performa01, thank you for answering this way.
You were right, the differences came from different dc offsets between the channels.
I run a self-calibration after the firmware update and the differences between rms and stdv reeadings vanished.
So the screenshot is not needed anymore.
I used the SDS2354x to get an exact noise level of 0dBm with 40 MHz bandwidth from an SDG6054X.
To set the exact level of 223.6 mV the Measure -> Tools -> Trend -> Trent-Stdv function of the scope
is very helpful to get a 'quiet' reading.
I adjusted the SDG to for a reading of 223.5 mV average(stdv) from the scope, the SDG showed 256.8 mV output at 50 Ohm then.
I verified the reading of the scope with an SSA3032X-R which should have showed a level in dBm/Hz which is 10 * log(40E6) dB lower
at -76.02 dBm/Hz.
The analyzer readout was -75.85 dBm/Hz (f=10MHz,Zero Span,BW RBW=1kHz,VBW=10Hz. AVG=200) which showes a very satisfying
accuracy of Siglent measurement devices.
With a noise-bandwith of 42 MHz the reading at the SSA was -76.0dBm/Hz!
The SDS2000Xplus scopes have more bandwith and therefore a better crest-factor than the old HP 3400A or 3400B rf-millivoltmeters.
Such a scope is a remunerative investment for hobbyists in many repects.
Can you post a similiar measurement with your SDS2000X-HD?