One metric for a design is how accurate the calculated angle from the sensor measurements are to the angle expected when the device is in the intended location. Offsets in placement leads to the calculated error being different than what is expected and warranting calibration depending on how large the error is and how much accuracy is desired. Figure 2-8 and Figure 2-9 indicate how many degrees of error result from the sensor being offset along the x and y axis. The ideal sensor position is considered to be centered on the magnet z-axis at the average of the z-offset for both sensors.

Figure 2-7 indicates that a ±1.5 mm offset along the x or y axis leads to <1° error for the stacked die device and that the error observed by both sensing elements is nearly identical which contrasts with what is observed for side-by-side sensing elements in Figure 2-8. Depending on the direction of offset, the deviation from the ideal measurement is 1.5°, additionally the two sensing elements exhibit mirror symmetry which has an impact on the standard deviation of measurement difference or max measurement difference for calculated angles between the two die, both of which are important metrics for redundancy of a dual sensor devices. In this application, the desire is that the measured angle, derived from the measured Bx and By fields, is identical. Any difference between measured angles constitutes error. Unless the error is uniform and applicable to all possible measurable points, the standard deviation is smaller than the max difference and a smaller standard deviation indicates that the difference in data tends to be smaller. Figure 2-10 shows the standard deviation of difference angle measurement for the stacked die, while Figure 2-11 and Figure 2-12 show the standard deviation for the side-by-side die.

Figure 2-13 shows the max difference in angle measurement for the stacked die while Figure 2-14 and Figure 2-15 show the max difference in angle measurements for the side-by-side die.

Figure 2-10 shows that the majority of the calculated angles over the full 360 degree rotation are within .01 degrees of the ideal angle even for the worst case x-offset. Alternatively for the side-by-side shown in Figure 2-13 and Figure 2-12, the standard deviation is nearly 1°.

Figure 2-13 indicates that the max difference in angle calculation for the stacked die is 100x smaller than that of the side-by-side die shown in Figure 2-14 and Figure 2-15.