Conclusion: RAW data is slightly better but using CSV isn't a real problem.
You mean if less than 3 channels are active?
I mean that using the CSV format is just as good as grabbing the data over LAN with DSRemote (or whatever).
The
real problem here is the assumption that you'll get the "best" data by zooming in, switching to dots mode then saving those 15 points of data.
That simply
isn't true.
If you don't understand why that is, watch the video again until you do:
My advice for recording time-aligned pulse inputs on 4 channels simultaneously would be:
Leave the DS1054Z in normal mode (not dots), zoom in, turn sin(x)/x off, then save the 'screen' data. By saving 1200 values you'll avoid the temptation to draw straight lines between sample points in Excel and trying to draw conclusions from that (take note, Wolfie!)
OTOH needing to record 4 time-aligned pulses simultaneously seems an unlikely scenario to me. Much better to use 2 channels (500MSamp./sec., Nyquist at 250Mhz) or 1 channel (Nyquist at 500Mhz, probably perfect reconstruction of everything that gets through the front end).
Why should you turn sin(x)/x off when 4 channels are enabled?Sampling theory tells us that 250MSamples/sec can only reproduce signals up to 125MHz. If there's anything above frequency that in the input signal then we start to see aliasing, Gibbs phenomenon, etc.
A Rigol DS1104Z doesn't have a brickwall filter at 125MHz, it has a -3dB point at about 130MHz (measured) and there's still visible signal well above 300MHz.
These high frequency components will distort the signal, as shown here:
Rigol has been a bit "naughty". When you push the button to turn off sin(x)/x it does
stop doing sin(x)/x, yes, but a
different reconstruction filter is used instead. This filter does a reasonable job of removing Gibbs artifacts from what you see on screen, as seen here:
What it does
NOT do is show the raw sample points. (Why not? Watch the video again....)