L
Admin
[Technical article / FAQ, rfforum Editorial Team]
mmWave and high microwave packages still live or die on bondwire / ribbon geometry. Simulators will happily report a beautiful match on an under-meshed wire that does not exist on the bench.
1. Failure modes that look like "model vs measurement"
- Coarse mesh on a thin wire under-reports inductance and misses local capacitance to the ledge / lid — S11 looks optimistic
- Loop height and touchdown position dominate L; a ±20 µm process window can move you off a narrowband match that the nominal model loved
- Ground return: a wire over a cutout or with a distant via fence is a different inductor than the same wire over continuous ground (package launch cousins: /t/329)
2. Freeze list before you trust a wire launch sim
- Mesh / adaptive refinement criteria on the wire and the first 2–3 via pitches of return path — not only on the 50 Ω trace
- Nominal vs corner geometry (height, length, die-pad offset) in the same study, or at least a sensitivity table
- Lid / cavity proximity if the package is shielded — floating lids change the return
- How the measurement reference plane is defined vs the sim port (wave port, lumped, de-embed length)
3. Habit
Archive mesh settings and wire cross-section tolerance next to every "bondwire match OK" slide. If measurement disagrees, check geometry + mesh before retuning the on-die match network.
Discussion: on your last wire-launch miss, was the root cause loop height, return-via placement, lid proximity, or an under-meshed model — and do you require a corner study before tape-out?
— rfforum 编辑部 / Editorial Team
mmWave and high microwave packages still live or die on bondwire / ribbon geometry. Simulators will happily report a beautiful match on an under-meshed wire that does not exist on the bench.
1. Failure modes that look like "model vs measurement"
- Coarse mesh on a thin wire under-reports inductance and misses local capacitance to the ledge / lid — S11 looks optimistic
- Loop height and touchdown position dominate L; a ±20 µm process window can move you off a narrowband match that the nominal model loved
- Ground return: a wire over a cutout or with a distant via fence is a different inductor than the same wire over continuous ground (package launch cousins: /t/329)
2. Freeze list before you trust a wire launch sim
- Mesh / adaptive refinement criteria on the wire and the first 2–3 via pitches of return path — not only on the 50 Ω trace
- Nominal vs corner geometry (height, length, die-pad offset) in the same study, or at least a sensitivity table
- Lid / cavity proximity if the package is shielded — floating lids change the return
- How the measurement reference plane is defined vs the sim port (wave port, lumped, de-embed length)
3. Habit
Archive mesh settings and wire cross-section tolerance next to every "bondwire match OK" slide. If measurement disagrees, check geometry + mesh before retuning the on-die match network.
Discussion: on your last wire-launch miss, was the root cause loop height, return-via placement, lid proximity, or an under-meshed model — and do you require a corner study before tape-out?
— rfforum 编辑部 / Editorial Team
