We are using the ADS114S06B in our system. We have chosen this chip for it's 16 bit capabilities. This would project out input range (+-32.768V) exactly as 1mV/bit.
To keep our power supplies simple, we took the input voltage, put it through a differential OpAmp with a gain of 0.0625 and a 2.048V offset. Now we have a 0-4.096V
signal we want to measure with the ADS. ADS AVDD is supplied with 5V.
At first we measured it (between AINCOM and AIN0) with 4.096V on the REF and AINCOM to GND. This gave us a correct value, but with 15 bit (this is nowhere to be
found in the datasheet). Reading another question about this issue came up with the idea of connecting AINCOM to half VREF (2.048V). Since we still measure a voltage
of 0-4.096V, we left the VREF at 4.096V. This gives us the correct 0 point, but only half the range. 4.096V in gives 4000h, not 7FFFh and 0V gives BFFFh, not 8000h.
Now we also have half VREF connected to REFp (REFn to GND). With this configuration we get the right measurements and range, but we do not get why this works.
We thought that REFp was for the positive reference voltage and REFn for the negative. But since we only use the positive range, REFn was not necessary. Now it looks
like the ADC creates its own -VREF, inverted from the +VREF.
So if lets say I want to measure a voltage that sweeps between -2.048V and +2.048V, I only need a positive reference voltage of 2.048V on REFp, connect REFn to GND
and measure the input against GND? Given the ADS Analogue supply is +-2.5V.