Minimizing the output ripple and transients of a switching regulator is important, especially when supplying noise-sensitive devices such as high-resolution ADCs, where the output ripple will appear as a unique stray in the ADC output spectrum.

To avoid reducing signal-to-noise ratio (SNR) and scatter-free dynamic range (SFDR) performance, switching regulators are often replaced with low-voltage differential regulators (LDO), sacrificing the high efficiency of switching regulators for cleaner LDO output. Understanding these artifacts allows designers to successfully integrate switch modulators into more high-performance, noise-sensitive applications.

This paper presents an effective method for measuring output ripple and switching transients in a switching regulator. Measurement of these parameters requires great care, as poor Settings can lead to incorrect readings, and loops formed by oscilloscope probe signals and ground leads can lead to parasitic inductance. This increases the magnitude of transients associated with fast switching transients, so shorter connections, efficient methods, and wide bandwidth performance must be maintained.

Here, ADP2114 dual-channel 2a/single-channel 4a synchronous step-down DC-DC converter is used to demonstrate the method of measuring output ripple and switching noise. This step-down regulator has high efficiency, switching frequency up to 2 MHz.

1. Output ripple and switching transients

The output ripple and switching transients depend on the topology of the regulator and the values and characteristics of the external components.

The output ripple is the residual AC output voltage and is closely related to the switching operation of the regulator. Its fundamental frequency is the same as the switching frequency of the regulator. Switching transients are high-frequency oscillations that occur during switching. Their amplitude is expressed as the maximum peak-to-peak voltage, which is difficult to measure accurately because it is highly dependent on the test setup. Figure 1 shows an example of output ripple and switch transients.

2. Influencing factors of output ripple

The inductance and output capacitance of the regulator are the main components that affect the output ripple. A smaller inductance produces a faster transient response at the cost of a larger current ripple; A larger inductance makes the current ripple smaller, at the cost of a slower transient response. Capacitors with low effective series resistance (ESR) minimize output ripple.

Ceramic capacitors with dielectric X5R or X7R are a good choice. Large capacitors are usually used to reduce output ripple, but the size and number of output capacitors are obtained at the expense of cost and PCB area.

2.1 measurement in frequency domain

It is useful for power engineers to consider the frequency domain when measuring unwanted output signals, which provides a better view of what discrete frequencies the output ripple and its harmonics are at, and what different power levels each corresponds to. Figure 2 shows an example of the spectrum. This information helps the engineer determine whether the selected switch regulator is suitable for its broadband RF or high-speed converter application.

For frequency domain measurements, a 50-ohm coaxial cable probe can be connected to both ends of the output capacitor. The signal passes through the isolation capacitor and terminates at the 50 ω terminal resistance at the input of the spectrum analyzer. The isolating capacitor prevents the DC current from passing through the spectrum analyzer and avoids the DC load effect. The 50 ω transmission environment minimizes high frequency reflections and standing waves.

The output capacitance is the main source of output ripple, so the measurement points should be as close as possible. The loop from the signal tip to the ground point should be as small as possible to minimize additional inductance that may affect the measurement results. Figure 2 shows the output ripple and harmonics in the frequency domain. ADP2114 generates 4 mV P-P output ripple at the fundamental frequency under specified operating conditions.

2.2. Time domain measurement

When oscilloscope probes are used, ground loops can be avoided by not using long ground leads because the loops formed by the signal tip and long ground leads can generate additional inductance and higher switching transients.

Low level output ripple is measured using a 1× passive probe or 50 ω coaxial cable, rather than a 10× oscilloscope probe, which will attenuate the signal by a factor of 10, reducing the low level signal to oscilloscope background noise. Figure 3 shows the suboptimal detection method. FIG. 4 shows the waveform measurement results with a 500 MHz bandwidth setting. High frequency noise and transients are measurement false signals caused by loops of long grounded leads and are not inherent in the switching regulator.

There are several ways to reduce stray inductance. One method is to remove the long grounding lead of the standard oscilloscope probe and connect its tube body to the grounding reference point. Figure 5 shows the tip and tube body method. However, in this case, the tip is connected to the wrong regulator output point rather than directly to the output capacitor; The correct method is to connect directly to the output capacitor. The ground lead has been removed, but the inductance caused by wiring on the PCB is still present.

Figure 6 shows the waveform results with a 500 MHz bandwidth setting. Because long ground leads are removed, high-frequency noise is reduced.

As shown in Figure 7, direct detection on the output capacitance using a grounded coil produces a near-optimal output ripple. The noise profile of switching transients has improved and the inductance of wiring on the PCB has dropped significantly.

However, low-amplitude signal contours are obviously superimposed on the ripple, as shown in FIG. 8.

The best way to detect the switch output is to use a 50 ω coaxial cable that is maintained at 50 ω and terminated by an optional 50 ω oscilloscope input impedance. A capacitor placed between the regulator output capacitor and the oscilloscope input can prevent direct current flow through. The other end of the cable can be welded directly to the output capacitor by a very short fly wire, as shown in Figures 9 and 10.

This can maintain signal integrity when measuring very low level signals over a wide broadband range. FIG. 11 shows the comparison of the tip and tube method with the 50 ω coaxial method at the output capacitor end at the 500 MHz measurement bandwidth.

These comparisons show that using coaxial cable at 50 ω yields more accurate results with less noise, even at a 500 MHz bandwidth setting. High-frequency noise can be eliminated by changing oscilloscope bandwidth to 20 MHz, as shown in FIG. 12. ADP2114 generates 3.9 mV p-p output ripple in the time domain, which is close to the frequency domain value of 4 mV p-p measured by using 20 MHz bandwidth.

3. Measure switch transients

Switching transients have lower energy but a higher frequency component than the output ripple. This happens during the switching process, usually normalized to peak-to-peak values that contain the ripple. Figure 13 shows a comparison of switching transients measured using a standard oscilloscope probe with a long ground lead versus a 50 ω coaxial terminating cable (500 MHz bandwidth). In general, ground loops caused by long ground leads produce higher switching transients than expected.

Output ripple and switching transient measurement methods are very important considerations when designing and optimizing system power supplies for low noise and high performance converters. These measurement methods can achieve accurate and reproducible time and frequency domain results. When measuring low level signals over a wide frequency range, it is important to maintain a 50 ω environment. A simple, low-cost way to make this measurement is to use a properly terminated 50-ohm coaxial cable. This method can be applied to all kinds of switching regulator topologies.

Source: ADI official website

Journal download: download.csdn.net/download/m0…