The Initial Idea

Our work on the MEVISYS platform began back in 2018. While developing a complex Embedded Linux device, we realized that Embedded Linux could actually have great potential as a cost-effective solution for data acquisition and measurement applications.

We aimed for a modular, brick-style system featuring a central unit and several additional I/O units, all designed to mount onto a standard DIN rail. With this in mind, we chose compact, space-efficient enclosures from Phoenix Contact, which take up minimal space on a DIN rail (just 25 mm). Our goal was to engineer an Embedded Linux module that would fit perfectly inside.

Our first solution utilized the well-known Raspberry Pi, specifically the CM3 module. The main reasons for this choice were its off-the-shelf availability and an unbeatable performance-to-price ratio.

The engineering layout was quite a puzzle, but we managed to fit everything in rather well. Unfortunately, it became clear almost immediately that its power consumption and overall design constraints made it less than ideal for industrial applications. As a result, we soon had to abandon the idea of using the Raspberry Pi.

Our second attempt was backed by thorough research; this time, we were looking for the optimal industrial solution. The results pointed us toward the i.MX6ULL processor, featuring an ARM Cortex-A7 architecture running at a maximum of 528 MHz. Compared to the Raspberry Pi CM3, this meant a significant drop in both performance and memory. On the other hand, it offered a significant reduction in power consumption, giving us the opportunity to integrate a short-term backup power supply using supercapacitors.

This solution proved to be more than adequate for data acquisition and measurement applications. Two additional units were engineered using the exact same form factor, and one of them even featured three Ethernet ports—thanks to the fact that the i.MX6ULL processor has two independent Ethernet MACs. We deployed a wide variety of experimental applications, mostly written as Python scripts. Working with the platform was seamless, and the device itself proved to be reliable.

Over time, however, a critical bottleneck emerged that held back broader adoption. Potential customers were pushing back on the performance and RAM capacity because their existing software stacks (like Python or Node.js) demanded significantly more processing power and memory. It became clear that this hardware setup was only viable for smaller applications or custom solutions where the product has a fixed feature set that leverages the benefits of Linux. On top of that, some prospects were put off by the limited software support or the fact that they couldn’t just flash a “standard” Linux distribution like Ubuntu onto the module.

After being sidetracked by other projects, we attempted to address these issues by moving to a standard x86 platform. We sourced an affordable single-board PC called LattePanda V1 and managed to get all the necessary components up and running. However, designing and manufacturing such a device proved to be rather difficult. In the end, our efforts were limited to working with off-the-shelf development kits; we never actually made it to the custom design stage, and mass production was out of the question.

On top of that, shifting market dynamics brought affordable, off-the-shelf x86 computers to the scene. Given how complex the whole endeavor was, it simply no longer made sense for us to pursue the hardware design and development of these devices.

In the end, our work with Embedded Linux solutions turned out to be a dead end. It’s not that Embedded Linux hardware is useless; it simply doesn’t make sense for projects centered around relatively straightforward data acquisition and measurement tasks. From this perspective, it turns out to be an unnecessarily expensive, over-engineered, and incomplete solution.

On the other hand, these projects provided us with the expertise to design, program, and manage Embedded Linux units. Applications will certainly come along where this approach proves advantageous, simply because Linux software support opens up far more possibilities.

From a purely economic standpoint, it is currently more viable to buy off-the-shelf products. Designing and manufacturing custom electronics only makes sense for very specific use cases that cannot be covered by standard, commercially available hardware.