eblaster, this will never end, because they cherry pick their demonstration cases. For example, here he went for a 25 Ohm signal generator output with settings beyond what most signal generators can do. Furthermore he selected an offset voltage above 2V, because that's the limit that most manufacturers will allow as offset voltage for the mV ranges (though there are easy tricks around that one). Then he selected a 10 mV pulse, well knowing that many signal generators can't generate clean 100 ns pulses to begin with, so what you get is noisy by default. The images he showed is mostly a demonstration of how good his signal generator is, it doesn't actually show the performance of the scope.
What are cherry pick.
Example 25ohm source impedance is standard method for feed signal to probe. When do probe tests it is standard method. And even more, if I simulate for teaching purposes how to look example 3 or 5V rail for its voltage, slow drifting and impulses there... do you know what is then signal source impedance... buuh...
Furhermore.
Offset. There is 1:10 probe !!! and scope set for x10.
3.3V is 330mV in scope input. What scope can not do this offset, even in mV range.
If setting is x1 then example Siglent have +/- 2V limit. 2v x 10 is 20V. This math works also in Siglent... buuh..
Why 10mV signal... just because I want walk around Rigol when it run. (and because this kind of ripple in 3.3V rail under test is not rare at all.
"Then he selected a 10 mV pulse, well knowing that many signal generators can't generate clean 100 ns pulses to begin with, so what you get is noisy by default."
Also 100ns pulse width is not at all unusual miracle, things what we meet in real life with many DUT's frequently. And if can not do pulse 200ns period 50% duty rectangle (square). 5MHz square.. buuh... oh but then... 2MHz square but 20% duty...buuh...
"noisy by default"... I think here scope front end noise is dominating clearly.
And if want low level from cheap generators like example Rigol DG1022. Drive out higher level and use normal external attenuator if you afraid its bottom level noise. Normal practice for low level signals.
This demonstration was made using cheap Siglent SDG1032X using just some BNC T, probe adapter feed thru's (for terminate it). Not even external attenuator. This is not how expensive equipments we have, it is how we use them. Example for detect well below 100nm anomalies in telescope main mirror shape need just one candle and razor blade.