Delete it if necessary, this capacitor should not exist

yvonneGan

Delete it if necessary, this capacitor should not exist [Copy link]

Author: Jiang Jie (Member of Mr. Gaosu's team of Yibo Technology's self-media)

2020 is halfway through, the domestic epidemic is gradually disappearing, and the situation is improving! The team led by Captain Gaosu has been running non-stop at the Shanghai Munich Electronics Show and seminars in various places. Wherever they go, the seats are full.

The R&D team in the back is also not relaxing for a moment, responding to various technical needs of domestic and foreign customers in a tense and orderly manner. The case that Mr. Gaosu encountered recently happened to echo the previous short video "How to hold Gbps high-speed signals with 100 MHz power decoupling", and by the way, answer the question at the end of the video.

Briefly introduce the background of the case: the customer provided power supply information and board-level capacitor configuration, and we need to do PDN impedance simulation of the power supply. Around the PDN impedance simulation report, we launched three rounds of friendly exchanges.

Customer question: The decoupling frequency band currently displayed is only up to 100MHz, which is not enough to show the role of planar equivalent capacitors. I hope to expand the frequency band to GHz.

Regarding the decoupling effect of planar equivalent capacitors, Mr. Gaosuo has actually done research before. In order to avoid the interference caused by the difference in decoupling frequency bands of capacitors with different capacitance values, the test board uses capacitors with the same capacitance, packaging and other parameters. Finally, the effect of the decoupling system at different frequency bands at the board level is shown in the following figure:

(, downloads: 0)

It can be found that the decoupling effect of planar equivalent capacitors is not as good as expected, and the planar capacitors and board-level capacitors usually have a relatively high parallel resonance point, which is also not conducive to filtering in the mid- and high-frequency bands. Of course, after explaining the principle to the customer, it is also necessary to provide data to be convincing. Therefore, we widened the frequency band of PDN impedance to GHz according to the customer's requirements.

(, downloads: 0)

The customer did not lose heart when he saw that the planar equivalent capacitors that he had high hopes for were not powerful, and proposed to add two pF-level capacitors to see if the impedance of the PDN high-frequency band could be suppressed.
(, downloads: 0)
For this solution, I have to give a thumbs up to the customer, because being able to come up with this solution shows that the customer is very familiar with the frequency band of capacitor decoupling. This solution is also feasible in theory, but another meaning of theoretical feasibility is that it is not practical. So, the competition entered the second round.

Let's take a look at the impedance curve of a 100pF capacitor in a 0402 package of a certain model of Murata:

(, downloads: 0)

The resonant frequency of this capacitor is 678MHz. Although it does not reach 1GHz, it is already a great improvement compared to uF-level capacitors. So, what is the PDN impedance after adding two pF-level capacitors to the board according to the customer's requirements? The result disappointed the customer again. The impedance curves before and after the capacitors were added basically overlapped, with almost no change!

(, downloads: 0)

The customer was silent for a while, and he said a word in anger: add capacitors! Seeing the customer's persistence, Mr. Gaosuo wanted to persuade him but couldn't bear it. In the end, he decided to use the simulation results to speak.

This time, we simply added 10 1nF and 100pF capacitors at a time to see what the effect would be.

(, downloads: 0)

The result made the customer collapse. After adding 20 capacitors, there was no splash at all, and the impedance curve remained calm. However, this was within Mr. Gaosuo's expectations. The key to the problem lies in the installation inductance of the board-level decoupling capacitor. Since the frequency band of pF-level capacitor decoupling is higher, and in this frequency band, the inductance will play a leading role, and the board-level capacitor is greatly limited by the installation inductance (see the yellow path in the figure below for details), it will eventually annihilate the decoupling effect of such capacitors, which is also the biggest obstacle for other board-level small-capacitance capacitors to play a decoupling role.
(, downloads: 0)
Back to the title of this article, as a rigorous engineer, it is necessary to add: delete it when necessary, and board-level decoupling capacitors with a capacitance of pF should not appear.

qwqwqw2088

Is the impedance curve in the article made by software simulation?

Actually solder or remove capacitors on the board, simulate with software, and then draw conclusions

Delete it if necessary, this capacitor should not exist [Copy link]

Latest reply