How to use capacitors appropriately? Read it in seconds and understand!
Starting from studying electronics, it is inevitable to deal with capacitors. As long as it involves filtering, decoupling, bypass, etc., capacitors are essential.
I remember when I first started working, referring to my teacher's 5V Buck circuit, the output end was a classic combination of 10uF+1uF+100nF in size and capacitance. At that time, I asked the master in charge, why is this combination like this? His explanation is that large capacitors filter out low-frequency interference while stabilizing voltage, while small capacitors filter out high-frequency interference.
Then I searched for information on Baidu again, and most of the responses were the same: a small capacitor filters out signals with higher frequencies, and the frequency of the signal passing through is in the interval between these two capacitors that are filtered out. A large capacitor is an electrolytic capacitor with polarity that does not act on alternating current, but filters out uneven direct current to make it smoother. A small capacitor is an infinite capacitor that filters out high-frequency AC noise.
These all seem very reasonable, after all, from the perspective of capacitance impedance characteristics, this explanation is very correct. A bypass is the practice of adding a low impedance path at the signal source to transfer the transient energy component to ground.
However, with the accumulation of work experience, I feel that this simple parallel connection of capacitance values can sometimes not only fail to filter, but also bring some negative effects. Especially when the system speed increases, it is suitable for circuits designed with lower system speed and slower logic requirements. Using parallel connections with different capacitance values can bring some problems, such as increasing RFI/EMI, reducing noise tolerance, and so on. Because the actual model of the capacitor is a series parallel combination of R-L-C, based on this actual capacitor model, it is known that the impedance curve of the capacitor has resonance points, which can also be obtained in the manual of the capacitor's Datasheet.
So, in actual circuit design, it is best to choose a capacitance value as much as possible. Even if multiple capacitors with different capacitance values need to be connected in parallel, it is necessary to select the resonant characteristics of the capacitors to match the frequency characteristics of noise filtering.
Capacitors are often used in EMI debugging. The following is a spectrogram tested in a 5V Buck circuit, and the bottom noise of the spectrograph is around 38dbuV. It is obvious that there is a peak of 59dbuV near 280MHz, and the amplitude of the noise is 1mV, which needs to be filtered out. The general practice is to parallel connect a capacitor of about 3.3nF to the ground on the input line of the power supply. For f=280MHz,
Its impedance Zc=1/(2 * pai * f * c)=1.7 Ω, and referring to the impedance curve of the capacitor, the resonant point is selected near 280MHz to form a low impedance path, so it should be effective.
Capacitor filtering noise signals is usually achieved in the form of RC low-pass filters. However, when selecting a 3.3nF capacitor, there is no resistance R. My understanding is that the internal resistance of the noise source acts as resistance R. If the internal resistance of the noise source is very small, selecting a 3.3nF capacitor will not work.
Similarities: The larger the internal resistance of the noise source, the greater the attenuation of the amplitude frequency curve.
Difference: The amplitude frequency curve does not continuously roll down at 20dB/10dec. This means that the actual capacitance model is not a single pole model. This is because the capacitor has ESR, which provides a zero point at high frequencies, so the amplitude frequency curve will not roll off again at high frequencies. In the improvement process of this EMI, it is fortunate to add a 3.3nF capacitor, which reduces the noise amplitude of 280MHz. Perhaps the source impedance of this noise source is about 10 Ω, so there can be a 20dB attenuation at 280MHz.
It should be noted that the source impedance of noise sources is my own understanding and may not necessarily be correct. Welcome everyone to leave a message at the end of the article and discuss together.
Due to the noise amplitude of 280MHz being 41dbuV, which has not yet reached the bottom noise level of 38dbuV, there is also a small convex hull. Replacing a capacitor with a different impedance curve should result in improvement. However, adding an inductor to the power line to form an LC second-order filter with a roll off slope of 40dB/10dec will result in better filtering performance: