AnsweredAssumed Answered

Noise bandwidth / measurement time

Question asked by emielb on Apr 17, 2015
Latest reply on Apr 28, 2015 by Dr_joel

Like • Show 0 Likes0
Comment • 3

What is the noise bandwidth for the MXA N9020A? We use the Average/RMS detector.
Is the Noise marker compensated for the noise bandwidth or the resolution bandwidth?
Does a SCPI command exist to read the noise bandwidth?
What is the actual measurement time for a frequency point? I now assume it to be equal to the sweep time divided by the number of points, but I am wondering if it includes any overhead?

No one else has this question

Outcomes

3 Replies

VANDERWAL Apr 22, 2015 3:57 AM
Mark Correct
Correct Answer
Related to this: We tried to measure the noise bandwidth for various resolution bandwidths. I noticed that the shape of the filter curve changes at the higher bandwidths, so I assume a single overall factor for noise bandwidth correction is not sufficient.

Furthermore, what is noticeable in the attached image is that strange 'noisy nulls' appear on the skirts of the filter. I do not know whether this is expected behavior, but it is not present at other settings. We could rule out the synthesizer, by using multiple brands of synthesizers. Also folding of other intermod products (in the first mixer stage) is not the case, since it is not dependent on the RF frequency. Changing the attenuator setting also does not change anything (except for a noise floor increase), so IMHO it is definitely not an intermod issue.
I do not care about the nulls, but the 'noisiness' seems strange.
Attachments
- Screen_0006.png88.6 KB
Like • Show 0 Likes0
Actions
- Dr_joel @ VANDERWAL on Apr 28, 2015 2:25 AM
  Mark Correct
  Correct Answer
  I can't say for sure but I believe there are different analog filters used in the IF before the digitizer depending on the RBW value chosen. In the narrowest RBW filters the shape is almoste entirely derived from the DSP filter and is essentially perfect (such as Gaussian or Flattop) but in the widest RBW, part of the shape comes from the hardware filter and part from the DSP filter, and so the shape is not numerically perfectly gaussian.
  
  I'm not sure what is causing the noisy nulls. Might be that the widest filter doesn't have as much isolation or I've seen (in VNA digital filters) where when the IF gets close to the ADC sample rate (what is the IF BW of your SA?) then the noise near the Fs/2 can be folded back into the measurement. If the wide IF filter doesn't have extreme isolation at Fs/2 maybe it can cause this, but again I'm not the expert on the SA. You might post back with model, IF BW, and FW rev of your SA to help other experts answer.
  Like • Show 0 Likes0
  Actions
- Dr_joel @ VANDERWAL on Apr 28, 2015 2:27 AM
  Mark Correct
  Correct Answer
  Also, how do you measure this? Do you put a CW signal into the SA input? If so, the "nosiness" of the null might indeed be due to a combination of the DSP filter sidelobs and the hardware filter sidelobes, that can phase in and out of phase to cause a strange response.
  Like • Show 0 Likes0
  Actions

Noise bandwidth / measurement time

Outcomes

Attachments

Related Content

Recommended Content