Noise figure measurements with spectrum analyser

[Ghost Tweaker]

Hi everyone,

does someone use spectrum analyser (no aditional noise source) for noise figure measurements? I'm trying to, use -174dBm/Hz noise density and given input level for input SNR and "measure" output noise density and signal. However, NFs are much larger than expected...

Best regard
Ghost

 

[marcomdd]

High uncertainties come from the use of SA as noise figure meter.

 

[Ghost Tweaker]

I agree with u but when I see the price of a noise figure meter, I think that there must be cheaper to evaluate it (not with the same accuracy but...). Therefore I was asking if someone knew a better way...

 

[egemini]

Yes,
I have acurately measured noise figure using spectrum analyser.
I have used R&S FSU 8 spectrum analyser.
DUT had a gain of around 25 dB and used frequency conversion.
We actually measure SNR at DUT out using spectrum and compute NF from that.
Hope you have understood.
thanks

 

[Ghost Tweaker]

I proceed the same way but I don't feel very confidant with it, seems like I'm trying to measure 2dB NF DUTs with an uncertaintity of 4dB...The NFs measured seem excessively high.

 

[maher]

I proceed the same way but I don't feel very confidant with it, seems like I'm trying to measure 2dB NF DUTs with an uncertaintity of 4dB...The NFs measured seem excessively high.

 

[djalli]

For a single element you can make accurate noise figure measurement. As you know if you have multiple elements in board and you try that in spectrum analyzer it means nothing. Devices may compesate each other, etc.

For multiple element board NF measurement is not volid.

 

[vfone]

Check if you can follow the steps shown at the end of this document:
https://zone.ni.com/devzone/conceptd.nsf/webmain/78A610B94390680486256D0B005403E8

 

[Element_115]

You can try the Hot/Cold Resistor method. I forget how to do it, but I'm sure you can do a google search and find the method. I remember we tried it once and the Design Engineer thought the value was too high so we rented a noise figure meter and the meter read a slightly higher value.
Hope this helps.

 

[FANT]

Look at this page:



argument already explained.

Mandi

 

[lguancho]

Hi, SA could be a quick way to evaluate NF of some components such as IF receiver whereby the NF is average about 8 dB or above.

This is how you do it:

[1] Set the SA Centre frequency (CF) to your measurement frequency. Set the Resolution BW to a value lower than the BW of your DUT. Set the averaging mode (if any) ON. Take note of the noise floor of the SA at the CF.

[2] Power up the DUT and connect it to the SA. Remember to terminate the DUT inputs with 50 ohm termination. If DUT is a receiver, make sure all LOs to the DUT are ON. Note the new noise floor on the SA at the CF


Now : Nout = K*T*RBW*F*G [ Linear]

Nout = Measured Noise floor in [2]
KT = -174dBm/Hz
RBW= SA Resolution BW in Hz
G= Gain of DUT ( You should know how to measure it using SA or VNA)

Some manipulation: => NF= Nout -(-174)-10logRBW- Gain [dB]
Whereby Nout is in dBm, Gain is in dB

Regards

 

[egemini]

to lguancho

This method requires the accurate measurement of gain.
If gain is in error by 1 dB , NF will be in error by 1dB. So care should be maximum if you go for this method

 

[sergio mariotti]

Sometime when you least expect it, this subj. appear.
Measuring NF with a SA is technically possible but often very unaccurate.
The question became: do you may accept 1 dB of uncertaity? If yes you may use a SA.
In this case, you may use a Precision SA, be far from the SA noise (add an Hi gain LNA) and set an adeguate video filter to reduce jiitter.

 

[lguancho]

Yes, I agree. NF measurement with SA is not so accurate. But I did mention that when this method should be used. When you are not so concerned about NF accuracy (ie. you know your NF is high!), but somehow you need to perform a measurement to obtain it.

8)

I fully agree Noise figure meter is the best, but what if you don't have it? Or say you need to characterise NF down to 1MHz region? Can your NF meter support it?

 

[Ghost Tweaker]

One last question. I have noticed that the input attenuator value has a great impact on the noise power measured. Should I therefore reduce the attenuation to its minimum value? (attenuation translates into noise figure degradation doesn't it?)

 

[sergio mariotti]

sure!
You should set RF Input attenuator to 0 dB
But 1st check the amount of power... Doing the measurement may be important but the SA safety is more important ;-)

 

[marcomdd]

The question became: do you may accept 1 dB of uncertaity? If yes you may use a SA.

Only 1 dB ?!!????!!???

 

[Mazi3]

it's no problem to do that. just use preamplifier, calibrate stuff somehow and don't forget to add some gain if video filter is used

 

[Ghost Tweaker]

Excuse but it looks like I still have a lot to learn (even though I've already learnt a lot here). What is this video filter compensation adjustement?