AN-1643| Application Note

Optimizing Feedforward Compensation In Linear Regulators

Optimizing Feedforward Compensation In Linear Regulators
Introduction
All linear voltage regulators use a feedback loop which controls the amount of current sent to the load as required to hold the output voltage at the correct regulated value. The feedback loop is compensated to provide adequate phase margin at the frequency where the loop gain crosses unity (zero dB). In cases where the output voltage is adjustable, and set by an external resistive divider, a compensation technique called feedforward can be employed which can increase phase margin by producing phase lead which cancels out some of the phase lag from the internal poles. It should be noted that not all linear regulators require feedforward compensation, but this application note explains the technique for those that will benefit from it (and most will).

National Semiconductor Application Note 1643 Chester Simpson May 2007

and R2 are external to the IC), a capacitor can be placed across R1 to add feedforward compensation which will also add phase lead. R1, R2, and CFF form a pole-zero pair, where the zero will always be at a lower frequency than the pole. The frequency of the pole-zero pair is given by: FZERO = 1 / 2 x x CFF x R1 FPOLE = 1 / 2 x x CFF x (R1 // R2) It should be noted that it is only the zero which adds beneficial phase lead, the pole adds phase lag which tends to cancel out the lead. To improve phase margin, the pole-zero pair must be positioned at the frequency where the zero adds maximum lead and the pole gives minimum phase lag at the unity gain frequency. The net positive phase lead obtained is the difference between these two values at the unity-gain point. It follows from this that maximum benefit is derived when the pole-zero pair are far apart (which occurs when R1 >> R2). As R1 gets smaller and smaller, the pole frequency moves closer to the zero frequency, eventually canceling out when R1 = 0. There
fore, the higher the ratio of R1/R2 is, the farther the pole and zero are separated and the more potential phase lead can be obtained. This means that feedforward compensation is more effective when the output voltage is set to higher ratios of the reference voltage (since it is the ratio of R1/R2 that is important, not the actual output voltage). A graphical illustration of this is shown in Figure 2 which shows the maximum positive phase lead which can be obtained for selected ratios of R1/R2:

Linear Regulator Loop Basics
All linear voltage regulators use a similar control topology (see Figure 1). A pass transistor device is used to source current to the load. The error amplifier controls this current in response to the output's feedback voltage appearing at the junction of R1 and R2. The action of the loop is such to always force the voltage at the error amplifier's input terminals to be equal to the fixed reference voltage. In this way, the control loop holds the output at the nominal voltage, which is given by: VOUT = VREF ( 1 + R1/R2)

Loop Compensation
The control loop is usually compensated locally at the error amplifier, and in some cases, also by using the ESR of the output capacitor to add some positive phase lead. In most "adjustable output" linear regulators (where the resistors R1

30022801

AN-1643

FIGURE 1. Basic Linear Regulator

2007 National Semiconductor Corporation

300228

www.national.com