By: Ohad Yaniv
CEO
Variscite

In the rush to put new electronic products on the market – edge/IoT devices, robotics, even specialty industrial or medical equipment – the choices manufacturers make early in the design process can have a large impact later.

The biggest choice in embedded computing design is, well, the computing that’s embedded, the processing subsystem that will perform the function the product is intended to do.

Given the explosion of devices, device developers, and applications, there’s no one best practice or perfect choice. There are, however, two primary schools of thought: chip-down design or System on Module/Computer on Module. They each have advantages and choosing the one that most suits your particular engineering effort may seem complicated.

Briefly: in chip-down design you’re starting from the ground up. Your design is unique and completely designed to your requirements. You select the components: processor, RAM, flash, networking, connectivity, power supplies, temperature controls, security features, drivers, and miscellaneous micro-hardware. You build the operating software to run it all and you manage the integration. The more in-house process you can afford, the more control you gain. This level of customization and control may be tempting but this approach comes with many disadvantages.

Time is money! The time it takes to achieve greenfield chip-down design may be enormous – R&D, trial and error, quality control. Your organization may not even have the in-house technical and human resources for chip-down design.

The real impact of this choice is go-to-market delay. Chip-down is a largely manual, hands-on process to design, engineer, prototype, source, and produce. Even a talented team may require many months and in some cases, even years, and that doesn’t consider supply chain delays (the global components crisis is not over yet) for all those myriad components you need to order.

The System on Module (SoM) approach is like starting at square twelve instead of square one. The design starts with a module that’s been pre-built with the core computing components accompanied by software. The device developer designs only the carrier board with the peripheral components and connectors the final product needs which can be outsourced as well. With this method, the R&D time and effort are focused on the specific IP of the company.

SoMs are off-the-shelf and ready for immediate implementation, obviously a very compelling option for compressed product development timelines but this isn’t the only advantage.

What about the cost optimization angle? Production volumes do play a big role in the choice between chip-down and SoM. If you produce consumer devices and expect over 100,000 units per year, economy of scale makes chip-down financially viable. However, if you produce industrial devices at a smaller scale, for example, there really is no cost advantage to using chip-down versus a SoM. In these volumes, a SoM is still more economically feasible because you don’t need a large pool of engineers during development or a maintenance team during the product life-cycle.

Some developers are tempted to choose chip-down for self-sufficiency, to avoid reliance on any one supplier. Turnaround time and longevity are often an existential concern, and there are few things more frustrating than a delay caused by a third party or end-of-life of components. The reality is that chip-down means relying on multiple third parties, each of which can let you down and cause a bottleneck in production or lead to redesigning in the worst case. SoM vendors typically purchase large volumes of components far in advance and can ship fully functional SoMs at small volumes within days – large volumes within weeks. Moreover, some SoM vendors have the privilege to get access to new technology, sometimes even before it is released to the market due to long-term partnership relations with the component vendors. Getting your hand on new technology can give you a significant competitive edge that couldn’t be reached without a highly maintained partnership.

There is the question of production quality. What’s ultimately higher quality, less error-prone, longer lasting? Bespoke chip-down products or those built on an off-the-shelf mass-produced SoM? The answer may surprise you. There is actually a greater probability of failures impacting product return rates in chip-down products because these are often first-time custom configurations that only been tested in a lab. On the other hand, SoMs are well tested in a vast number of different environments and configurations in the embedded market. Behind each SoM there is an engineering team that consistently work on updates.

Finally, when would a company still prefer a chip-down design? Today, top SoM manufacturers offer a dizzying array of interfaces and features, processor architectures, speeds and price points, enough to meet the requirements of 99.9 percent of the market. There are niche companies out there making niche devices that can’t be built on a SoM due to the need for a special features and components that don’t exist in a pre-built platform, so I won’t claim 100 percent.

About the Author
Based in Lod, Central District, Ohad Yaniv is the Chief Executive Officer at Variscite. Previously, Ohad was the Vice President of R&D for Micronet and before that, Software Group Manager for Fourier System.