Memory is not the problem.
Problem is that you have 4GS/s that is not 4GS anymore as soon as you use any timebase longer than 50us/div. And then drops rapidly. If you set timebase for 2ms/div (to look at somethin 20 ms long) it takes nosedive to 40MS/s.. At that timebase, if you have anything higher than 20 MHz in signal it will alias and show weird artefacts in signal that aren't there.
Well, seeing how the actual sample rate ultimately determines the maximum frequency you can reconstruct anyway, wouldn't they use a low-pass filter with a variable upper frequency on the input in order to eliminate the aliasing you're talking about? I mean, this is an issue you'd have even on the SDS2000X series. Once you make the timebase long enough you can't maintain the native sample rate. The additional memory helps, of course, but doesn't eliminate the issue.
Some users who are more familiar with other brands of scopes fear aliasing as its pretty easy to produce. But... its very hard to get the Keysight/Agilent meagzoom scopes to show aliasing artefacts as their front end filtering and decimation "just work" (invisibly and without any controls, prioritising maximum sample rate possible at all times).
You are correct, Keysight uses all kinds of neat tricks, and it kinda deals with some aliasing artefacts. For instance it will silently use Peak detect to extract outside envelope of AM modulated signal that should alias. But that will be only screen display, if you press zoom, zoom will show nonsense. But, still, better than others.
Depends which acquisition mode the scope is set in, as whats on the screen and whats in the waveform acquisition memory are often completely different!
Acquisition Mode | Plotter (screen) | Memory (decimated) |
Normal | Full Bandwidth | Random |
Peak Detect | Full Bandwidth | Min/Max |
High Resolution | Anti-aliased | Anti-aliased |
None of those modes "alias", you're probably recalling zooming in on a signal outside the memory sample rate bandwidth in normal mode where the random decimation makes it look like noise. But the realtime/screen display always shows an accurate picture. The point of the phosphor emulation is to show the full path of the incoming waveform, so of course that includes the min/max points, but its not throwing away all the other points and showing solid blocks (as in the peak detect memory acquisition), it shows the relative likelihood of the waveform in the much desirable (and expensive to produce) 2D histogram.
You're simplifying it down way too much, and trying to minimise the effects with sideways statements like "kinda deals with some aliasing artefacts", perhaps some examples? I've never(!) seen aliasing on those scopes, ever, even when trying to provoke it, so more like "always eliminates aliasing".
Whats important to remember is that the acquisition waveform and screen display are decoupled and not produced from each other.
And if you want to have the screen and/or acquisition antialiased, turn on the high resolution. Nothing silent or hidden or sneaky, all the users choice.