In our organization, we have the following specification NOISE FIGURE ANALYSER

*N8973A (10MHz - 3GHz) NOISE FIGURE ANALYSER and NOISE SOURCE is 346B (10MHz-18GHz) nominal ENR 15dB.*

My question is, from the above specification NOISE FIGURE ANALYSE R and NOISE SOURCE. We are able to measure the 0.1 dB or 0.2 dB or 0.3 dB or 0.4 dB or even less than 0.2 dB noise figure. Because We have designed the Low Noise Amplifier (LNA) with 0.2 dB noise figure.

If it is not possible, which number NOISE FIGURE ANALYSER and NOISE SOURCE is used for measuring the Noise figure less than 0.2 dB.

*N8973A (10MHz - 3GHz) NOISE FIGURE ANALYSER and NOISE SOURCE is 346B (10MHz-18GHz) nominal ENR 15dB.*

My question is, from the above specification NOISE FIGURE ANALYSE R and NOISE SOURCE. We are able to measure the 0.1 dB or 0.2 dB or 0.3 dB or 0.4 dB or even less than 0.2 dB noise figure. Because We have designed the Low Noise Amplifier (LNA) with 0.2 dB noise figure.

If it is not possible, which number NOISE FIGURE ANALYSER and NOISE SOURCE is used for measuring the Noise figure less than 0.2 dB.

First and foremost, you must understand that the uncertainty of a noise figure measurement depends on many things, it is not a simple measurement. Here are the key contributors:

1) Noise source uncertainty (if you are using a noise source -- in fact there are other methods than using a noise source to measure noise figure).

This is typically on the order of 0.2 dB for most common noise sources, without a special calibration.

2) Match of the test system, and it's effect on the DUT, through the DUT's noise parameters.

As an LNA designer you surely must know that the noise figure of the DUT depends upon it's noise parameters, and if the match of at the input is not the optimum match (gamma opt) then the noise figure will not be minimum. Most designs try to acheive minimum noise figure at 50 ohms. Most noise sources are NOT 50 ohms, but typically range from 45-55 ohms. So, depending upon the "r-factor" in the noise parameters, the mismatch error of the noise source can pull the input match of the DUT and change it's apparent noise figure. This is one reason why, if you try different noise sources, you will get different noise figure results. The lowest result may not be correct, in that it might be the DUT was initially designed and tested wtih a 45 ohm actual noise source, and if you also use a 45 ohm noise source for verification you will get a low reading. But if you happen to use a noise source that is exactly 50 ohms, it will read slightly higher, but this higher reading might be the correct 50 ohm noise figure. Mismatch effect can be quite high, on the order of 0.2-0.5 dB.

3) Gain of the DUT (or rather, excess noise at the output).

All noise measurements system have their own internal noise. If the the gain+noise figure of the DUT produces excess noise (noise above the kTB level) at the output of the amplfier that is similar to or lower than the noise figure of the NFA system, then there will be significant error and jitter in the measurement. If the DUT is high gain, then the noise power coming out of the amplifier is significantly higher than the NFA input noise and this source of error is not a big effect.

For the best measurement accuracy, in fact you should use a noise parameter based noise figure measurement system. The noise figure application on the PNA-X does in fact perform noise parameter extraction, around 50 ohms, and computes the 50 ohm noise figure from this. This eliminates to a great extend the error #2 above.

For error #1, the PNA-X supports two calibration methods: using a noise source or using a power meter. The power meter method has slightly lower theoretical error based on the fact that the uncertainty of the best power sensors is a little better than the uncertainty of the best noise source.

For error #3, the PNA-X has a built in LNA for low noise (in fact the SA based and NFA based systems do as well) but only the PNA-X uses the noise parameters of it's own receiver to properly de-embed the receiver's noise power from the overall noise measurement. SA and NFA based systems measure the 50 ohm noise power of their receivers, and subtract that from the overall noise, but if the output match of the DUT being measured is not exactly 50 ohms, then by the nature of the noise parameters of the SA or NFA, the noise power created when the DUT is connected is NOT the same as the noise power during calibration, and so the subtraction of the internal noise will not be correct. This error can either add or subtract from the overall noise figure.

In conclusion, in my opition, the PNA-X noise figure measurement application is the best commercially available noise figure measuerment system to determine the 50 ohms noise figure of a DUT, and is comparible to the best methods used at national laboritories (such as NPL and NIST). Here is a reference paper (co-authored by NIST and Keysight including myself) which describes this correlation effort:

http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5755200&tag=1