L
Admin
[Technical article / FAQ, rfforum Editorial Team]
A cavity or enclosure simulation can show a sharp resonance that survives every DUT tweak because the resonance lives in the truncated domain or the absorbing boundary, not in the product. PML, Mur, and "absorbing" foam in a measurement box all have residual reflection; at the wrong frequency that residual looks like a Q you will never solder away.
1. What the boundary actually owns
- Numerical absorbing layers have a reflection coefficient that depends on incidence angle and frequency. A mode that grazes the boundary can bounce enough to fake a cavity peak.
- Domain size and the distance from the DUT to the absorber set which modes fit. Growing the box by λ/4 and watching the peak move (or die) is cheaper than respining the shield.
- In the lab, absorber foam and liner gaps create the same class of artifact: a "box mode" that tracks lid torque and foam compression more than DUT geometry.
2. Freeze list before you redesign the shield or the board
- Boundary type, thickness, and distance from the nearest metal of the DUT
- Whether the peak frequency moves when you enlarge the domain or thicken the absorber (sim) / add foam or change lid gap (lab)
- Field plots: energy piled on the boundary or in the air volume, versus on the DUT seams
- Mesh/convergence: a under-resolved PML can invent lossy resonances that vanish with tighter mesh
3. Habit
Never open a shield ECO for a resonance that moves when only the box or the PML changes. Archive domain size, boundary settings, and a "larger box" check next to every unexplained cavity peak. If the peak tracks the lid gap and not the board stackup, stop etching the PCB.
4. Practical gate
- Sim: re-run with +λ/4 domain and thicker absorber; lab: change foam or lid gap with the same DUT
- If the peak moves more than the DUT geometry did, the box owns the ticket
- Name the boundary in the same sentence as the resonance frequency
Discussion: on your last false cavity peak, did the fix land in domain size, absorber, or lid gap — and do you require a larger-box check before freezing a shield redesign?
— rfforum 编辑部 / Editorial Team
A cavity or enclosure simulation can show a sharp resonance that survives every DUT tweak because the resonance lives in the truncated domain or the absorbing boundary, not in the product. PML, Mur, and "absorbing" foam in a measurement box all have residual reflection; at the wrong frequency that residual looks like a Q you will never solder away.
1. What the boundary actually owns
- Numerical absorbing layers have a reflection coefficient that depends on incidence angle and frequency. A mode that grazes the boundary can bounce enough to fake a cavity peak.
- Domain size and the distance from the DUT to the absorber set which modes fit. Growing the box by λ/4 and watching the peak move (or die) is cheaper than respining the shield.
- In the lab, absorber foam and liner gaps create the same class of artifact: a "box mode" that tracks lid torque and foam compression more than DUT geometry.
2. Freeze list before you redesign the shield or the board
- Boundary type, thickness, and distance from the nearest metal of the DUT
- Whether the peak frequency moves when you enlarge the domain or thicken the absorber (sim) / add foam or change lid gap (lab)
- Field plots: energy piled on the boundary or in the air volume, versus on the DUT seams
- Mesh/convergence: a under-resolved PML can invent lossy resonances that vanish with tighter mesh
3. Habit
Never open a shield ECO for a resonance that moves when only the box or the PML changes. Archive domain size, boundary settings, and a "larger box" check next to every unexplained cavity peak. If the peak tracks the lid gap and not the board stackup, stop etching the PCB.
4. Practical gate
- Sim: re-run with +λ/4 domain and thicker absorber; lab: change foam or lid gap with the same DUT
- If the peak moves more than the DUT geometry did, the box owns the ticket
- Name the boundary in the same sentence as the resonance frequency
Discussion: on your last false cavity peak, did the fix land in domain size, absorber, or lid gap — and do you require a larger-box check before freezing a shield redesign?
— rfforum 编辑部 / Editorial Team
