AD620 In-Amp, single 5v supply, reference pin affecting gain

Question

Given the datasheet here, I can't explain my observations with my AD620AN test circuit.

The circuit:

• Supply: +5v / 0v (Well regulated)
• 1k potentiometer across the supply, with the wiper going to both the reference pin and the negative input
• +ve/-ve inputs: K-type thermocouple. (41uV/K)
• Gain resistor: 100 Ohm (Gain of 500).

It seems that the gain is strongly dependent on the voltage of the reference pin. If I set the pot at precisely 2.5v, I get roughly the expected gain. But even slighly higher or lower and the gain reduces significantly. (For example, just +/- 0.05V change in reference reduces gain by about 10% for output signals of +/- 100mV). I've been taking measurements with a standard multimeter.

I haven't done meticulous testing with this, but none the less have lost a fair bit of hair. Is there something obviously wrong with what I'm doing? I must have run through the figures in the datasheet two dozen times. Is it the circuit at fault? Or my measurements? Or both?

Answer 1

Notice the small block diagram in Fig.2 if the AD620 datasheet.

Here's a larger diagram from a datasheet of a similar InAmp with a 3-OpAmp topology (LT1920).

Notice that VREF is not a high impedance input. In order not to skew the differential output stage, VREF should be connected to a low impedance output. A voltage divider by itself is not a low impedance output. However, the voltage divider can be buffered.

I have built and tested a similar circuit. (In my case, the VREF came a source other than a potentiometer, though.)