Attached is an old screenshot showing the phase error for all three Bode Plot channels with Ch.1 being the reference.
We can see that the actual skew is about 0.926 ns (40° phase shift at 120 MHz).
A skew less than 1 ns at a sample clock of 1 GHz for each ADC (500 MHz for each channel) isn't that bad after all - the only question remains why there is no deskew like during normal operation?
Of course I've brought this to Siglent's attention back then and the fact that it isn't fixed yet is either because of the very low priority of the Bode Plot or there are actually some technical reasons that we aren't aware of.
The 2nd screenshot shows the same scenario when viewed in traditional Y-t mode. Here we can see that the skew is well compensated (the remaining skew is due to the setup, not the scope).
Well, quick update.
It looks like Deskew *IS* used in BodePlot II.....
but I have discovered that there are problems with Deskew that made it just appear by chance like it wasn't being used
.
Deskew seems to work fine at 1 ns/div timebase, but is very broken at other timebases -- and those other timebases happen to be what Bode plots use while being generated.
For example:Using the same ADC, e.g. Chans 1+2.
If I have two signals that are about 780 ps skewed, I can go to 1 ns/div timebase and adjust Deskew to get them to match perfectly.
However, when I then change to, say, 50 ns/div the phase measurement suddenly jumps back to what seems to be the orginal phase difference! (The Deskew value is still set) Now, further adjusting the Deskew up or down doesn't affect the skew (visually or auto-measured) at all until 1 ns thresholds are reached -- at which point the skew measurement jumps another full ns to a new value and stays there until the next full ns higher/lower Deskew increment is reached.
If I change the Deskew enough to hit a threshold that it feels like changing the phase measurement at,
*then* it will affect the Bode plot traces.
I see this weird Deskew problem below 50 ns/div as well. 20 and 10 ns/div are easy to see, while at 5 ns/div it is more subtle, but noticeable when changing from 0 to +10ps Deskew.
Hopefully this description makes sense.
I have been making a write-up to break down and describe the different issues I'm seeing, but it's not ready to post tonight (the testing is kind of tedious
). I'll add this to the list.