we will address them accordingly.