The layout and value of a small decoupling capacitor

First of all, the layout of the decoupling capacitor

In the picture below, ae is wrong? what reason?

As shown above, the capacitance at this location generally has two effects.

One is to provide the IC power supply with the large current required for instantaneous operation (also called bypass).

The second is to act as a decoupling, that is, to suppress the internal noise of the IC, such as multiple harmonics of the oscillator, from being transmitted to the power supply and interfering with other circuits , that is, the noise is not transmitted to the power layer or the ground layer.

However, this capacitor has been marked as a decoupling capacitor in this figure.

For the first case, it is not necessary to pass the capacitor before receiving the power or ground pin of the IC, but try to be as close as possible. A typical example is a decoupling capacitor for a BGA package, which is typically placed on the back side. When you are as close as possible, pay attention to the wiring of the capacitor to the power supply and the ground plane. The shorter and thicker the better, otherwise the wiring inductance will be introduced. Because the instantaneous power supply is also the shortest impedance path, too large distributed inductance will bring disadvantages.

For the second case, the power supply of the IC first passes through the capacitor and then to the power supply or the ground plane. This is the best, so that the noise is removed from the capacitor first, and it will not go to the power supply or the ground.

In the picture a to e, after the noise comes out from the ground and power lines, before reaching the decoupling capacitor, it has already passed through the branch of the via to other circuits. The f in the figure is better. . It’s not that it’s not going to be a problem. It’s definitely not going to be a problem. When it’s more than the actual pcb, it’s not always possible to meet this standard because of various objective reasons. Come.

In addition, in this case, especially pay attention not to introduce excessive inductance in the wiring, because of high frequency noise, and its higher harmonics, the frequency is very high, and at high frequencies, small inductance will bring For larger impedances, the high-frequency noise can't be coupled to the ground by the capacitor with low resistance, which reduces the decoupling effect from the surface.

Let's talk about why the decoupling capacitors are mostly worth 104 capacitors?

This is called decoupling rather than filtering and is used to deal with high frequency noise in the power supply loop. For conventional low-speed digital circuits and general analog circuits, the operating frequency is not high, the spectral characteristics of 104 capacitors can be satisfied, and when the spectrum characteristics can be satisfied, the larger the capacity, the better, so these circuits use 104 decoupling capacitors. However, the larger the capacitance, the larger the ESL, and the worse the high-frequency characteristics. When decoupling the high-frequency circuit, a small-capacity decoupling capacitor is needed. For the GHz-level circuit, the decoupling capacitor is even needed. When used on the order of 10pF, a plurality of capacitors of different capacities are often used in parallel to decouple.

It can also be extended to the capacitance of each node on various circuits. For the value of the capacitor, there is no need to care about the difference between decoupling and filtering. This is a decoupling capacitor or a filter capacitor to filter out noise on the power supply.

For low frequencies such as audio, it is generally necessary to parallel a larger capacity electrolytic capacitor.

For digital circuits, 104 is fine.

For high-frequency circuits, multiple capacitors such as 104, 102, 100P, and 10P are generally required in parallel.

It is related to the equivalent circuit of the capacitor and is related to the frequency. For example, 39P is suitable for 900MHz, and 10P is suitable for 1800MHz.