LNA Noise Figure: Y-Factor vs. Cold Source, and When the Difference Actually Matters

I see a lot of confusion about which noise figure method to trust, so here is my bench-level take after measuring dozens of LNAs with both approaches.

**How they differ (briefly):**
- **Y-factor** uses a calibrated noise source (hot/cold states) and computes NF from the power ratio. It is fast and the classic approach, but the result inherits the ENR uncertainty of your noise source — typically ±0.1 to 0.2 dB in practice.
- **Cold source** (vector method) measures gain and noise power with the source at ambient, using a VNA-characterized source match to correct for mismatch. No noise source needed, and it handles high-gain, low-NF devices well.

**When the difference matters:**
1. **Sub-1 dB LNAs**: here ±0.15 dB ENR uncertainty is a huge fraction of what you are trying to verify. Cold source with a well-characterized tuner/source match usually gives tighter, more repeatable numbers.
2. **Poor input match**: Y-factor assumes the DUT sees the noise source impedance it was calibrated for. If your LNA input VSWR is bad, mismatch error silently biases the Y-factor result. Cold source explicitly corrects for it.
3. **Production speed**: Y-factor wins. One noise source, one power meter or NFA, done in seconds. For pass/fail screening at 2 dB NF spec, the extra rigor buys you nothing.

**Practical checklist before you believe any NF number:**
- Verify the noise source ENR cal is current (they drift).
- Measure with the exact source impedance the DUT will see in the real circuit.
- Repeat the measurement after a full power cycle — thermal drift in the first 30 minutes fools everyone once.

My rule of thumb: Y-factor for screening and datasheets comparisons, cold source when you are signing off a sub-1 dB design or arguing with a vendor.

Which method does your lab default to, and have you ever caught a mismatch-driven NF error in the wild?

Welcome, RF_Bench_Notes — a very clear comparison, and exactly the question people ask us about NF measurements.

A quick illustration of why sub-1 dB LNAs are hard with Y-factor: at NF = 0.6 dB (F ≈ 1.148) a ±0.15 dB ENR error maps almost one-to-one onto the NF result, i.e. ±25% of the noise you are trying to measure. Also worth mentioning: the noise source's ENR changes between its on and off states' match, so a low-ENR (≈6 dB) source with better match is often preferred over a 15 dB ENR source for low-NF DUTs.

Two practical additions to your checklist:
- Remember the second-stage correction: if the receiver's NF is high and the DUT gain is low, the measured result is very sensitive to the receiver calibration.
- Account for any adapter or cable before the DUT: 0.1 dB of loss at the input adds 0.1 dB to NF directly.

Question: in your experience, how close do cold-source and Y-factor agree on a well-matched LNA (say input return loss better than 15 dB)?

— rfforum 编辑部 / Editorial Team