The SECO Modular Vision 15.6 QCS6490 presents developers with a full-featured HMI solution capable of implementing real-time AI vision at the edge. For PCB assembly and inspection systems, this can dramatically streamline operations for engineers tasked with maintaining manufacturing quality, diagnosing machine errors, and manually reworking boards.
For many decades, automated optical inspection (AOI) has helped mitigate the risk of manufacturing large batches of unusable printed circuit boards (PCBs), or ones that require manual correction by skilled soldering engineers. In particular, the widespread domination of surface-mount technology (SMT) by the late 1990s made AOI essential for quickly detecting minute manufacturing defects—such as improper solder pad alignment or missing components—so that machine operators could pause automated processes and address upstream problems like pick-and-place arm issues. Similarly, when AOI flags a false positive—i.e., a fault where there isn’t one—operators have been able to override individual notifications to minimize unnecessary downtime and improve production efficiency.
For performance monitoring and equipment control, machine operators require a sophisticated human-machine interface (HMI) with high image quality for manual visual inspection, in addition to a robust build that can withstand the environmental demands of industrial settings. With the introduction of low-power edge artificial intelligence (edge AI) across the electronics industry, developers have been presented with an interesting opportunity:
By combining industrial HMIs with edge-based AI vision systems, PCB assembly machine developers can streamline AOI infrastructure while reducing the requirement for separate industrial PCs. This consolidation also allows operators to bring the AOI benefits of the main PCB assembly line to repair or rework stations, enabling faster defect detection and correction that increases manufacturing throughput. However, edge-AI vision requires powerful on-board computing resources.
An Off-the-Shelf HMI Solution with Built-in AI Acceleration
Recognizing this demand, SECO created the Modular Vision 15.6 QCS6490, a high-end 15.6-inch HMI that is well suited for a wide range of industrial applications. This robust solution features a 1920×1080 capacitive touch display for clear, high-definition images and direct operator control. 3 mm of chemically strengthened cover glass protects this display during use, and a brightness of 400 cd/m² ensures readability even under intense factory lighting. Additional HMI features include an audio output for connecting to a mono speaker, enabling sound-based notifications that increase operator awareness of events that require immediate attention, such as the AOI system flagging a fault.
Alongside these operator-facing features, the Modular Vision 15.6 QCS6490 includes a range of interfaces for connecting the HMI to wider industrial systems. For example, the Gigabit Ethernet interface supports networking with reporting platforms for quality control (QC) and productivity monitoring. The peripheral-focused interfaces like USB, RS-232 and RS-485 serial ports, CAN Bus, and digital I/Os facilitate connections to cameras, lighting rigs, and other essential components of an AOI system.
At the heart of SECO’s modular HMI solution lies the Qualcomm® Dragonwing QCS6490, a full-featured processor with 8x 64-bit Arm instruction set cores spanning three performance/efficiency levels, ideal for driving complex machine application software while simultaneously managing high-resolution video data for AOI. This task is further supported by an integrated Qualcomm® Adreno™ GPU for sophisticated graphics that aid operators in sample examination and improve HMI usability.
Nevertheless, the standout feature of the Qualcomm® Dragonwing QCS6490 for edge-AI vision is the Qualcomm Hexagon digital signal processor (DSP) and neural processing unit (NPU). Offering up to 12 TOPS (trillion operations per second) of energy-efficient, real-time AI performance, this dedicated hardware accelerator is the key to enabling next-level component inspection at the edge.
These processing resources are supported by ample memory for high-throughput video inspection, with up to 8 GB LPDDR5-6400 RAM and up to 64 GB eMMC 5.1 storage. Both memories are soldered directly on board to promote reliable operation in vibration-prone settings, with an externally accessible microSD slot for easily removable storage.
Leveraging the Qualcomm® Dragonwing QCS6490 for Vision AI at the Edge
To make full use of the Modular Vision 15.6 QCS6490’s extensive feature set, SECO’s Clea OS—part of SECO’s ready-to-go Clea software framework—enables developers to quickly create an application environment based on the open-source Yocto Project. For more advanced development, Clea OS can also incorporate other software development kits (SDKs)—such as Qualcomm AI Engine Direct SDK—that enable developers to deploy various AI models on their choice of processing resource, whether that be the Hexagon hardware accelerator, Adreno GPU, or CPU cores.
SECO has already tested the real-time AI vision performance of the Qualcomm Dragonwing QCS6490 as part of its SOM-SMARC-QCS6490 computer-on-module (CoM), which runs the Modular Vision 15.6 QCS6490. For an automated quality inspection model based on LiteRT and PyTorch frameworks for resource-constrained hardware, the SOM-SMARC-QCS6490 delivered an inference latency of 112.44 ms when using the CPU cores to detect surface defects and pattern deviations from a reference. This corresponds to an inspection throughput of just under 9 images per second, which makes for a promising benchmark for high manufacturing productivity, even without further acceleration via the dedicated NPU.
By streamlining the foundational software implementation for the Modular Vision 15.6 QCS6490 hardware platform, as well as portability to other SECO platforms across different processor families, Clea OS enables developers to focus on the user-facing features of an HMI, such as tweaking application software and AI models for maximum effectiveness. To further support developers, SECO products also offer security and compliance by design, with features like secure boot, over-the-air (OTA) updates, and device monitoring facilitating cybersecurity certifications in accordance with EU directives.
A Highly Integrated Road to Smarter PCB Inspection
When overseeing manufacturing processes, PCB assembly machine operators face ongoing challenges as they inspect faults on increasingly complex boards with shrinking components that require ever more precise alignment. With the help of AI-accelerated AOI systems and more sophisticated HMIs, this task can be made significantly easier for greater operational efficiency.
SECO’s Modular Vision 15.6 QCS6490 presents a full-featured computer and HMI platform that supports these technological advancements while streamlining AOI system development, with the on-board Qualcomm Dragonwing QCS6490 processor supporting energy-efficient, real-time AI inference at the edge. In combination with SECO’s flexible Clea OS software stack, this solution offers industrial OEMs a faster path to AI vision development, with a robust software and hardware build to match.
Discover SECO’s full range of edge computing and HMI solutions at seco.com and visit the SECO App Hub for a wider range of tested AI vision workloads.