PGA 10x, weird behavior when working near top rail and grounds connected

[dusan]

I'm using PGA specifically AD8253 (https://www.analog.com/media/en/technical-documentation/data-sheets/ad8253.pdf). It is powered by +/-12V. When I measure voltage near upper rail like in this image and gain is 10x, everything works fine:


   
    R1 has 0.9V on it, PGA gain is 10x, battery ground is not connected anywhere. The PGA output has 0.9 * 10 = 9.0V, as expected.
    But when I connect battery ground to analog ground, like this:


   
    Then the output of PGA drops from 9.0V to 3.88V. I get the same behavior (3.88V on output) when I power the resistor divider from +12V rail of PGA (which is what I normally do, I don't use battery, I only used battery to show that 10x gain works). Same behavior on the negative rail. It stops being problem below 7.5V.

    Here's actual schematic:
   

   
What's causing it and how to solve this problem?

[eblc1388]

You have exceeded the common mode input voltage limit under the condition of Gain=10x. The Ad8253 is of the classic three-op amps design and one of the two front op amp's output is saturated with the high(10V) common mode voltage at gain=10x. According to the tool on Analog site, its output needs to go to 14.1V to match the calculation which it simply can not because of the 12V supply voltage. Therefore the output voltage is wrong.

The following shows the url of the tool and the results. To solve the problem, you have to reduce the common mode mode voltage seen by the op amp.

Tools at Analog Devices for calculation of input limits: https://tools.analog.com/en/diamond/#difL=0.9&difR=0.9&difSl=0.9&gain=10&l=9.1&pr=AD8253&r=9.1&sl=9.1&tab=2&ty=2&vn=-12&vp=12&vr=0