Advertisement

# Noise Figure One and Two, Friis and IEEE

Click here to go to our original page on noise figure

Click here to go to a "A Note on Noise"

New for February 2019: This page was written by John W McCorkle, who has also written extensively elsewhere, including in peer-reviewed IEEE journals, and who holds many patents. Thanks, John, we really appreciate your expertise and contribution!

There are two entirely different definitions for noise-figure in common use, the original from Friis [ [1] , [2] , [3] ], and the other from the IEEE [ [4] ]. By Friis's definition, noise figure (*NF*) and noise factor (*F*) are measures of degradation of the signal-to-noise ratio ( *SNR*), between the input and output of a component or an entire signal chain. *F* is the ratio of input to output SNR. NF is the number of dB that the SNR has dropped by. By the IEEE's definition, F and NF are not SNR measures, but are measures that are nonlinearly proportional to the noise temperature of a port on a device. They are numbers by which the noise made by an antenna, noise source, amplifier, or a radio receiver can be specified, which can equally be specified with an effective noise temperature. Each definition must be used consistently with its definition. The values cannot be interchanged―the IEEE NF cannot be substituted into Friis's NF = SNR_{in-dB}*−*SNR _{out-dB} formula.

First Definition, From Friis. 2

Second Definition, From IEEE.. 3

In Practice, Does The Difference In Definitions Matter?. 5

*T _{e} For An Attenuator Or Cable. 8 *

Rules For Using The Two Definitions. 9

Appendix 1 - A Walk Through Of The IEEE-Standard Source Document 11

Appendix 2 - Definition of underlying terms. 12

What is equivalent noise bandwidth *B*.. 12

What Is Effective Input Noise Temperature *T _{e}* 13

## Common Nomenclature

In both definitions, the nomenclature used is *F* for "noise factor", which is related to power, and NF for "noise figure", which expresses *F* in dB, so,

Noise Figure =* NF =* 10 log_{10}(*F*) and

Noise Factor = *F* = 10^{NF}^{/10}

A few terms are required to clearly state and understand the noise-figure definitions. The terms are: (1) an equivalent noise bandwidth (*B*) of a block, (2) terms for signals, noise and signal-to-noise ratio (*SNR* ), (3) an effective noise temperature of a source (*T _{s}*) ("s" for source) that drives a block's input (which is defined by the system within which a block is used), and (4) an effective input noise temperature (

*T*) ("e" for effective) of a block. Appendix 2 defines these. If they are familiar, great; if not, jump to Appendix 2 when you need clarification.

_{e}## First Definition, From Friis

In 1942 [1,2] Friis, working at Bell Labs in Holmdel NJ, invented the concept of, and terms, "noise-figure" and "noise-factor". Reference [1] is the original and authoritative source the definition of noise-figure. The Friis definition of Noise Figure (*NF*) will be referred to as*NF _{gen}* ("gen" for general). Fundamentally, Friis's

*NF*"noise-figure" is a function of two variables (two temperatures,

_{gen}*T*and

_{s}*T*) that is a dB metric that answers the question, "If I connect a particular block (like a single transistor amplifier, or an entire complex receiver chain) to a source (like an antenna) that is putting out noise power in addition to the signal, how much worse (lower in dB) will the block's output SNR be, relative to the SNR going into the block? Or in other words, how much worse (lower in dB) will the output SNR be from this particular block relative to an identical but perfectly noiseless block? Friis's

_{e}*NF*"noise-figure" is a system specific metric (i.e. it depends on the application and the SNR coming into that system) allowing different amplifiers and different receiver configurations to be compared based on a metric that captures exactly how many dB the SNR drops by in that system with that incoming SNR.

_{gen}The function of two variables Friis developed is,

*f* (*T _{e}*,

*T*) = 1 +

_{s}*T*

_{e}/T_{s}such that,

Friis's Noise Factor = *F _{gen}* =

*f*(

*T*) = 1 +

_{e},T_{s}*T*

_{e}/T_{s}*=*

*SNR*

_{in}/SNR_{out}*=*(

*s*(

_{in}/*kT*))

_{s}B*/*

*SNR*

_{out}Noise Factor *F _{gen}* is a ratio of power-ratios, and represents the factor by which the

*SNR*is reduced across a device. Since

*F*is a power ratio (not a voltage ratio), the conversion to dB is 10 log (not 20 log), and therefore Friis's noise-figure

_{gen}*NF*can be stated these ways,

_{gen}*NF _{gen}*

*=*10 log

_{10}(

*F*)

_{gen}_{ _ }

*=* *10* log_{10} *(SNR _{in}/SNR_{out})*

*=* *SNR _{in-dB}*

*−*

*SNR*

_{out-dB}_{ _}

Solving for *T _{e}*, when using Friis's

*F*definition, we get,

_{gen}*T _{e}* = (

*F*− 1)

_{gen}*T*

_{s}Below is a pictorial illustrating the derivation of Friis's NF and SNR formula. It is important to grasp that these formulas do not allow one to set *T _{s}* to a particular temperature, like 290°

*K*on one side of the equation (i.e. on the

*F*=1+

*T*/

_{e}*T*side), and at the same time let

_{s}*SNR*or

_{in}*SNR*on the other side of the equation be variables that are allowed to take on an arbitrary values- i.e. values based on a different

_{out}*T*.

_{s}*T*is on both sides of the equation and is buried within both the SNR

_{s}_{in}term and the SNR

_{out}term. All the terms must stay consistent.

Friis's *f*(*T _{e},T_{s}*) function allows engineers to calculate the impact (or degradation) that an arbitrary block has on the signal-to-noise ratio (SNR) of a signal passing through it.

*NF*cannot be applied to a receiver alone or an amplifier alone because the SNR degradation across it depends on what the amplifier or receiver is connected to―i.e. how noisy the source is. In order to compute "

_{gen}*SNR*reduction", the "noise-figure" function

*be a function of the two variables*

__must__*T*, and

_{s}*T*. From these inputs, Friis's noise-figure

_{e}*NF*characterizes how much a block,

_{gen}*, degrades the signal-to-noise ratio (SNR). The same cannot be said of the second definition.*

__as used in the system it is connected to__## Second Definition, From IEEE

On June 11, 1959, the IRE published a standard [4] for measuring and marking devices like amplifiers, with a "noise figure" (which we will call *NF _{std}*), even though the amplifier was

*connected to, or being used in, a system where*

__not__*T*is known and could be factored in. On January 1, 1963, the AIEE and the IRE merged to form the Institute of Electrical and Electronics Engineers (IEEE), and this second definition for "noise-figure" remains an IEEE standard. Reference [4] remains the authoritative source document for the IEEE's noise-figure definition or standard. This IEEE "noise-figure",

_{s}*NF*, is the number you will find listed on a receiver's or amplifier's specification sheet.

_{std}Fundamentally, rather than a function of two variables, *F _{std}* and

*NF*are functions of only one variable, a single temperature in degrees

_{std}*K*, which may be either an effective input noise temperature, or an effective output noise temperature. The function is,

*h* (*T _{e}*) = 1 +

*T*290°

_{e}/*K*=

*F*.

_{std}To get back to an effective noise temperature with the above IEEE definition,

*T _{e}* = (

*F*+ 1) 290°

_{std}*K.*

Appendix 1 walks through the source document [4] to show how the function *h*(*T _{e}*) comes out of its language.

A common use for this IEEE "noise-figure" number, is for it to be listed on a data sheet for, or stamped on, a block like an amplifier, or a receiver, where its effective input noise temperature is simply converted to an IEEE "noise-figure" *NF _{std}* number. Another common use is for it to be stamped on, or listed on a specification sheet for, a 1-port device, like an antenna or a noise source that only has an output. In this case, the effective output noise temperature is simply converted to an IEEE "noise-figure"

*NF*number.

_{std}The Friis definition of noise-figure *NF _{gen}* cannot apply to 1-port devices, because it captures the difference, in dB, between an input

*SNR*and an output

*SNR*, which requires the device being characterized to have an input and an output―as is shown in the above pictorial derivation. But the IEEE definition of "noise-figure"

*NF*can be applied to a 1-port networks because the IEEE's noise-figure does

_{std}*represent an*

__not__*SNR*reduction. Instead, the IEEE definition simply defines

*. It can be applied to blocks with one port (like a noise source or antenna) or blocks with any number of ports, where each port can have its own*

__a forward and backward transform of a temperature__*NF*"noise-figure".

_{std}The IEEE's noise-figure *NF _{std}* carries no more information than simply a temperature―as is evident in the defining function, where

*T*is its lone argument. Laboratory measurements of an amplifier's effective input noise temperature

_{e}*T*and the amplifier's

_{e}*NF*use the exact same steps, apart from the final step of using the

_{std}*h*(

*T*) function to convert

_{e}*T*to

_{e}*F*and

_{std}*NF*.

_{std}Clearly the functions *f* and *h* are not the same―i.e. the two "noise-figures" are not the same. Functionally, and in general,

*h* (*T _{e}*) ≠

*f*(

*T*,

_{e}*T*)

_{s}Unlike the first definition from Friis, the IEEE "noise-figure" *NF _{std}* definition has nothing to do with SNR or a ratio of SNR's.

*NF*is simply a number that is monotonically representative of a noise temperature. It is simply a non-linear mapping function or transform that changes or warps an effective input noise temperature of a device, (

_{std}*T*, with units of temperature in degrees

_{e}*K*), into other numbers that have no fixed meaning relative to an arbitrary system's SNR. The values are

*a dB value representing a ratio of SNR power ratios―which is what*

__not__*NF*is. A 1 dB change in

_{gen}*NF*carries no quantitative meaning regarding how much an amplifier would change the SNR coming out of an arbitrary system―i.e. it does not mean there is a 1 dB change in

_{std}*SNR*like

*NF*does. This fact will be illustrated by way of example later.

_{gen}On systems that happen to have a source with a temperature of*T _{s}* = 290°

*K*, the IEEE's non-linear

*h*(

*T*) mapping makes the values of

_{e}*F*and

_{std}*F*equal at that singular source temperature

_{gen}*T*= 290°

_{s}*K*point. But the formula, or fundamental equation,

*h*(

*T*) = 1 +

_{e}*T*290°

_{e}/*K*=

*F*has nothing to do with SNR reduction. In general,

_{std}*F _{std}* ≠

*F*

_{gen}and as for formulas,

*F _{gen}* =

*SNR*,

_{in}/SNR_{out}but

*F _{std}* ≠

*SNR*.

_{in}/SNR_{out}## Mathematical Language Rigor

The fact that *F _{std}* ≠

*SNR*may surprise many readers given the existence of voluminous literature that begin by saying, correctly, that

_{in}/SNR_{out}*F*=

_{gen}*SNR*and then go on to erroneously say

_{in}/SNR_{out}*F*=

_{std}*SNR*. Let me try to illustrate the problem of starting IEEE noise-figure descriptions with the erroneous

_{in}/SNR_{out}*F*=

_{std}*SNR*formula by way of an example. Suppose

_{in}/SNR_{out}*T*and

_{F}*T*are temperatures in Fahrenheit and Celsius respectively. Mathematical formulas like

_{C}*T*= 1.8

_{F}*T*− 32 or

_{C}*F*=

_{gen}*SNR*do not just imply, but are always and inherently understood to mean that the terms can take on arbitrary values and moreover, that slopes and curvatures and partial derivatives associated with the various terms are accurate. Operations must be applied to both sides of the equations to keep them valid. These concepts are fundamental to mathematical language.

_{in}/SNR_{out}To start a description of IEEE standard noise figure by saying

*F _{std}* =

*SNR*(which is wrong)

_{in}/SNR_{out}and then qualifying that equation by saying that *n _{in}* is equal to

*kT*where

_{ref}B*T*= 290

_{ref}*°*

*K*, is like starting an explanation of the Fahrenheit and Centigrade temperature scales by saying

*T _{F}* =

*T*(which is wrong)

_{C}and then adding qualifying words that *T _{C}* = −40 degrees―where the qualifying words, in both cases, is supposed to make the erroneous equation be ok. Regardless of the qualifying words, the formula

*T*=

_{F}*T*is simply not a valid mathematical formula to convert these temperatures and will most definitely lead users of the formula to make erroneous derivations and calculations.

_{C}Similarly, by fundamental mathematical language, just like the formula*T _{F}* =

*T*is wrong, the formula

_{C}*F*=

_{std}*SNR*is also wrong. Neither of these formulas work when the formula's terms take on arbitrary values. Both are only true at a single point. They express a single unique condition or solution and are not a general formula. Application notes, book chapters, and literature that imply the erroneous

_{in}/SNR_{out}*F*=

_{std}*SNR*formula in a general formula context has lead many trained engineers and physicists as well as laymen to create and use derivative formulas that are erroneous―simply because they assumed that fundamental mathematical language was operative.

_{in}/SNR_{out}*F*≠

_{std}*SNR*because the two definitions,

_{in}/SNR_{out}*f*(

*T*

_{e},*T*) and

_{s}*h*(

*T*), are fundamentally different and cannot be interchanged.

_{e}## In Practice, Does The Difference In Definitions Matter?

Yes―the difference matters in two ways. First, it matters when the wrong formula is used to derive erroneous derivative formulas. Second, the values themselves can be significantly different in practice.

The text in [4], immediately following its definition of *NF* (quoted above), understates the discrepancy between the original Friis *NF _{gen}* and the new

*NF*definition when it states,

_{std}"The standard noise temperature 290°*K* approximates the actual noise temperature of most input terminations. An alternative but related measure of performance useful for very low-noise transducers designed to operate from input terminations with noise temperatures substantially below 290°*K* is the "effective input noise temperature.""

Let's look at the claim that implies *NF _{std}* (which defines

*T*to be 290°

_{s}*K*regardless of reality)

__approximates__*NF*(where

_{gen}*T*is the actual source temperature). In practice, in applications where

_{s}*T*is far from the IEEE's 290°

_{s}*K*the IEEE standard

*NF*stamped on a receiver or an amplifier, or calculated for a component like a lossy cable, is not even close to being a metric of SNR reduction. In low noise applications like satellite communications, radio astronomy, and applications using tiny "active antennas" where the tiny antenna puts out very little energy, a 0.1 dB change in the IEEE noise figure (

*NF*) can result in over 10 dB of change in SNR. That is far from being approximately equal. It is not just low temperatures that are a problem, as suggested in this text in [4], it is a high temperature problem as well. In high temperature applications, like short wave antennas operating at low single digit MHz frequencies, antenna noise temperatures, like from a ¼ wave monopole, exceed 1 million degrees. In this case a 10 dB change in the IEEE noise figure (

_{std}*NF*) can result in less than 0.1dB change in the actual system's SNR. That too is far from being approximately equal.

_{std}In the above practical examples, in one case *F _{std}*<<

*SNR*/

_{in}*SNR*i.e.

_{out}*F*<<

_{std}*F*, while in another case, just the reverse,

_{gen}*F*>>

_{std}*SNR*/

_{in}*SNR*i.e.

_{out}*F*>>

_{std}*F*. Based on these real-world examples, the first sentence in the above statement from [4] is shown to be far from true for many systems. One cannot simply assume or assert that

_{gen}*NF*"approximates"

_{std}*NF*. Often,

_{gen}*NF*and

_{std}*NF*are in fact, dramatically different.

_{gen}Moreover, the text does not highlight or communicate the importance of the fact that while the Friis definition is a function that is fundamentally and exactly, an SNR degradation, contrarily, the new definition is based on a function that is fundamentally related to noise power alone and not SNR.

Continuing reading in the text from [4] above, we have,

"An alternative but related measure of performance useful for very low-noise transducers designed to operate from input terminations with noise temperatures substantially below 290°*K* is the "effective input noise temperature.""

which acknowledges that there is a problem, suggests that the problem matters, and offers a way to deal with the problem. It gives the solution when it says, "An alternative but related measure … is the effective input noise temperature." Indeed, this assertion takes the reader back to the original Friis definition for noise-figure that had been in use for 15 years, and back to the "Friis formula for noise temperature" (described below). This sentence from [4] tells a careful reader that to get accurate SNR-reduction values, they should make their noise calculations based on the actual source temperature *T _{s}* and the

*T*'s and G's of the components in the signal chain. The text is telling readers that to guarantee accuracy, either use

_{e}*T*directly, or convert

_{e}*NF*back to an effective noise temperature

_{std}*T*, and then make all SNR calculations using the "Friis formula for noise temperature" and use Friis's noise-figure

_{e}*NF*formula, which uses

_{gen}*T*and

_{s}*T*and calculates the SNR reduction.

_{e}## Key Formulas

### *T*_{e} For A Signal Chain

_{e}

The total effective input noise *T _{e}* from a series of cascaded devices with gains

*G*and effective noise input temperatures of

_{n}*T*is easily derived as it is simply the solution to an equation for the noise coming out of a string of amplifiers, where the left side of the equation has a single effective noise temperature for the chain, and the right side has the individual temperatures of each stage in the chain, and the gains on the two sides match. Pictorially, it looks like this,

_{n}The equation to be solved is:

*T _{e}G*

_{1}

*G*

_{2}

*G*

_{3}∙∙∙ = [(

*T*

_{1}

*G*

_{1}+

*T*

_{2})

*G*

_{2}+

*T*

_{2}]

*G*

_{3}∙∙∙

The solution is called the "**Friis** **formula** **for** **noise** **temperature**", and is,

*T _{e }= T*

_{1}+

*T*

_{2}

*/G*

_{1}+

*T*

_{3}

*/*(

*G*

_{1}

*G*

_{2}) +

*T*

_{4}

*/*(

*G*

_{1}

*G*

_{2}

*G*

_{3}) + ∙∙∙ +

*T*

_{n}

*/*(

*G*

_{1}

*G*

_{2}

*G*

_{3}∙∙∙

*G*

_{n-1})

This formula is completely general and applies to all blocks and all systems. The terms are simple and well defined. Importantly, there are no hidden assumptions―i.e. it is risk free, it simply always works. The *SNR _{dB}* loss metric (i.e. Friis's

*NF*defined noise-figure) can be computed by using the result of this formula (

_{gen}*T*for the system) together with the actual source temperature

_{e}*T*.

_{s}Easily derived from the above Friis formula for noise temperature, there is also the **"****Friis formula for noise factor"** , which is,

*F _{total}* =

*F*

_{1}+ (

*F*

_{2}−1)

*/G*

_{1}+ (

*F*

_{3}−1)

*/*(

*G*

_{1}

*G*

_{2}) + (

*F*

_{4}−1)

*/*(

*G*

_{1}

*G*

_{2}

*G*

_{3}) + ∙∙∙ + (

*F*

_{n-1}−1)

*/*(

*G*

_{1}

*G*

_{2}

*G*

_{3}∙∙∙

*G*

_{n-1})

Just be aware that this formula has a buried assumption that all the *F*_{n} have been computed using the same source noise temperature *T*_{s}. Because of its buried and potentially illegitimate assumption, the author recommends simply always using the risk free and completely general "Friis formula for noise temperature".

### *T*_{e} For An Attenuator Or Cable

_{e}

An attenuator stage, like a cable, with a physical temperature *T _{a}* (

*a*for ambient) and a gain of

*G*=

*P*(ratio of output power to input power), has an effective input noise temperature of,

_{out}/P_{in}*T* _{e} = *T _{a}* (1/

*G*- 1).

## Example Usage

We will evaluate a system with a series connected antenna, lossy cable, and receiver and look at the SNR impact of the components and from the system. Suppose we have a room temperature (290°*K*) cable with 0.4dB of loss. In this case, *G* = 10^{−} ^{0.4/10} = 0.912, so (1/*G* − 1) = 0.096 and the cable would have an effective temperature *T _{e}* = 290°

*K*(0.096) = 28°

*K*. As such, it would have and IEEE standard

*NF*= 10 log

_{std}_{10}(1+28/290) = 0.4 dB. Note that

*NF*equals the loss if and only if the cable (or attenuator) temperature equals the IEEE standard's 290°

_{std}*K*.

Next we will use the "Friis formula for noise temperature" to look at combinations of components. Since the above 0.4 dB loss cable is the first component in the chain, *T*_{1} = 28°*K* and all the other noise contributions are amplified by the cable loss 1/*G* _{1}, or ~1.1―i.e. relatively speaking, the noise from other components is effectively raised up 10% due to the loss in the cable.

Suppose this cable is in front of a receiver, and the receiver has a* T*_{e} = 10°*K*, then the overall * T _{e}* is

*T _{e} = T*

_{1}+

*T*

_{2}

*/G*

_{1}= 28°

*K*+ 10°

*K*/0.912

*=*39°

*K*

Assuming *T _{s}* is the noise temperature of the source for the system, the SNR degradation caused by the system is,

*F _{gen}* = 1 +

*T*

_{e}/T_{s}Continuing the example, suppose the small antenna in our system has an output noise temperature of *T _{s}* = 2°

*K*, and it connects to the previous 0.4 dB loss room temperature cable, and then to the previous

*T*= 10°

_{e}*K*receiver. Since the system comprised of the cable and receiver has a combined

*T*= 39°

_{e}*K*, the total SNR degradation from the system would be computed as,

*F _{gen}* = 1 +

*T*1 + 39/2 = 20.5, and

_{e}/T_{s}=*NF _{gen}* = 10 log

_{10}(

*F*) = 10 log

_{gen}_{10}(20.5)= 13.1 dB

Without the 0.4 dB loss cable, the SNR degradation from the *T _{e}* = 10°

*K*receiver alone would be reduced to,

*F _{gen}* = 1 +

*T*1 + 10/2 = 5, and

_{e}/T_{s}=*NF _{gen}* = 10 log

_{10}(

*F*) = 10 log

_{gen}_{10}(5) = 7 dB

which is 13.1 − 7 = 6.1 dB less degradation. Note that this 6.1 dB is the SNR reduction caused by adding the cable to the above system, and that this 6.1 dB loss is different from the loss in SNR across the cable alone, which, from the* T _{e}* = 28°

*K*cable, is,

*F _{gen}* = 1 +

*T*1 + 28/2 = 15, and

_{e}/T_{s}=*NF _{gen}* = 10 log

_{10}(

*F*) = 10 log

_{gen}_{10}(15) = 11.8 dB

The effective noise temperature, *T _{co}*, at the cable output, of the

*T*= 28°

_{e}*K*cable is

*T _{co}* = G(

*T*) = 0.912(2°

_{s}+T_{e}*K*+ 28°

*K*) = 27.3°

*K*

The additional SNR reduction caused by the *T _{e}* = 10°

*K*receiver is

*F _{gen}* = 1 +

*T*1 + 10/27.3 = 1.37, and

_{e}/T_{co}=*NF _{gen}* = 10 log

_{10}(

*F*) = 10 log

_{gen}_{10}(1.37) = 1.3 dB

So the total SNR reduction from the cable (11.8 dB) and the receiver (1.3 dB) is

11.8 dB + 1.3 dB = 13.1 dB,

which matches the 13.1 dB we initially calculated for the entire system.

Takeaways from this example/illustration are:

a) The 0.4 dB loss cable, with an*NF _{std}* = 0.4 dB, actually loses

*NF*= 11.8 dB of SNR between its input and its output in this example.

_{gen}b) From (a), *NF _{std}* is grossly erroneous as an SNR loss metric! It is off by 11.4dB!

*NF*is tremendously misleading and of no value. Again, yes, the difference between these definitions matter.

_{std}c) In this particular system, the cable causes an extra 6.1 dB of SNR loss relative to having no cable at all. If I worked hard to reduce the 0.4 dB cable loss in this system (i.e. same receiver), the best possible improvement I could make in the SNR is 6.1 dB, even though the initial SNR loss across the cable is 11.8 dB. In other words, the more I reduce the SNR loss in the cable, the more SNR loss there is in the receiver. Only by improving both the receiver and the cable can the SNR be improved by more than 6.1 dB.

d) If I worked hard to improve the receiver in this system (i.e. with no cable change), the best possible improvement I could make in the SNR is only 1.3 dB. Only if the cable loss were reduced would the way be opened for a receiver improvement to make more than 1.3 dB of SNR improvement.

e) Only* NF _{gen}* provided reliable and accurate information regarding each component's impact to the system's SNR. It exactly captured the SNR reductions caused by the components, and as a result, gave clear insight into which components matter most and how to improve the system.

## Rules For Using The Two Definitions

Rule Summary: __(1) c____onvert all____ __ __NF _{std}__

__values a__

*T*and (2) then use Friis's formulas everywhere_{e}__else for SNR calculations__. Given the existence and common use of the two entirely different definitions for noise-figure, it is vital that the reader (a) keep this fact in mind when reading, and (b) keep values and equation derivations coordinated and consistent. This section gives two simple rules that if followed, will always result in sound calculations. You will ace every test. You will notice these two rules were followed in the above example.

__Rule-1__ , Only use the IEEE standard to do the only thing it can do, to go back and forth between a temperature and an *F _{std}* or

*NF*value―and nothing else

_{std}Anywhere you have a component with an *NF _{std}* , convert it back to a

*T*. And otherwise do not use

_{e}*NF*for anything else.

_{std}*Never* use *NF _{std}* directly in any SNR calculation or to represent an SNR reduction.

__Rule-2__ , Base all SNR degradation calculations on (a) the actual input source noise (*T _{s}*) for the system being analyzed, (b) the component noise temperatures (

*T*'s) and gains (

_{e}*G*'s), and (c)

*Friis's*function to compute

*NF*. In other words, Rule-2 is, always start with temperatures (which is what you have if you have followed Rule-1), if necessary use the "Friis formula for noise temperature" to find the combined

_{gen}*T*for a block to be assessed, then use the system

_{e}*T*, plus the

_{e}*T*driving the block, with Friis's

_{s}*NF*formula, to compute SNR reductions.

_{gen}Corollary - Always follow the rules. Do not be tempted to use special cases or non-general formulas. To reduce risk of errors, it is much safer to never depend on special cases but instead, simply always use the same reliable and generic steps captured in the 2 rules―(1) convert all *NF _{std}* values a

*T*and (2) then use Friis's formulas everywhere.

_{e}## Antenna Noise Figure

Antennas sometimes come with a noise figure specification. It is easy to see how an antenna can have an effective output noise temperature *T _{out}* that would be the source temperature to a receiver connected to the antenna. Whatever noise an antenna picks up appears as some power coming out of its terminals. We can measure the noise coming out of the antenna through a system's equivalent noise bandwidth

*B*and compute an effective noise temperature

*T*such that k

_{out}*T*produces the measured noise power. But how can an antenna have a classical "Friis" noise figure, i.e. an

_{out }B*SNR*−

_{in-dB}*SNR*, or

_{out-dB}*NF*, since:

_{gen}(1) it only has an output;

(2) it does not have a unitless power-ratio gain G that is required by the "Friis formula for noise temperature" (It is not like an amplifier where the gain G is a unitless power ratio―such as 10 watts out per 1 watt in making G=10. An antenna has watts out, per watts/m ^{2} power density impinging upon it, which means the gain number G needed, has units of m^{2}―which is not a simple unitless number); and

(3) It does not have a single gain number. (It does not have the same gain all directions. The gain number varies depending on where the signal is coming from, so there cannot be a single SNR reduction metric that applies to the antenna).

Excellent question. The answer is that it cannot have a classical (Friis)*NF _{gen}* noise figure, but it can have an IEEE

*NF*"noise-figure" because the IEEE standard has nothing to do with an SNR and it also applies to 1-port networks. The antenna's output noise temperature

_{std}*T*is simply converted to an IEEE noise factor and noise figure with the IEEE standard's equations:

_{out}*F _{std}* = 1 +

*T*290°

_{out}/*K*and

*NF _{std}* = 10 log

_{10}(

*F*)

_{std}Given an antenna's* NF _{std}*, a system SNR analysis process starts in the standard way, by converting the antenna's

*NF*back to an effective noise temperature, and then doing all the SNR calculations based on based on Friis's formulas―i.e. the antenna's effective output temperature

_{std}*T*and the

_{out}*T*'s and gains of the signal chain components. In this case, the antenna's

_{e}*T*simply becomes the source noise to the signal chain―i.e.

_{out}*T*

_{s}*=*

*T*.

_{out}## Summary

In 1959 the IRE standards group created an opposing definition for the term noise-figure, a term which was invented by Friis and universally understood to be his SNR metric since its publication in 1944. Errors continue to be made due to having two different definitions in common use at the same time. Given that we have and must live with two different definitions for Noise Figure, the best way to deal with this situation is to clearly illuminate and teach the differences so that usage is kept consistent with each definition. This article is intended to serve this purpose.

The two noise-figure metrics were precisely defined, and rules were given to make sure each metric was used consistently with its definition. An example case was given to illustrate the proper use of both definitions and how different their resulting numbers are. Hopefully this article will help electronics enthusiasts, engineers, and physicists to properly use both definitions and keep their use consistent with their definition.

An effective input noise temperature can be stamped on an amplifier or receiver, or listed on its data sheet to give the exact same information as stamping or listing the IEEE's *NF _{std}* on them. As such, nothing is gained by using the IEEE's noise-figure. On the other hand, much is lost by the IEEE's standard―an unambiguous, unconfused, unique, definition―the historical Friis definition that defines noise figure to be the SNR reduction across a block as used in an arbitrary system, which is not duplicated by any other metric or standard.

## Appendix 1 ― A Walk Through Of The IEEE-Standard Source Document

Reference [4] says,

"The noise factor, at a specified input frequency, is defined as the ratio of 1) the total noise power per unit bandwidth at a corresponding output frequency available at the output port * when* the noise temperature of the input termination is standard (290°

*K*) to 2) that portion of 1) engendered at the input frequency by the input termination."

The use of the terms "output frequency" and "input frequency" are there to allow blocks to use frequency conversion stages like mixers.

The "noise temperature of the input termination" is the source noise *T _{s}* applied to (or going into) the input, and is defined to be

*T*= 290°

_{s}*K*. Literature sometimes refers to this as the reference temperature, or

*T*= 290°

_{ref}*K*.

Assuming the block has a gain of *G*, and an effective input noise temperature of *T _{e}*, the "total noise power per unit bandwidth…available at the output port" [i.e. part "1)…"] is

*G*(

*kT*+

_{s}*kT*) =

_{e}*G*(

*k*290°

*K*+ k

*T*).

_{e}The ratio of "1)…" (i.e. the above) to "2)…" is,

[*G*(*k* 290°*K* + k *T _{e}*)] / [

*Gk*290°

*K*] = 1+

*T*/290°

_{e}*K*

So the IEEE "noise-figure" definition is

*F _{std}* =

*h*(

*T*)= 1+

_{e}*T*/290°

_{e}*K*

And the invers to find *T _{e}* is,

*T _{e}* = (

*F*− 1) 290°

_{std}*K*

## Appendix 2 ― Definition of underlying terms

Definitions of equivalent noise bandwidth (*B*) , SNR, effective noise temperature of a source (*T _{s}*), and effective input noise temperature (

*T*) are provided here.

_{e}### What is equivalent noise bandwidth *B*

The equivalent noise bandwidth *B* in Hz, of a block is defined as follows. First, the block has a transfer function of *H*(*f*), where the maximum value of *H*(*f*) is *H*(*f*) _{max}. This function might have a very steep skirts on a passband, or might be a slow rolloff 2-pole filter. Either way, there is a maximum value of linear power gain *P _{out}*(

*f*)

*/P*(

_{in}*f*) which is

*G*= |

_{max}*H*(

*f*)|

^{2}

_{max}. A block so defined has an equivalent noise bandwidth

*B,*in Hz, of

*B* = (1/|*H*(*f*)|^{2}_{max})∫|*H*( *f*)|^{2}*df,*

where the integration limits cover frequencies from 0 to ∞*.*

When wideband noise with a power of *n* Watts/Hz is applied to the input of this block, using this definition for the noise-bandwidth, the measured output noise power will follow the simple formula *P _{out}* =

*nBG*.

_{max}Another way *B* is defined is as follows. Suppose we substitute for the above block, a "brick wall" vertical skirt filter having a frequency range *f* covering (*f _{0}* −

*B*/2) <

*f*< (

*f*+

_{0}*B*/2)―i.e. a bandwidth that is the same as the noise-bandwidth of the above block. And suppose within this frequency range, the gain of this substituted filter is a flat

*G*

_{max}(i.e. the same as the above block), and the gain is zero outside this range. When wideband noise with a power of

*n*Watts/Hz is applied to the input of this substituted "brick-wall filter", the output noise power will be

*nBG*, same as the block above.

_{max}### SNR Terms

Standard terms representing signals and noise that will be used are

*s _{in}* ,

*n*,

_{in}*s*, and

_{out}*n*

_{out}which represent signal and noise powers going into an input, or coming out of an output, respectively, as limited by the equivalent noise bandwidth ( *B*) of the block.

Using these terms, the input and output signal to noise power ratios (SNR) will be

*SNR _{in}* =

*s*/

_{in}*n*, and

_{in}*SNR*=

_{out}*s*/

_{out}*n*.

_{out}These power ratios are expressed in dB as,

*SNR _{in-dB}* = 10 log

_{10}(

*SNR*) and

_{in}*SNR*= 10 log

_{out-dB}_{10}(

*SNR*).

_{out}### What Is *T*_{s}

_{s}

*T _{s}* is an effective source temperature. The concept of an effective noise temperature is simply that the power of the noise going into or coming out of a circuit node, say

*P*watts, with a noise-bandwidth of

*B*, can be characterized as coming from heat quantified as an effective noise temperature

*T*. In this case, the effective noise temperature

*T*is simply

*T* = *P* / *kB*

where *k* is Boltzmann's constant.

When the effective noise temperature *T* is known, the noise power, *P* watts, can be computed as

*P* = *kTB*.

In any particular system, the energy from a source (like an antenna) driving a block will be some signal power *s _{in}* accompanied by some noise power

*n*.

_{in}*n*is the noise power in watts coming out of the source, going into the input of the block. It is exactly like the

_{in}*P*term above. The power of interest is limited by the noise-bandwidth

*B*of the system, as that is the power that will pass through the system. As such, the source noise (e.g. noise coming out of the antenna) can be characterized by an effective noise temperature, just like the

*T*above, but we call it

*T*, where the subscript s is for "source". The noise power coming out of an antenna over a noise-bandwidth of

_{s}*B*, might be

*n*watts. If

_{in}*n*and

_{in}*B*are known, the effective source temperature

*T*=

_{s}*n*/(

_{in}*kB*). To allow an analysis over arbitrary bandwidths, generally sources are characterized as having a source temperature

*T*such that

_{s}*n*can be calculated as

_{in}*n _{in}* =

*kT*.

_{s}B### What Is Effective Input Noise Temperature *T*_{e}

_{e}

The "effective input noise temperature" *T _{e}* of a block or device under test (DUT) is defined this way. Assume a noise free signal

*s*going into the block and that we measure the output SNR of the block. It is the

_{in}*T*that generates a noise power of

_{e}*kT*that when added to the input signal

_{e}B*s*would make the input SNR,

_{in}*s*/

_{in}*kT*, equal to the measured output SNR. Effectively, all the noise sources that might be within the block are lumped together to form an equivalent "effective input noise" equal to

_{e}B*kT*, where

_{e}B*T*is the "effective input noise temperature" of the block.

_{e}Another way of expressing this concept is to assume the energy driving the block's input is known and comprised of *s _{in} *and

*n*. With these known quantities, the output SNR is measured. In this case, it is the temperature

_{in}*T*such that the block's measured output SNR equals

_{e}*s*/(

_{in}*n*+

_{in}*kT*). The effective input noise from the block,

_{e}B*n*=

_{block}*kT*, is added to

_{e}B*n*so that the output SNR can be computed directly from the input signal level relative to the sum of the noise of the source plus the equivalent input noise of the block.

_{in}[1] . H. T. Friis, "Noise Figures of Radio Receivers," in *Proceedings of the IRE*, vol. 32, no. 7, pp. 419-422, July 1944 https://doi.org/10.1109/JRPROC.1944.232049 . http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=1695024&isnumber=35709

[2] . Friis, H. T. *Seventy Five Years in an Excitng World*. San Francisco: San Francisco Press, 1971

[3] . Friis Biography at IEEE http://www.ieeeghn.org/wiki/index.php/Harald_T._Friis

[4] ."IRE Standards on Methods of Measuring Noise in Linear Twoports, 1959," in *Proceedings of the IRE*, vol. 48, no. 1, pp. 60-68, Jan. 1960. https://doi.org/10.1109/JRPROC.1960.287380 . http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=4065915&isnumber=4065901

**Author : **John W. McCorkle © 14 Jan 2019

Advertisement