2 Total Output Error of CSA at Near-Zero V_sense

To include zero V_sense in the linear input range, use the bidirectional CSA. A bidirectional CSA is equipped with one or two reference pins. The output of such CSA can be biased such that the output is no longer masked by the swing to rails limitation. As a result, as soon as the differential input changes by a small amount, the output changes by the same amount, multiplied by the device gain.

As illustrated by Figure 1-4, the real transfer curve deviates from theoretical due to device errors such as offset, gain error, and nonlinearity. One figure of merit commonly used in evaluating the accuracy of a system is total output percentage error, and is defined as the amount of deviation of actual output relative to the theoretical output.

Figure 2-1 shows the typical error versus current plot of a CSA. Since V_sense equals to the product of current and shunt resistance, the x-axis can be changed to V_sense and the curve maintains shape.

In the higher current range (for example, > 2 A) of this particular example, the percentage error is near flat and is dominated by gain error; in the lower current range, offset has an out-sized impact as current decreases toward zero while offset itself remains constant. At zero current, the percentage error approaches infinity, and the effect from device imperfections becomes more pronounced.

The total error plot reflects the CSA specification only, and does not consider other components at the system level. For example, shunt resistor tolerance has an impact on total error that is similar to gain error. This plot also assumes worst-case device specification. Individual device performance is likely to be much better, due to the fact that device parameters typically follow normal distribution. The likelihood of running into samples with the worst performance for all parameters is low.