How to use long memory on Rigol DS1052E?

[Lion251]

I received my brand new Rigol DS1052E last week. Firmware 2.05 SP2.
So far, it looks really nice. Intuitive in operation, not too much noise.
But I have a problem with using the 'long memory' option.
I press 'Acquire', and under MemDepth, I select 'Long Mem'.
But when I then take a measurement, I don't know how to access the long memory that has supposedly be stored.
When I scroll with Horizontal - Position or zoom in with Horizontal - Scale buttons, I get to see only a little of my signal, and at a low resolution.
I'm trying to debug a serial communications problem.
Data is at 10 Mbps, and the whole exchange lasts only tens of us.
With 250 Msps (2 channels, long memory), and 1 million points, it should be easy to capture everything in sufficient detail, but I don't seem to know how to.
I have attached two captured images to show what I'm seeing.
Can anyone help me with this?

-- Update: OK, I've found one way to do it: set triggering to 'Single Sweep' instead of 'Normal'.
Not a big problem, just remember to press Run/Stop before you run the experiment.
I have added a new screenshot taken in Single Sweep mode.

[Mechatrommer]

you are capturing at 2us/div (newfile1) and stop... zooming to 200ns (newfile0) and expect it to high fidelity? no, my scope did this too. the only thing you can get like in the (singleshot) is capturing and viewing at the same s/div, 200ns in your case. its digitized resolution issue, cant get rid of it.

[Psi]

from memory.... Set the scope so it triggers at the start of the signal and have the timebase set so it gives a clear image of the individual pulses. Then, once captured press stop and zoom out / move left or right / zoom in etc..

[Lion251]

you are capturing at 2us/div (newfile1) and stop... zooming to 200ns (newfile0) and expect it to high fidelity? no, my scope did this too. the only thing you can get like in the (singleshot) is capturing and viewing at the same s/div, 200ns in your case. its digitized resolution issue, cant get rid of it.

Ehhh.... the 'singleshot' file was taken at 200 us/div(!!!), then zoomed in, so I could see the overview of the whole exchange, then zoomed in to 200 ns/div to look at the individual bytes.  The key was setting the trigger to 'single sweep' instead of 'normal' or 'auto'.

Psi: It appears that, if triggering is set to 'normal' or 'auto', long memory isn't used at all.
I only get it to work in 'single sweep' mode'
If I set the timebase so it gives a clear view of the pulses during capture, and use 'normal' or 'auto' triggering mode, I can zoom out only one extra screen or so.
The pulses are 100 ns, one exchange takes 20 us or so, and two exchanges are approx 2 ms apart.
By starting at 200 us/div, my sample rate is 100 Msamples/sec, so I have 10 data points per pulse.
Then, I can zoom in to the level of singleshot.png, or zoom out so I can see the second exchange as well, and zoom in on that.
This is the way I expected long memory to work, but it would be useful if the manual would mention that it only works in single sweep mde.