Computer-on-module
A computer-on-module (COM), also referred to as a system-on-module (SOM), is a compact, self-contained circuit board that integrates core computing components—including a processor, memory, storage, and essential interfaces—into a single modular unit designed to plug into a larger carrier board for customized application-specific functionality.[1][2][3] These modules separate the standardized computing hardware from the application-specific elements on the carrier board, enabling developers to focus on tailoring peripherals, power distribution, and I/O expansions without redesigning the core system.[1][3] Key components typically include a central processing unit (CPU) such as ARM or x86 architectures, dynamic random-access memory (DRAM) like DDR4, non-volatile storage options such as eMMC or NAND flash, power management integrated circuits (PMICs), and interfaces for connectivity including USB, Ethernet, HDMI, I2C, SPI, and UART.[2][4] Many COMs also incorporate graphics processing units (GPUs), wireless modules for Wi-Fi, Bluetooth, or GNSS, and support for expansion via standardized connectors to ensure interchangeability across vendors and processor generations.[3][2] COMs adhere to industry standards that define form factors, pinouts, and electrical interfaces to promote compatibility and longevity, with prominent specifications including COM Express (offering sizes like Basic at 125 mm × 95 mm for versatile applications from mobile devices to servers), SMARC (for compact, high-performance needs), Qseven (suited for low-power, rugged industrial use), and COM-HPC (targeting high-bandwidth computing).[1][3] These standards facilitate scalability, allowing modules to be upgraded for improved performance—ranging from low-power edge devices to high-end systems—while maintaining backward compatibility and extending product lifecycles beyond 10 years in many cases.[2][3] The primary advantages of COMs lie in their ability to accelerate development, reduce costs, and minimize risks in embedded systems design by providing pre-validated, production-ready hardware that shortens time-to-market by up to 9–12 months and lowers bill-of-materials expenses through scalable production.[2][4] They enhance flexibility for engineers to iterate on carrier boards for specific requirements, improve reliability via rigorous industry testing, and support sustainability by enabling hardware upgrades without full system overhauls, thereby maximizing return on investment.[3][1] COMs find widespread application in demanding sectors requiring robust, modular computing, such as industrial automation for control systems, healthcare devices like patient monitors, transportation and smart mobility solutions, Internet of Things (IoT) edge nodes for real-time data processing, telecommunications equipment, medical imaging, robotics, security and defense systems, and scientific research platforms.[1][2][3] Their modular nature makes them ideal for low-latency environments like warehouse tracking, smart cities, and energy management, where customization and performance scalability are critical.[4][2]Overview
Definition and Core Concept
A computer-on-module (COM), also known as a system-on-module (SoM), is a compact, self-contained computing subsystem that integrates essential components such as a central processing unit (CPU), memory, and basic peripherals onto a single, small circuit board without built-in user I/O connectors.[5] This design allows the module to plug into a custom carrier board via standardized connectors, enabling application-specific interfaces to be developed separately.[6] By encapsulating the core computing elements in this modular form, COMs provide a reusable foundation for embedded systems, distinct from full single-board computers that include all I/O on the board itself.[7] The primary purpose of a COM is to facilitate rapid prototyping and customization in embedded applications by decoupling the core processing hardware from the input/output (I/O) and expansion requirements tailored to specific use cases.[8] This separation reduces development time and costs for hardware engineers, as the standardized module handles the complex integration of processors and memory, while the carrier board focuses on peripheral connectivity, thereby accelerating time-to-market and enhancing design flexibility.[6] In practice, COMs support scalability across industries like industrial automation, medical devices, and transportation, where frequent upgrades to processing power are needed without redesigning the entire system.[5] In the basic operational model, the COM module manages core functions such as data processing, storage, and initial peripheral control, drawing power and signals from the carrier board to which it connects.[2] The carrier board, in turn, supplies regulated power, additional expansion slots, and customized interfaces—such as Ethernet for networking, USB for peripherals, or display outputs for human-machine interfaces—tailored to the application's needs.[6] This architecture ensures the module remains compact and focused, while the carrier enables seamless adaptation to diverse environments without altering the core compute unit.[7] COMs emerged in the 1990s as a response to the growing demand for standardized, upgradable computing solutions in embedded systems, where proprietary designs previously hindered scalability and longevity.[9]Key Components and Functionality
A computer-on-module (COM) typically integrates a central processing unit (CPU) as its core, supporting architectures such as ARM for low-power embedded applications, x86 for compatibility with PC ecosystems, and emerging RISC-V for open-source flexibility.[10][11][12] These processors handle computation, with examples including NXP i.MX series for ARM-based modules and Intel Core Ultra for x86 variants, enabling tasks from real-time control to AI inference.[13][12] Memory in COMs consists of dynamic random-access memory (DRAM), commonly DDR4 or DDR5 variants, with capacities scaling up to 128 GB in modern designs.[12][14] Storage interfaces include embedded MultiMediaCard (eMMC) for onboard flash up to 128 GB and support for NVMe over PCIe for high-speed solid-state drives, providing reliable boot and data persistence without relying on external media.[15][10] Peripheral components enhance functionality, including integrated graphics processing units (GPUs) optimized for AI acceleration, such as those in NVIDIA Tegra or NXP i.MX modules, which offload parallel computations from the CPU.[10] Power management integrated circuits (PMICs) regulate voltage and current for efficient operation across varying loads, often supporting features like dynamic scaling to extend battery life in portable systems.[16] Essential buses on the module include PCIe for expansion, USB for connectivity, and I2C for low-speed sensor interfacing, ensuring foundational communication without application-specific tailoring.[10][17] The software stack begins with firmware for hardware initialization, followed by bootloaders like U-Boot, which loads the operating system kernel and configures peripherals during startup.[18] COMs support diverse operating systems, including Linux distributions via Yocto Project for customizable embedded use, Windows IoT for x86 compatibility in industrial settings, and real-time OS like VxWorks for deterministic performance in safety-critical applications.[19][20] These layers abstract hardware details, allowing developers to focus on application logic. Functionally, COMs expose standardized interfaces, such as MXM-style connectors with up to 314 pins, for power delivery and high-speed data transfer to a carrier board, which then adds application-specific I/O like displays or sensors to maintain modularity.[21][22] This boundary ensures the module remains a reusable compute core, independent of end-use peripherals. The carrier board extends these interfaces to full system functionality in one integration step.[10]History and Evolution
Origins and Early Development
The concept of the computer-on-module (COM) emerged in the late 1990s as a response to the limitations of proprietary embedded boards prevalent in industrial computing, where custom designs led to high costs and inflexibility for original equipment manufacturers (OEMs). These early modules aimed to separate core computing functions from application-specific hardware, allowing easier upgrades without full system redesigns, particularly in sectors like automation and telecommunications that required reliable, scalable solutions amid the transition from ISA to PCI bus architectures.[23] Pioneering efforts began around 1998–2000, with companies such as Kontron (formerly JUMPtec) and ADLINK developing initial concepts to evolve standards like PC/104, which had been introduced earlier for compact embedded systems. Kontron's ETX specification, launched in January 2000, marked one of the first formalized COM approaches, featuring a compact form factor of approximately 95 mm × 114 mm that integrated x86 processors, memory, and essential I/O interfaces like serial ports, USB, Ethernet, and graphics, connected via custom high-density connectors to support both ISA and PCI buses on carrier boards.[23][24] ADLINK, building on its PC/104 heritage from Ampro, contributed to these pre-standard modules by focusing on rugged, x86-based designs for industrial PCs, typically sized around 100 mm × 100 mm, to address the need for off-the-shelf components in harsh environments.[24] Early implementations faced significant challenges, including a lack of interoperability between vendors' custom connectors and pinouts, which perpetuated vendor lock-in and increased development risks for OEMs in fast-evolving markets. This spurred collaborative efforts toward open standards, such as the formation of the ETX Industrial Group in 2001 by JUMPtec and Advantech, laying groundwork for broader adoption.[23] By the mid-2000s, these issues drove the transition to unified specifications like COM Express, enhancing compatibility across the industry.[24]Major Milestones and Standards Adoption
The introduction of the COM Express specification in 2005 by the PCI Industrial Computer Manufacturers Group (PICMG) marked a pivotal milestone in standardizing computer-on-module (COM) designs for embedded systems. This initial ratification defined a modular architecture with four module sizes—Mini, Compact, Basic, and Extended—and established eight pinout types, including Types 1, 2, 3, and 10, which separated processor and I/O functions to support high-performance applications like edge processing and networking through standardized interfaces such as PCI Express and Serial ATA.[25] During the 2010s, complementary standards addressed the demand for more compact and energy-efficient modules. The Qseven specification, first released in 2008 by the Qseven Consortium and subsequently maintained by the Standardization Group for Embedded Technologies (SGET), introduced a 70 mm x 70 mm form factor with a high-speed MXM connector, enabling low-power designs up to 12 W for mobile and IoT applications while supporting both x86 and ARM architectures.[26][27] In 2013, SGET ratified the SMARC (Smart Mobility ARChitecture) standard, which utilized 82 mm x 50 mm or 82 mm x 80 mm modules with a 314-pin edge connector to facilitate ARM-based mobile computing and low-power SoC integration, incorporating emerging MIPI interfaces for camera and display connectivity in portable embedded systems.[28][29] The 2020s brought updates focused on advanced computing demands, including AI and high-bandwidth connectivity. PICMG released COM Express Revision 3.1 in summer 2022, enhancing Type 6, 10, and 7 modules with support for PCIe Gen 4 (up to 16 lanes), USB 4.0, MIPI-CSI 2.0, and a 16 Gbps connector upgrade, which facilitated the integration of AI accelerators like neural processing units (NPUs) for edge AI processing in industrial applications.[25][30] Concurrently, RISC-V-based COM modules gained traction starting in the early 2020s, with vendors developing compliant designs leveraging the open-source instruction set architecture for customizable, royalty-free embedded solutions.[31] Standardization efforts also advanced for 5G and edge AI, as seen in SGET's SMARC 2.2 maintenance update in 2024 and the release of the SMARC Design Guide v2.2 in June 2025, alongside PICMG's COM-HPC specification (ratified 2021), which supports PCIe Gen 5 lanes for server-class performance.[32][33][34] By 2025, COM standards had achieved widespread adoption in industrial embedded systems, with the global COM market reaching approximately USD 1.7 billion as of 2024 and projected to grow at a CAGR of 4.22% through 2033, largely coordinated by organizations like SGET and PICMG to ensure multi-vendor interoperability and scalability.[35]Design and Architecture
Module Design Principles
The design of computer-on-module (COM) systems fundamentally relies on modularity principles that separate the core compute elements—such as the processor, memory, and essential interfaces—from application-specific I/O and expansion capabilities provided by a carrier board. This separation enables CPU and memory upgrades without necessitating a complete system redesign, reducing development time and costs for embedded applications. High-density connectors, typically ranging from 220 to 400 pins in standards like COM Express and COM-HPC, facilitate this interface by supporting high-speed data transfer while maintaining a compact form factor.[25][34] Reliability in COM design emphasizes robust environmental tolerance to ensure operation in demanding industrial settings. Thermal management incorporates heat spreaders and optimized stack heights to dissipate heat effectively, supporting industrial temperature ranges from -40°C to 85°C for extended operational stability. Modules are engineered for shock and vibration resistance compliant with IEC 60068-2-27 and -2-6 standards, often featuring soldered components like memory and storage to enhance durability against mechanical stresses. Additionally, manufacturers commit to long lifecycle support exceeding 10 years, minimizing obsolescence risks through standardized components and extended availability.[36][37][38] Scalability is achieved by accommodating multi-core processors and heterogeneous computing architectures that integrate CPU with GPU or other accelerators, allowing modules to evolve with advancing silicon technologies. Power efficiency targets typically range from 5W to 50W TDP, balancing performance needs across low-power edge devices and higher-throughput applications while adhering to wide input voltage tolerances like 4.75V to 20V.[34][39] Customization in COMs is inherently limited to board-level assembly during manufacturing, with all critical components soldered in place to guarantee consistency, reliability, and compliance with form factor specifications; post-manufacture, modules offer no user-modifiable parts to prevent compatibility issues or void warranties.[25]Integration with Carrier Boards
Computer-on-module (COM) systems achieve complete functionality by integrating the module with a carrier board, which provides the necessary mechanical support, electrical connectivity, and application-specific interfaces. This integration follows standardized specifications such as COM Express, SMARC, and Qseven, enabling modular designs where the COM handles core computing while the carrier extends I/O capabilities. The process ensures reliable signal transmission and power delivery, typically through high-density connectors that support both low-speed control signals and high-bandwidth data lanes.[25][29][27] Mechanical integration primarily relies on board-to-board connectors and mounting mechanisms to secure the COM to the carrier board. In COM Express, dual 220-pin connectors (for Basic and Extended types) or a single connector (for Mini types) facilitate attachment, with standardized mounting holes allowing screws or clips for fixation; torque specifications of 0.5 Nm are recommended for M2.5 screws to ensure stability in rugged environments. SMARC modules use a single 314-pin, 0.5 mm pitch MXM-compatible connector with standoffs and thermal pads for mounting, supporting form factors of 82 mm × 50 mm or 82 mm × 80 mm and enabling direct soldering or clipping for vibration resistance. Qseven employs a 230-pin MXM edge connector with 5.0 mm board spacing, secured via M2.5 standoffs (≤5.6 mm diameter), which suits compact, low-profile assemblies in industrial settings. These approaches prioritize alignment precision and heat dissipation, often incorporating aluminum heat spreaders or clips to maintain thermal contact without compromising structural integrity.[40][41][42] Electrical interfaces are defined by standardized pinouts that allocate pins for power, control signals, and high-speed communications, ensuring compatibility across vendors. Power delivery typically includes 3.3 V, 5 V, and 12 V rails; for instance, COM Express specifies a 12 V main rail (up to 5.5 A for PCIe x16) with 3.3 V standby (375 mA) via connector pins, while SMARC uses a 3.0–5.25 V input (up to 5 A total) without dedicated standby rails, and Qseven relies on a single 5 V rail with optional 3 V RTC backup. Signal pinouts support low-speed interfaces like LVDS for displays (e.g., 4 differential pairs in COM Express, up to 24-bit in SMARC), SPI/I²C for sensors (multiple channels at 3.3 V or 1.8 V, 100–400 kHz), and GPIO (up to 8 in COM Express). High-speed connectivity includes up to 16 PCIe lanes in COM Express (Gen 4, 92 Ω differential impedance), 4 PCIe lanes in SMARC (Gen 3), and x4 in Qseven, routed with length matching (e.g., <5 mils mismatch) to preserve signal integrity over carrier traces up to 15.85 inches. These allocations allow carriers to route signals to external ports while adhering to impedance controls (100 Ω for LVDS, 92 Ω for PCIe).[25][40][41] Customization of carrier boards involves designing application-specific I/O using vendor-provided tools and reference schematics, which accelerate development and reduce time-to-market. For example, Toradex offers design kits with schematics for adding interfaces like CAN bus (via dedicated pins in SMARC/Qseven) or HDMI (converted from SDVO/LVDS signals using chips like Silicon Image SIL1364 in COM Express). Reference designs from PICMG and SGET include pre-validated layouts for USB, SATA, and Ethernet expansion, allowing engineers to integrate features such as MIPI CSI for cameras or Gigabit Ethernet without altering the COM. This process typically starts with CAD libraries for pinouts, followed by simulation for routing, enabling tailored solutions for sectors like automotive or medical while maintaining standard compliance.[40][29] Testing and validation focus on ensuring reliable operation post-integration, encompassing signal integrity, electromagnetic compatibility (EMC), and environmental robustness. Signal integrity checks involve verifying differential pair impedances and eye diagrams for high-speed lanes (e.g., PCIe insertion loss budgets <2.5 dB in Qseven), often using tools like oscilloscopes to confirm no exceedance of trace length limits. EMC compliance requires filters (e.g., PI-filters on video lines) and grounding planes to meet CE/FCC standards, with clocks routed internally to minimize emissions. In industrial applications, hot-swap limitations are tested via bus switches (e.g., 74CBT3306 for SMBus isolation), preventing damage during module exchanges, while overall validation includes power sequencing simulations and thermal cycling to validate ruggedness. These steps, guided by specifications, confirm the assembled system's performance under operational stresses.[40][42][41]Standards and Form Factors
Primary Standards
COM Express, developed by the PCI Industrial Computer Manufacturers Group (PICMG), is a widely adopted standard for computer-on-module (COM) designs, defining pinouts for types 1 through 10 to accommodate various application needs, with types 6, 7, and 10 emphasized in recent revisions for modern interfaces.[25] It supports x86 and ARM architectures, enabling flexible processor integration across performance levels.[43] Modules connect via 220-pin (Type 10, mini form factor) or 440-pin (types 1-9, using dual 220-pin connectors) gold-finger edge interfaces, providing high-bandwidth signals like PCIe, Ethernet, and display outputs.[25] The latest revision, 3.1 released in 2022, introduces support for USB4 with DisplayPort alternate mode and 10GbE, enhancing connectivity for edge computing and networking applications.[44] SMARC (Smart Mobility ARChitecture), standardized by the Standardization Group for Embedded Technologies (SGET), targets low-power embedded systems with a 314-pin, 0.5mm-pitch MXM connector for reliable signal integrity in compact designs.[29] It is optimized for ARM and RISC-V processors, emphasizing energy-efficient SoCs suitable for battery-powered devices.[29] Modules adhere to a maximum size of 82mm x 50mm (short form factor), though a long variant up to 82mm x 80mm is defined for additional expansion.[29] The standard focuses on applications in mobile and consumer electronics, such as portable medical devices and infotainment systems, with power consumption typically under 6W to support extended operation in space-constrained environments. The SMARC Module Specification v2.2 was released in June 2025, providing updates for improved functionality and compatibility.[29][33] Qseven, also governed by SGET, employs a 230-pin MXM connector to deliver a compact, standardized interface for entry-level to mid-range COMs in graphics-intensive scenarios.[27] Its square 70mm x 70mm or rectangular 40mm x 70mm form factors enable integration of core components like CPU, GPU, memory, and I/O, supporting up to 12W TDP for balanced performance in embedded graphics applications.[27][45] The design prioritizes modularity for sectors like digital signage and industrial visualization, where direct graphics acceleration via LVDS or HDMI is essential.[27] Governance for these standards ensures industry-wide consistency: PICMG, an international consortium of over 200 member companies, oversees COM Express through collaborative subcommittees focused on open specifications for high-performance computing.[46] SGET, a German-based non-profit standards organization, manages SMARC and Qseven, promoting interoperability across vendors via detailed pinout definitions and design guides.[47] Both bodies facilitate certification programs, where modules undergo compliance testing to verify adherence to electrical, mechanical, and functional requirements, thereby guaranteeing seamless integration with carrier boards from multiple suppliers.[25][29]Form Factor Comparisons
Computer-on-module (COM) form factors vary significantly in physical dimensions and power consumption, influencing their suitability for different embedded applications. The COM Express standard, governed by PICMG, primarily utilizes a basic form factor of 95 mm × 95 mm for its compact variant, with options extending to 95 mm × 125 mm for higher-performance needs, supporting power budgets up to 137 W to accommodate demanding processors and peripherals.[25] In contrast, the SMARC standard from SGET features a smaller footprint of 82 mm × 50 mm (full-size) or 82 mm × 80 mm, optimized for ultra-low power scenarios with typical consumption under 6 W, enabling efficient deployment in battery-constrained devices.[29] Qseven, also under SGET, adopts a square 70 mm × 70 mm layout with power envelopes generally limited to 6–12 W, striking a balance between compactness and moderate processing capabilities.[27][45] These differences highlight trade-offs: larger COM Express modules offer greater computational density at the cost of higher power draw and space, while SMARC and Qseven prioritize portability and energy efficiency, often at reduced peak performance.[48]| Form Factor | Dimensions (mm) | Typical Power (W) | Key Trade-offs |
|---|---|---|---|
| COM Express | 95 × 95 (compact) to 95 × 125 | Up to 137 | High capability vs. larger size and power |
| SMARC | 82 × 50 or 82 × 80 | Typical <6 | High density vs. limited bandwidth |
| Qseven | 70 × 70 | 6–12 | Balanced vs. fewer high-speed pins |