
A Review of Power Distribution Architecture Solutions for Servers within Data Centers by Dorin Neacsu
- When
- Length
- 1 hr 30 min
- Platform
- GoTo Webinar
- Topic
- Technology
- Price
- Free
About this webinar
Abstract:
Servers and data centers continue to grow the demand for higher efficiency and higher power density. This lecture focuses on defining their power distribution systems in respect to standards, regulation, big-corporation plans, and power system realities.
First, definitions for general servers, rack servers, and blade servers are provided. Requirements for reliability, availability, and installed power are explained using Uptime Institute’s and TI’s classifications. A decision between low-voltage and medium-voltage supply is herein recommended based on power level by means of regulation for installed power as defined within various countries.
Solutions towards intermediary bus of either 48/54 V or 12 V are discussed for both AC and HVDC supply. The “80 Plus” efficiency standards used by industry are included for qualification of the AC-to-12 V conversion units (a.k.a. PSU). Solutions for 48/54 V or 400/800 V DC distribution bus inside the cabinet are explained, as the Open Compute Project is currently standardizing 800 V. A variant based on a series connection of servers is discussed.
The lecture continues with various on-the-board circuits for power conversion from intermediary DC bus towards a low-volt-range digital load – the voltage regulators (VR) or point-of-load (PoL) converters, in either lateral, vertical, or on-chip location. Circuits are classified and compared for their benchmark efficiency and power density performance, represented with either contemporary industrial products or research results. The improvements with GaN transistors and serial communication (PMBus) are demonstrated.
The lecture is rich in examples and represents a great introduction to the amazing opportunity for grid-to-chip funded research.


