Guys, this really is a solved problem. No need to speculate about cable delays (negligible) or pulse durations (ton4eff uses a 50% duty cycle square wave). It really seems entirely clear to me that the SDG ignores trigger edges which arrive while it is still performing an ongoing sweep.
The first time I tested it even worked with 7.1ms. Now, for some reason it wont - it just jumps forward and backward at his will.
A trigger signal with 7.1 ms period must work as well. In that case, every other trigger is safely ignored because the SDG is still sweeping. So it is equivalent to a 14.2 ms period.
If it "jumps forward and backward", that will be just on your oscilloscope: Because you trigger from the 7.1 ms signal, your scope will sometimes trigger on the active trigger edge, sometimes on the one that gets ignored. Set a trigger holdoff time of e.g. 10 ms on your scope to avoid this.
And I still don't understand why it takes in consideration the sweep signal with external triggering source. It should trigger completely independently, that's why it's "external", right? [...] So, I'm sorry, but for now, I still think that there is a problem...
The one thing I agree with is the head-banging smiley...
The external trigger means that you can determine externally
when a sweep starts, namely at the edge of your trigger signal. But you cannot determine externally
how long the sweep takes -- that is still defined in the SDG2042X. And the SDG is designed to
always complete a sweep once it was started. If another trigger edge arrives while the sweep is still ongoing, it is ignored. After the sweep is completed, the SDG will be ready for the next trigger edge to start the next sweep. It's really quite simple.
There is another way one could implement this, and that is probably what your Rigol generator does: If another trigger arrives while the sweep is still ongoing, stop the current sweep and start a new one immediately. This will avoid missed triggers, but will give you incomplete sweep ranges instead. I think there is no "right" or "wrong" implementation.
Edit: There is a third way, and maybe that is what you are imagining: Lock the internal ramp generator which creates the sweeps to the external trigger signal via a PLL. In that case, the external signal could define the starting time (phase)
and the duration (period) of the sweep. The sweep range (center and span, or min and max frequency) would still be defined in the function generator, but the rate (slope) of the sweep would adapt to the external trigger. I am sure such sweep generators exist, but the SDG is not one of them -- and neither is the Rigol, I believe.