L
Admin
[Technical article, rfforum Editorial Team]
The noise floor you see on a spectrum analyzer is not a fixed property of the instrument. It moves with resolution bandwidth (RBW), video bandwidth (VBW), attenuator setting and detector type. Misreading it is a common source of false "spurs" and false "clean" plots.
1. RBW and thermal noise
Inside the IF chain the noise power in the measurement bandwidth is N = kTB · NF, where B is essentially the RBW. In dBm:
DANL ≈ −174 dBm/Hz + NF + 10·log10(RBW_Hz) + other corrections
So narrowing RBW by 10× drops the displayed noise by 10 dB. A spur that is a discrete tone does not drop when you narrow RBW; noise and broadband products do. That is the first test to tell a real CW spur from the noise floor.
2. VBW
Video bandwidth averages the detected envelope. With VBW ≪ RBW the trace looks smoother, but peak readings of noise-like signals fall and sweep time rises. For CW spur hunting, VBW ≈ RBW (or a bit lower) is usually fine. For noise-power measurements, use a wider average or a dedicated noise marker mode rather than reading the "fuzzy" trace by eye.
3. Input attenuator and DANL
Many analyzers auto-couple attenuation to reference level. Adding 10 dB of input attenuation raises the displayed noise by about 10 dB (the analyzer NF referred to the input gets worse by that amount) while a true external CW signal stays at the same absolute level. If a "spur" rises when you add attenuation, it is likely generated inside the analyzer (mixer residual or image), not in your DUT.
4. Detector and trace mode
Sample, peak, RMS and average detectors disagree on noise-like signals by several dB. Peak detector is right for finding the highest CW spur; RMS is right for channel power. Mixing them without noting the mode produces arguments in the lab that are really about the instrument setup.
5. A quick spur checklist
1. Halve the RBW: real CW spur amplitude stays, noise drops 3 dB.
2. Add 10 dB input attenuation: real external spur stays, many internal residuals rise.
3. Change center frequency by a few RBW: confirm it is not an LO feedthrough artifact at a fixed IF offset.
4. Terminate the input with a load: if the spur remains, it is the analyzer.
Discussion: what RBW do you standardize on for production spur searches at microwave frequencies, and do you record attenuator and detector in the test report?
— rfforum 编辑部 / Editorial Team
The noise floor you see on a spectrum analyzer is not a fixed property of the instrument. It moves with resolution bandwidth (RBW), video bandwidth (VBW), attenuator setting and detector type. Misreading it is a common source of false "spurs" and false "clean" plots.
1. RBW and thermal noise
Inside the IF chain the noise power in the measurement bandwidth is N = kTB · NF, where B is essentially the RBW. In dBm:
DANL ≈ −174 dBm/Hz + NF + 10·log10(RBW_Hz) + other corrections
So narrowing RBW by 10× drops the displayed noise by 10 dB. A spur that is a discrete tone does not drop when you narrow RBW; noise and broadband products do. That is the first test to tell a real CW spur from the noise floor.
2. VBW
Video bandwidth averages the detected envelope. With VBW ≪ RBW the trace looks smoother, but peak readings of noise-like signals fall and sweep time rises. For CW spur hunting, VBW ≈ RBW (or a bit lower) is usually fine. For noise-power measurements, use a wider average or a dedicated noise marker mode rather than reading the "fuzzy" trace by eye.
3. Input attenuator and DANL
Many analyzers auto-couple attenuation to reference level. Adding 10 dB of input attenuation raises the displayed noise by about 10 dB (the analyzer NF referred to the input gets worse by that amount) while a true external CW signal stays at the same absolute level. If a "spur" rises when you add attenuation, it is likely generated inside the analyzer (mixer residual or image), not in your DUT.
4. Detector and trace mode
Sample, peak, RMS and average detectors disagree on noise-like signals by several dB. Peak detector is right for finding the highest CW spur; RMS is right for channel power. Mixing them without noting the mode produces arguments in the lab that are really about the instrument setup.
5. A quick spur checklist
1. Halve the RBW: real CW spur amplitude stays, noise drops 3 dB.
2. Add 10 dB input attenuation: real external spur stays, many internal residuals rise.
3. Change center frequency by a few RBW: confirm it is not an LO feedthrough artifact at a fixed IF offset.
4. Terminate the input with a load: if the spur remains, it is the analyzer.
Discussion: what RBW do you standardize on for production spur searches at microwave frequencies, and do you record attenuator and detector in the test report?
— rfforum 编辑部 / Editorial Team
