L
Admin
[Technical article, rfforum Editorial Team]
A converged |S11| does not mean your edge currents are trustworthy. Many antenna and filter models look "done" because the reflection coefficient stopped moving, while far-field tilt, ohmic loss, and near-metal hot spots are still mesh artefacts.
1. What S11 misses
Port reflection is a global integral. Local current singularities at thin metal edges, slots, and via pads can keep changing under refinement even after S11 has settled within 0.1 dB. If your figure of merit is pattern, efficiency, or power handling, S11-only stopping criteria are incomplete.
2. Edge and skin effects
At microwave frequencies the current piles at conductor edges. Under-resolved edges underestimate loss and can invent fictitious reactance. Check edge line length vs skin depth and local λ; force denser mesh or edge seeding along radiators, slot lips, and ground cutouts.
3. Practical convergence checklist
- Refine until S11 and the quantity you ship (gain, axial ratio, isolation, loss) both stop moving.
- Compare at least two mesh strategies (length-based vs curvature/edge seeding).
- Spot-check surface current plots for hot pixels that jump when you re-mesh.
- For lossy metals, confirm the solver is using a consistent surface-impedance or volume model, not a perfect-E shortcut left on from an early draft.
4. Bench habit
If measured efficiency or pattern disagrees with a "converged" S11 model, re-open the mesh at edges before blaming the connector or the chamber.
Discussion: what secondary metric do you freeze as a go/no-go besides S11 when you sign off an antenna or filter mesh?
— rfforum 编辑部 / Editorial Team
A converged |S11| does not mean your edge currents are trustworthy. Many antenna and filter models look "done" because the reflection coefficient stopped moving, while far-field tilt, ohmic loss, and near-metal hot spots are still mesh artefacts.
1. What S11 misses
Port reflection is a global integral. Local current singularities at thin metal edges, slots, and via pads can keep changing under refinement even after S11 has settled within 0.1 dB. If your figure of merit is pattern, efficiency, or power handling, S11-only stopping criteria are incomplete.
2. Edge and skin effects
At microwave frequencies the current piles at conductor edges. Under-resolved edges underestimate loss and can invent fictitious reactance. Check edge line length vs skin depth and local λ; force denser mesh or edge seeding along radiators, slot lips, and ground cutouts.
3. Practical convergence checklist
- Refine until S11 and the quantity you ship (gain, axial ratio, isolation, loss) both stop moving.
- Compare at least two mesh strategies (length-based vs curvature/edge seeding).
- Spot-check surface current plots for hot pixels that jump when you re-mesh.
- For lossy metals, confirm the solver is using a consistent surface-impedance or volume model, not a perfect-E shortcut left on from an early draft.
4. Bench habit
If measured efficiency or pattern disagrees with a "converged" S11 model, re-open the mesh at edges before blaming the connector or the chamber.
Discussion: what secondary metric do you freeze as a go/no-go besides S11 when you sign off an antenna or filter mesh?
— rfforum 编辑部 / Editorial Team
