30 Power Switch Design Tips 2 - Controlling Noisy Power Supply

Noise-free power supplies are not designed by accident. A good power supply layout is designed to minimize experimentation time. Minutes or even hours spent carefully reviewing the power supply layout can save days of troubleshooting.

A block diagram of some of the key noise-sensitive circuits within a power supply is shown in Figure 1. The output voltage is compared to a reference voltage to generate an error signal, which is then compared to a ramp to generate a PWM (pulse-width modulation) signal that drives the power stage.

Power supply noise comes from three main places: the error amplifier input and output, the reference voltage, and the ramp. Careful electrical and physical design of these nodes can help minimize troubleshooting time. Generally speaking, noise is capacitively coupled to these low-level circuits. A good design ensures that these low-level circuits are closely laid out and away from all switching waveforms. The ground plane also acts as a shield.

Figure 1 Many noise generation opportunities in low-level control circuits

The error amplifier input is probably the most sensitive node in the power supply because it usually has the most connected components. If you combine this with the very high gain and high impedance of this stage, the situation is very complex. During layout, you must minimize the node length and place the feedback and input components as close to the error amplifier as possible. If there is a high-frequency integrating capacitor in the feedback network, you must place it close to the amplifier, followed by other feedback components. And, a series resistor-capacitor may also form a compensation network. The ideal result is to place the resistor close to the error amplifier input, so that if a high-frequency signal is injected into this resistor-capacitor node, it will have to see the high impedance of the resistor—while the capacitor will have little impedance to the high-frequency signal.

Ramps are another potential source of noise problems. Ramps are usually generated by charging a capacitor (voltage mode) or by sampling the current from a power switch (current mode). Voltage mode ramps are usually not a problem because the capacitors present little impedance to high frequency injected signals. Current ramps are more problematic because of the peak rising edge, relatively small ramp amplitude, and power stage parasitics.

Figure 2 shows some of the problems with current ramps. The first figure shows the rising edge peak and the resulting current ramp. The comparator (due to its different speeds) has two potential trip points, resulting in out-of-order control operation that sounds more like bacon frying.

This problem is best solved by using rising edge blanking in the control IC, which ignores the initial part of the current waveform. High frequency filtering of the waveform can also help solve this problem. Also place the capacitor as close to the control IC as possible. Another common problem is subharmonic oscillation, as shown in these two waveforms. This wide-narrow drive waveform shows inadequate slope compensation. Adding more voltage ramp to the current ramp can solve this problem.

Figure 2 Two common current-mode noise issues

Even though you have designed your power supply layout very carefully, your prototype power supply is still noisy. What to do? First, you need to make sure there is no problem with the response of the loop that is eliminating the instability. Interestingly, a noise problem may look like an instability at the power supply crossover frequency. But what is really happening is that the loop is correcting the injected error at its fastest response speed. Again, the best approach is to identify that the noise is being injected into one of three places: the error amplifier, the reference voltage, or the ramp. You can just solve it step by step!

The first step is to check the nodes to see if there is significant nonlinearity in the ramp or high frequency changes in the error amplifier output. If that doesn't reveal anything, remove the error amplifier from the circuit and replace it with a clean voltage source. You should then be able to vary the output of that voltage source to smoothly vary the power supply output. If that works, you've narrowed the problem down to the reference voltage and the error amplifier.

Sometimes the reference voltage in the control IC is susceptible to the switching waveform. Adding more (or appropriate) bypassing may improve the situation. Also, using gate drive resistors to slow down the switching waveform may help solve this problem. If the problem is in the error amplifier, then reducing the impedance of the compensation components can help because it reduces the amplitude of the injected signal. If all else fails, remove the error amplifier node from the printed circuit board. Air wiring the compensation components can help identify where the problem is.