Friis Noise Formula: Why the First Stage Dominates
Each stage after the first adds its excess noise (F − 1) divided by the total gain ahead of it. A 2 dB NF LNA with 20 dB gain ahead of a 10 dB NF stage gives 2.24 dB overall; reverse the order and the chain rises to about 10.03 dB.
Engineer’s note: Everyone learns in school that the first stage, the LNA, sets the system noise figure. What bites in practice is everything in front of it. Each 1 dB of loss between the antenna pad and the LNA input pin, in routing, switches and filters, adds exactly 1 dB to the system noise figure. Defend that stretch hard: shortest path, lowest loss.
For practitioners
Noise factor, noise figure and the 290 K reference
Noise factor F measures how much a two-port degrades signal-to-noise ratio. You compare the SNR at the input with the SNR at the output, with the source held at the reference temperature T0 = 290 K:
A noiseless network has F = 1 (NF = 0 dB). A real one has F > 1. The 290 K reference matters. Noise factor is only defined for a source at T0, so a quoted NF assumes a source whose available noise power is k T0 B. Here k is Boltzmann's constant and B is the noise bandwidth.
The noise a stage adds on its own is described by its excess noise factor, F − 1. All of the cascade analysis comes down to tracking where that excess noise enters the chain and how much gain sits ahead of it.
Deriving the cascade formula
Take two stages with noise factors F1 and F2 and available power gains G1 and G2. Drive the input from a source at T0, which delivers noise power k T0 B.
Stage 1 amplifies the source noise and adds its own. By the definition of F, its total output noise is k T0 B · F1 · G1. This is the source noise times G1, plus the stage's own contribution k T0 B (F1 − 1) G1.
Stage 2 amplifies everything it receives by G2. It also adds its own noise, which referred to its input is k T0 B (F2 − 1). The total output noise is:
The overall noise factor is the output noise divided by the source noise times the total gain, k T0 B G1 G2:
Adding stages repeats the same step. Each new stage's excess noise is referred back to the input by dividing by all of the gain ahead of it:
This is the Friis formula. Every F and G in it is a linear ratio, not a dB value. G is the available power gain of each stage. The structure shows the main result: the first stage's noise factor enters in full, and every later term is scaled down by the gain that comes before it.
Worked example: LNA first, then the order reversed
Use two stages:
- An LNA with NF = 2 dB (F = 1.585) and G = 20 dB (G = 100).
- A second stage with NF = 10 dB (F = 10).
LNA first. The LNA's full noise factor enters, and the second stage's excess noise of 9 is divided by 100:
F = 1.585 + (10 − 1)/100 = 1.585 + 0.09 = 1.675, so NF = 10 log10(1.675) = 2.24 dB.
The 10 dB stage adds only 0.24 dB to the LNA's 2 dB. Of the chain's total excess noise of 0.675, the second stage supplies 0.09, about 13 percent.
Order reversed. Now the 10 dB NF stage comes first, with G = 10 dB (G = 10), and the 2 dB LNA follows:
F = 10 + (1.585 − 1)/10 = 10 + 0.0585 = 10.06, so NF ≈ 10.03 dB.
Both chains use the same parts, yet the results are about 8 dB apart. In each case the first-stage term sets the answer almost alone. The divided term is a correction. It is small when the first stage has a low noise factor and high gain, and it is irrelevant when the first stage is noisy, because nothing later can remove noise that has already been added.
The same result in noise temperature
Equivalent noise temperature often reads more naturally for low-noise front ends:
Since Te is proportional to F − 1, the cascade has the same shape:
For the example, the LNA has Te1 = 290 × 0.585 ≈ 170 K and the second stage has Te2 = 290 × 9 = 2610 K. Behind 100 times of gain, the second stage's 2610 K turns into 26.1 K at the input. The total is about 196 K, which matches 290 × 0.675. Seen this way, the high gain of the first stage turns a noisy later stage into a small additive term.
What the formula assumes, and where it bites
Three practical consequences follow from the derivation.
Passive loss is a stage too. A matched passive loss at T0 has F = L and G = 1/L. Put it ahead of the LNA and it becomes the first stage. Its noise factor enters in full, and its fractional gain then multiplies every later term instead of dividing it. A loss of L dB in front of the LNA adds L dB straight to the system noise figure. The same loss placed after a 20 dB stage adds only (L − 1)/100 to F. The page on front-end loss before the LNA works through this case.
First-stage gain matters, not only first-stage NF. The second term is (F2 − 1)/G1. A low-noise first stage with modest gain still lets a noisy second stage show through. To keep the chain noise figure close to the LNA noise figure, G1 has to be large compared with F2 − 1.
Available gain and source impedance. G is available power gain, and each stage's F is specified for a particular source impedance. In a real chain, stage 2's F depends on the output impedance of stage 1. If interstage matching differs from the conditions behind the data-sheet numbers, the inputs to the formula are wrong even though the formula itself is still exact.
In plain words · for students and beginners
The Friis noise formula in plain words
An everyday analogy, and where it breaks down
Picture recording a quiet speaker. The first microphone preamp adds a little hiss. Everything after it amplifies the voice and that hiss together. Later equipment adds hiss of its own, but by then the voice is already loud, so the extra hiss hardly matters by comparison. The first piece of equipment decides how clean the recording is.
The analogy breaks down in two places. First, amplifying cannot improve the signal-to-noise ratio (SNR, signal power divided by noise power). Gain raises signal and noise by the same amount. Gain only helps by making the next stage's noise small in comparison. Second, in RF the "hiss" is defined against a reference source at 290 K (about room temperature), and the formula works in power ratios, not in loudness as you hear it.
The same example, step by step
Noise figure (NF) is noise factor (F) expressed in dB. The formula needs plain ratios, so convert first.
- LNA (low noise amplifier, the first amplifier after the antenna): NF 2 dB → F = 1.585. Gain 20 dB → 100 times.
- Second stage: NF 10 dB → F = 10.
- Take the second stage's extra noise: F − 1 = 9.
- Divide by the gain ahead of it: 9 / 100 = 0.09.
- Add the LNA's F: 1.585 + 0.09 = 1.675.
- Convert back to dB: 10 log10(1.675) = 2.24 dB.
Now swap them. The 10 dB NF stage comes first, with 10 times gain:
- Its F of 10 goes straight in.
- The LNA's extra noise is 0.585. Divide by 10 to get 0.0585.
- Total: 10.06, which is about 10.03 dB.
Same parts, different order, about 8 dB worse.
Mistakes beginners make
- Adding dB values. 2 dB + 10 dB is not the answer. Convert to ratios, use the formula, and convert back.
- Using F instead of F − 1 for later stages. Only the first stage uses its full F. Later stages contribute only their extra noise.
- Forgetting that loss is a stage. A cable, filter or switch in front of the LNA has a noise factor equal to its loss. A 3 dB loss ahead of the amplifier adds the full 3 dB to the noise figure. The same loss after a 20 dB amplifier hardly shows.
- Ignoring first-stage gain. A quiet LNA with little gain still lets a noisy next stage leak through, because the dividing number is small.
Remember
- The first stage's noise factor counts in full. Each later stage adds (F − 1) divided by all the gain before it.
- Always calculate with linear ratios, never with dB.
- Put the lowest-noise, highest-gain stage first, and keep loss out of the path ahead of it.
Related pages on RF View
Frequently asked questions
What is the Friis formula for cascaded noise figure?
F_total = F1 + (F2 − 1)/G1 + (F3 − 1)/(G1 G2) + ..., with every F and G as a linear ratio and G as available power gain. Convert the result to noise figure with NF = 10 log10(F).
Why does the first stage dominate receiver noise figure?
The first stage's noise factor enters the sum in full. Every later stage's excess noise is divided by the total gain ahead of it, so a first stage with high gain makes later stages count for little.
Can I add noise figures in dB to get the cascade noise figure?
No. Convert each noise figure and gain to a linear ratio, apply the Friis formula, then convert the total back to dB. Adding dB values gives the wrong answer.
How does noise temperature relate to noise factor?
Equivalent noise temperature is Te = T0 (F − 1) with T0 = 290 K. In temperature form the cascade is Te_total = Te1 + Te2/G1 + Te3/(G1 G2) + ..., which carries the same information as the Friis formula.