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