Would the best approach be to use a unipolar to bipolar converter as described here? The op-amp approach would also allow the output from the hall effect sensor to be tailored to the input of the ADC.
Or, as an alternative approach, use the LTSR-25, which offers access the to its reference output. This sensor would seem superior to the S22P for a bipolar input because the voltage coming out of an external series reference != reference voltage inside the S22P. With access to the reference voltage inside the LTSR25, one would always have a truely differential output.
The Analog statement inside the data sheet
"..the maximum signal level on analog inputs for V1P/V1N and V2P/ V2N is ±0.5 V with respect to AGND." (p.16 of datasheet)
suggests that the common-mode voltage has to be 0V. Thus, even if I use the LTSR-25, it would seem that a differential op-amp would have to return the signal to 0 +/- 0.5V to make it palatable for the AD7753?