Noise figure measurements

[Leiser]

Hi all,

I am doing some noise figure measurements for LNAs. I am using a network anylizer, ESA4402 with the noise figure suite installed and a noise figure source, the agilent 346A with around 5.5dB ENR.

Well, I have experienced regularly that after doing the calibration, just connectig the noise source straight to the ESA, the noise figure plotted oscillates between -1 and 1dB the worse case.

I would like to know if this is the accuary expected when using these equipments to measure noise figure.

I contacted Agilent and checked the specs for the uncertainty of the measurement when using these equipment, but still the results are out of that. (I played with the BW, internal att, etc)

Does anyone have experience measureng NF with this setup? what is the best results you get after doing a calibration?

Thanks!

 

[Skogsgurra]

I do not know about your application, but in general, the uncertainty is rather big.

The standard deviations expected according to CISPR/A291/CDV are: 4.5 dB in the 30 - 300 MHz range and 5.2 dB in the 300 - 1000 MHz range.


An audit measurement carried out in Germany, where around 100 laboratories took part, shows that measurements differ a lot even when the devices, measurement conditions and measurement set-up are identical.

The standard deviation for 1589 measurements was 7.2 dB. Assuming standard distribution, that means that around 60 of the measurements were 14-15 dB or more off the mean (and probably "most correct") value.

I cannot help thinking about this when EMC labs hunt tenths of decibels. Either to make or break a device.

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[VEBill]

We have a Noise Figure meter here which does the job. Simply set it up, connect the device under test and read the results. Since the basic method of determining noise figures is just measuring and comparing amplitudes (noise generator on and off), there must be a problem in the technique if you can't get reliable and repeatable results.


"standard deviations ... 5.2 dB in the 300 - 1000 MHz range"

Don't modern UHF 'Low Noise Amplifiers' have NFs of significantly less than 1.0 dB these days (like 0.3dB)?

This sort of limits the expected variability of such measurments to a tiny fraction of 1 dB.

 

[Leiser]

Hi again,

yes, the LNA I´m measuring has 1dB noise figure so I´m a little confuse about the 5.2dB deviation in the 300-1000MHz range.

I am using the Y-method to measure NF but with an automatic suite from the SA.

Then my question is how do people measure noise figures of amplifiers below 1dB if the calibration itself has an uncertainty of +-1dB???

I am not sure if I´m doing something wrong, or it is just that measuring NF with a SA has this kind of uncertainty and to measure NFs below 1dB an NFA (noise figure analyzer) must be used???

Thanks again

 

[Skogsgurra]

I guess that I misunderstood the noise you are measuring. I referred to EMC measurements. Another thing, obviously.

Would LNA stand for Low Noise Amplifier?

Gunnar Englund

http://www.gke.org

--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...

 

[Leiser]

Hi,

yes, that´s correct. I am measuaring low noise amplifiers for radio receivers so typical noise figure values are less than 3dBs.

 

[bhawkins]

while i haven't used that exact equipment, i'd expect an overall accuracy of +/- 0.1db

bob hawkins

http://www.smelectronicsllc.com