While I agree that a "rail splitter" could well be useful, the fact that the ones "officially" designated as rail splitters and which which I'm familiar are not sufficiently flexible. OTOH, a power op-amp to generate the pseudo-GND would allow the negative rail to be set a 5 volts below GND and, likewise, the remaining rails to be generated from a zener-based or even voltage divider op-amp as well. While the original query set the total current at ~300 mA or so, I'd guess most of that current would flow into the 3V3 supply.

The TI TLE-whatever rail splitters do exactly that, i.e. they split the input supply into positive and negative rails. If they'd split 12 into 7 and -5 rather than 6 and 6, that might work out better, allowing for some headroom in the positive regulators. With clever op-amp selection, I suspect a really well-regulated, low-noise linear supply would result.

If you, Oliver, know of a splitter that would provide both the low-impedance GND and one useable rail, and it's possible the O/P could live with +/- 6 Volts rather than 5, perhaps that could be made to work.

RE