Read the top line on your first screenshot:
5 Ms/s
So, that's good till 2.5MHz, that is to say meaningless (a minor miracle something vaguely resembling the real signal is shown at all)
If there's no modulation, then not necessarily - it is ultimately limited by the analogue front end.
To put it simply, if I have a 1MHz bandwidth signal on a 1GHz carrier, then I only need to sample at >2MS/s. However, the sampling head has to have useful behaviour at 1GHz.
On paper, yes, no doubts about it, and that's, in fact the principle ETS DSO are based upon, if I'm not mistaken:
if the signal is repetitive, and band limited you can under sample it and obtain a correct reconstruction.
But here we have a signal that varies in time and I can expect the result to depend on interference between the signal
period, the waveform update rate and the scan blind time.
So I would not trust this kind of results.
Am I completely off-track? I admit my digital signal theory is a bit rusty, to say the least, so thanks for any clarification.
ETS works well for simple signals but for complex ones.it falls flat on its face.
The classic case,which I quote
ad nauseum is that of trying to look at one field of a PAL video signal using an early '90s DSO.
To see one field,it is necessary to display 20ms worth of signal on the screen.
At the required time/div settings,the sample rate on these devices reduces to one not quite sufficient to reproduce a 15kHz line rate signal,& totally insufficient to reproduce 4.434MHz colour information,or 5MHz HF luma.
The resulting display is totally unuseable.
In the OP's case,using a far more capable instrument,the DSO sees the signal as a very low frequency one,with a huge range of frequencies superimposed upon it.
It is stretching the bounds of friendship to ask the poor DS1054Z to sample & store at its native maximum sampling rate over such a large time.
It will run out of memory!
It does the obvious thing & reduces the sample rate to one which is still quite reasonable for most signals that would be viewed at this time/div setting.
The anti-aliasing filters don't do anything until the top end of the range,so "funnies" happen after the sweep approaches & then passes 5MHz.
People have checked the frequency response of their DS1054Z's "retro-style" using a series of "spot " frequencies,without the result looking anything like the OP's screenshot.