Hooked the Ultegra DI2 up to the oscilloscope this morning. My scope at home is an old analog Philips, so my possibilities are quite limited compared to the modern digital scopes.
I got some results. Before I connected the scope I attached a shifter to the wire normally going into the junction box. The result, nothing happened. Quite strange, since the rest of the group seems to be a true bus protocol. When I unplug the rear-derailleur and attach the front shifter, I can shift the front-derailleur. Connecting to rear shifters also works flawless.
With the oscilloscope I was able to measure at least a part of the communication. When a button is pressed, the DC level of the bus raises slightly, this doesn't seem to be really important. The interesting part is the signal which is modulated on top of the DC signal. It has a frequency of around 1 MHz and a peak to peak voltage of 0.3V. There is a couple of times a phase shift in this signal. But this signal isn't constant. After something like 30 periods it stops and then the signal is repeated. The repeating happens with a frequency of around a couple kHz. My first though was they use On-Off keying, however the time between the high frequency waves seem to be constant. Second idea was PSK, but whatever button I press I get the same phase shifts.
So I gained some information, but I still don't know how it works. I can do some more measurements (for example with unplugging parts), but then I'd have to take everything to the university to measure with a better oscilloscope. Maybe I'll give it a try to make the group work without the battery holder tomorrow.