With the exponential growth of AI computing demands and the critical need for operational sustainability, energy management systems have become the core of modern data center infrastructure. The power conversion and distribution systems, serving as the "vascular and neural network" of the facility, provide precise and robust power delivery to key loads such as high-voltage DC distribution, server power supplies (PSUs), and cooling fan arrays. The selection of power MOSFETs directly determines system efficiency, power density, thermal management overhead, and overall reliability. Addressing the stringent requirements of data centers for maximum power usage effectiveness (PUE), fault tolerance, and high availability, this article focuses on scenario-based adaptation to develop a practical and optimized MOSFET selection strategy.

I. Core Selection Principles and Scenario Adaptation Logic

(A) Core Selection Principles: Three-Dimensional Optimization

MOSFET selection requires coordinated optimization across three key dimensions—Voltage & Current, Losses, and Thermal & Package—ensuring precise alignment with the high-demand, continuous operation of data centers.

Voltage & Current Ruggedness: For high-voltage DC distribution (e.g., 380V/400V HVDC) and PSU bus lines (e.g., 12V/48V), prioritize devices with sufficient voltage margin (≥30% for primary side, ≥50% for secondary side) to handle transients and ensure long-term reliability. Current rating must support sustained and peak loads with ample derating.

Loss Minimization as Priority: Focus on ultra-low Rds(on) to minimize conduction loss in high-current paths and optimize Qg & Coss for switching loss reduction in high-frequency converters. This is paramount for improving energy efficiency and reducing cooling requirements.

Thermal & Package Suitability: Select packages (TO-247, TO-262, DFN) that balance current handling, thermal impedance (RthJC), and power density. Ensure the package can be effectively coupled to heatsinks or system cooling to maintain safe junction temperatures under 24/7 full load.

(B) Scenario Adaptation Logic: Categorization by Power Chain Segment

Divide applications into three core scenarios based on their position and function in the power chain:

1.  High-Voltage Power Distribution & PSU Primary Side: Requires high-voltage blocking capability and robust switching performance.

2.  High-Current, Low-Voltage Power Switching (PSU Secondary Side, VRM): Demands extremely low Rds(on) and high current capability for efficiency.

3.  High-Efficiency Power Conversion Modules: Requires an optimal balance of voltage rating, low loss, and package compactness for high-frequency operation.

II. Detailed MOSFET Selection Scheme by Scenario

图1: AI数据中心能耗管控系统方案与适用功率器件型号分析推荐VBP16R64SFD与VBMB195R09与VBA2309B与VBN1105与VBQF1104N与VBGQE11506产品应用拓扑图_en_01_total

(A) Scenario 1: High-Voltage Distribution & PSU Primary Side (380V/400V HVDC)

This scenario involves switching and protection in high-voltage rails, demanding high voltage withstand and good switching characteristics.

Recommended Model: VBP16R64SFD (Single-N, 600V, 64A, TO-247)

Parameter Advantages: Super-Junction (SJ_Multi-EPI) technology enables an excellent balance of high voltage (600V) and low Rds(on) (36mΩ). The 64A continuous current rating provides high power handling. TO-247 package offers superior thermal performance for heatsink mounting.

Adaptation Value: Ideal for active PFC stages, DC-DC converter primary sides, and high-voltage bus switches in 380V HVDC systems. Its low Rds(on) reduces conduction loss significantly at high currents, contributing to higher PSU efficiency (e.g., Titanium/Platinum levels).

Selection Notes: Ensure sufficient margin for voltage spikes. Pair with gate drivers capable of driving the moderate Qg. Careful PCB layout for high-voltage clearance is critical. Requires proper heatsinking.

(B) Scenario 2: High-Current, Low-Voltage Power Switching (12V/48V Server Bus, VRM)

This scenario focuses on the highest current paths where conduction loss dominates, requiring minimal Rds(on).

Recommended Model: VBN1105 (Single-N, 100V, 100A, TO-262)

Parameter Advantages: Trench technology achieves an ultra-low Rds(on) of 9mΩ (at 10V). An exceptional continuous current rating of 100A handles severe load currents. The 100V rating is well-suited for 48V bus applications with margin.

Adaptation Value: Perfect for synchronous rectification in high-current DC-DC converters, server 12V/48V bus distribution switches, and high-power point-of-load (POL) converters. Drastically reduces I²R losses, directly lowering energy waste and thermal load.

Selection Notes: Requires meticulous attention to PCB copper area and layout to minimize parasitic resistance and inductance. A robust gate drive is necessary for fast switching. Thermal management via PCB pours or a heatsink is essential.

(C) Scenario 3: High-Efficiency Power Conversion Modules (High-Frequency 48V-12V/5V Converters)

图2: AI数据中心能耗管控系统方案与适用功率器件型号分析推荐VBP16R64SFD与VBMB195R09与VBA2309B与VBN1105与VBQF1104N与VBGQE11506产品应用拓扑图_en_02_hv

This scenario demands devices that excel in both conduction and switching performance for high-frequency, efficient conversion.

Recommended Model: VBGQE11506 (Single-N, 150V, 100A, DFN8x8)

Parameter Advantages: SGT (Shielded Gate Trench) technology delivers an outstandingly low Rds(on) of 5.7mΩ combined with excellent switching figures (low Qg, Coss). The 150V rating is optimal for 48V-12V intermediate bus converters (IBCs). The DFN8x8 package offers low parasitic inductance and good thermal performance in a compact footprint.

Adaptation Value: Enables the design of ultra-high-efficiency, high-power-density IBCs and LLC resonant converters. Facilitates switching frequencies in the several hundred kHz range, allowing for magnetics size reduction and higher system power density.

Selection Notes: The DFN package requires a carefully designed PCB thermal pad with abundant vias to an internal plane for heat dissipation. High-speed gate drive design is critical to leverage its fast switching capability.

III. System-Level Design Implementation Points

(A) Drive Circuit Design: Matching Device Characteristics

VBP16R64SFD: Pair with isolated gate driver ICs (e.g., Si823x) with adequate drive current. Implement miller clamp functionality to prevent parasitic turn-on.

VBN1105: Use high-current, low-impedance gate drivers placed very close to the device. Optimize gate loop inductance.

VBGQE11506: Employ drivers optimized for high-speed switching. Use a small gate resistor to minimize switching times, balanced against EMI concerns.

(B) Thermal Management Design: Tiered and Aggressive Cooling

VBP16R64SFD & VBN1105: Mandatory use of aluminum heatsinks. Apply thermal interface material (TIM) properly. Consider forced air cooling from system fans.

VBGQE11506: Implement a large, multi-via thermal pad on the PCB connected to internal ground/power planes. Use thick copper (≥2oz). System airflow should be directed over the PCB area.

General: Implement NTC thermistors for temperature monitoring. Design control algorithms for fan speed modulation based on MOSFET temperature.

图3: AI数据中心能耗管控系统方案与适用功率器件型号分析推荐VBP16R64SFD与VBMB195R09与VBA2309B与VBN1105与VBQF1104N与VBGQE11506产品应用拓扑图_en_03_lv

(C) EMC and Reliability Assurance

EMC Suppression: Use snubber circuits (RC/RCD) across drain-source for high-voltage switches (VBP16R64SFD). Employ low-ESR decoupling capacitors very close to the drain and source pins of high-current devices (VBN1105, VBGQE11506). Implement proper input EMI filtering for each power stage.

Reliability Protection:

Derating: Adhere to strict derating guidelines (e.g., voltage ≤80% of rating, current derated based on case temperature).

Overcurrent Protection: Implement hardware-based desaturation detection for VBP16R64SFD. Use precision shunt resistors or inductor DCR sensing for VBN1105/VBGQE11506 circuits.

Overvoltage/Transient Protection: Utilize TVS diodes on gate pins and bus voltages. Implement Vds clamping circuits.

IV. Scheme Core Value and Optimization Suggestions

(A) Core Value

Maximized Energy Efficiency: Directly contributes to lower PUE by minimizing conversion losses across the power chain, potentially reducing facility energy consumption by 2-5%.

Enhanced Power Density and Scalability: The combination of high-performance devices enables smaller, more efficient power shelves, supporting higher compute density per rack.

Superior Reliability for Critical Infrastructure: Selected devices with robust ratings and recommended protection schemes ensure the high availability required for 24/7/365 AI data center operation.

(B) Optimization Suggestions

For Higher Voltage/Medium Power: Use VBMB195R09 (950V) for 3-phase AC input stages or boost PFC.

For Space-Constrained, High-Current Apps: Consider VBQF1104N (100V, 21A, DFN8) for distributed POL converters on server boards.

For Redundant Power Bus Control: Utilize VBA2309B (P-MOS, SOP8) for high-side switching in OR-ing controllers, saving space.

Integration Path: Explore power stage modules that integrate drivers and MOSFETs (like VBGQE11506 with a driver) to further simplify design and improve performance.

图4: AI数据中心能耗管控系统方案与适用功率器件型号分析推荐VBP16R64SFD与VBMB195R09与VBA2309B与VBN1105与VBQF1104N与VBGQE11506产品应用拓扑图_en_04_ibc