O.K., I have a dumb question here, and I guess the answer is 'No', but is there such a thing as 'low noise resistors'? It appears from Tim's equation that 'a resistor is a resistor' and noise is independent of actual resistance (I would not have guessed that). How are they getting there phenominally low noise floors on modern LNA's used in GPS and space communication? My only guess is chilling of the input stages to get the best possible SNR at the antenna input and first amplifier stage. Reminds me of the old analog T.V. days where folks had fringe area reception and tried to amplify the signal which gave a stronger signal with stronger noise PLUS added amplifier noise and the result was worse not better. The very reason good amplifiers were attached outdoors right onto the antenna. Thinking about space communications, what is considered the ambient noise floor and how is it noted? In R.F. work I think we couldn't see below -140Db or was that Dbm? That was the noise floor of our instruments and signals of interest needed to be about 6 to 10 Db higher. What would be the absolute noise floor for audio at room temperature? What would be the measurement term DBV? Dbm at 600 ohms? I remember the Telcos used the term DBRN. Db referenced to noise, but I don't know what level they assumed for 'noise'. Thank-you in advance to anyone who can refresh my mind on these terms and measurements!!! Cheers!!
"Low noise resistors": Johnson's noise theorem refers to ideal resistors, and is very good for resistors without DC current flowing through them, and real resistors also have parasitic inductance and capacitance.
For audio and similar purposes, some physical resistors also show "excess noise" due to DC current, which is often called "1/
f" noise (due to its non-flat spectrum at low frequencies), "pink noise" (also spectral), or "flicker noise".
Old-style carbon composition resistors, where the physical path for current is tortuous, have this as a serious problem, and the best ones are bulk foil and good wirewound resistors, where the path is smoother.
This is covered in the Motchenbacher books I cited, and 1/
f noise shows up in semiconductor devices as well.
Noise power and voltage: the theorem states that the available noise power is given simply by
N =
kT x BW. This means that if you connect a resistor
R at temperature
T into a matched load
R, with the load in a cryogenic dewar to reduce its own noise contribution to a negligible level, the power delivered to that load is
kT x BW.
For most circuit analysis purposes, it is easier to model the resistor noise as a voltage source in series with its resistance. Since only half of that voltage appears across the matched load, simple algebra gives us a value
Vnoise2 = 4 x
kT x BW, which is often given as "volts per root Hz".
My mnemonic is that the noise in a 50\$\Omega\$ resistor at room temperature is 0.9 nV/Hz
1/2. You can scale that for other resistances by the square root of the resistance.
If you use a parallel model instead, the noise current decreases with increasing resistance, while the series model's noise increases with increasing resistance (keeping the noise power constant, independent of resistance).
For inductors and capacitors, the reactance does not contribute noise, but the parasitic resistance of the physical part does contribute noise into the series or parallel model of the component.