Switched-filter bank ahead of an LNA: IL passed, NF died on a reflective off-state

Started by LunaWei · 2026-11-17 01:48 UTC · 18 views · 3 replies
L Admin
[Technical article / FAQ, rfforum Editorial Team]

A switched-filter bank in front of an LNA can pass insertion-loss and return-loss in the selected path, then still wreck noise figure. The unused filters, or the off throw of the switch, often present a reflective stopband right at the LNA input. The LNA sees a bad source match out of band (and sometimes in band), and NF or stability moves even though the "on" path IL looks fine.

1. What the off-state owns
- Many PIN / solid-state switches look open or poorly terminated in the off throws. A filter hanging on that throw reflects stopband energy back toward the LNA.
- NF is measured with a noise source at the bank input. If off-path reflections or switch leakage change the source Gamma at the LNA, the NF number includes the bank, not just the filter IL.
- A short or open a few centimeters away can also land a resonance in-band after a cable or housing change.

2. Freeze list before you swap the LNA
- On-path IL and RL, and NF, with the other throws terminated in 50 ohm vs left as in the product
- LNA source-pull or a simple pad at the LNA input: if 1-2 dB pad kills the NF surprise or the oscillation, stop the transistor ECO
- Switch control and bias: a half-on throw is a different reflection than a hard off
- Housing and cable length A/B; a resonance that moves with a jumper is a reflection, not an LNA wafer issue

3. Habit
Never close "LNA NF is 1 dB worse in the box" from an on-path IL budget alone. Archive off-throw termination and a terminated-vs-open A/B next to every bank+LNA NF claim. If terminating the idle filters restores NF, the LNA was never the patient.

4. Practical gate
- Terminate unused throws; re-measure NF and S11 at the LNA plane
- Add a temporary pad; note NF / stability change
- Then decide between absorptive switches, terminations on unused filters, or a real LNA change

Discussion: on your last bank-plus-LNA NF miss, did the fix land in off-throw termination, switch type, or a real LNA - and do you forbid an NF sign-off without an unused-path A/B?

— rfforum 编辑部 / Editorial Team
L Admin
Addendum — terminate unused throws before an LNA ECO. Measure NF and S11 at the LNA plane with the idle filters (or off throws) terminated in 50 ohm versus left as in the product. If the terminated case restores NF or kills a start-up oscillation, the transistor was never the patient.

A 1-2 dB pad at the LNA input that buries the surprise is the same diagnosis: source Gamma, not a wafer lot. Archive switch control/bias (a half-on throw is a different reflection) and a housing/jumper A/B; a resonance that walks with a cable is a reflection, not an LNA change.

Open: on your last bank-plus-LNA NF miss, did the fix land in off-throw termination, switch type, or a real LNA — and do you forbid an NF sign-off without an unused-path A/B?

— rfforum 编辑部 / Editorial Team
G Member
Reflective off-state is the whole story. An open shunt branch in the off path turns the filter bank into a stub that sits at whatever electrical length the layout gives it, and that stub transforms to a low impedance right at the LNA input at the band edge. Measure the off-state S11 of each unselected branch into a matched load, not into a short, then re-check the cascaded NF with the bank in place.
E Member
How's the power consumption on this? Our project has a tight power budget.
广告 AdRenesas 瑞萨微控制器芯片产品照片Renesas Electronics微控制器、模拟与射频/时钟方案 · MCUs, analog, RF and timing访问官网 →免费展示 · 资料来自公开信息 · 如需撤下请联系 ads@rfforum.com
广告 Ad广告位招商 / Advertise hereads@rfforum.com · 刊例 Rate cardT2 · 728×90
Log in or register to reply.