I'm using an ADA4528 to make a voltage amplification and I'm not able to explain why I have an offset on my AOP. I explain :
I have a voltage divider : 0V5 - 470000 ohms - 8200 ohms - 0V. Without load (ADA4528), the measured voltage on the 8200 ohms resistor is 8.5mV (with tolerances of the resistors it's the expected voltage). Next I load the voltage divider with the ADA4528 in non-inverter configuration (G = 56 / R = 8K2/150R) and the voltage divider ups to 8.8mV. It's a very high error because the signal is digitalized under 16 bits on a range of 4mV to 130mV so a LSB of 192nV. I can't explain this behaviour.
Second test, when I remove the voltage divider to link the positive input directly to the ground, the AOP read an offset (10uV approximatively, the datasheet show a Vos of 4uV max!) and it's worst when the positive input is linked to the ground through a resistor : higher the resistor is, higher the offset is.
The PCB is double side face with one for ground plane.
The copper is visible, it's a home-made PCB.
So, what can cause this offset error ?