MOSFET/IGBT Selection Strategy and Device Adaptation Handbook for AI Data Center Cooling Systems with High-Efficiency and Reliability Requirements

With the rapid growth of AI computing demands, data center cooling systems have become critical for maintaining operational stability and energy efficiency. The power conversion and motor drive systems, serving as the "heart and muscles" of cooling units, provide precise power delivery for key loads such as compressors, fans, and pumps. The selection of power MOSFETs/IGBTs directly determines system efficiency, thermal performance, power density, and reliability. Addressing the stringent requirements of data centers for high power, continuous operation, low noise, and fault tolerance, this article focuses on scenario-based adaptation to develop a practical and optimized device selection strategy.

### I. Core Selection Principles and Scenario Adaptation Logic

 

 

图1: AI数据中心制冷系统方案与适用功率器件型号分析推荐VBP16I60与VBGE1805与VBP16R07与VBP115MR03与VBK362K与VBM2603产品应用拓扑图_en_01_total

 

(A) Core Selection Principles: Four-Dimensional Collaborative Adaptation

Device selection requires coordinated adaptation across four dimensions—voltage, loss, package, and reliability—ensuring precise matching with system operating conditions:

- Sufficient Voltage Margin: For high-voltage buses (e.g., 400V AC-DC links), reserve a rated voltage withstand margin of ≥50% to handle surges and transients. For low-voltage buses (e.g., 12V/48V DC), ensure margin for peak stresses.

- Prioritize Low Loss: Prioritize devices with low Rds(on) or VCEsat (reducing conduction loss) and low switching losses (e.g., low Qg, Coss), adapting to 24/7 operation, improving energy efficiency, and reducing thermal load.

- Package Matching: Choose TO247/TO263 packages with low thermal resistance for high-power loads (e.g., compressors). Select compact packages like TO252 or DFN for medium-power fans, balancing power density and heat dissipation.

- Reliability Redundancy: Meet mission-critical durability requirements, focusing on thermal stability, wide junction temperature range (e.g., -55°C ~ 150°C), and robustness against ESD/surge events.

(B) Scenario Adaptation Logic: Categorization by Load Type

Divide loads into three core scenarios based on function: First, compressor drive (power core), requiring high-voltage, high-current handling. Second, fan/pump drive (efficiency-critical), requiring medium-power, high-efficiency operation. Third, power switching/auxiliary control (reliability-critical), requiring fast response and fault isolation. This enables precise parameter-to-need matching.

### II. Detailed Device Selection Scheme by Scenario

(A) Scenario 1: Compressor Drive (1-10kW) – High-Power Core Device

Compressors in chillers or refrigerant systems require handling high voltages (400V+), continuous currents, and startup peaks, demanding robust and efficient switching.

- Recommended Model: VBP16I60 (IGBT+FRD, 600/650V, 60A, TO247)

- Parameter Advantages: Fast-switching (FS) technology achieves low VCEsat of 1.7V at 15V, reducing conduction loss. Integrated FRD enhances reverse recovery performance. TO247 package offers low thermal resistance (RthJC ≤ 0.5°C/W) for heat dissipation. Voltage rating suits 400V AC-DC buses with ample margin.

- Adaptation Value: Enables efficient PWM-driven compressor control, reducing loss by 15-20% versus standard IGBTs. Supports frequency modulation up to 20kHz, improving temperature regulation precision. Meets 24/7 operation with junction temperature up to 150°C.

- Selection Notes: Verify compressor power and peak current (derate ICE to 70% at high ambient). Use with isolated gate drivers (e.g., IR2110) featuring desaturation protection. Ensure PCB creepage for high-voltage isolation.

(B) Scenario 2: Fan/Pump Drive (200W-2kW) – Efficiency-Critical Device

 

 

图2: AI数据中心制冷系统方案与适用功率器件型号分析推荐VBP16I60与VBGE1805与VBP16R07与VBP115MR03与VBK362K与VBM2603产品应用拓扑图_en_02_compressor

 

Fans and pumps for airflow/liquid cooling require medium power, high efficiency, and low noise, often operating on 48V or 24V DC buses.

- Recommended Model: VBGE1805 (N-MOS, 80V, 120A, TO252)

- Parameter Advantages: SGT technology achieves ultra-low Rds(on) of 4.6mΩ at 10V, minimizing conduction loss. 120A continuous current (peak ≥240A) suits 48V buses. TO252 package balances thermal performance (RthJA ≤ 50°C/W) and compactness.

- Adaptation Value: For a 48V/1kW fan (20.8A), single device loss is only 0.2W, increasing drive efficiency to >97%. Supports 10kHz-30kHz PWM for smooth speed control, reducing acoustic noise below 40dB. Aligns with data center PUE optimization goals.

- Selection Notes: Match with BLDC driver ICs (e.g., DRV8323) for sensorless control. Provide ≥150mm² copper pour and thermal vias for heat dissipation. Add snubber circuits to limit voltage spikes.

(C) Scenario 3: Power Switching/Auxiliary Control (50W-500W) – Reliability-Critical Device

Power distribution units (PDUs) or auxiliary loads (sensors, controllers) require reliable high-side switching, fault isolation, and fast response for system safety.

- Recommended Model: VBM2603 (P-MOS, -60V, -120A, TO220)

- Parameter Advantages: Trench technology yields low Rds(on) of 3mΩ at 10V, enabling minimal drop in high-current paths. -60V withstand voltage suits 48V DC buses with margin. TO220 package offers easy mounting and thermal management (RthJA ≤ 40°C/W).

- Adaptation Value: Enables intelligent power sequencing and fault isolation for cooling subsystems, with response time <5ms. Reduces standby loss by 30% through efficient switching. Supports hot-swap applications with current limiting.

- Selection Notes: Use NPN transistor or dedicated high-side driver for gate control. Add reverse-polarity protection for inductive loads. Ensure current derating to 80% of ID in continuous mode.

### III. System-Level Design Implementation Points

(A) Drive Circuit Design: Matching Device Characteristics

- VBP16I60: Pair with isolated gate drivers (e.g., ISO5852S) providing ≥2A peak current. Add miller clamp circuits to prevent shoot-through. Use 10Ω gate resistor and 100nF bootstrap capacitor.

- VBGE1805: Direct drive by PWM controllers (e.g., UCC27714) with 4.7Ω gate series resistor. Add 1nF gate-source capacitor for stability. Implement current sensing via shunt resistors.

- VBM2603: Use level-shifting circuits (e.g., NPN transistor with 1kΩ pull-up) for high-side drive. Add 100pF RC snubber across drain-source to dampen oscillations.

(B) Thermal Management Design: Tiered Heat Dissipation

- VBP16I60: Mount on heatsink with thermal grease (RthSA ≤ 1°C/W). Use 2oz copper PCB with thermal vias. Monitor junction temperature via NTC sensors.

- VBGE1805: Provide ≥200mm² copper pour on PCB, augmented with aluminum spreader if needed. Ensure ambient airflow from server fans.

- VBM2603: Attach to chassis or heatsink via TO220 tab. Use insulating pads for isolation. Keep power loop inductance minimal to reduce switching losses.

- Overall: Place devices near cooling exhausts. For liquid-cooled racks, integrate cold plates for direct cooling.

(C) EMC and Reliability Assurance

- EMC Suppression:

 

 

图3: AI数据中心制冷系统方案与适用功率器件型号分析推荐VBP16I60与VBGE1805与VBP16R07与VBP115MR03与VBK362K与VBM2603产品应用拓扑图_en_03_fanpump

 

- VBP16I60: Add 2.2nF CMC filter at input and RC snubbers (10Ω + 470pF) across IGBT terminals.

- VBGE1805: Use shielded cables for fan connections and ferrite beads on gate lines.

- VBM2603: Incorporate Schottky diodes (e.g., MBRB20100CT) for inductive load freewheeling.

- Implement star grounding and separate analog/digital zones on PCB.

- Reliability Protection:

- Derating Design: Operate VBP16I60 at ≤75% VCE rating; derate VBGE1805 current by 30% above 85°C ambient.

- Fault Protection: Use desaturation detection for IGBTs, overcurrent comparators (e.g., TLV1701) for MOSFETs, and overtemperature shutdown via driver ICs.

- Surge/ESD Protection: Add TVS diodes (e.g., SMCJ48A) at power inputs and gate resistors with ESD suppressors (e.g., SMF05C).

### IV. Scheme Core Value and Optimization Suggestions

(A) Core Value

- High-Efficiency Operation: System efficiency reaches >96% under typical loads, reducing data center PUE by 0.05-0.1 and lowering operational costs.

- Scalability and Reliability: Devices cover 400V-1500V ranges, supporting modular cooling designs. Robust packages ensure MTBF >100,000 hours.

- Cost-Effectiveness: Mature silicon technology balances performance and cost, suitable for large-scale deployment versus GaN/SiC alternatives.

(B) Optimization Suggestions

 

 

图4: AI数据中心制冷系统方案与适用功率器件型号分析推荐VBP16I60与VBGE1805与VBP16R07与VBP115MR03与VBK362K与VBM2603产品应用拓扑图_en_04_powerswitch

 

- Power Adaptation: For >10kW compressors, choose VBP115MR03 (1500V/3A) for HV links. For low-power fans (<100W), use VBK362K (dual-N-MOS) for space savings.

- Integration Upgrade: Adopt IPM modules (e.g., with integrated drivers) for compressor drives. Use VBP16R07 (600V/7A) for auxiliary PFC stages.

- Special Scenarios: Select automotive-grade variants (e.g., VBGE1805-Auto) for edge data centers in harsh environments. Implement predictive maintenance via current monitoring with VBM2603.

- Cooling Synergy: Pair VBGE1805 with temperature-controlled fan curves using MCU algorithms, enhancing energy-aware operation.

### Conclusion

Power MOSFET/IGBT selection is central to achieving high efficiency, reliability, and intelligence in data center cooling systems. This scenario-based scheme provides comprehensive technical guidance for R&D through precise load matching and system-level design. Future exploration can focus on SiC devices and digital power management, aiding in the development of next-generation sustainable cooling solutions to support AI infrastructure growth.