Specifically, the LTC1150, but have also seen it with another similar opamp.
Powered from +/-9V DC-DC converter with a startup ramp of about 15ms.
It is in unity gain inverting configuration (10K resistors) with non inverting input tied to analog ground. Osc input and output floating.
The problem is that after powerup the virtual ground at the inverting terminal is about 7mV away from analog ground. After a few seconds it sometimes corrects itself to zero. Sometimes not, with the resulting output becoming noisy.
We are using this to invert a 2V5 reference, going into a 24bit ADC. Since we are measuring better than 1ppm this is a serious problem.
Has anyone heard of similar behaviour? Or solutions that do not require a redesign and layout of the board?
Best Answer
The answer turns out to be simple. The output of the LTC1150 was not only being used as a reference but as the negative power supply of the ADC itself. It was sucking too much current from the opamp, and the resulting symptom was an offset voltage 1000x worse than it should have been. Unfortunately the LTC1150 data sheet does not indicate max current one can draw from it (other than short circuit current), or the degradation this causes.