Noise Figure Question

[sunwave]

Which Noise Figure is better?

18db@fmhz or 14db@fmhz

 

[dlwtrunked]

Putting "Is smaller or larger noise figure better?" will yield various results on this and "Basically, a lower figure value means the network adds very little noise (good) and a higher noise figure value means it adds a lot of noise (bad). " Google can be a quick way to get answers.
Me: A higher noise figure NF (the 10*log of the noise factor F, thus dB), means a higher noises factor. As the noise factor is SNR_in/SNR_out, that means a higher ratio SNR_in/SNR_out. As SNR_in is assumed the same, for that division to be larger, the number divide by, SNR_out would have to be less, which and that means a smaller "worse" SNR_out.

 

[Ubbe]

Noise figure are the amount of noise that are added, and you probably want as little noise as possible and 0dB are zero noise added.

/Ubbe

 

[dlwtrunked]

Noise figure are the amount of noise that are added, and you probably want as little noise as possible and 0dB are zero noise added.

/Ubbe

 

[dlwtrunked]

Putting "Is smaller or larger noise figure better?" will yield various results on this and "Basically, a lower figure value means the network adds very little noise (good) and a higher noise figure value means it adds a lot of noise (bad). " Google can be a quick way to get answers.
Me: A higher noise figure NF (the 10*log of the noise factor F, thus dB), means a higher noises factor. As the noise factor is SNR_in/SNR_out, that means a higher ratio SNR_in/SNR_out. As SNR_in is assumed the same, for that division to be larger, the number divide by, SNR_out would have to be less, which and that means a smaller "worse" SNR_out.

And of course a 0 dB noise amplifier is impossible (everything has a temperature above absolute zero and that itself generates noise). And adding any amplifier with a noise figure worse (higher) than the current receiver, regardless of the amplifier's gain, worsens the sensitivity to true signals. People often wrongly only think about gain when it is usually more an issue of noise and signal-to-noise.

 

[prcguy]

Which Noise Figure is better?

18db@fmhz or 14db@fmhz

Those are big NF numbers in the wrong direction, are they posted numbers or something made up?

 

[bob550]

Putting "Is smaller or larger noise figure better?" will yield various results on this and "Basically, a lower figure value means the network adds very little noise (good) and a higher noise figure value means it adds a lot of noise (bad). " Google can be a quick way to get answers.
Me: A higher noise figure NF (the 10*log of the noise factor F, thus dB), means a higher noises factor. As the noise factor is SNR_in/SNR_out, that means a higher ratio SNR_in/SNR_out. As SNR_in is assumed the same, for that division to be larger, the number divide by, SNR_out would have to be less, which and that means a smaller "worse" SNR_out.

 

[Ubbe]

When molecules move around they create both heat and noise. When cooling them down they move more slowly and their creation of noise are reduced. Maybe that fmhz are some kind of Fahrenheit to MHz relationship?

Or is it simply the gain figures from different amplifiers without their noise figure being stated.

/Ubbe

 

[dlwtrunked]

When molecules move around they create both heat and noise. When cooling them down they move more slowly and their creation of noise are reduced. Maybe that fmhz are some kind of Fahrenheit to MHz relationship?

Or is it simply the gain figures from different amplifiers without their noise figure being stated.

/Ubbe

(I think he mean frequency in MHz.) Not sure how you mean to define "heat". In ecense, the temperature of the molecules movement causes electromagnetic radiation which for our purpose will noise. Moving molecules (which all do as all in this universe are above absolute zero) generate signals throughout the electromagnetic spectrum. At temperatures that we are living in, that temperature is near 300 K (Kelvin scale used in physics). The peak for thiose temperatures will be in what is called the infrared band (around a wavelength of 8 microns) but there is a probability distribution of the wavelengths (and thus frequencies produced). And that means though the peak was in the infrared band, there are emissions also in the RF bands (though much reduced). Cooling not only reduces the the electromagnetic emissions ("noise") but also moves the peak to longer wavelengths (lower frequencies). And heating has the reverse effect, "red hot" being when heated so that the curve has moved the peak in the visible red wavelengths (frequencies). Note that related conducted heat is something different and is moving molecules loosing energy to nearby molecules causing them to move more. It is for the same reason as cooling an amplifier reduces this thermal noise, that cooling a digital camera will do the same thing. Sensitive astronomical cameras are often cooled and very sensitive infrared camera must be greatly cooled (though there are uncooled infrared camera, they are not near as sensitive).

[Relevant added note: I retired from working 20 years in an infrared imaging laboratory where this was the main focus of the research in which I was involved. So I could not help elaborating.]

 

[RFI-EMI-GUY]

You can use fMhz to determine the noise figure at narrower or wider bandwidth in a logarithmic function. Example 14db@fMHz would equate to 11dB@500 KHz.

Noise figure - Wikipedia

en.wikipedia.org

 

[Ubbe]

You could say that when molecules bump into each other it causes friction and friction produce heat. More movement more heat.

/Ubbe

 

[prcguy]

You can use fMhz to determine the noise figure at narrower or wider bandwidth in a logarithmic function. Example 14db@fMHz would equate to 11dB@500 KHz.

Noise figure - Wikipedia

en.wikipedia.org

You need more than frequency and band width to determine noise figure. Once you determine noise figure then you can calculate a change in noise figure with a change in BW.

 

[RFI-EMI-GUY]

You need more than frequency and band width to determine noise figure. Once you determine noise figure then you can calculate a change in noise figure with a change in BW.

I should have stated "Relative" NF.

 

[xmo]

"You can use fMhz to determine the noise figure at narrower or wider bandwidth in a logarithmic function. Example 14db@fMHz would equate to 11dB@500 KHz."

AND

"You need more than frequency and band width to determine noise figure. Once you determine noise figure then you can calculate a change in noise figure with a change in BW."

No, No, No, No ....

You guys are confusing noise figure with noise power.

For example, thermal noise power is given by N = kTB where k is Boltzmann's constant, T is temperature, and B is bandwidth. Thus, noise power can only be stated correctly as a Spectral Power Density (e.g. -174 dBm/Hz) or as an absolute power in a known or specified bandwidth - for example, the following is an accurate statement: "the thermal noise floor of a receiver having a 12 kHz ENBW is -133.2 dBm.

The relationship between bandwidths is 10 log (BW2/BW1), thus for 12 kHz ENBW, 10 log (12000/1) = 40.8 and 174 + 40.8 = -133.2 dBm

Noise figure, as previously stated, is a measure of how much noise a device such as an amplifier or receiver adds to the noise at the input, thereby degrading the output S/N compared to the input S/N. Noise figure does not depend on bandwidth nor is it measured or expressed in terms of bandwidth.

A device such as an amplifier could have a relatively constant noise figure across a wide range of frequencies whereas another device such as a receiver could have a noise figure that changes significantly with frequency, thus the specific measurement frequency or range of frequencies should be known when considering noise figure.

Noise figure can be estimated by circuit simulation but is normally measured.

 

[prcguy]

"You can use fMhz to determine the noise figure at narrower or wider bandwidth in a logarithmic function. Example 14db@fMHz would equate to 11dB@500 KHz."

AND

"You need more than frequency and band width to determine noise figure. Once you determine noise figure then you can calculate a change in noise figure with a change in BW."

No, No, No, No ....

You guys are confusing noise figure with noise power.

For example, thermal noise power is given by N = kTB where k is Boltzmann's constant, T is temperature, and B is bandwidth. Thus, noise power can only be stated correctly as a Spectral Power Density (e.g. -174 dBm/Hz) or as an absolute power in a known or specified bandwidth - for example, the following is an accurate statement: "the thermal noise floor of a receiver having a 12 kHz ENBW is -133.2 dBm.

The relationship between bandwidths is 10 log (BW2/BW1), thus for 12 kHz ENBW, 10 log (12000/1) = 40.8 and 174 + 40.8 = -133.2 dBm

Noise figure, as previously stated, is a measure of how much noise a device such as an amplifier or receiver adds to the noise at the input, thereby degrading the output S/N compared to the input S/N. Noise figure does not depend on bandwidth nor is it measured or expressed in terms of bandwidth.

A device such as an amplifier could have a relatively constant noise figure across a wide range of frequencies whereas another device such as a receiver could have a noise figure that changes significantly with frequency, thus the specific measurement frequency or range of frequencies should be known when considering noise figure.

Noise figure can be estimated by circuit simulation but is normally measured.

Yes on this. I used to perform noise figure measurements on microwave receivers I build for Hughes Aircraft using the Y factor method. In this case you need a calibrated noise source at the receiver input and spectrum analyzer at the IF output turning the noise source on and off measuring the difference in noise then performing a simple calculation. The noise power must into the spectrum analyzer be at least 10dB above the noise floor of the spectrum analyzer for the results to be accurate. The spectrum analyzer band width has little bearing on the measurement.

 

[dlwtrunked]

"You can use fMhz to determine the noise figure at narrower or wider bandwidth in a logarithmic function. Example 14db@fMHz would equate to 11dB@500 KHz."

AND

"You need more than frequency and band width to determine noise figure. Once you determine noise figure then you can calculate a change in noise figure with a change in BW."

No, No, No, No ....

You guys are confusing noise figure with noise power.

For example, thermal noise power is given by N = kTB where k is Boltzmann's constant, T is temperature, and B is bandwidth. Thus, noise power can only be stated correctly as a Spectral Power Density (e.g. -174 dBm/Hz) or as an absolute power in a known or specified bandwidth - for example, the following is an accurate statement: "the thermal noise floor of a receiver having a 12 kHz ENBW is -133.2 dBm.

The relationship between bandwidths is 10 log (BW2/BW1), thus for 12 kHz ENBW, 10 log (12000/1) = 40.8 and 174 + 40.8 = -133.2 dBm

Noise figure, as previously stated, is a measure of how much noise a device such as an amplifier or receiver adds to the noise at the input, thereby degrading the output S/N compared to the input S/N. Noise figure does not depend on bandwidth nor is it measured or expressed in terms of bandwidth.

A device such as an amplifier could have a relatively constant noise figure across a wide range of frequencies whereas another device such as a receiver could have a noise figure that changes significantly with frequency, thus the specific measurement frequency or range of frequencies should be known when considering noise figure.

Noise figure can be estimated by circuit simulation but is normally measured.

What is being left out that needs clearly emphasized in much of this thread is the fact that *some* noise measurement quantities (including noise figure) are measured *per unit bandwidth* which is why it does not depend on bandwidth. Then if one wants the total noise, one has to multiply by the bandwidth. That is of course seen in the above post as the thermal noise power multiplied kT by B, the bandwidth. But in the above

noise power can only be stated correctly as a Spectral Power Density (e.g. -174 dBm/Hz) or as an absolute power in a known or specified bandwidth

Power is power, not the density of the power.
If you said something "noise figure" at the start, I would agree.---did you mean to say that instead?

 

[xmo]

"did you mean to say that instead?"

No, I meant exactly what I said which is completely factual.

On the other hand, the assertion that noise figure is measured "per unit bandwidth" is erroneous and therefore cannot be supported by citation of a credible reference.

The definition of noise figure is well documented in multiple authoritative references, none of which make any claim that noise figure is measured per unit of bandwidth.

Noise figure is figure is expressed in db, never dB per Hz.

Expressions in dB simply indicate a difference. In the case of noise figure, the difference is the degradation of the signal-to-noise ratio that occurs in a receiver or amplifier due to the noise contribution of the device's circuitry.

dB is often followed by another letter that indicates a reference. For example, expressions in dBm indicate a difference relative to a one milliwatt reference, hence, dBm is an absolute power value.

Noise in the context of noise figure is thermal noise which is uniform across the spectrum. As previously stated, Thermal Noise Power is given by N = kTB where the "B" in kTB is bandwidth, thus to state a noise power, one must express it as the power density per unit bandwidth or the total power in a specified or known bandwidth.

 

[dlwtrunked]

"did you mean to say that instead?"

No, I meant exactly what I said which is completely factual.

On the other hand, the assertion that noise figure is measured "per unit bandwidth" is erroneous and therefore cannot be supported by citation of a credible reference.
...

You can start here:
"The Noise Factor, at a specified input frequency, is defined as the ratio of the total Noise
Power per unit bandwidth available at the output port when noise temperature of the input
termination is standard (290 K) to that portion of engendered at the input frequency by the
input termination." from

http://www.submm.caltech.edu/kids_html/DesignLog/DesignLog179/MillerMUSICReadoutDocs/HEMT%20Power%20Supply/Data%20Sheets/Understanding%20Noise%20Figure.pdf

and Noise Figure is just Noise Factor converted to dB. So you can go argue with Caltech if you argue with what they and I just said.
And I never actually said db per Hz. It would be best here that we talk about Noise Factor. Trying to make English statements after taking the 10*log10 of that to talk about Noise Figures leads to very imprecise English-something I learned long ago as a mathematician. Would you agree that Noise Factor is measured per unit bandwidth or believe the CalTech person is wrong? I hope you realize your kT units are include "per Hz". But even then, some will write that the units for the dB of just that part (before multiplying by the bandwidth) as "dB/Hz" which is understood to actually mean dB of something that is per Hz, though that is a little abuse of the notation but understandable and often used both in the CalTech and other papers.

 

[dlwtrunked]

You could say that when molecules bump into each other it causes friction and friction produce heat. More movement more heat.

/Ubbe

But one could also say that heat is the motion of the molecules causing the bumping. This is like "Which came first, the chicken or the egg?) But then is infrared radiation heat. It warms thing up and is the result of things being above absolute zero. I think heat in essence really means just anything that increases temperature.

 

[dlwtrunked]

You can start here:
"The Noise Factor, at a specified input frequency, is defined as the ratio of the total Noise
Power per unit bandwidth available at the output port when noise temperature of the input
termination is standard (290 K) to that portion of engendered at the input frequency by the
input termination." from

http://www.submm.caltech.edu/kids_html/DesignLog/DesignLog179/MillerMUSICReadoutDocs/HEMT%20Power%20Supply/Data%20Sheets/Understanding%20Noise%20Figure.pdf

and Noise Figure is just Noise Factor converted to dB. So you can go argue with Caltech if you argue with what they and I just said.
And I never actually said db per Hz. It would be best here that we talk about Noise Factor. Trying to make English statements after taking the 10*log10 of that to talk about Noise Figures leads to very imprecise English-something I learned long ago as a mathematician. Would you agree that Noise Factor is measured per unit bandwidth or believe the CalTech person is wrong? I hope you realize your kT units are include "per Hz". But even then, some will write that the units for the dB of just that part (before multiplying by the bandwidth) as "dB/Hz" which is understood to actually mean dB of something that is per Hz, though that is a little abuse of the notation but understandable and often used both in the CalTech and other papers.

I want to add, you say "Noise figure does not depend on bandwidth nor is it measured or expressed in terms of bandwidth." I agree and what I said is necessary for that to be true. By being independent of bandwidth, it MUST be per unit bandwidth.