The rapidly expanding AI industry, with its escalating power demands, is driving significant innovation in power delivery infrastructure for data centers. A major advancement in this field is the increasing adoption of 800V HVDC (High Voltage Direct Current) power distribution, which requires highly efficient conversion to the 12V levels used by components such as GPUs. Fixed-ratio converters play a central role in this transition, providing several key benefits:
– Reduced resistive losses: 800V distribution lowers current, significantly minimizing resistive losses and improving overall system efficiency.
– Enhanced power density: Smaller component sizes increase power density.
– Lower costs: Reduced copper utilization and simplified maintenance decrease operating costs.
One of the critical elements in fixed-ratio converters is the magnetic structure. At the 2014 APEC Conference, Rompower introduced the concept of the multi-leg magnetic structure, later referred to as the matrix transformer. Rompower patented this design (U.S. Patent 10,937,590) several years before other similar patents emerged, establishing a pioneering position in this technology. The multi-leg magnetic structure is particularly well-suited for high-power fixed-ratio converters.
In addition, Rompower has developed a proprietary LLC topology that substantially reduces the RMS current through output capacitors, improving both performance and reliability. Together, these innovations provide highly effective solutions for high-power fixed-ratio conversion.
Figure 1
In Figure 1 is presented the general concept of fixed ratio converter having the input connected to a high voltage bus ( 400V or 800V) and producing an isolated output of (48V or 12V)
In this architecture, the output voltage is a fixed fraction of the input, determined by the turns ratio or K factor. Depending on application requirements, the power level can reach several kilowatts.
With its proven expertise, Rompower offers advanced engineering solutions tailored for these demanding applications, ensuring high efficiency, scalability, and cost-effectiveness in AI data center power delivery.
