L
Admin
[Technical article, rfforum Editorial Team]
A bias tee that looks fine at DC and at the carrier can still host a parallel resonance where the choke meets board parasitics. That resonance shows up as a kink in |S22|, a stability circle that suddenly kisses Γ=1, or a load-pull contour that refuses to make sense near one weird frequency.
1. Where the tank comes from
The choke is a large L with self-capacitance; the tee’s blocking C and the PA board’s bypass network add more C. Together they form one or more high-Q poles, often below or beside the band. At that pole the bias node is no longer a short for RF — it is a reactive termination hanging on the PA.
2. Symptoms on the bench
- Narrowband gain peaking or oscillation only when the tee is attached
- Load-pull "islands" that move when you swap tees or cable length on the DC side
- K-factor / μ that looks fine in a bare-die model but fails with the real bias network
3. What to do
- Sweep S-parameters of the tee alone (RF port, DC port terminated as in use) and mark SRF / suck-outs
- Damp the bias node (series R, lossy ferrite, or a well-placed bypass to a solid RF ground) so the pole is low-Q
- Keep DC cable shields and return paths short; a long DC pigtail is part of the resonator
- Re-check stability with the production bias network, not only with a lab tee
4. Habit
Treat the bias tee as an RF component with a datasheet curve, not as invisible plumbing.
Discussion: at what frequency did your last bias-network surprise appear relative to the PA band, and did you damp it or change the choke?
— rfforum 编辑部 / Editorial Team
A bias tee that looks fine at DC and at the carrier can still host a parallel resonance where the choke meets board parasitics. That resonance shows up as a kink in |S22|, a stability circle that suddenly kisses Γ=1, or a load-pull contour that refuses to make sense near one weird frequency.
1. Where the tank comes from
The choke is a large L with self-capacitance; the tee’s blocking C and the PA board’s bypass network add more C. Together they form one or more high-Q poles, often below or beside the band. At that pole the bias node is no longer a short for RF — it is a reactive termination hanging on the PA.
2. Symptoms on the bench
- Narrowband gain peaking or oscillation only when the tee is attached
- Load-pull "islands" that move when you swap tees or cable length on the DC side
- K-factor / μ that looks fine in a bare-die model but fails with the real bias network
3. What to do
- Sweep S-parameters of the tee alone (RF port, DC port terminated as in use) and mark SRF / suck-outs
- Damp the bias node (series R, lossy ferrite, or a well-placed bypass to a solid RF ground) so the pole is low-Q
- Keep DC cable shields and return paths short; a long DC pigtail is part of the resonator
- Re-check stability with the production bias network, not only with a lab tee
4. Habit
Treat the bias tee as an RF component with a datasheet curve, not as invisible plumbing.
Discussion: at what frequency did your last bias-network surprise appear relative to the PA band, and did you damp it or change the choke?
— rfforum 编辑部 / Editorial Team
