I am trying to read the differential voltage from a pair of hall sensors using the AD4022. The AD4022 seems to read fine until a certain positive or negative voltage is read. It doesn't seem to be a data unpacking issue, as I have two separate boards, and the problem happens at slightly different voltages between the two. In the figures below, the data read by the AD4022 is shown in the top plot, and in the bottom plot I have the voltage recorded by a "ground truth", which is an AD7609 that I have verified works.
It seems to me that the AD4022 is mixing bits after a certain point, and so the voltage goes up in "steps", which is shown best by this first figure.
Again, it doesn't seem to be a data unpacking issue, as when the voltage spikes and then goes down in the second figure from 1.85 to 1.15 V, the first byte of the data from the AD4022 goes from 0x47 to 0x31, as verified on the Oscilloscope. However, the ground truth voltage is steadily increasing in that entire time, so the first byte should not decrease.
This third figure shows that this happens for negative voltages as well, around -0.8 Volts.
The AD4022 is on a custom PCB as shown in the schematic below.
I am communicating with the device as specified on page 31 of the datasheet. I have copied page 31 below, as well as a scope probe showing my communication.
Please let me know if there is any other information that would be useful in figuring this out.