Technical Design Proposal: Split-Type Passive Phase-Change Thermal Armor for PCIe 5.0 SSD
Technical Design Proposal: Split-Type Passive Phase-Change Thermal Armor for PCIe 5.0 SSD
Document No.: ENG-SSD-TH-2026-004
Version: Rev 6.0 (NPI Approved)
Confidentiality: Proprietary
Date: 2026-08-21
1. Project Overview
1.1 Background
A PCIe 5.0 SSD dissipates approximately 14 W under full load. In a fanless, passive cooling environment, the controller junction temperature (Tj) can easily exceed 80°C and trigger thermal throttling. This proposal adopts a "sandwich structure + HDT dual-pipe parallel + split extended fin array + Z-axis adaptive suspension" architecture to achieve fully passive, silent cooling while ensuring mass production yield and mechanical reliability.
1.2 Design Objectives
- Thermal Performance: At 35°C ambient (local in-chassis environment) and 14 W full load, controller junction temperature Tj ≤ 76°C (based on actual power distribution); also meet derating requirements for harsh chassis environments up to 45°C.
- Mechanical Compatibility: Compatible with single/double-sided M.2 2280 SSDs; no interference with adjacent PCIe slots or mainstream graphics cards; Z-axis height adaptive; pass system-level shock/drop validation to ensure M.2 connector solder joint reliability.
- Cost Control: Mass production unit cost ≤ 25 RMB.
- Noise Level: 0 dB.
2. Thermal Design
2.1 Thermal Resistance Network and Junction Temperature Prediction
The thermal resistance chain is based on the actual power dissipation path of the controller, with missing items such as spreading resistance, contact resistance, and fin efficiency now included.
Design Condition: Controller power 7.5 W, NAND/DRAM total power 6.5 W, total power 14 W; ambient temperature 35°C.
| Link Node | Physical Description | Parameter Setting | Thermal Resistance (K/W) | ΔT (Controller 7.5 W) |
|---|---|---|---|---|
| θjc | Junction → Package Case | Vendor typical value | 0.80 | 6.0°C |
| R_tim | Case → Heatpipe Base | 0.2–0.25 mm phase-change material | 0.83 | 6.2°C |
| R_spread | Spreading resistance (small-area heat source) | Copper/aluminum base spreading | 0.40 | 3.0°C |
| R_contact | Heatpipe-to-base expansion contact | Expansion bonding + thermal adhesive | 0.10 | 0.8°C |
| R_hp | Heatpipe phase-change transport | Dual D4 sintered heatpipes in parallel (HDT) | 0.15 | 1.1°C |
| R_conv | Fins → Ambient air | Natural convection, corrected fin efficiency | 1.10 | 8.3°C |
| ΣR | Total controller path resistance | — | 3.38 K/W | 25.4°C |
Controller Junction Temperature (35°C Ambient):
Tj = 35°C + 7.5 W × 3.38 K/W = 60.3°C
45°C Harsh Chassis Environment Derating Check (for high-end platforms with high GPU heat output):
Tj = 45°C + 7.5 W × 3.38 K/W = 70.4°C
Still below the 76°C design target and 80°C throttling threshold, with ample thermal margin.
Worst-Case Concentrated Power Check (14 W entirely applied to controller path, not encountered in practice):
Tj_max = 35°C + 14 W × 3.38 K/W = 82.3°C, exceeding the 80°C throttling threshold.
Note: This condition is for extreme verification only and is not an acceptance criterion. Under actual full load, controller power is approximately 7–8 W, and NAND/DRAM heat is conducted through multiple paths (base and backplate), keeping controller Tj stable at 60–65°C.
2.2 HDT Heatpipe System and Base Optimization
- Dual-Pipe Parallel Architecture: Two D4 sintered copper heatpipes in parallel, with a combined heat transport capacity of 16–20 W, ensuring no dry-out at 14 W full load. After flattening/bending, capacity may degrade by ~30%; measured parallel capacity must be ≥ 21 W (1.5× design margin).
- Heatpipe Direct Touch (HDT): Heatpipe bottoms are precision die-flattened and then directly contact the PCM, eliminating the intermediate copper baseplate thermal resistance. Minimum wall thickness after flattening ≥ 0.25 mm to prevent puncture.
- Z-Axis Stepped Profile: The aluminum extruded base bottom is CNC-machined to create a stepped structure—the controller area protrudes to match chip height, while NAND/DRAM areas are recessed accordingly. Different thickness interface materials ensure uniform contact pressure on all heat-generating components, eliminating contact failure due to height differences.
- HDT Bottom Surface Flatness Control (New in NPI): Die flattening alone cannot reliably meet the ≤ 0.05 mm inter-pipe gap tolerance. In NPI, a "precision flattening + micro fly-cutting" composite process is adopted: leave 0.1 mm allowance after flattening, then perform a minimal fly-cutting pass to achieve coplanarity. Ultrasonic thickness measurement is introduced to ensure the thinnest wall after machining is ≥ 0.25 mm, guaranteeing both flatness and integrity.
2.3 Fin Geometry and Convection Optimization
- Specifications: 6063 aluminum extrusion, dimensions 120 mm × 60 mm × 40 mm.
- Fin Spacing Optimization: The original 15-fin design had 3.64 mm spacing, causing boundary-layer overlap in natural convection. Revised to 12 fins, 0.6 mm thick, average spacing 4.8 mm, satisfying non-overlapping boundary-layer requirements and verified by CFD.
- Total Heat Exchange Area: Approximately 0.121 m² (including exposed base area). After fin efficiency correction, effective area ≈ 0.115 m².
- Natural Convection Coefficient: h = 7.5 W/m²·K (typical for in-chassis natural convection). R_conv = 1/(7.5 × 0.115) ≈ 1.16 K/W; with fin efficiency, 1.10 K/W is used.
3. Mechanical Design and Compatibility
3.1 Sandwich Structure
To address the single-point mounting weakness of M.2 interfaces, a symmetric clamping scheme is adopted:
- Top Heatsink: Carries the HDT heatpipes and fin array.
- Bottom Backplate: 0.6 mm 304 stainless steel stamped plate that grips the SSD rear edge.
Insulation Design: A 0.1 mm PI film or anodized insulation layer is applied to the backplate contact surface; chip locations have relief cutouts filled with thermally conductive insulating pads, eliminating short-circuit risk. - Fastening: Two M2 stepped screws pass through the top heatsink and bottom backplate, forming a closed clamping loop.
3.2 Z-Axis Adaptive Suspension System
- Floating Mechanism: M2 stepped screw shoulders limit at the motherboard standoff; spring-loaded screw ends provide 1.5 mm Z-axis adaptive travel.
- Pressure Control: Spring preload 18 N, total pressure 0.07–0.1 MPa, well below the BGA solder joint fracture threshold (0.5 MPa), while ensuring full wetting of the PCM interface.
Stress Verification: Spring force application point is near the controller top surface; PCB bending stress simulation confirms no excessive bending moment on the M.2 connector.
3.3 Spatial Interference Avoidance
- Offset Extension: The cooling tower is offset 20 mm toward the chassis front/memory area to avoid PCIe slot centerlines.
- Keep-Out Verification: Via 3D CAD assembly model, maintain ≥ 5 mm clearance to thick graphics card backplates such as RTX 4090; ≥ 3 mm to memory modules and their heatsinks; avoid motherboard capacitors, inductors, and other components.
- Center of Gravity Offset and Torsional Moment: Total heatsink weight is approximately 150–200 g; with 20 mm offset in a vertically mounted motherboard, torsional moment on the M.2 connector is possible. System-level drop/shock testing will be conducted in EVT. If micro-crack risk is detected, an optional auxiliary support bracket (connecting heatsink side to motherboard screw holes) will be designed to distribute torque.
4. Manufacturing Process and Workflow
4.1 Heatpipe Manufacturing Process (Revised Sequence)
Heatpipes must be manufactured and sealed before mechanical forming to prevent wick oxidation, working fluid leakage, and tube collapse.
Correct Sequence:
Raw tube → Cleaning → Sintered wick → Working fluid charging → Evacuation and sealing → Bending → Die flattening / machining of bottom surface → Leak test
- Flattening preferably by die pressing, leaving 0.1 mm allowance, followed by CNC fly-cutting for coplanarity; inter-pipe gap ≤ 0.05 mm.
- Minimum wall thickness after flattening and machining ≥ 0.25 mm; leak test required.
4.2 Interface Materials and Pump-out Control
- Controller Area: 0.2–0.25 mm high-thermal-conductivity phase-change material (PCM) to minimize thickness and inhibit pump-out after liquefaction. A containment dam (dispensed adhesive or gasket frame) or reservoir groove is added to further prevent PCM migration.
- NAND/DRAM Area: 0.5–1.0 mm standard high-thermal-conductivity gap filler pad to absorb thickness tolerances.
- Tolerance Requirement: After HDT bottom machining, inter-pipe gap depth ≤ 0.05 mm.
4.3 Assembly Flow
- Base Machining: Extruded aluminum base CNC-machined with stepped relief → embed dual D4 heatpipes → expansion bonding.
- Fin Attachment: Heatpipes pass through aluminum fin holes → hydraulic expansion bonding → black anodizing.
- Final Assembly: Apply PCM/thermal pads → place SSD → install stainless steel backplate (with insulation) → torque-lock stepped screws.
5. BOM Cost Estimate (10k Volume)
| Item | Specification / Process | Cost (RMB) | Remarks |
|---|---|---|---|
| Heatpipe Assembly | Dual D4 sintered heatpipes (HDT), sealed then flattened + fly-cut | 5.8 | Includes leak test & ultrasonic thickness |
| Aluminum Extruded Base | 6061 aluminum + CNC stepped milling | 2.0 | Includes stepped relief |
| Aluminum Fin Array | 6063 extrusion + expansion + anodizing | 7.0 | 120 × 60 × 40 mm, 12 fins |
| Fastener Set | M2 stepped screws + compression springs | 2.5 | 2 sets |
| Stainless Steel Backplate | 0.6 mm 304 SS stamped + PI insulation | 1.0 | Includes relief cutouts & insulating pads |
| Interface Materials | PCM (0.2 mm) + gap filler + containment dam | 3.0 | Area-specific materials |
| Assembly & Testing | Automated expansion + torque verification + initial yield monitoring | 3.7 | Includes QC & NPI yield buffer |
| Total | 25.0 RMB | Meets ≤ 25 RMB target |
Cost Control Contingency: PCM dam dispensing, stepped spring screws, and HDT leak testing carry hidden costs. First 1000-unit pilot run scrap rate will be monitored. If overall cost exceeds limit, supplier bidding and process simplification plans will be activated to keep mass production cost within target.
6. Technical Specification Summary
| Parameter | Specification | Verification Method |
|---|---|---|
| Controller Tj | ≤ 60.3°C (@35°C ambient, controller 7.5 W) | 14 W full load 30 min, IR thermal imaging |
| 45°C Harsh Environment Tj | ≤ 70.4°C (@45°C ambient, controller 7.5 W) | Wind tunnel / thermal chamber simulation |
| Total Controller Path ΣR | 3.38 K/W (incl. spreading, contact, fin efficiency) | Theoretical calculation + measured fitting |
| Fin Spacing | 4.8 mm (12 fins) | Caliper measurement |
| Heatpipe Configuration | Dual D4 sintered heatpipes (HDT), wall ≥ 0.25 mm | Cross-section / ultrasonic thickness / ΔT test |
| Compatibility | Compatible with single/double-sided M.2 2280 SSD, insulated backplate | Stepped screw travel verification + insulation withstand test |
| Clamping Pressure | 0.07–0.1 MPa | Pressure sensor measurement |
| Mass Production Cost | 25.0 RMB | BOM at 10k volume, incl. NPI yield buffer |
| EVT Validation Items | Pump-out / expansion peel force / vibration / high temp high humidity / heatpipe aging / pressure decay / insulation withstand / multi-orientation thermal / system-level shock | 2000 thermal cycles / peel ≥ 100 N / 5–500 Hz random vibration / 85°C 85% RH 1000 h / 1000 h ΔR ≤ 5% / 1000 h pressure decay ≤ 10% / 500 VDC insulation ≥ 100 MΩ / horizontal, vertical ±90° / 50 G 11 ms shock |
7. Validation and Test Plan (EVT/DVT)
| Test Item | Method / Condition | Acceptance Criteria |
|---|---|---|
| Thermal Performance | 14 W full load, 35°C ambient, 30 min | Controller Tj ≤ 76°C (actual power distribution) |
| 45°C High-Temp Derating | 14 W full load, 45°C ambient, 30 min | Controller Tj ≤ 76°C, no throttling |
| Multi-Orientation Thermal | Horizontal, vertical +90°, vertical −90° | Tj ≤ 76°C in all orientations, no heatpipe dry-out |
| Pump-out Effect | PCM area, 2000 thermal cycles (−20–85°C) | No visible PCM loss, ΔR ≤ 10% |
| Expansion Peel Force | Heatpipe-fin/base peel test | Peel force ≥ 100 N |
| Vibration Test | 5–500 Hz random vibration, 3 axes, 1 h each | No loosening, no heatpipe shift, no weld cracks |
| System-Level Shock/Drop | 50 G / 11 ms shock with motherboard, or equivalent drop | No M.2 solder joint micro-cracks, functional |
| High Temp High Humidity | 85°C / 85% RH, 1000 h | ΔR ≤ 5%, no corrosion |
| Heatpipe Aging | 1000 h full-load accelerated aging | ΔR ≤ 5%, no dry-out |
| Pressure Decay | Spring preload 18 N, 1000 h | Pressure decay ≤ 10% |
| Insulation Withstand | Backplate vs. SSD metal surface, 500 VDC | Insulation resistance ≥ 100 MΩ |
| Interference Verification | 3D CAD + mainstream motherboard fit test | ≥ 5 mm to GPU backplate, ≥ 3 mm to memory |
| HDT Wall & Flatness | Ultrasonic thickness + CMM | Wall ≥ 0.25 mm, gap ≤ 0.05 mm |
8. Risk and Mitigation
| Risk | Impact | Mitigation |
|---|---|---|
| Heatpipe capacity drop after flattening | Dry-out, increased resistance | Measure flattened capacity; require ≥ 21 W; upgrade to D5 if needed |
| PCM long-term pump-out | Increased interface resistance | Thin PCM + containment dam + reservoir groove |
| Backplate short circuit to SSD | Functional failure | PI film / anodized insulation + relief cutouts + withstand voltage test |
| Fin spacing too dense | Reduced natural convection | Revised to 12 fins, 4.8 mm spacing, CFD verified |
| Offset interference with memory/motherboard | Assembly interference | 3D CAD verification, ≥ 3 mm clearance |
| Torsional moment causes M.2 solder cracking | Long-term reliability | System-level shock/drop; optional auxiliary bracket |
| HDT flatness vs. wall thickness conflict | High contact resistance or puncture | Precision flattening + micro fly-cutting + ultrasonic thickness, wall ≥ 0.25 mm |
| Cost overrun | Production loss | Supplier bidding + process simplification + scale purchasing; monitor first 1000-unit scrap |
9. Conclusion
controller Tj ≤ 60.3°C (35°C ambient) or ≤ 70.4°C (45°C ambient), mass production cost 25.0 RMB, and complete mechanical compatibility and reliability validation plans. This proposal passes design review and is ready for NPI pilot production.
Prepared by: Thermal Engineering Department
Reviewed by: Technical Committee
Approved by: Program Director
Date: 2026-08-21