Analysis of bypass capacitors in high-speed PCBs

With system volume The reduction of the operating frequency and the increase of the operating frequency make the functions of the system complicated, so that multiple different embedded functional modules are required to work at the same time. Only each module has good EMC and low EMI, can guarantee the realization of the whole system function. This requires that the system itself not only needs to have good performance of shielding external interference, but also requires that it cannot produce serious EMI to the outside world when it works with other systems at the same time. In addition, switching power supplies are more and more widely used in the design of high-speed digital systems, and a system often needs to use multiple power supplies. Not only is the power supply system susceptible to interference, but the noise generated during power supply can cause serious EMC problems to the entire system. Therefore, in high-speed PCB design, how to better filter out power supply noise is the key to ensuring good power integrity. This paper analyzes the filtering characteristics of capacitors, the influence of the filtering performance of the parasitic inductance and capacitance of the capacitors, and the current loop phenomenon in the PCB, and then makes some conclusions on how to choose bypass capacitors. This paper also focuses on analyzing the generation mechanism of power supply noise and ground bounce noise, and makes an analysis and comparison of various placement methods of bypass capacitors in the PCB on this basis.

The ability of EMI power filter to suppress interference noise is usually measured by insertion loss (Insertion Loss) characteristics. Insertion loss is defined as the ratio of the noise power P1 transmitted from the noise source to the load when the filter is not connected to the noise power P2 transmitted from the noise source to the load after the filter is connected, expressed in dB (decibel). Figure 1 shows the insertion loss characteristics of an ideal capacitor. It can be seen that the slope of the insertion loss curve corresponding to a 1μF capacitor is close to 20dB/10 times the frequency.

Observe one of the insertion loss characteristics. When the frequency increases, the insertion loss value of the capacitor increases, that is to say, the value of P1/P2 increases, which means that the system filters through the capacitor Later, the noise that can be transmitted to the load is reduced, and the ability of the capacitor to filter out high-frequency noise is enhanced. From the analysis of the ideal capacitance formula, when the capacitance is constant, the higher the signal frequency, the lower the loop impedance, that is, the capacitor is easy to filter out high-frequency components. The conclusions drawn from both sides are the same.

Then observe the curves corresponding to different capacitors. When the frequency is very low, the insertion loss values ​​of various capacitors are approximately the same, but as the frequency increases, the insertion loss value of small capacitors The larger the capacitance, the slower the increase, and the slower the increase in the P1/P2 value, which means that the larger capacitance is easier to filter out low-frequency noise. Therefore, when we design high-speed circuit boards, we usually place a 1-10μF capacitor at the power access end of the circuit board to filter out low-frequency noise; place a 0.01-0.1 μF capacitor between the power supply and the ground wire of each device on the circuit board μF capacitor to filter out high frequency noise.

The impedance of the capacitor connected between the power supply and the ground can be calculated by the following formula: The purpose of capacitor filtering is to filter out the AC component superimposed in the power supply system. It can be seen from the above formula that when the frequency is constant , the larger the capacitance value, the smaller the impedance in the loop, so that the AC signal is easier to flow to the ground plane through the capacitor. Real capacitors do not have all the characteristics of ideal capacitors. The actual capacitance has parasitic components, which are formed when the capacitor plates and leads are constructed, and these parasitic components can be equivalent to the resistance and inductance connected in series on the capacitor, which are usually called equivalent series resistance (ESR) and equivalent A series inductor (ESL), modeled as shown in the left half of Figure 2. If the parasitic resistance of the capacitor is ignored, the model can be equivalent to the right half of Figure 2. In this way, the capacitor is actually a series resonant circuit. In the actual circuit or PCB design, the existence of the parasitic inductance of the capacitor will have a great impact on the filtering performance of the capacitor, so the capacitor with a relatively small parasitic inductance should be selected in the system design.

From Section 2.1, we know that the actual capacitor has parasitic inductance during operation, making the capacitor loop a series resonant loop. The resonant frequency is, where: L is the equivalent inductance; C is the actual capacitance. As shown in Figure 3, when the frequency is less than f0, it appears as a capacitance; when the frequency is greater than f0, it appears as an inductance. So, the capacitor is more like a band stop filter than a low pass filter. The ESL and ESR of the capacitor are determined by the structure of the capacitor and the dielectric material used, and have nothing to do with the capacitance of the capacitor. The ability to suppress high frequencies will not be enhanced by replacing large-capacity capacitors of the same type. The impedance of a capacitor of the same type with a larger capacity is smaller than that of a small-capacity capacitor when the frequency is lower than f0, but when the frequency is greater than f0, ESL determines that there is no difference in the impedance between the two. It can be seen that in order to improve the high frequency filtering characteristics, capacitors with lower ESL must be used. The effective frequency range of any kind of capacitor is limited, and for a system, there are both low-frequency noise and high-frequency noise, so different types of capacitors are usually used in parallel to achieve a wider effective frequency range.