Who is giving the Rigol vs Siglent noise floor verdict based on the screenshots?
It is not aat all so simple that we can set SA for some settings and then other spectrum and look images and then give simply answer.
There is many things what affect to displayed noise level. It is also important to understand what is noise and how spectrum analyzer work and so on. Basic thing is that original noise is random and chaos. Now if we use say example peak detector, what is mostly as default when we use SA. Because noise is chaotic random freq and random level peaks, it depends how long time we collect data using some limited filter. Spectrum is sweeping (or sequantially fast hopping). More fast sweep and less collected data for every frequency point. So propability to just detect highest peak with this freq spot is low. If stop (or slower speed) and collect more data it takes some amount of time before there exist most high peak. Simply it can say, more slow sweep and more high noise level you see until reach peak max noise with this noise what is there. (it is never theoretical pure random white noise, it exist only in theory books)
Also it need note that different spectrum analyzers may have different BW filter shape. Also if go to more deep also detectors may have differencies.
If I have here class room I can demonstrate it easy. Using different speeds and using different Detectors and different trace draw modes. If start with P-Pk detector and then use Max-Hold. Start and look how this noise level start from low and start rise. Wait a 10s it is more high, go to coffee break and come back it is sure reached top. Just because wo collect more and more data from random noise
Peaks.For random noise level: Peak Detector is (mostly) wrong. But, because mostly for sinusoidal type of signals we use peak detector it is of course important because noise peaks are what we then see in practice. But, we need also know that many things affect what is level what we see. For characterize SA base noise most good is VideoAverage and level setting Log Power. It give same average if use detector mode Sample. But if usse this it may need least 50 time Trace Averaging for reliable tell noise averace. DANL is Displayed(!) Average Noise Level.
But before write one book about noise and SA lets stop it. Real apple to apple comparation between different equipments is easy but also difficult. many things affect some amount. So nitpicking with some 1-3 dB's is waste of time. Until we start doing science and really have instrumets for make science. In many cases I can see Siglent is bit better if look overall noise levels but
I want take here now one example. @egonotto first image.
There is 1MHz RBW, 0-1500MHz range and sweep time 50ms (bit fast, perhaps compromize with accuracy and speed so that speed have more weight)
Hi
attached some pictures from DSA815
As it is very slow i changed sweep speed
Best regards
egonotto
Here same with Siglent. (first image)
But now also, using different sweep speeds and detector mode P-Peak as used also in Rigol.
I have used
24ms, 96ms (UNCAL is for these speeds), 1s and 10s sweep time (over 1500MHz Span)
I have added 434MHz -40dBm signal so that it can also see that SA detect it.
(There read UNCAL because SA normal speed is over ~140ms for this span for this accuracy what designer have wanted to reach when he select nominal speed for default)
In this case UNCAL is because 24ms is really too fast for reliable peak measurements. There exist also nearly blind spots in freq span with this speed but, this is just for demonstrate how speed affect to visible noise level)
Then
There is also other thing . There is two sweep modes in Siglent, Sweep and FFT sweep.
There may be (and also there is) noise level differencies between these two modes.
Example with Sweep mode FFT there is 10Hz RBW and then look noise level using 30Hz RBW with Sweep mode Sweep. RBW 30 show more low noise average than FFT mode RBW 10Hz.