Published Nov 7, 2026 · Last updated Nov 8, 2026 · 6 min · IndieRF

Cascaded Noise Figure: Gain in Front Hides Stage Two

A 10 dB NF second stage adds 2.3 dB to a 1 dB NF LNA with 10 dB of gain, and 0.3 dB with 20 dB. Friis divides the second stage's noise by the gain in front of it, so with enough gain the first stage sets the receiver's noise figure.

Put a 1 dB NF LNA with 20 dB of gain in front of a 10 dB NF second stage and the cascaded noise figure is 1.3 dB. Give the same LNA 10 dB of gain and it is 3.3 dB. The second stage is the same part both times. The gain in front of it decides how much of its noise reaches the answer: about 0.3 dB in the first case and about 2.3 dB in the second.

Friis divides each stage by the gain ahead of it

Noise factor FF is noise figure in linear terms, F=10NF/10F = 10^{\mathrm{NF}/10}. For a chain of stages with available gains G1,G2,…G_1, G_2, \ldots (also linear), Friis's 1944 result is

F=F1+F2−1G1+F3−1G1G2+⋯F = F_1 + \frac{F_2 - 1}{G_1} + \frac{F_3 - 1}{G_1 G_2} + \cdots

The first stage counts in full. Every later stage counts only its excess noise, F−1F - 1, divided by all the gain ahead of it. By the time the signal reaches the second stage it is G1G_1 times bigger, and the second stage's own noise is measured against that. The second stage got no quieter. It got outvoted.

The worked case is a 1 dB NF LNA (F1=1.259F_1 = 1.259) ahead of a 10 dB NF second stage (F2=10F_2 = 10, excess 99), in a 1 MHz channel from a 290 K source:

  • With 10 dB of LNA gain, the second stage adds 9/10=0.99/10 = 0.9. F=2.159F = 2.159, or 3.34 dB.
  • With 20 dB, it adds 9/100=0.099/100 = 0.09. F=1.349F = 1.349, or 1.30 dB.
ChainGain ahead of stage 2 (dB)Cascaded NF (dB)Added by stage 2 (dB)Total gain (dB)Noise floor in 1 MHz (dBm)
Second stage alone010.00020−103.98
LNA, 10 dB gain, then stage 2103.3422.34230−110.63
LNA, 20 dB gain, then stage 2201.3000.30040−112.68
LNA, 30 dB gain, then stage 2301.0310.03150−112.94
1 dB cable, 20 dB LNA, stage 2192.3000.30039−111.68

Each row is hand Friis, and the RF Cascade Analyzer reports the same cumulative NF on the last stage of each chain. The noise floor column uses kT0=−173.975 dBm/HzkT_0 = -173.975\,\mathrm{dBm/Hz}, as Cascade's noise-floor row does. Open the two main cases and change the LNA gain yourself:

In the clip, the LNA's gain goes from 10 to 20 dB and the budget table's cumulative NF at the last stage falls from 3.3 to 1.3 dB, the same 3.34 and 1.30 rounded to the table's one decimal.

The 10 dB chain from the first link in the RF Cascade Analyzer, with the LNA's gain typed up to 20 dB.

How much gain is enough

The rule of thumb: with 20 dB of gain ahead of it, a 10 dB NF second stage adds about 0.3 dB to a 1 dB NF first stage, and with 10 dB ahead it adds about 2.3 dB. The general form is that the second stage fades once G1G_1 is large compared with F2−1F_2 - 1. Gain ahead equal to the second stage's NF plus 10 dB holds its share to about 0.3 dB behind a 1 dB LNA. It breaks when the first stage is itself noisy or lossy, which is the cable row below.

I size LNA gain to that number and stop. The 30 dB row shows why: ten more dB of gain buys 0.27 dB of NF, and it costs the same ten dB of headroom at the second stage, whose input IP3 referred to the antenna drops dB for dB with the gain in front of it. Friis rewards gain for noise and the IP3 cascade charges for it.

What 2 dB of noise figure costs one level up

A receiver's noise floor is −174 dBm/Hz+10log⁡10B+NF-174\,\mathrm{dBm/Hz} + 10\log_{10}B + \mathrm{NF} at 290 K. In 1 MHz that is −174+60=−114-174 + 60 = -114 dBm before the receiver adds anything, so the 10 dB gain chain sits near −110.7 dBm and the 20 dB chain near −112.7 dBm. That is 2.0 dB of sensitivity at the same SNR. The rule holds for a matched source at 290 K, and Cascade's exact kT0kT_0 reads 0.025 dB higher, which nobody will measure.

In free space, received power falls 6 dB per doubling of distance, so 2 dB is about a fifth of the range. A link that closes at 10 km with the 20 dB LNA closes at about 7.9 km with the 10 dB one. That is the price of saving one gain stage, and I'd buy the stage.

A lossy part in front counts in full

A passive part at 290 K has a noise figure equal to its loss. Put it first and Friis counts it at full weight, then divides everything after it by its gain, which is less than one. The last row is the 20 dB chain with 1 dB of cable in front of the LNA: 1.30 dB becomes 2.30 dB. The cable cost a full 1 dB. The whole 10 dB NF second stage cost 0.3. That's why the LNA goes at the antenna, and the filter in front of it is the one loss you budget on purpose.

Where Friis stops

The formula adds available gains, with each stage's NF measured from the source impedance it sees in the chain and a 290 K source. Cascade makes the same assumptions: every stage sits at a matched 50 Ω plane. A real LNA's NF depends on the impedance at its input (its FminF_\mathrm{min}, RnR_n and Γopt\Gamma_\mathrm{opt}), so an LNA behind a 2:1 antenna reads worse than its 50 Ω datasheet by an amount those three numbers set, and the datasheet's noise circles are where to look it up. An antenna looking at cold sky sits well under 290 K, and then system noise temperature, Te=(F−1)⋅290 KT_e = (F - 1) \cdot 290\,\mathrm{K}, is the better unit to budget in.

References

  • H. T. Friis, "Noise Figures of Radio Receivers," Proceedings of the IRE, vol. 32, no. 7, July 1944. The cascade formula and the 290 K reference.
  • Keysight Application Note 57-1, "Fundamentals of RF and Microwave Noise Figure Measurements." Noise factor, noise temperature, and the cascade, with measurement caveats.
  • D. M. Pozar, Microwave Engineering, the noise chapter: cascaded systems and the noise figure of a lossy line.

Questions? Contact

Discussion

No comments yet. Start the thread with a measurement, a correction, or a worked example.