14dB ENR noise source for NF measurement of LNA
Thanks.
Your NF of simulation is 2dB, it's very low ,so I think noise source with ENR 14dB is sufficient for your measurement .
I'll measure my LNA on-wafer. Before that, I need to verify the NF and gain measurement using 2 port losses passive device (on wafer also). Ideally, for passive device F=1/G.
The G i measure is similar to S21, but the NF is jumping around, don't follow the relationship F=1/G. Is it due to the ENR? I use noise source with nominal ENR 14dB for measurement.
Thanks again.
The reason can be:
- Noise measurement setup or measuring equip. problem, or
- Your 14 dB ENR source is not precise enough. Do you correct your results with the calibration of the noise source ?
- Probably better is to use an about 5 dB ENR noise source for measuring low noise figures. In this case generally the measured noise uncertainity is less. (in case of about 10dB NF, some .5 dB uncertainity may be accptable, but in th case of 1.5 dB NF it is unacceptable
Signals in the air can affect NF measurment. Is you test setup properly sheilded. I have had problems at 800MHz with NF measurments.
A 14dB ENR is plenty of ratio to make hot/cold NF measurments for an expected 2dB NF, but how much gain do you have? If your noise masurment device has a NF close to the gain of you DUT you will not get reliable measurments.
My LNA has gain around 15dB at 2.4GHz.
[/quote]If your noise masurment device has a NF close to the gain of you DUT you will not get reliable measurments.
NF is the ratio Sig_out/Noise_out over Sig_in/N_thermal.
If the amp input is terminated the output would be G*Noise_thermal+NF, by the above definition.
The hot cold method of NF measurment assumes you don' t know G so two levels of noise are injected to the DUT. Noise_thermal and a hotter noise, 14dB higher in your case. For these two inputs, the two outputs should be
G*Noise_thermal+NF
G*(Noise_enr=14dB) ;assuming G*(Noise_enr=14dB)>>G*Noise_thermal+NF
With the two measurments you can then solve for the two unknowns, G and NF. That is why automatic NF meters often display gain as well. Power_out to Powe_in has a slope of G until it gets close to the NF and then the slope goes to zero.
If you look at the first output, G*Noise_thermal+NF, you would expect this to be -174+15+2=-157dBm/Hz. My HP8563 has a floor of about -150dBm/Hz. What do you have that can measure that accurately? You might have to put a low noise gainstage between your DUT and the Spectrum analyzer. This is what is often done, and you will have to know the characteristics of that amp inorder to extract you measurment.
Good luck
C