Hi Max,
There are a couple of ways to look at this and I will try to keep it simple. The first viewpoint is with respect to the sensor and the second is with respect to the ADC input. A single ended measurement usually refers to one end of the sensor connected to a common. This is usually ground, and signal swing is only in the positive direction. A differential measurement is one input with respect to the other and can be bi-directional.
There are a number of advantages to a differential measurement with respect to the input stage. This is the method of measurement of the ADS1147 in all cases (independent on the type of sensor connected) as the input stage is similar to differential amplifier. One of the key advantages of this kind of input stage is high common mode rejection (or noise common to both is canceled.)
With the RTD you are making a measurement with one side of the RTD relative to the other. As I've already stated this is how the ADS1147 measures anything even though the current path is flowing through the RTD in a single direction making it appear as though it is single ended. In the case where one side of the RTD is connected to AIN0 and the other side is connected to AIN1, you might assign AIN0 to the + side of the input and AIN1 to the - side. For this example let's say current is exciting the RTD so that a voltage drop can be measured across the RTD. Let's also say that current is being injected closest to the AIN0 side of the RTD and a Rbias resistor is nearest to the AIN1 side. It is much easier to see that in this configuration that neither AIN0 or AIN1 is connected to ground nor is the measurement taken with respect to ground. It is taken with respect to the voltage induced across the RTD element and is the difference of the voltage developed across AIN0 to AIN1.
Let's take a look at a different type of configuration where instead of using the Rbias resistor we use bipolar supplies (+/-2.5V) for AVDD and AVSS. This allows for the sensor to be directly connected to ground (as it is now within the common mode input range.) If we connect the AIN- side to ground it appears to be more like a single ended measurement. We are now making the measurement with respect to a common point which in this case is analog ground. However, the ADS1147 input is still taking a measurement with AIN0 to AIN1 and is measuring the difference between the two.
If the ADC of the ADS1147 measures the input signal the same way, why use one method over the other? The answer to this is written in books and I really can't go deeply into this, but the end result of the measurement has a lot to do with the actual sensor output (which may be referred to ground for single ended measurements) or the necessity for high common mode rejection.
The actual filtering structure changes as well depending on whether one of the inputs connects directly to ground.
Best regards,
Bob B