Today, I tried to measure the NF of RF power amplifer with Aglient Spectrum analyzer and Signal generator. I know there is three ways to measure NF,
1. use NF analyzer (I don't have one)
2. Q-factor (I don't have two noise source)
3. Gain method. NF = Pn(tot)@output+174dbm-Gain

The problem as following,
1. The Pn(tot)@output is ~-70dbm[pwr in 0dBm, Gain(RFPout-RFPin):32dB, PA is working(Vcc(3.5v), Vpc(2.2v, 25%DC)), center frequency 1710MHz]. So base on the result, the NF of PA should be ~68dB:-(. it is impossible.

My question are
1. Did I do something wrong during test NF?
2. Did the formula right for cal NF by Gain method?
3. Do I need to set the fixed RBW for Spectrum Analyzer? Is there any relationship between NF and RBW setting for SpecAnalyzer?
4. Could anyone provide me some other ways to test it with my available equipment? as well as any helpful appnote.

Simon Liu

First let me correct one statement:

This should be Y-factor.

I was going to mention that to use the last method (which you call Gain Method) you have to make sure you're most definitely above the Analyzer Noise Floor. But it seems that in your application (PA) you may well be.

With this said, this is how I would go by doing it:

1) Set the analyzer to the proper settings (i.e. ref level, minimum required attenuation, freq span, RBW/VBW). Note that attenuation and frequency span will affect the analyzer noise floor (RBW/VBW will as well, but the Analyzer can take these effects out). So the narrower the freq span the better resolution you'll have and the lower the internal attenuation (this will be set by the incident power into the Analyzer) the better resolution as well.

2) Look at the noise floor before and after transmission. You should see at least a 10dB jump while transmitting. This will indicate that the noise generated by the transmitter (i.e. noise you want to measure) is at least 10dB above the analyzer noise floor.

3) Measure the noise power (Pn) while transmitting. Now you can use your equation, where the thermal noise power is -174dBm/Hz. Which means that you have to either state the measured noise power in a 1Hz BW, or state the thermal noise in the measured BW. To make it clearer, if your measuring BW (i.e. RBW) is 1kHz, the thermal noise will be 30dB higher (i.e. -174dBm/Hz + 10*log(1kHz) = -144dBm). Most Analyzers come with a noise marker which will allow you to measure noise power in a 1Hz BW directly, then you can apply your equation directly.

Note:

1) A notch filter maybe useful to notch transmitter power and reduce level into Analyzer. This will allow to go with minimum attenuation (0dB?) which will give the best Analyzer noise floor (w/o the use of an external LNA).
2) Remember when measuring noise to always list your measuring bandwidth.

Hope this helps.

Thanks

I tried today, it can works.