Accurately measure power supply ripple

Measuring power supply ripple accurately is an art in itself. In the example shown in Figure 1, a junior engineer used an oscilloscope all wrong. His first mistake was using an oscilloscope probe with a long ground lead; his second mistake was placing the loop formed by the probe and the ground lead close to the power transformer and switching components; and his last mistake was allowing excess inductance between the oscilloscope probe and the output capacitor. This problem manifests as high-frequency pickup in the ripple waveform. In a power supply, there are a number of high-speed, large-signal voltage and current waveforms that can easily couple to the probe, including magnetic fields coupled from the power transformer, electric fields coupled from the switch nodes, and common-mode currents generated by the transformer mutual winding capacitance.

Figure 1. Incorrect ripple measurement leads to poor measurement results.

The measured ripple results can be greatly improved by using the correct measurement techniques. First, ripple is often specified using bandwidth limits to prevent picking up high frequency noise that is not really there. The correct bandwidth limit should be set for the oscilloscope used for the measurement. Second, by removing the probe "cap" and forming a pickup (as shown in Figure 2), we can eliminate the antenna formed by the long ground lead. Wrap a small piece of wire around the probe ground connection point and connect that ground to the power supply. Doing this will reduce the tip length exposed to the high electromagnetic radiation near the power supply, further reducing the pickup.

Finally, in an isolated power supply, a large amount of common-mode current flows through the probe ground connection. This creates a voltage drop between the power supply ground connection and the oscilloscope ground connection, which appears as ripple. To prevent this problem, we need to pay special attention to the common-mode filtering of the power supply design. In addition, wrapping the oscilloscope leads around a ferrite core can also help minimize this current. This forms a common-mode inductor, which does not affect the differential voltage measurement while reducing the measurement error caused by common-mode current. Figure 2 shows the ripple voltage of the exact same circuit, using the improved measurement method. In this way, the high-frequency peak is virtually eliminated.

Figure 2: Four minor changes significantly improved the measurement results.

In fact, once integrated into a system, the power supply ripple performance can be even better. There is almost always some inductance between the power supply and the rest of the system. This inductance may be in the wiring or just etched onto the PWB. In addition, there are always additional bypass capacitors around the chip that act as a load to the power supply. Together, these two form a low-pass filter that further reduces the power supply ripple and/or high-frequency noise. In the extreme case, when current flows briefly through a one-inch conductor with a 15 nH inductor and a 10 μF bypass capacitor, the filter has a cutoff frequency of 400 kHz. In this case, this means that high-frequency noise will be greatly reduced. In many cases, the cutoff frequency of this filter will be below the power supply ripple frequency, which can significantly reduce the ripple. Experienced engineers should be able to find ways to use this method in their testing process.