A problem across power partitions
This is a power split problem of PCBBBS network...
It may be that cross-zone will cause a surge of energy, you can use add cap to balance the noise of the two parts...
On the other hand, it may be: when high-speed line cross cutting, its loop will be destroyed, and the path will become farther, so some people will suggest adding a capacitor to shorten its loop.
Personal understanding: When the high-speed line crosses the partition area, if no capacitor is added, then the return flow of these high-speed lines will bypass the partition area, which will form a large loop . This has the following problems:
One is to seriously increase the inductance of the signal path;
The second is that it may overlap with the return path loops of other signals, causing mutual inductance between the traces and increasing crosstalk;
The third is that a large loop will also form a large external radiation, which is not good for EMI; the fourth is that a large loop is easy to receive external interference.
But if a small capacitor connecting the two partitions is placed next to the high-speed line, then these high-speed lines will add a loop through the capacitor, because the loop area of this loop is greatly reduced, and the impedance is also small, so the signal This is the return path (the water flows to a low place, and of course the high-speed signal has to pick up the low-impedance path and run). In this way, some of the problems mentioned above will naturally be much better.
I think about EMI purely on the second floor (because I do EMI)... In fact, the ultimate goal should also be considered in terms of EMI. Some cross-cutting places will also add Bead and 0ohm resistors. The specific requirements are See what kind of cut it is across.
Are the two ends of the capacitor, one end is electricity, one end is ground or other? In my opinion, this capacitor functions as a "bridge". Generally, high-speed differential lines must have a complete ground plane. When this cannot be guaranteed, of course, a capacitor must be bridged. This is how I analyzed it.
Both ends are connected to the power supply.
Positive solution: When the high-speed line crosses the partition area, if no capacitor is added, then the return flow of these high-speed lines will bypass the partition area, which will form a large loop.
But I think it doesn’t matter whether the two ends of the capacitor are power or ground. For high-frequency signals, power and ground are the same. , Relative to the ground is a constant potential, you can regard it as the ground when doing AC analysis), so the power supply can also be used as a return path, but the effect may be slightly worse than that of the ground, right? For high-speed digital signals, due to the characteristics of RF, the nearest reference plane is always selected as the ground loop, regardless of whether this plane is ground or power. Therefore, when the signal crosses this dividing line, the reference loop becomes discontinuous, and the signal loop If it becomes larger, it is prone to EMI problems. Adding capacitors can make high-frequency signals have loops, reduce loop area and reduce EMI problems.
For the discontinuity of the ground plane, you can add BEAD and 0 OHM resistors to realize it. For the PCB board with the layer structure of TOP signal layer, power layer, ground layer, and BOTTOM signal layer. This applies to the partitioning groove where the bottom high-speed signal crosses the ground.
While the power supply plane discontinuous process, a high-frequency circuit can increase the capacitance, isolated DC circuit . For the above layer structure, the signal on the top layer crosses the power plane, and a capacitor is needed to provide an RF loop.
Mainly from the current loop consideration, in the case of cross-segmentation, the loop current will not pass, so it may go around very far, or even radiate into the air. But because it is a high-frequency circuit, it can be connected with a capacitor across the division (the capacitor is equivalent to a short circuit with respect to high frequency), and the loop current can easily pass through the capacitor, thus making the loop continuous.
As for the choice of capacitor, theoretically, of course, the larger the better, but the larger the capacitance, the larger the equivalent resistance and equivalent inductance, which is extremely unfavorable for high-frequency signals, and from the perspective of manufacturing , The large capacitance is worth the capacitance, and it takes up a lot of space, which often causes great difficulties. In short, to choose a cross-segment capacitor, it is necessary to choose a capacitor with a large capacitance value, but with a small equivalent resistance and inductance, a small volume, and a reasonable price.
I personally think that the larger the capacitance, the better, but it is determined according to fmax="1/3".14*tr, where 3.14 is pai
In high-speed circuits, if 0 ohm resistors or BEAN magnetic beads are used to straddle the division, basically two grounds are used for single-point connection, but sometimes it may not be directly connected with copper. .
When the capacitor is connected to two power supplies, the two ends of the capacitor are best connected to the corresponding two power supplies.
The calculation method is the capacitor with the resonance point at 50MHz. Take Murata0402 Y5V capacitor as an example. A capacitor of about 22nF should be used. The accuracy requirement is not high.
In fact, I find it very strange. Since it is a high-speed signal, why not give him priority and give him a complete ground plane for reference, which is much better than spanning two power planes.
As for the capacitance value of the capacitor bridge, it is better to be smaller, but ESL is the main factor that affects the high-frequency characteristics of the capacitor, so it is more important to choose a capacitor with a low ESL.
The key is to look at the reference layer. If the power layer is not used as a reference, but the complete ground layer is used as a reference, the ground layer can be used as a return loop.
Shouldn't a capacitor be connected across the power supply?
One of my understanding is that the ground corresponding to the power supply here may also be divided, for example, analog ground and digital ground. Then another solution is to use magnetic beads or 0 ohm resistors to connect between the ground .
Another understanding is that the complete ground layer and the signal layer to be reflowed are far apart, not adjacent. This method is also worth considering and is only a personal understanding.
I am a motherboard. Our company’s four-layer boards are TOP, POWER, GND, BOTTOM. High-speed lines generally go to the bottom layer. Gnd will only cut into agnd and gnd. Generally, there will be no cross-cutting problems, but other line crosses. The power supply will add a jumper capacitor
It depends on your signal frequency and the resonant frequency of the capacitor.
There is one additional point. The premise that the AC signal regards the power supply and the ground as the same plane ground is that the capacitors formed by the power supply and the ground and the added decoupling capacitors make the impedance between the power supply and the ground very high at this frequency point. Only low. .
Yes, it is to provide a complete return path for the tell signal and reduce EMI problems
103 experience points