Actual effect of lossy component on physical noise power.

[morangie18]

Scenario to help with my question:

A flat broadband noise spectrum well above the thermal noise floor existis. A fixed CW carrier is introduced into the spectrum.

My question:

Assuming a 1Hz measurement bandwidth, the ratio of the carrier to noise level will be called our SNRin. Now introduce an -3dB attenuator, flat across an infinite bandwidth. I think that if I were to physically measure this on a specturm analyzer (unfortunately not an option for me right now) I would expect to see both the noise floor and carrier level drop by 3dB if measuring at the attenuator output. So if both dropped by 3dB, SNRout is the same as my SNRin isn't it?

So how can that -3dB attenuator have a 3dB noise figure if it appears my signal to noise ratio didn't actually change. Assuming no matter what that the carrier will experience a full 3dB drop through the attenuator, I would assume the noise level could then not change at all to achieve the 3dB NF I know that attenuator to have.

I fear the explanation will lead to a discussion on noise temperatures which is hazy for me since I've never understood how to relate the mathematical device of 'noise temperature' to a physical noise power (say in a 1Hz BW for simplicity).

Thanks in advance for straightening me out (if you can).

 

[barry14]

Re: Actual effect of lossy component on physical noise power

It is true that a 3 dB attenuator will attenuate both signal and noise present at its input by the same amount (3 dB). However any noise introduced after the attenuator will now have a 3 dB advantage over the signal, thus the potential drop in signal-to-noise ratio. For example, many advanced short wave radios have a switchable attentuator at their inputs. This is to allow reducing the level of strong signals which can overload the radio's front end. However, for normal strength signals, the attentuator is not used because it reduces the incoming signal level which effectively amplifies the effect of the radio's internal noise (which is not important for very strong signals).
Theory says that any resistance generates noise which, among other variables, varies as the square root of the absolute temperature. For any given noise source, an equivalent noise temperature is the temperature of a resistance equal to the noise source which theoretically generates the same level of noise.