It may be interesting to check if other HDO800/900 specimen perform better, especially since the test is rather easy to perform.
On my DHO804, using your parameters (20us/ timebase, 1.25GSa/s) the noise of the 1V/div stopped 50ohm capped signal I can measure (as std-deviation or AC-RMS) is 9.5mV at 20mv/div, 1.1mV at 2mV/div and 260uV at 500uV/div. (I prefer to use std-dev to measure noise, rather than Vpp, since Vpp could be reflecting real tail excursions in the sampled distribution, which can make the signal Vpp noise look "better" in less capable scopes.)
This might point to a bad design. This scope seems to (again like DS1000Z did) derive measurements from screen decimated data. MSO5000 had capability to work from full data, much better design, despite being a bit noisy..
To explain: If you capture data at 1 V/div and do AC RMS or P-P measurement then stop the scope, changing V/div on stopped data
must not change any measurement. Only thing that could change will be number representation (on different ranges it will have different display).
For instance, on different scope that I have, If I capture at 1V/div I get 5,0484 mV AC RMS. If I view that at 10 mv/div I get 5,048367 mV, and at 10V/div I get 5,083 mV RMS. That is with 200 MHz BW. If I go to 500 µV/DIV it show over-range as it should.
If DHOX00 changes value it means that is another thing done really wrong....
@ebastler, it is not visual trace thickness that is wrong but automated measurements.
@TurboTom, time base settings are only one that in theory should change AC RMS of noise (BW changes). Rest of changes are because front end creatively combines ADC, VGA and attenuators for whole range of input sensitivities and not all combinations are best case. But S/N ratio should stay same order of magnitude except for most sensitive ranges....
P.S. did anybody notice in that video how Dave compares how blocky Siglent trace looks despite using 3 Bit ERES? Problem is that Siglent is set to 0.5 Bit ERES not 3 Bit.... If it where 3 Bits it would look much better, if not the same in reduced BW that gives...