Single to split rail converter current draw

[XTEC]

Hi. I'm building a mic pre-amp, and require a split rail from a single rail 12v supply. So I'm using a typical op-amp virtual ground circuit, with a TL071 and a couple small BJT's. My pre-amp circuit will only draw about 20mA standard, worst case <100mA if the output is shorted.

I get a nice clean +5v / -5v with reference to the virtual ground... except the current consumption is huge! The shown circuit is drawing about 100mA with no load. Adding a load makes the current increase as expected (ie; add 10mA load makes the total current consumption 110mA). Needless to say the transistors and regulator are cookin'! Any idea why so much current? We did find one error on the PCB, but have corrected that and installed all new components, and still this big current draw.

I modelled this in LTSpice, and I get nowhere near this sort of current. Hummm.... 

Would much appreciate any suggestions, insight or education anyone can provide. Thank you in advance! 

EDIT: Added PCB layout. Q3 collector and emitter are backwards, we have bodged this on PCB for now.

[XTEC]

This never got a comment, but FYI for anyone in future:
Turns out the op-amp was self-oscillating. See https://electronics.stackexchange.com/questions/557532/why-is-this-buffered-rail-splitter-oscillating-and-how-can-i-compensate-it

We added the 10nF cap between output and inverting input, and a 2K4 resistor in the feedback loop, and now it runs well.

[XTEC]

This:

[MarkT]

The resistor normally goes between the opamp output and transistor emitters, which then act as current dumpers.  A much better circuit is to use a proper current buffer in place of those transistors, or at least to bias them to class B rather than class C as they were in your original circuit.

[Zero999]

+ 1 regarding the resistor in between the op-amp's output and the emitters of the transistors.

That's hardly surprising. Are you even sure you need C1 and C2? I would just put them across R1 and R2.

[Smokey]

I'm not sure if this is relevant here, but higher current "rail splitters" are probably classAB amplifiers and they will always have some quiescent current due to biasing.  The more bias current the lower the crossover distortion will be so sometimes they set it higher than you would expect.