Issue link: https://iconnect007.uberflip.com/i/1540184
30 DESIGN007 MAGAZINE I OCTOBER 2025 T he evolution of adaptive system-on-chip (SoC) devices has brought unprecedented comput- ing capabilities to modern electronic systems. How- ever, with this advancement comes increasingly complex power delivery challenges. In particular, as organizations strive to accelerate time-to-mar- ket while managing costs, optimizing power deliv- ery networks (PDNs) has become a critical factor in successful product development. This article demonstrates advanced PDN optimi- zation techniques through a real-world case study of AMD's Versal adaptive SoC platform. Using the VCK190 evaluation kit featuring the Versal AI Core F E AT U R E A RT I C L E by Za c h C a p ra i , S i e m e n s E DA PDN Optimization: Balancing Performance and Cost in SoC Designs series VC1902 device, I'll explore how effective PDN design and optimization can help meet demanding technical specifications while addressing essential business goals. Understanding Modern PDN Challenges Modern SoCs require precise power delivery across a broad frequency spectrum, from DC to the giga- hertz range. The PDN must maintain stable volt- age levels while managing current transients and minimizing impedance across all frequencies. For configurable devices like the AMD Versal adaptive SoC, this challenge is particularly complex because power requirements vary sig- nificantly based on device configuration and application. The SoC's core voltage sup- ply V CCINT , with a voltage of 0.8 V, must maintain stability within a 3% operating range, with 1% allocated for DC tolerance and 2% for AC voltage ripple. An underperforming PDN can lead to severe consequences when subjected to increased current loads. Excessive voltage ripple, caused by the PDN's inabil- ity to maintain a stable voltage level, can introduce noise and distortions that compromise signal integrity. Furthermore, unacceptable voltage drops can starve critical components of the required power, leading to erratic behavior, data corrup- tion, and even complete sys- tem failure.