September 1, 1998
Optimize output-voltage accuracy of adjustable low-dropout regulators
An adjustable LDO is a good choice when you need a nonstandard
voltage, but getting the highest accuracy from one requires a few circuit tricks.
Paul Paglia, Telcom Semiconductor
Low-dropout regulators (LDOs) have become common in portable electronic equipment.
Their small size, high noise rejection, and low cost make them an attractive solution to a
range of system problemsfrom active power-supply decoupling to local voltage
regulation and control. Although LDO suppliers offer several fixed output-voltage
settings, some applications require settings that you can't get in an off-the-shelf
product.
Your solution to this problem is often a special order for a custom output-voltage LDO,
but the disadvantages of longer leadtime, availability from only one vendor, and a
possible increase in cost can make an adjustable-output LDO more attractive. Adjustable
LDOs, however, lack the tight voltage accuracy of their fixed-output counterpartsa
problem that is difficult to remedy in a cost-effective manner. However, you can improve
the accuracy of adjustable LDOs to match the accuracy of their fixed-output counterparts
by using a few circuit tricks.
The output-voltage accuracy of an adjustable LDO depends on the initial accuracy,
stability, and temperature coefficient of its internal bandgap reference and the external
feedback resistors. Rather than specifying VOUT accuracy on adjustable
regulators, the vendor specifies the initial accuracy and temperature coefficient of their
internal reference; the vendor does not specify VOUT accuracy because that
accuracy depends on the external feedback resistors. In a typical adjustable-LDO feedback
circuit (Figure 1), resistors R1 and R2
set the output voltage, as determined by the following formula:
VOUT=VREF[(R1/R2)+1],
where VREF=1.23V.
The adjust pin (ADJ) is a high-impedance CMOS input. Consequently, resistor values can
range from 300 kOhm to 1 MOhm to minimize the current through R1 and R2.
When VOUT equals VREF (by making R1), the tolerance of VOUT
is approximately that of VREF. Also, when VOUT is greater than VREF
(which occurs when R1/R2>0), the tolerance of VOUT is
a function of both the tolerance of VREF and the tolerance of the R1/R2
ratio. In a worst-case analysis, the tolerances of R1 and R2 are
additive; if both R1 and R2 are 1% resistors, the maximum tolerance
of the R1/R2 ratio is 2%.
By re-examining the effect of tolerances on the equation, you can see that the
tolerance of VOUT worsens proportionally as the VOUT setting departs
the value of VREF. Stated another way,

Table 1 shows the percentage of total output-voltage
error contributed by the tolerances of VREF and R1/R2 for various values of VOUT.
The output-voltage accuracy of the adjustable regulator improves with tighter tolerance
resistors. However, accuracy is limited to ±2% due to the accuracy of the reference. Table 2 shows output-voltage accuracy for the adjustable LDO
using resistors with 1, 0.5, and 0.1% tolerance.
The only way to improve the output accuracy to better than ±2% is to employ an
external reference, such as Telcom Semiconductor's (www.telcom-semi.com)
TCN4041, (Figure 2). By definition, the voltage at the
adjust pin equals VREF. Consequently, you can express VOUT as:

With this configuration, the effect of the LDO's internal reference on output error is
kept at 2% (24.6 mV) absolute, as opposed to being multiplied by the feedback resistance
ratio. The external reference also improves output accuracy because, unlike the internal
reference, it is not exposed to self-heating during high load conditions. You use
resistors R1x and R1y to program the output voltage of the
reference.
Don't neglect initial accuracy
Initial accuracy, in addition to the tolerances of programming resistors R1x
and R1y, also affects the voltage reference. The relationship
between the VOUT of the voltage reference and the programming resistors is
given by:

Note that VREF4041 is the value of the TCN4041's internal-reference voltage,
which depends on the output-voltage setting of the reference. You can express this
reference voltage as

The quantity (DVREF/DVOUT) is a reference-voltage error term. In
this case, internal reference voltage V9REF4041 is specified as 1.233V.
Substituting this equation into the previous equation gives you the overall expression for
the voltage reference's output voltage:

This equation shows that the tolerances of the internal reference, the
reference-voltage error term, and the tolerance of the (R1y/R1x)
ratio combine to determine the overall reference-output- voltage tolerance. Also, there is
a small amount of bias current (<100 nA) that flows from the adjust input through R1y
to ground. You must add the resulting voltage built up across R1y to VOUT4041
as an error voltage. Table 3 shows the effect of these
tolerances on the TCN4041 output-voltage tolerance.
The voltage reference is not the only contributor to the LDO's output voltage. The
LDO's internal reference also plays a major role in determining the LDO's output (see
Equation 1). The LDO's internal reference has a ±2% accuracy, which adds 24.6 mV of
absolute error. Consequently, the internal reference's effect on error decreases with
increasing VOUT (Table 4). R2 has no
effect on accuracy; it is used only to set the operating current of the voltage reference
and to provide a current path for the (R1x/R1y) resistor divider.
The total error that VREF4041 and VREFLDO contribute to VOUT
is:

You can generate the total output error for the circuit in Figure
2 by substituting output-error values from Tables 3
and 4 into Equation 2. The findings are in Table 5, which uses a reference with 0.5% reference error for
tabulation. Note the significant improvement in output error when you use 0.1% resistor
tolerances. Table 6 compares the total output error of the
two LDO feedback circuits of Figures 1 and 2 and shows that an external reference, such as the TCN4041,
significantly improves accuracy, particularly as VOUT and ohmic tolerances
increase. |