|Publisher DOI:||10.1109/SPI54345.2022.9874946||Title:||Parametric S-Parameters for PCB based Power Delivery Network Design Using Machine Learning||Language:||English||Authors:||Schierholz, Christian Morten
|Keywords:||Artificial neural network; PB-Modeling; PCB; power delivery network||Issue Date:||May-2022||Source:||26th IEEE Workshop on Signal and Power Integrity (SPI 2022)||Abstract (english):||
In this contribution, a methodology using ANN techniques is presented for the analysis and design of power delivery network (PDN) in printed circuit board (PCB) design. The trained artificial neural networks (ANNs) are applied to answer relevant PDN design questions such as PCB resonance frequency and target impedance (TI) violations. Based on PCB geometry and material variations a PDN design space is defined. To train the ANNs, inside the design space a sparse sampling with 10000 physics-based (PB) simulations is performed. The S-parameter database is created using physics based via models which are validated by a commercial full-wave finite element method (FEM) solver in the frequency spectrum of 1MHz to 1GHz. The S-parameters are available in the SI/PI-Database. The flexibility of the unterminated S-parameters is demonstrated by a post processing termination using decoupling capacitor (decap) distributions. Limitations of the data generation are discussed with respect to computational resources and disk space.
|Conference:||26th IEEE Workshop on Signal and Power Integrity, SPI 2022||URI:||http://hdl.handle.net/11420/13756||ISBN:||9781665486255||Institute:||Theoretische Elektrotechnik E-18||Document Type:||Chapter/Article (Proceedings)|
|Appears in Collections:||Publications without fulltext|
Show full item record
checked on Feb 7, 2023
Add Files to Item
Note about this record
Cite this record
Items in TORE are protected by copyright, with all rights reserved, unless otherwise indicated.