# noise floor vs. noise figure

#### ismellsmoke

Joined Feb 10, 2009
12
Am i understanding this correctly?? the noise floor would be the sum of all the unwanted signal, and the noise figure would be the ratio of input total noise, to output total noise??
--if i'm wrong here, can anybody explain it to me??

#### Papabravo

Joined Feb 24, 2006
16,509
I think the noise floor is the minimum level at which all the noise is below that level. Signals are above the noise floor and noise is below the noise floor.

Noise figure is a measure of noise energy per unit of bandwidth. It is measured in volts squared per Hertz or equivalently in Volts per root Hertz.

#### davebee

Joined Oct 22, 2008
540
The noise floor can also be dependent on the bandwidth - a narrower bandwidth will have a lower noise floor (at least for thermal noise).

A good example is the GPS signal. In the satellite, a signal is mixed with a pseudo-random series of pulses which spreads the bandwidth of the power of the signal. When that signal reaches the Earth, it's power is below the relatively high "noise floor" based on the spread bandwidth.

But in the receiver, after the signal has been despread by mixing it again with the identical local pseudo-random pulse signal, at the correct time synchronization, the signal is restored to a much more narrow bandwidth that has power well above the much lower "noise floor" based on the narrow bandwidth of the unspread signal.

Noise figure describes how components degrade a signal by adding noise to it. But I think its not just the ratio of noise in to noise out; its the ratio of signal-to-noise in to signal-to-noise out. So if you have two amplifiers that add the same amount of noise, but one amplifies the signal much more, then that one will have a lower noise figure.

#### ismellsmoke

Joined Feb 10, 2009
12
thanks for the replies guys.....i think i've got it now...

#### TechRat

Joined Oct 20, 2010
5
So, when the amplifier says it has a noise figure (NF) of 3dB - how do I figure out what that means to my circuit and my signal?

(I know that's a vague question - I'm not clear on this)

#### Papabravo

Joined Feb 24, 2006
16,509
3 dB is a relative measurement that means something is twice as big as something else or half as big as something else if there is a negative sign because
Rich (BB code):
10 * log ( 2 / 1) = 10 * 0.3010 ≈ 3
Try the wiki for noise figure for a deeper understanding
http://en.wikipedia.org/wiki/Noise_figure

which talks about (NF) in terms of signal to noise ratio.

Try also this wiki which talks about Volts per root Hertz
http://en.wikipedia.org/wiki/Noise_(electronics)

#### golferkid

Joined Oct 22, 2010
2
An Amplifier with a 3 dB Noise Figure would degrade (reduce) a signal's signal-to-noise ratio (SNR) as it amplifies the signal by an amount of 3 dB (reduces by a factor of 2) - that's the definition of Noise Figure - the amount of reduction of SNR through a device. This applies when the noise at the input of that amplifier has not already been amplified above the level of thermal noise, such as would be the case when the amplifier is the first active stage in a receiving system.

#### fffdddooo

Joined Nov 7, 2013
3
If F=SNRi/SNRo then in a 2 blocks cascade system why Ftotal isn't F1*F2 and is instead Ftotal F1+(F2-1)/G1 ???

#### shteii01

Joined Feb 19, 2010
4,644
If F=SNRi/SNRo then in a 2 blocks cascade system why Ftotal isn't F1*F2 and is instead Ftotal F1+(F2-1)/G1 ???
My textbook on Software Radio references: H. T. Friis, "Noise Figure of Radio Receivers," Proceedings of the IRE, vol. 32, pp. 419-422, July 1944.

The actual formula in my textbook for a chain is:
Noise Figure total=1+(NF1-1)+(NF2-1)/G1+(NF3-1)/(G1G2)+(NF4-1)/(G1G2G3)+...

The textbook explains that the very first component (NF1) will have the greatest influence on the NFtotal because all the later components have their noise figure reduced by the gain of the previous components. Also, you want the component with greatest gain as close to the start of the chain as possible because the gain will appear in the denominator of all the later following components, which will reduce their contribution to the NFtotal.

#### fffdddooo

Joined Nov 7, 2013
3
My textbook on Software Radio references: H. T. Friis, "Noise Figure of Radio Receivers," Proceedings of the IRE, vol. 32, pp. 419-422, July 1944.

The actual formula in my textbook for a chain is:
Noise Figure total=1+(NF1-1)+(NF2-1)/G1+(NF3-1)/(G1G2)+(NF4-1)/(G1G2G3)+...

The textbook explains that the very first component (NF1) will have the greatest influence on the NFtotal because all the later components have their noise figure reduced by the gain of the previous components. Also, you want the component with greatest gain as close to the start of the chain as possible because the gain will appear in the denominator of all the later following components, which will reduce their contribution to the NFtotal.
I know, What you say is in every text book about noise, but in every book is that noise factor is F=SNRi/SNRo and NF=10logF so in a chain of blocks the total NF might be F1+F2+F3+...... and it's not and I would like to know why. What is wrong about F=SNRi/SNRo

regards

#### shteii01

Joined Feb 19, 2010
4,644
I know, What you say is in every text book about noise, but in every book is that noise factor is F=SNRi/SNRo and NF=10logF so in a chain of blocks the total NF might be F1+F2+F3+...... and it's not and I would like to know why. What is wrong about F=SNRi/SNRo

regards
You would have to go back to what Friis published. I think that is where all the textbooks got their info.

#### fffdddooo

Joined Nov 7, 2013
3
I think you mean this document by Friis. 1944

I've read it, but (4) says F=SNRi/SNRo so I don't know at what point between (4) and (15) I get lost (by the way, (15) was not supposed to be Fab=Fa+(Fb-1)/Ga ?????? )

regards