Noise Figure / Y-Factor Disagreement Conundrum - Solved
Smorgasbord / Kirt's Cogitations™ #350

<Previous                     Next>

Website visitor Jared Finan contacted me a few days ago to ask whether I had any idea why noise figure and Y-factor measurements made with his HP/Agilent/Keysight 8975A Noise Figure Meter were not in agreement. The same measurements made with his HP/Agilent/Keysight 8970A - a much older piece of equipment - agreed very well. My suggestion turned out to not be the answer. Jared wrote back later saying he found the cause of the problem - a real exercise in troubleshooting! He gave his permission to post our dialog here so in case someone else might have been plagued with a similar issue.

First contact --

Hello Mr. Blattenberger: I am an RF Engineer, and long time RF Cafe visitor (since 2003-ish). I came across excellent NF resources on your website. But, I still have an issue that has been a problem for years and no one at my work has been able to explain or solve it. The issue is a pretty big discrepancy (2-5 dB) differences in NF readings between the 8975A and our 8970A and Y Factor reading. Y Factor and 8970A give NF readings we expect. The 8975A gives abnormally high readings. This only happens when we use a downconverter, I have seen this in V, E, and W bands. If you are willing/interested in shedding some insight from your experience, I'd love to hear it. Thanks and best regards, Jared Finan

My response --

Greetings Jared: I remember using the HP8970 noise figure meter back in the 1980's and 90's both with and w/o the internal downconverter. To be honest, I don't think I ever attempted to correlate NF and Y-factor numbers from it. I am not familiar with the 8975A. You would expect the newer instrument with improved architecture and software would give better readings. My barely-educated guess is that the mixer circuit in the 8975A is injecting excess noise (discrete and/or distributed) into a wideband LO input that, depending on the filtering, could be included in the measurement while not appearing in the output band. I haven't done any engineering work on systems above S-band, but have read that measurements in the realms where you work are very sensitive to setup imperfections (dirty connectors, poor grounding between components, etc.), and have caused guys all manner of headaches trying to hunt down offenders. That's my 1 cent worth (or two cents with inflation of late) without a lot of additional cogitating. Sincerely, Kirt Blattenberger

Jared's response --

Thanks for the reply. I agree with the noise getting into the 8975A mixer idea. We are trying to get Keysight support to help us out, so I hope to figure it out soon. Although personally, I prefer just doing Y Factor with a spectrum analyzer, almost just as fast once you have a spreadsheet set up. Jared

Jared's solution --

Hi Kirt: Just an update on this, I was able to figure out the problem. The 8975A NFA has a 28 VDC source to power the noise source, and it is on at most for 25 ms. The 8970A NF meter is on for close to 100 ms. The problem turned out to be that the 8975A was not powering on the noise source long enough for the analyzer to get a valid noise measurement. I modified our noise sources to always be powered on, and then used the 28 V from the NFA to switch on/off the noise power output, and now the NF readings are within 0.2 dB of the Y factor measurements. I guess Agilent designed the 28 V to work with their noise sources, but others (such as the ones we sell and the Elva-1 models) take longer to reach the peak noise level. Newer analyzers now have a "noise source settling time" that allows change for this, but the old analyzers do not. Anyway, just thought I'd share the solution with you. Take care, Jared

Noise figure measurement is a technique used to quantify the noise performance of radio frequency (RF) systems or devices. The noise figure (NF) is a measure of how much additional noise is added to the signal by the system or device, compared to the theoretical minimum noise that could be added based on physical limits.

There are several methods for measuring noise figure, including the Y-factor method, the cold source method, and the hot/cold load method. The Y-factor method is the most widely used technique for noise figure measurement.

In the Y-factor method, a noise source is connected to the input of the system or device, and the output power is measured using a power meter. The same measurement is repeated with a hot and cold noise source at the same frequency, and the ratio of the output powers with the hot and cold sources is defined as the Y-factor. The noise figure can then be calculated from the Y-factor, the noise temperature of the hot and cold sources, and the system gain.

The cold source method involves measuring the output power of the system or device with a known input signal level and a low temperature noise source at the input. The hot/cold load method involves measuring the output power of the system or device with a known input signal level and either a hot or cold termination at the input.

Noise figure measurement is important for RF system and device designers because it determines the sensitivity and dynamic range of the system or device. A lower noise figure indicates better noise performance, which is important for applications such as satellite communication, radar, and cellular networks where weak signals must be detected in the presence of noise.

Posted March 16, 2023