As the switch is turned on and off, the current in inductor L fluctuates around the effective value of the output current. Therefore, a ripple with the same frequency as the switch frequency will appear at the output. This is what is generally referred to as ripple, and it is related to the capacitance and ESR of the output capacitor. The frequency of this ripple is the same as that of the switching power supply, ranging from tens to hundreds of kHz.
Furthermore, switches typically use bipolar transistors or MOSFETs. Regardless of the type, there is a rise time and a fall time during their conduction and cutoff. This results in noise in the circuit that has the same frequency as or an odd multiple of the switch's rise and fall times, typically in the tens of MHz range. Similarly, during the reverse recovery of diode D, its equivalent circuit is a series connection of a resistor, capacitor, and inductor, which can cause resonance, generating noise at a frequency also in the tens of MHz range. These two types of noise are generally called high-frequency noise, and their amplitude is usually much larger than the ripple.
In addition to the two types of ripple (noise) mentioned above, AC/DC converters also have AC noise, which has a frequency of approximately 50-60Hz, the same as the input AC power supply frequency. There is also common-mode noise, caused by the equivalent capacitance generated when the power devices in many switching power supplies use their casings as heat sinks.
Measurement of switching power supply ripple
Basic requirements: Use an oscilloscope with AC coupling and a 20MHz bandwidth limit; disconnect the probe's ground wire.
1. AC coupling removes the superimposed DC voltage to obtain an accurate waveform.
2. Enabling the 20MHz bandwidth limit is to prevent interference from high-frequency noise and avoid erroneous measurement results. Because high-frequency components have large amplitudes, they should be removed during measurement.
3. Disconnecting the oscilloscope probe's grounding clip and using a grounding ring for measurement is to reduce interference. Many departments do not have grounding rings, and if the error is permissible, they may directly use the probe's grounding clip for measurement. However, this factor should be considered when determining whether the measurement is合格 (qualified/acceptable).
Another point is to use a 50Ω terminal. The oscilloscope's documentation states that the 50Ω module removes the DC component and accurately measures the AC component. However, few oscilloscopes come with this dedicated probe; most use standard 100KΩ to 10MΩ probes. The impact of this is currently unclear.
The above are basic precautions for measuring switch ripple. If the oscilloscope probe is not in direct contact with the output point, a twisted pair cable or a 50Ω coaxial cable should be used for measurement.
When measuring high-frequency noise, the full passband of an oscilloscope is used, typically in the range of several hundred megahertz to GHz. Other aspects are the same as described above. Different companies may have different testing methods. Ultimately, the first thing is to understand your own test results.
Some digital oscilloscopes, due to interference and memory depth limitations, cannot accurately measure ripple. In this case, the oscilloscope should be replaced. Sometimes, even older analog oscilloscopes with bandwidths of only tens of megabits per second can perform better than digital oscilloscopes in this regard.
Suppression of switching power supply ripple
Switching ripple is both theoretically and practically guaranteed to exist. There are generally three ways to suppress or reduce it:
1. Increase the inductance and output capacitor for filtering.
According to the formula for switching power supplies, the magnitude of current fluctuation within the inductor is inversely proportional to the inductance value, and the output ripple is inversely proportional to the output capacitance value. Therefore, increasing the inductance and output capacitance values can reduce ripple.
The current waveform within the inductor L of a switching power supply, and its ripple current ΔI, can be calculated using the following formula: It can be seen that increasing the value of L, or increasing the switching frequency, can reduce the current fluctuation within the inductor. Similarly, the relationship between output ripple and output capacitance is: Vripple = Imax / (Co × f). It can be seen that increasing the output capacitance value can reduce ripple.
The usual practice is to use aluminum electrolytic capacitors for output capacitors to achieve a large capacitance. However, electrolytic capacitors are not very effective at suppressing high-frequency noise and have a relatively high ESR. Therefore, a ceramic capacitor is connected in parallel with it to compensate for the shortcomings of aluminum electrolytic capacitors.
Meanwhile, when the switching power supply is working, the input voltage Vin remains constant, but the current changes with the switch. In this case, the input power supply will not provide a good current supply, so a capacitor is usually connected in parallel near the current input terminal (near the SWITCH terminal in the BucK type, for example) to provide the current.
The above approach has limited effectiveness in reducing ripple. Due to size limitations, the inductor cannot be made very large; increasing the output capacitor to a certain extent no longer has a significant effect on reducing ripple; increasing the switching frequency will increase switching losses. Therefore, this method is not ideal when requirements are stringent. For more information on the principles of switching power supplies, please refer to various switching power supply design manuals.
2. Second-stage filtering involves adding another LC filter.
LC filters are effective at suppressing noise and ripple. By selecting appropriate inductors and capacitors to construct the filter circuit based on the ripple frequency to be removed, ripple can generally be reduced significantly. In this case, the sampling point of the feedback comparison voltage needs to be considered.
Selecting the sampling point before the LC filter (Pa) will result in a decrease in output voltage. This is because any inductor has a DC resistance, and when current is output, a voltage drop occurs across the inductor, causing the power supply's output voltage to decrease. Furthermore, this voltage drop varies with the output current. Selecting the sampling point after the LC filter (Pb) will produce the desired output voltage. However, this introduces an inductor and a capacitor into the power supply system, potentially leading to system instability.
3. After the switching power supply output, connect an LDO filter.
This is the most effective way to reduce ripple and noise, and the output voltage is constant. It does not require changing the original feedback system, but it is also the most expensive and consumes the most power.
To reduce ripple, the PCB layout of the switching power supply is also crucial. For high-frequency noise, due to the high frequency and large amplitude, although subsequent filtering has some effect, the effect is not significant. There is specialized research in this area; a simple approach is to connect a capacitor C or RC in parallel with the diode, or to connect an inductor in series.
The equivalent circuit of a practical diode. When a diode is in high-speed conduction and cutoff, parasitic parameters must be considered. During the reverse recovery period of the diode, the equivalent inductance and equivalent capacitance form an RC oscillator, generating high-frequency oscillations. To suppress this high-frequency oscillation, a capacitor C or an RC snubber network needs to be connected in parallel across the diode. The resistor is typically 10Ω-100Ω, and the capacitor is 4.7pF-2.2nF. The value of the capacitor C or RC network connected in parallel across the diode must be determined through repeated experimentation. Improper selection can actually cause more severe oscillations.
summary
The above is a summary of some points regarding switching power supply ripple. Adding waveform data would be even better. While it may not be comprehensive, it is sufficient for most applications. Regarding noise suppression, not all methods are necessarily applied in practice. The key is to choose the appropriate method based on your design requirements, such as product size, cost, and development cycle.
