4 Options for Driving Reference Pins and Input Referred Reference Error

Three methods are commonly used to drive the reference pin, namely with a reference voltage IC, with a supply divider, or with a supply divider followed by a buffer. These options are illustrated in Figure 4-1.

Each of these has its own pros and cons. A high-level comparison is show in Table 4-1.

From a performance perspective, a reference IC is the best choice. The electrical characteristics of a reference are well defined and production tested. Key parameters such as offset, drift, and noise are excellent. Such reference provides a stable, clean output voltage that is ideal for data acquisition. Sometimes a reference already exists because it is required by other components such as ADC. It might be possible to use the existing reference without incurring additional cost.

The most affordable reference is the supply divider, where the reference voltage is derived from the device power supply with a resistor divider. Any value in between 0 V and supply is possible. It is a natural tendency to use large resistors so that power dissipation is kept to a minimum. However, a large divider adds to the internal resistor of the reference pin and breaks the balance of the resistor network. To reduce such impact, The Thevenin's equivalent resistance should be kept as small as possible. But it may become impractical if the resistance values get too low and consequently the power dissipation becomes too great. In theory, an ideal point between the two extremes is possible, which may be suitable for certain applications. The following sections look deeper into the performance impact of such dividers. Another aspect that should not be ignored is the quality of the power supply itself.

For applications that are power sensitive, the third approach might be a good option. A large resistor divider followed by a general-purpose buffer can provide a good reference at a reasonable cost. The buffer isolates the divider from the internal resistor network and provides a virtual ground for the reference pin. The buffer amplifier does not have to be high performance. Because its non-idealities are added to the CSA output, they get divided down by the CSA gain when referred to input. Figure 4-2 illustrates the idea of referring reference error to equivalent input error.

Regardless of the method chosen, in reality there is going to be some error introduced by the reference source. It may be necessary to account for this error if it is not negligible. To do this, it is important to keep the calculation consistent with that for the rest of the error sources. In other words, either input referred or output referred calculation can be used. But the two methods should not be mixed.

Reference source errors are directly added to the output. To compare or combine with other error sources, the reference errors are normally referred to input. As an example, the external reference source in Figure 4-2(a) has an offset V_{os_x} in addition to its ideal value of V_{ref_x}. when calculating input referred error contribution, V_{os_x} needs to be divided by the CSA gain. Figure 4-2(b) shows the equivalent circuit after V_{os_x} is referred to input.

Even though offset of the reference source is used as an example, the same principle applies to other non-idealities, including gain error, temperature drift, and noise, to name a few.