日本語
How to Balance Computing Performance and Thermal Management in Edge Computing Gateway Substrate?
As the core node between IoT terminal devices and the cloud, an edge computing gateway must support data acquisition, local computing, multi-protocol wireless communication, and secure transmission. This places extreme demands on the substrate design. Achieving the right balance between computing density and thermal management is now one of the most critical challenges in next-generation edge gateway hardware.
Industry pain points remain significant:
Nearly 30% of edge gateways underperform because substrate-level computing chip placement is suboptimal, limiting CPU/FPGA performance to 70% of their designed capability.
About 35% of gateways lack adequate thermal design, causing CPU/FPGA temperatures to exceed 105°C, triggering throttling or system crashes.
Many gateway substrates suffer from unstable power delivery, leading to signal dropouts when multiple modules operate simultaneously.
PCBGOGO specializes in high-density computing substrates for edge computing gateways, offering advanced thermal solutions such as copper-based thermal-electric separation and embedded copper blocks. Its gateway substrates are already deployed across industrial IoT, intelligent transportation, smart campus systems, and other environments.
This article focuses on the dual requirements of high computing performance and efficient thermal management, providing substrate design strategies to help R&D teams build high-reliability, high-stability edge computing devices.
1. Technical Requirements for Edge Computing Gateway Substrates
1.1 Industry Standards and Key Performance Metrics
An edge computing gateway substrate must comply with:
IPC-2221 (high-density PCB design)
PCI Express 4.0 (high-speed interface requirements)
IPC-2152 (PCB thermal performance design)
EMC/EMI standards such as IEC 61000
Core technical requirements include:
Minimum 8-layer stack-up
Minimum trace/space: 0.076 mm
PDN impedance: ≤ 50 mΩ
Substrate thermal conductivity: ≥ 1 W/(m·K)
Maximum core chip temperature: ≤ 95°C
Industrial temperature range: –40°C to 85°C
1.2 Key Design Challenges
High-density computing integration vs limited space
CPUs, FPGAs, memory, and wireless modules create routing congestion and raise signal interference risks.High-power thermal load
FPGA power consumption may reach 50–100W, creating severe thermal hotspots that demand fast heat conduction.High-speed signal integrity
PCIe 4.0 and DDR4 require strict impedance control to prevent attenuation or timing failures.Power delivery stability
Different modules require different voltages, making PDN design crucial for system stability.
PCBGOGO addresses these challenges through high-density placement simulation, advanced thermal-conduction structures, and high-speed signal optimization. Its gateway substrates achieve 99.3%+ yield and reduce chip temperatures by up to 30°C.
1.3 PCBGOGO’s Technical Capabilities
PCBGOGO utilizes:
Materials:
Shengyi S1130 (multilayer)
Rogers RO4350B (high-speed signal layer)
Copper-core thermal-electric separation substrates
Equipment:
LDI exposure (±0.01 mm trace accuracy)
6-axis drilling (±0.01 mm blind/buried via accuracy)
X-ray inspection (5 μm precision)
Manufacturing scale:
Two advanced production bases (Guangde & Shangrao)
Supports 8–32 layers
Rapid prototyping in 24 hours
Mass production within 3 days
AI-driven modeling:
AI-MOMS system for thermal simulation and computing-block layout optimization.
2. Practical Optimization Framework for Edge Gateway Substrate Design
2.1 Computing Module Layout: High Density With Low Interference
Chip placement guidelines:
Place CPUs (e.g., ARM Cortex-A55) and FPGAs (e.g., Xilinx Artix-7) near the substrate center.
Maintain ≥10 mm spacing between high-power chips to avoid overlapping thermal zones.
Keep DDR4/eMMC within ≤50 mm of the CPU to maintain signal integrity.
Functional zoning:
Divide the substrate into:
Computing zone (CPU, FPGA, memory)
Wireless zone (Wi-Fi, LoRa, Ethernet PHY)
I/O zone (USB, HDMI, serial ports)
Maintain ≥5 mm isolation space between zones.
High-speed routing:
Design PCIe 4.0 and DDR4 as 100Ω differential pairs
Length skew ≤ 3 mm
Use stripline (inner-layer) routing to minimize EMI
PCBGOGO achieves ±0.01 mm width tolerance and ±3% impedance tolerance.
2.2 Thermal Design: Efficient Heat Extraction From Core Chips
Materials selection:
For ≥50W chips: copper-core thermal-electric separation substrate
Copper thickness: 1.0–2.0 mm
Thermal conductivity: 50–200 W/(m·K)
For ≥100W FPGA workloads: embedded copper block
Copper block thickness: 2.0 mm
Conductivity: 385 W/(m·K)
Thermal structure guidelines:
Create a solid copper thermal pad at least 2× chip size
Use thermal vias (0.4 mm diameter, 5 mm pitch) to connect heat to backside copper
Use ENIG or bare copper to reduce heat resistance
System-level heat dissipation:
Pre-allocate mounting holes (2.5 mm) for heatsinks or heat pipes
Distance between mounting holes and chip center ≤ 10 mm
2.3 Power Delivery Network (PDN): Stable Power for Multi-Module Operation
Architecture:
Use a “main power + DC-DC conversion” structure
Example: 12V → CPU 0.8V / FPGA 1.2V / wireless module 3.3V
Power layer design:
Use an independent power plane (≥2 oz copper)
Couple power and ground layers tightly to reduce noise
PCBGOGO’s AI-CAM can simulate voltage drop and optimize layer distribution.
Decoupling strategy:
Place 0.1 μF + 10 μF + 100 μF capacitors within 3 mm of chip power pins
Add EMI filters at the power input
Ensure stable supply during peak loads
2.4 Interfaces and Expandability
I/O arrangement:
Place USB, HDMI, and Ethernet at the board edge for enclosure alignment
Keep ≥10 mm distance from the computing zone
Add TVS for ≥8 kV ESD protection
Future expansion:
Reserve PCIe slot, GPIO headers, sensor interfaces
Maintain industry-standard pad dimensions for compatibility
Testability:
Reserve test pads (≥0.8 mm) for high-speed signals and power rails
Keep ≥0.5 mm clearance from components
3. Case Study: Thermal and Signal Integrity Optimization for an Industrial Edge Gateway
3.1 Initial Issues
A client’s industrial IoT gateway substrate (CPU + FPGA + Wi-Fi 6 + Ethernet) suffered from:
FPGA temperature reaching 115°C, triggering severe throttling
DDR4 error rate 5%, causing processing delays
Power droop 0.3V, causing wireless module disconnections
3.2 PCBGOGO’s Optimization
Thermal enhancement:
Used copper-core thermal-electric separation substrate (1.5 mm copper core)
Added 40×40 mm copper thermal pad (3 oz copper)
Added 12 thermal vias (0.4 mm)
Attached backside heat pipe aligned with copper core
Signal/layout enhancement:
Moved DDR4 within 30 mm of CPU
Redesigned all DDR4 traces to 100Ω differential pairs, skew ≤2 mm
Enlarged isolation zones between computing, wireless, and I/O blocks to 8 mm
PDN enhancement:
Power plane increased to 2 oz
Added two additional 100 μF capacitors at CPU and FPGA
Switched to 95%-efficiency DC-DC converters
3.3 Results
Temperature reduced to 82°C, no more throttling
DDR4 error rate reduced to 0.1%, 40% faster processing
Power droop reduced to ≤0.05V, eliminating communication dropouts
Stable 99.4% manufacturing yield
Conclusion
Edge computing gateway substrate design must balance maximum computing integration, optimized thermal conduction, and stable power delivery. R&D teams should focus on:
Selecting appropriate thermal-management structures (copper-core separation / embedded copper block) based on chip power.
Optimizing high-speed routing to ensure signal integrity for PCIe 4.0 and DDR4.
Building a robust PDN architecture that maintains stability across all modules.
With advanced thermal materials, high-precision manufacturing, and AI-assisted layout simulation, PCBGOGO provides high-performance, production-ready substrates for edge computing gateways in industrial, transportation, and smart-city applications.
