Product Integration Guide

SOM / SBC adoption guide

Adopting a SoM (System on Module) lets you hand off the difficult parts around the SoC — high-speed memory routing, power sequencing, and BSP maintenance — to a module that is already a finished product. On the other hand, there are SoM-specific considerations, such as carrier board design and the outlook for long-term availability. Here we explain the flow from selection to mass production in practical order.

01 / Overview

Overview of the setup

Development with a SoM goes through four broad stages. Each stage involves different decisions.

  1. SelectionNarrow down candidates by application, performance, interfaces, operating temperature, and supply period. Decisions made here affect every later step.
  2. EvaluationVerify on actual hardware using an evaluation kit or evaluation unit. This is the stage to check heat generation, real-world performance, and BSP maturity, which catalog values do not reveal.
  3. DesignDesign the carrier board, following the SoM specifications (pinout, power requirements, and mechanical dimensions).
  4. Mass productionConfirm the supply plan and EOL (end of life) handling. In industrial equipment, supply over a decade or more is not unusual.
Common mistakes

Skipping evaluation and going straight to design is a common case. Even if the catalog says requirements are met, problems such as insufficient heat dissipation or immature drivers for specific peripherals surface later on actual hardware. Once the board has been fabricated, the cost of fixes rises sharply. Please make use of our evaluation unit loans.

02 / Selection

How to approach product selection

Starting from “which SoC is best?” leaves too many options to decide. It is more practical to narrow down from your requirements.

Items to check in order of priority

Items to check in order of priority
OrderChecklistDecision points
1Required interfacesNumber of Ethernet ports, CAN, camera inputs, and so on. SoCs with too few can be ruled out immediately
2Operating temperature rangeIndustrial applications often require -40 to 85℃, and few products support it
3Processing performanceFor AI inference, check NPU performance (TOPS); for video processing, check VPU codec support
4Supply periodHow many years after the start of mass production do you need? For 10 years or more, you are limited to industrial SoCs
5Power consumptionFor battery-powered devices and sealed enclosures, this directly affects thermal design
6OS / BSPWhether the manufacturer provides a BSP for the OS you want to use

The pin-compatible family option

Our Chinchilla series is designed to be Pin2Pin compatible. You can swap in SoMs with different performance while keeping the same carrier board.

The pin-compatible family option
ProductsSoCDimensions
SSOM-i.MX8MP-ChinchillaNXP i.MX 8M Plus20 x 50 mm
SSOM-i.MX8MM-ChinchillaNXP i.MX 8M Mini20 x 46 mm
SSOM-i.MX935-ChinchillaNXP i.MX 9320 x 50 mm
SSOM-PX30-ChinchillaRockchip PX3020 x 50 mm
SSOM-RK3588-Chinchilla (V1)Rockchip RK358860 x 45 mm
SSOM-RK3588-Chinchilla (V2)Rockchip RK358839.4 x 47 mm
SSOM-RK3588S-ChinchillaRockchip RK3588S39.4 x 47 mm
SSOM-RK3576-ChinchillaRockchip RK357647 x 39.4 mm
SSOM-i.MX95-ChinchillaNXP i.MX 9540.0 x 28.05 mm

This structure has two practical advantages. One is product lineup expansion: higher-end and lower-end models can be built on the same carrier board. The other is diversification of SoC supply risk: even if the adopted SoC becomes hard to obtain, you can migrate to another SoC without redesigning the carrier board.

03 / Carrier Board

Carrier board design

The biggest advantage of using a SoM is that the difficult routing is confined to the module. Length-matched DDR memory routing and the SoC power sequencing are handled entirely within the module and are not dealt with on the carrier board.

Documents to check during design

Handling differential signals

Differential pairs such as USB 3.0, PCIe, MIPI, and Ethernet also require trace length and impedance control on the carrier board. Match lengths within each pair, match impedance through the layer stackup, and avoid unnecessary vias — miss any of these three and operation becomes unstable.

Easily overlooked points

Thermal design should not be left until later. Because SoMs are small, their heat density is high, and getting full performance out of them assumes heat conduction to a heatsink or enclosure. When using one in a sealed enclosure, be sure to measure the temperature during continuous operation at the hardware evaluation stage. If thermal throttling occurs, catalog performance will not be achieved.

About design reviews

We review carrier board schematics and artwork. Many problems can be prevented by checking before the board is fabricated, so please consult us, especially if you are working with a SoM for the first time. We also offer contract design.

04 / Bring-up

Bring-up procedure

When the board arrives, do not boot the OS right away; verify step by step. This makes it easier to isolate problems.

  1. Check supply voltagesWithout the SoM mounted, check the voltage and power-up sequence of each power rail with an oscilloscope. Any abnormality here could destroy the SoM.
  2. Secure the serial consoleSet up the debug UART first. It will be your only way to see what is happening in every subsequent step.
  3. Confirm the bootloader startsMount the SoM and check whether bootloader messages appear. If you get this far, the SoM side is working correctly.
  4. Kernel bootConfirm that the OS boots and that you can log in on the console.
  5. Check peripherals one by oneCheck Ethernet, USB, cameras, and so on one at a time. Enabling them all at once makes it difficult to isolate the cause of problems.

Device tree adjustment

In Linux, the carrier board–specific configuration is described in the device tree. The usual approach is to start from the manufacturer’s device tree for the evaluation kit and rewrite it for your own carrier board configuration.

# Typical items that need to be changed
- Enable/disable the UART/I2C/SPI in use
- Ethernet PHY address and connection type
- Camera and display connections and lane counts
- GPIO function assignments
- Power regulator definitions

Because this work is a common stumbling point, we also offer bring-up support on a contract basis.

05 / BSP & OS

Choosing the BSP and OS

SoM manufacturers typically provide a BSP (Board Support Package) for several operating systems. Which one you choose affects how smoothly development proceeds.

Choosing the BSP and OS
OptionsBest suited forNotes
YoctoFor mass production. Lets you build a minimal configuration with only the functions you needThe build environment takes time to learn
Debian / UbuntuPrototyping and evaluation. Wide range of packages lets you try things right awayLarge images and relatively slow boot
BuildrootSmall, lightweight systemsManagement becomes complex in large configurations
AndroidDevices with a touch UIWide customization scope makes effort hard to estimate
FreeRTOS / ZephyrReal-time control. Sub-CPU side of the heterogeneous coresRequires co-design with the Linux side

In practice, the approach of Debian-based for prototypes, Yocto for mass production is common. It prioritizes speed at the prototype stage, then optimizes the configuration for mass production.

What to check

Check the BSP’s update frequency and kernel version. For products in long-term operation, whether security fixes will continue to be provided is critical. Whether the kernel is an LTS (long-term support) version is also a deciding factor.

06 / Longevity

Long-term availability and EOL handling

Industrial equipment may require supply for 10 years or more after the start of mass production. Here we summarize the points to check when selecting a SoM.

As a distributor, we continuously track EOL information from manufacturers. For products you have adopted, we notify you of changes in supply status early.

07 / Pitfalls

Common pitfalls

Performance falls short

In most cases, heat dissipation is the cause. Under continuous load, thermal throttling kicks in and lowers the clock. Measure the temperature first. The next most common cause is saturated memory bandwidth, in which case you need to rethink how data flows.

A specific peripheral does not work

Most cases are an omission in the device tree or an incorrect pin multiplexer setting. Because a single pin has multiple functions, the intended function may not be enabled.

Won’t boot / unstable

Check the power-on sequence and reset timing. Especially in configurations with multiple power supplies, the power-on order may be specified. The boot media (eMMC/SD/USB) selection pin settings should also be checked.

Worked in prototyping but fails in mass production

Phenomena caused by unit-to-unit variation or temperature conditions are suspected. This is a pattern where a lack of margin that did not appear in a few prototype units surfaces at mass production volumes. The countermeasure is to secure temperature and voltage margins at the design stage.

08 / Support

Evaluation unit loans and support

Positive ONE supports you all the way from selection to mass production.

Contact us first

We can help from stages such as “the application is decided but we cannot narrow down the product” or “we are considering replacing another company’s product”. It is fine if the specifications are not yet finalized.

Get advice on configuration and product selection View Chinchilla products Sensor development guide