Based On The Integrity Of The Power Supply Decoupling Network Design Method Research

The trend of power supply of the IC device in today’s digital system is towards higher speed, lower supply voltage, larger transient current. Therefore, in the board-level power distribution network, the design about decoupling network has been gradually transformed from traditional approach based on experience into the methodology based on power integrity, which means designing a reasonable decoupling network to control the impedance of the power distribution network under the target impedance from frequency domain can make the power supply noise to meet the requirements of the noise margin, and thus, achieve the decoupling purpose for power supply noise.The state-of-the-art design method of decoupling network is to use PI simulation software to do what-if analysis by two aspects. One is the combination of decoupling capacitors, the other is their location.For the designers who lack the rich experience, there are two problems of this method. Firstly, when calculating combination of decoupling capacitors, they need to do manually repeatedly adjustment and verification to be able to find a satisfactory solution and are failure to be confronted with the complex situation of using versatile kinds of capacitors. Secondly, when doing the analysis of the location of decoupling capacitors, they lack definite principle.For the two problems, three aspects follow:Firstly, analyze the model basis of decoupling network design. This part includes the lumped model and the distributed model of power distribution network. On one hand, research about the lumped model including analyzing the property and formula principle of every part, illustrating a simple method to calculate the effective impedance of lumped model, and using a recent free sotware to simulate the impedance profile of freuquence. On the other hand, research about the distributed model, use a open-source software about resonat cavity model, and analyze the impedance of power plane under the distributed model.Those resear about model not only illustrate some principle of doing decoupling network design,and also provide a platform to auto calculate the combination of decoupling capacitor.Secondly, analyze the method about the combination of decoupling capacitors. Introduce the method to select the capacitor. Introduce a simple feasible method to auto calculate the combination of decoupling capacitors, and point the limition of the method that it is unable to provide versatile option to select different kinds of capacitors and is unable to get the accurate result for the reason of calculation aobut mounting inducatance. We propose a is given improved method on the basis of the original method, which not only make a rapid accurate simulation result and also flexibly analyze the selection of capacitors using multiple capacitance value by decades. Finally, According to the improved method, we analyze combinations about decoupling capacitors under different objectives and requirements, and provide more design options to meet different design requirements about the types of capacitors and the total number of capacitors.Finally, analyze the design method about the location of decoupling capacitors. Introduce the concept about self impedance and transfer impedance under the distributed analysis, and analyze their pricinple and physical background, and propose using transfer impedance and self impedance to design the location of decoupling capacitor respctively.Combined with specific example, we show the procedure and the result of design.All relative facets of the decoupling network analysis and design are studied comprehensively in this dissertation, and the results are validated by the commercial software of Hyperlynx PI. And, the design method and the experience which we summaried are already applied to a practical project about decoupling network design for a single power supply.