to create and rate content, and to follow, bookmark, and share content with other members.
Resolution bandwidth for Noise Figure Measurements:
Question asked by
on Feb 7, 2007
on Feb 15, 2007 by joegorin
Show 0 Likes
How do I decide the correct bandwidth settings for my NFA. My DUT is a band pass filter with a 25 MHz band pass and 1 dB noise figure?
No one else has this question
Mark as assumed answered
This content has been marked as final.
Show 2 comments
(Required, will not be published)
Mar 5, 2007 7:12 PM
The lower bandwidth setting produces the highest uncertainty for low-gain devices (with 0dB DUT attached such as immediately after calibration with the noise source connected directly to the analyzer). Intuitively, the detector receives very low noise power at the lowest bandwidth setting. Therefore, the measurement noise (amplitude jitter) is highest at this bandwidth. The noise power being measured by the analyzer at its highest bandwidth of 4 MHz is 16dB above the noise power measured at 100kHz (4000kHz/100kHz); 10 log (40) =16dB. The detector receives 16dB higher noise power at 4MHz than at 100kHz.
In addition, since the NFA noise figure NF2 generally increases with measurement frequency, the effect of measurement at 100kHz IF bandwidth is multiplied above 3GHz measurement frequency because the analyzer's noise floor is higher.
If you truly need to measure a lossy DUT, then it is important to reduce the system noise figure by using a system LNA.
Low-cost LNA's are available here...
for under $150.
Once the higher noise power is measured out of a higher-gain DUT, the analyzer's uncertainty (amplitude jitter component) is much lower, and more reasonable. It is the same phenomenon in using the DUT amplifier as a system LNA. The analyzer is measuring noise power well above its noise floor, and this is much more stable measurement.
If your DUT has high gain, they do not need a system LNA. But they should ignore the jitter until the DUT is connected.
If you need to have more stable readings while measuring a low-gain or lossy DUT, then they must use a system LNA. This is true when measuring passive mixers for example.
Good characteristics of a system LNA include very low noise figure (1dB or lower) and moderate to high gain (15--25dB). Choosing the most appropriate system LNA depends on the gain or loss of the DUT. Use this tool to determine required system LNA noise figure and gain:
Selection of NFA bandwidth for a given measurement should consider DUT characteristics. Always keep NFA bandwidth lower than DUT bandwidth. For DUT's with narrow bandwidth (below 10MHz bandwidth), or when measuring near the rolloff, consider using NFA bandwidth below 4MHz. Most often the 100kHz bandwidth setting should not be used unless the DUT bandwidth is below 1MHz. I have seen cases where DUT is so narrow-banded that even our 100kHz BW is too large. This results in negative noise figure results. If the NFA bandwidth and DUT bandwidth are nearly the same, frequency accuracy becomes a bigger concern. You never want to measure noise figure on the skirt of a filter. This will result in excessive errors.
Hope the above explanation helps.
Show 0 Likes
Feb 15, 2007 3:33 PM
Let me give a different perspective on the choice of RBW.
RBW choice is a tradeoff, so let's discuss what constrains it from getting too large, and too small.
Effects that make narrow RBWs undesirable
: The variance of a noise measurement depends on the product of the measurement time and the bandwidth. Therefore, if we choose a bandwidth that works acceptably and then reduce it by, for example, a factor of 10, then the measurement time must increase by a factor of ten unless the user is willing to live with more variance. In low Y-factor applications (cases where the noise measured by the analyzer with the noise source on is not much higher than the noise measured with the noise source off), the variance of the noise measurement can be greatly multiplied into the variance of the measured noise figure.
Effects that make wide RBWs undesirable
: If the RBW is wider than the passband of the DUT, AND the "second stage noise" (the noise of the analyzer) is a significant contributor to the measurement (this usually means low Y factor as expressed in the previous paragraph), the use of the wide RBW increases the amount of noise measured during the calibration phase, but does not increase the amount of noise measured from teh noise source/DUT combination. This situation unnecessarily reduces the Y factor and thus unnecessarily increases the uncertainties of the measurement.
: I would recommend using an RBW that is similar to the DUT bandwidth as a good tradeoff between these effects.
About S/N Ratio
: When measuring CW signals, the narrower the RBW gets, the better the signal to noise ratio, and thus the more accurate the measurement can be. When measuring noise figure, reducing the RBW reduces the noise added by the analyzer but similarly reduces the noise of the DUT/noise source combination that we want to measure. So there is no S/N ratio advantage in reducing RBW. Wider RBWs are not usually inherently better or worse than narrow ones, just inherently faster.
Show 0 Likes
Retrieving data ...
Agilent 34970A and Windows 10
measurement with 3458a with power line frequency set a 60Hz but actually 50
What is the confidence level for Keysight DMM specifications?
Small-Signal HB Simulation issues