Differential and single ended noise figure ?

[aware_boy]

differential noise figure measurement

Hi there,

Is there anyone who knows if there is any different between noise figure of a LNA taken as single ended output and differential outputs?

In my measurement, I did a single ended measurement and I would like to convert this single ended noise figure into a differential one.

Any material about this is welcome.

Thanks

 

[gszczesz]

differential amplifier noise nf 3db

Are you saying you had a differential output amplifier and you took 2 noise-figure measurements on each port, while the other one was properly terminated?

If so, then there are 2 extreme situations: the noise is fully uncorrelated between the 2 ports, and the other one is fully correlated.

If it is uncorrelated, then the analysis is as follows: The differential signal voltage doubles, but so does the impedance. This results in a 3dB increase in signal power. For a low impedance output stage, the noise voltage out of each stage stays the same, but the impedance doubles. Hence each noise power goes down by 3dB but combine to keep the net noise power the same. This means the net noise figure is improves by 3dB. On the other hand, if the output stage is high impedance, then the analysis gets a little trickier. Only the differential portion of the current can make it across the load, this lowers the net noise current by 3dB. However, the increase in impedance increases the noise power by 3dB. The net is a wash, and the overall NF still improves by 3dB.

If the noise is correlated, then there are two conditions: fully in phase or out of phase. If the voltages/currents are 180 degree apart, then the noise power is no different from the signal and the NF stays the same.

The problem you will have is that you don't know the degree of correlation, and there is no way of finding out practically (I'm assuming 1GHz up operation). It is most likely that the first few stages in your amplifier are producing the bulk of the noise, and hence the differential output drives the amplified noise 180 degrees out of phase, therefore by far the most likely scenerio is that the final NF will be the same.

 

[vfone]

differential input noise 3db worse

The differential topology has a better immunity to bond wire and other parasitic interconnect inductors, since ideally no signal current flows through them. This results in a higher gain and reduced tendency to unwanted oscillations.
The second advantage is its lower sensitivity to noise in the supply and substrate voltages (this would be hard to show in your simulation).
**broken link removed**

 

[flyhigh]

single-ended rf signal

Hi,

well, I also think that differential topology do not improve noise figure in its basic definition, i.e. it does not improve signal to noise ratio. It is beneficial if implemented in "poluted" environment, i.e. when high-speed digital board is close, or in mixed-signal systems on chip or systems in a package. Than some external signal can interfere with the RF signal and build equivalent noise at RF inputs and output. This is why differential topology offers advantage, this external signal comes as common mode signal and is basically rejected while useful signal comes as differential and is being amplified (in ideal case, when rejection of common mode is high, with S/N ratio degrades only proportional to the noise figure of the single amplifier, and external noise is ignored).

flyhigh

 

[estradasphere]

noise power single ended differential

using a low-loss balun at the output of your amplifier could solve your problem.

 

[aware_boy]

differential single end noise

Thank you all,

Now I have some material to begin my investigation on noise.

 

[aware_boy]

differential noise floor measurement technique

If the differential output signals are really 180 out of phase. Can we say that the differential NF is 3dB lower than the noise we measure at only one ended port of this differential output ?

 

[biff44]

differential stage noise figure

When you have a published noise figure on an amplifier input, it is measured with a differential signal input, ie a noise figure meter and a low loss balun. If you did something like grounding one of the two input terminals, you would have half the signal voltage swing available, but still have the same amplifier input noise floor on BOTH terminals. So, since the signal to noise is worse, your NF should be worse. Don't know if it would be 3 or 6 dB, have to think about it.