Voltage-Regulator Droop

The circuit model in Figure 1 captures the important low-frequency behavior of most voltage regulators (see "Voltage-Regulator Model"). Parameters C2, R2, and L2 represent a typical bulk decoupling-capacitor array. If you are modeling the response only at frequencies below 100 kHz, you can just set L2 to 0 as its effect will be negligible in that low-frequency range. Parameter R1 models the regulation resistance, or stiffness, of a small switching-regulator module. Parameter L1 models the response time.

Now try something outrageous. Add some series resistance to the regulator output by increasing the value of R1.

When the load draws current, the new, larger value of R1 increases the droop measured at VCC. That scenario sounds bad, but in some special circumstances, it is actually good for your circuit.

Figure 2 shows the regulator's voltage response for values of R1 ranging from 3 to 12 mΩ. The circuit is subject to an 8A step load with a maximum dI/dt of 2.5A/μsec. The plot shows the load current at the bottom and the collection of response curves at the top. It offsets each curve horizontally to visually separate them.

Concentrate on the red waveform (minimum value of R1). Beginning from rest at Point A, the VCC output sits at its nominal, midlevel value. When the load turns on, VCC responds with a downward glitch.

Only components C2 and R2 limit the initial amplitude of this glitch, because the regulator can't respond instantaneously; it takes a few switching cycles to respond.

Once the regulator wakes up, it drives the voltage back to a new operating Point B. The sluggish response of the regulator mimics the action of an inductor, which is why Figure 1 makes such an effective model.

When the load switches off, the response pops back high at Point C. (Inductors do that, and so do regulators.) The overall peak-to-peak response of the red waveform equals nearly twice the amplitude of the initial glitch.

Now, increase R1 to 0.012 Ω and reapply the load. The black waveform goes down and stays down, displaying more long-term droop because of its larger series resistance. When the load cuts off, the positive glitch that inductor L1 causes begins at a lower level. Beginning lower, this glitch does not reach as high as the glitch on the red waveform. As a result, you obtain the smallest peak-to-peak response with R1=0.012 Ω.

If you use this method, offset the resting voltage of your regulator toward the high end of its range to best center the overall step-response waveform.

Some switching regulators let you lower the gain of the control loop, effectively increasing the regulation resistance, R1, without dissipating additional power. This cool trick makes no difference to a linear regulator circuit.

Whatever you do, beware of the tolerances associated with all components in this circuit. Do not design something so tricky and intricately balanced that small changes in the component values throw the whole system out of whack.