Have you tried self-cal?
Yes, and more than once. No effect.
Haven't tried self-cal with PE connected, though, but not sure if it'll help (or rather I expect it to have no effect). Will try this when I make an extension cord to connect to mains earth.
If you "gnd" the probe input in the app probe-setting, still an offset?
Yes, about one pixel above zero. And when I rotate the vertical scale encoder, it jumps to exactly zero when it switches and stays there for a fraction of second, then moves back one pixel above.
What if you put a terminator on the BNC?
That's what I tried it with. Not really a terminator (don't have one), just a 51 Ohm resistor across the input BNC.
Using the PE cable to connect chassis to the earth ground directly reduces noise, but does nothing to the offset.
Same behavior on Ch2.
Here's what I see on the screen (probe set to 1x in channel settings, 51 Ohm resistor in the input jack).
1. 10mV, below zero.
2. 5mV, above zero this time.
(5mV to 500uV -- pretty much the same up and down, never exactly zero)
3. 500uV
4. 200uV, the offset is very clear now
5. One clockwise step of the vertical scale encoder -- still 200uV, but now the offset becomes negative!
This offset might at least be understandable, but what the heck happens when I switch from 200uV to 200uV? It's supposed to do nothing in this case, and yet there is something mysterious going on under the hood. That's why I suspect some kind of firmware bug here.
Of course it must be made clear that it's a 804 which thinks that it's a 924. This may lead to some side effects. That's why I wonder if anyone else can reproduce this with the same setup.