Noise figure is a single number that tells you how much a device degrades the signal-to-noise ratio of whatever passes through it. Every amplifier, filter, cable, and mixer adds some noise of its own on top of the signal, and noise figure quantifies exactly how much. It’s expressed in decibels (dB), and lower is always better: a perfect, noiseless device would have a noise figure of 0 dB.
Noise Factor vs. Noise Figure
These two terms describe the same thing in different scales. Noise factor (F) is the linear ratio of the signal-to-noise ratio at a device’s input to the signal-to-noise ratio at its output:
F = SNR_input / SNR_output
Because the device always adds some noise, the output SNR is always worse than the input SNR, which means noise factor is always 1 or greater. A noise factor of 1 means the device adds zero noise. A noise factor of 2 means the device doubles the noise power relative to the signal.
Noise figure (NF) is simply noise factor converted to decibels: NF = 10 × log₁₀(F). So a noise factor of 2 becomes a noise figure of 3 dB. Engineers almost always work in dB because it makes cascaded calculations easier, so “noise figure” is the term you’ll encounter most often in datasheets and system specifications.
Why Noise Figure Matters for Receivers
In any radio receiver, radar system, or sensitive measurement instrument, the noise figure of the front end sets a hard floor on how weak a signal the system can detect. The relationship is direct: the receiver sensitivity equation adds noise figure (in dB) to the thermal noise floor to determine the minimum detectable signal level. Every extra dB of noise figure raises that floor by 1 dB, meaning you lose 1 dB of sensitivity. For systems trying to pick up faint signals from satellites, distant cell towers, or deep-space probes, even a fraction of a dB matters.
Noise Figure of Passive Components
For any passive device (a cable, filter, switch, attenuator, or coupler), the noise figure equals its insertion loss. A cable with 2 dB of loss has a noise figure of 2 dB. A bandpass filter with 3.5 dB of insertion loss has a noise figure of 3.5 dB. This is why putting a long cable between an antenna and a receiver directly hurts sensitivity, and why system designers try to place the amplifier as close to the antenna as possible.
The Friis Formula for Cascaded Stages
Real systems chain multiple components together: an antenna connects to a cable, then a low-noise amplifier, then a filter, then a mixer, and so on. The overall noise figure of this chain isn’t simply the sum of each stage’s noise figure. Instead, it follows the Friis cascade equation:
F_total = F₁ + (F₂ − 1)/G₁ + (F₃ − 1)/(G₁ × G₂) + …
Here, F₁, F₂, and F₃ are the noise factors of each stage, and G₁, G₂ are the gains of the preceding stages. The critical insight is that each stage’s noise contribution gets divided by the total gain of everything before it. This means the first stage dominates. If the first amplifier has high gain and low noise, it effectively masks the noise added by everything downstream. If the first component is lossy (like a long cable), it degrades the entire system’s noise performance before the signal ever reaches the amplifier.
This is why low-noise amplifiers are placed at the very front of receiver chains, often mounted directly at the antenna.
Typical Noise Figure Values
Modern low-noise amplifiers (LNAs) achieve remarkably low noise figures. An amplifier is generally classified as “low noise” if its noise figure falls below 3 dB. In practice, the best devices do much better than that:
- Below 3 GHz: Narrowband LNAs can achieve noise figures under 1 dB. Some models reach as low as 0.36 dB at 1.5 GHz.
- Wideband (0.5 to 8 GHz): A typical wideband LNA sits around 1.5 dB, with minimums near 1.3 dB.
- Millimeter-wave (10 to 45 GHz): Noise figures range from about 1.6 dB at 20 GHz up to 5.2 dB at the highest frequencies, reflecting the increasing difficulty of building low-noise circuits as frequency rises.
For comparison, a typical cable or connector might add 0.5 to 2 dB, a bandpass filter 1 to 4 dB, and a mixer 5 to 8 dB. These numbers explain why system designers obsess over the order of components in a signal chain.
Noise Temperature: An Alternative Way to Express It
In satellite communications and radio astronomy, engineers often express noise performance as an equivalent noise temperature in Kelvin rather than a noise figure in dB. The conversion is:
T_noise = 290 × (10^(NF/10) − 1)
A noise figure of 0.5 dB corresponds to a noise temperature of about 35 K. A noise figure of 3 dB corresponds to 290 K, meaning the device adds as much noise as a resistor at room temperature. Noise temperature is preferred in these fields because it makes it easier to add up noise contributions from the antenna, the atmosphere, and the receiver into a single system temperature.
How Noise Figure Is Measured
The most common technique is the Y-factor method. It works by connecting a calibrated noise source to the device under test and switching that source between two known power levels, called “hot” and “cold.” A spectrum analyzer or noise figure meter records the output power at each level. The ratio of those two output readings is the Y factor, and from it, along with the known noise levels of the source, the device’s noise figure can be calculated.
The two biggest sources of measurement uncertainty in this method are the calibration accuracy of the noise source and the precision of the power ratio measurement. For demanding applications, an alternative approach uses a “cold source” technique. One version employs a specially designed diode-based noise generator that produces noise temperatures in the range of 150 to 200 K, well below room temperature, which can be precisely characterized through simple electrical measurements. This tighter calibration reduces measurement uncertainty compared to traditional hot-source methods.
Putting It All Together
Consider a simple satellite TV setup. The dish antenna feeds into an LNB (low-noise block converter) mounted right at the focal point. That LNB contains an LNA with a noise figure around 0.5 to 1 dB and enough gain (typically 50 to 60 dB) to ensure that the long cable run into the house, which might have 5 to 10 dB of loss, barely affects the overall system noise figure. If you moved the LNB indoors and ran the cable first, the system noise figure would jump by 5 to 10 dB, and you’d lose most of your channels.
The same logic applies at every scale, from cell phone base stations to deep-space communication networks to laboratory test equipment. Noise figure gives engineers a universal, comparable number to evaluate how much noise any device adds, and the Friis formula tells them exactly where in the chain to focus their optimization effort.