Why ? That is a rather useless test, If I have much RFI here then it will show more noise as in a surroundig where no other gear is powered on.
right, such test (assuming no signal connected to DSO, no external termination nor bnc shielding = open inputs ) is only
usefull when comparing DSOs at same day/location/temp. and even operator. Therefore a compare like "my DSO is having this,
so better than TEK but not such good as Agilent" is complettly useless, unless someone have of of these models from
Agilent app note and have exact same readings on such DSO - then, and only then, a compare to XYZ models make sense.
Btw, your proposal to test with signal is exact that bad as with no signal, as there is no situation (except channel grounded)
with no signal. And when using "defined" signal source you have still lot of differences due location/temp./operator etc.
and of course model.
So whatever you choose it's ok, as long you test more than one DSOs at time and compare results, a single measur of
something done somewhere by somebody at some time is useless, it is like "my dick is longer" (and everybody on earth
know that i have the longest
).
RF-loop: I think you do not understand what they write. They do not use the 50 Ohm open input,
They terminate the the inputs with 50 Ohm.
from the Agilent document :
"Each scope was terminated into 50-? and was set up to acquire waveforms with no input signal connected,
using each scope’s maximum specified sample rate"
hmm, terminated into can be as well "internal 50R termination enabled", but anyway, as said above this didn't really
matter as we can't compare other DSOs to their in Agilent app note.
"using each scope maximum sample rate" is as well trap, however Agilent did it properly and (at least for peak-to-peak measurments)
did it 10 times and used avg vales. But still, this is 2.5 to 10GS/s span and 8 to 32Mpoint of data, you can see here everything
but not what really the noise is (the DSO with 10GS/s and 32M will pickup more peak-to-peak events than a 8M/ 2.5GS/s DSO).
It is as ell important, when comparing this way, to check what is the memory depth at specific timebase when specific sample
rate enabled.
Orange: I did not use avarage. The Hamegs have one of the lowest noise floors.
oh well, Hameg is at least using this as sales argumentation. I give you example of cheap chinese DSO (Tekway/Hantek)
set to not filter anything up to 500MHz (yes, the firmware can be forced to not use any filter), with 250MHz bw at 2mV/DIV.
The signal didn't look bad (except the interleave distortion), not that far from what on Hameg.
But there is small extra difference, you would have to zoom to see it, Hameg is using the same amount on vertical DIV
as this chinese DSO, the resulting line looks thiner because Hameg is drawing half of the dots bit darker.
I like it, it looks "analog" to me.
Agilent/Rigol is using completly different way to draw waveform, their signal can be compared to Hameg
with persistency full on. But here again chinese DSO example, the baseline (not open input but external terminated with 50R)
looks thin. With infinite persistency it looks however thick, with bit of persistency (0.2s) it looks still thicker as without any persistance, and that is what you can see on Rigol/Agilent (intensity grading / shading combination looks similar to other
DSOs with persistency on).
So it's not (only-if any and not only marketing thing) lowest noise floor, but as well question how the UI designer decided
to draw waveform. Therefore it is better to compare Vrms of noise than line thickness.
TO: The whole question "How Thick is Your Baseline on Your Digital Oscilloscope?" with combnation of "i wish to use it as DMM"
is anyway crazy, who is using DSO as DMM? How can i casre about line thickness when the DSO inaccuracy is by 2-3% of full scale?