A low-noise amplifier quoted at 0.8 dB NF and 18 dB gain can still go deaf next to a handset PA or a Wi-Fi radio. Desense is usually compression or intermodulation from an out-of-band blocker, not a mysterious rise in NF. A compact way to budget it.

1. Mechanisms
- Gain compression: a strong blocker reduces small-signal gain for the desired weak tone; SNR falls even if the NF meter looks fine without the blocker.
- Reciprocal mixing: blocker × LO phase noise folds noise onto the IF / baseband (more of a mixer/LO problem, but the LNA passes the blocker).
- IM3 / cross-modulation: two blockers or blocker × TX leakage create in-band junk.
- Supply / bias pulling: large RF swings modulate the bias network and create AM sidebands.

2. Numbers that matter
IIP3 and IP1dB from the datasheet are starting points at a stated frequency and tone spacing. For a blocker at offset Δf you also need:
- Front-end filter rejection J(Δf) in dB before the LNA
- Blocker power at the antenna P_b
- Power at LNA input: P_b - J(Δf) (plus antenna-to-LNA cable loss)

Rule of thumb: keep blocker power at the LNA input several dB below IP1dB if you need near-full gain for the desired signal. For IM3-limited cases, use the two-tone IIP3 and your blocker pair levels to estimate in-band IMD.

3. Worked sketch
Desired signal -100 dBm at LNA input. Blocker +10 dBm at antenna. Preselect gives 40 dB rejection at the blocker offset → -30 dBm at LNA. If IP1dB = -15 dBm, you have 15 dB margin to 1 dB compression; desense from compression should be small. If the preselect is only 25 dB, blocker at LNA is -15 dBm and you are on the edge; expect desense.

4. Measurement that matches the field
- Inject desired CW or modulated tone at the sensitivity level
- Sweep blocker power at the relevant offsets (TX band, Wi-Fi, cellular harmonics)
- Record SNR or BER vs blocker power; mark where you lose 3 dB SNR (common desense metric)
- Repeat with the real filter and shield, not only a conducted SMA into a bare LNA eval board

5. Mitigation order
1) Filtering and duplex / isolation before the first nonlinear stage
2) Higher IP1dB / IIP3 LNA if current and NF budget allow
3) Automatic gain / attenuator ahead of the LNA in pathological coexistence cases (NF penalty is real; use sparingly)
4) Bias network with enough headroom and decoupling so large signals do not modulate the supply

6. Datasheet gaps to watch
NF measured without blockers. IIP3 at 1 MHz spacing when your blocker is 100 MHz away. OIP3 quoted after a fixed gain assumption. Always re-state the tone spacing and match condition you care about.

Budget the blocker at the LNA pin, not at the antenna connector marketing slide.