Hi @rf-loop and @tautech,
I've been playing with the previous firmware, see the picture old_fw, there you have a few chars @ 115200, horizontal is set to 50ms/div and sample rate was 2MSa/s (*), that gives us 2e6(Sa/s)*50e-6(s) = 100 Sa/div (in the zoomed window @ 50µs/div). At that baud rate there are less than 6 bit-slots per div (again looking at the zoomed window), and that means slightly more than 16 samples per bit: that ought to be enough to decode properly (it was enough 40 years ago with a 6551), but it isn't.
WRT the new fw update, please see the picture new_fw, horizontal was set 10 ms/div, sample rate was 10MSa/s, so we've got 5 times more samples per div, or about 10e6(Sa/s)*50e-6(s)/6= 83.3 samples per bit, but again it isn't decoding properly.
So there is something wrong with the decode algorithm they are using, or there are still bugs somewhere.
I'm returning this scope, but I were the owner of an SDS1000X I wouldn't be happy at all if in order to "fix" a serial decoder bug, the solution was to remove functionality as in shutting down for good the serial decoder above 20 ms/div.
I have tested this with 2 different serial sources, and the Agilent next to the Siglent has been able to decode everything always, so I don't think the problem is in the serial streams. I don't have the Rigol 1000Z at hand right now, but I found no problems whatsoever when it was here, doing the same.
You or tautech or somebody should escalate this, ain't chooch, not skookum!
Thanks,