DIY DAC - Offset issue

[LLL003]

Hi,
(Not sure if this is quite the correct place to post this, but here we go...)
I'm designing a dual 8 bit, discrete DAC. Nothing groundbreaking, just a couple of LM358 s.
I have an output buffer consisting of a LM358 and an IGBT ( this one http://goo.gl/T90Hzv ) tagged on the end as well.
Here's my Schematic -

 (ignore the IGBT symbol, its just got the right footprint)

I'm having an issue where by there is an offset of around 3.6V when all of the input pins are disconnected, on every node ( except Vcc and GND). This seems to be invariant of Vcc, ive tried from 3v upto 15v, all the same. When the input pins of one DAC are pulled to ground, there is still an offset of ~1.2V on the non-inverting input. As far as i was aware, isnt that point supposed to be a virtual ground?

Honestly i have no idea what i've done wrong. I can include some of the test data as well if thats any use to anyone.

Any help would be much appreciated.

[Kleinstein]

It is well possible the circuit is oscillating. Usually there is extra fast feedback around the OP needed. Also the LM358 has quite some internal offset - so something like 10 mV offset is expected from this.

[LLL003]

It is well possible the circuit is oscillating. Usually there is extra fast feedback around the OP needed. Also the LM358 has quite some internal offset - so something like 10 mV offset is expected from this.

I couldn't see any oscillation on the scope (a small amount of noise from my power supply, but nothing major) and there's far more than 10mv offset, that's for sure. about 1.2V at  pin 3 on IC1A in the schematic when the inputs are 0v, and around 3.6V when unconnected.

Is there any way i can check if my OPamp is fried without removing it from the circuit?

[Kleinstein]

At least in the drawing, the inputs of the first OP are reversed. Feed back should go to negative input.
Depending on OP internal offset the output would this way be at its upper limit all the time.

[LLL003]

At least in the drawing, the inputs of the first OP are reversed. Feed back should go to negative input.
Depending on OP internal offset the output would this way be at its upper limit all the time.

Yerp. I see it now...
 
Excuse me whilst i try and hack the thing together XD

[CaveMannDave]

Hi:

Yeah, buddy: SLAM that first OPA stage into whatever rail it first drifts towards. HOOEE

Positive only feedback, much?

Also, don't they make DAC08s, or some more recent equivalent, any longer?

Much tidier.

Cheers,

Dave