(Editor's note: there is a complete, linked list of all previously published Power Tips! articles below the About the author box at the bottom.)

Output voltages are falling and voltage regulation requirements are getting tighter. However, your job may not be as difficult as it might seem on the surface. Even though you are forced to design with resistors with tolerance of one percent or worse, you may still be able to provide very precise output voltages.

Figure 1 shows a typical regulation circuit for a power supply. The output is divided down and compared against a reference. The difference is amplified and used to drive the regulation loop. At first glance, you might think that this scheme is limited to an accuracy of twice the resistor tolerances. Luckily, this is not true; the accuracy is also a strong function of the ratio of the output voltage to the reference voltage.

Figure 1: Output accuracy is a function of the divider ratio, reference accuracy and error amp offset.
(Click on image to enlarge)

Three different scenarios are fairly easy to visualize for this ratio:

The first scenario is if there is no division at all. In other words, the output voltage is equal to the reference voltage. Obviously, there is no resistor division error in this case.

The second situation is if the output voltage is much larger than the reference. In this case, R1 is much larger than R2. The divider error is twice the tolerance of the resistors, providing the value of R1 shifted in one direction and R2 shifted in the other direction.

A third point that is easy to visualize is if the output voltage is twice the reference voltage. In this case, the nominal value of the resistors is equal. So if the resistor's tolerance shifted in opposite directions, the top of the divider equation shifts by the tolerance value, while the denominator shift is zero.

Figure 2 shows the output accuracy as a function of the relationship of the reference voltage to the output voltage. (See the detailed derivation in the Appendix.) The simplified answer is that the divider accuracy is:

(1 – Vref/Vout) × 2 × Tolerance,

which correlates with our three data points we derived by inspection. This equation is simplified somewhat, but should be sufficiently accurate for most resistor tolerances.

Figure 2: Output accuracy is simply: (1 – Vref/Vout) × 2 × tolerance (1% resistors shown).
(Click on image to enlarge)

Interestingly, this allows more accuracy for lower voltage outputs. Many IC references are in the 0.6-to-1.25 volt range, which allows one percent accuracies or better as the output voltage falls into these ranges.

Table 1 presents some information you may not want to see. This is a compilation of resistor error terms from a typical resistor data sheet.

Table 1: Resistor tolerances can add up
(Click on image to enlarge)

This list can be hard to comprehend in a design. Most engineers stop at the initial tolerances, but there are error terms in this list that probably should not be ignored. Each one of these elements is subtle in its impact. For instance, there is no range specified on the temperature coefficient while, in reality, both resistors probably will shift in the same direction with temperature and will not be at opposite ends of the extremes. After a brief survey of well seasoned design engineers, it was concluded that assuming 2.5% accuracy for a one percent tolerance resistor provides a reasonable tradeoff between worst case and reasonable costs.

To summarize, providing good accuracy on low-voltage outputs will not be a daunting task, as low divider ratios are inherently accurate.

Please join us next month when we will discuss an interesting power-supply topology for making negative voltages.

For more information about this and other power solutions, visit www.ti.com/power-ca.

About the author