The pros and cons of computer-on-modules
Computer-on-modules are world-leading platforms for embedded system designs. What makes them so attractive and what are their limitations?
Studies from IHS Markit state that computer-on-modules are leading the global ranking of embedded form factors followed by standalone boards and VME/VPX solutions. They also forecast a growth of 8.6% CAGR during the period 2015–2020, which is impressive as market-leading technology is generally well established and market volume tends to be stable rather than dynamic. Similar studies from Research and Markets paint a much healthier growth perspective, forecasting that the global computer-on-module market will be growing at a CAGR of 17.97% during the period 2016–2020.
The large difference between these two forecasts may be caused by the highly uncertain market dynamics in the IoT segment, where these modules will get massively deployed. IHS Markit identifies the needs of industrial automation and Industry 4.0 as a major driver of growth in coming decades. So from a bird’s-eye view of the market perspectives, there is no doubt that computer-on-modules are suitable candidates to evaluate for embedded system designs. But what makes them so attractive?
Made for customisation
Computer-on-modules are building blocks for custom system designs. Custom designs are quite often demanded in the embedded computing area as off-the-shelf motherboards cannot be used for all embedded applications. The available space may not be sufficient. Interface demands are almost always individual in terms of number, configuration and location on the boards. Also, a motherboard with expansion cards may just not offer the required resistance against mechanical or thermal stress.
All these individual demands lead to the question: shall I build my own design from scratch with all efforts and costs involved, or are there other options available that can help me design my dedicated system faster and more efficiently? Computer-on-modules were invented exactly to help with this ‘build or buy’ question and the intention to simplify the use of embedded processor technologies in customised designs.
Application-ready super components
Computer-on-modules are application-ready super components that offer engineers high design efficiency. One benefit for purchase departments is the fact that the bill of material is reduced from many components to a single module for the processing core — but this is only the smaller part of the efficiency gain. More important are the reduced efforts required to design-in the processor, RAM and high-speed interfaces on the one hand, and to build the entire board support package with all the necessary drivers, libraries and APIs on the other. All this work is already done and modules can be deployed nearly as simply as a new processor on a motherboard — but there is a huge difference between the switch of a processor and a computer-on-module.
Nearly endless scalability
Computer-on-modules offer nearly endless scalability. While a processor change can only be executed with pin-compatible processors that are generally only available within a certain processor generation, computer-on-modules can basically host all processors from all leading embedded processor vendors. One example is the update from the 5th generation of Intel Core processors to the 6th generation, where the grid array and memory interface changed. When leveraging a customised board, designers would have to redesign their PCB. With computer-on-modules, a switch between processor generations and vendors is much simpler and always possible. A new product generation can be launched just by switching the module.
Modules also make a design vendor-independent, with the benefit of higher design security. Another advantage of this scalability feature is that it extends the long-term availability of applications as when the seven- or 10-year product life cycle of an embedded processor ends, a successor is often available that can be used as a retrofit. If the design is based on modules, again only a switch of the module is required.
The benefits of standardisation
But this scalability can only be secured by interface standardisation. Computer-on-modules achieve this by standardising the footprints as well as the interface to the custom-designed carrier boards of the modules. While such standardisation can be a limiting factor that makes it necessary to have several specifications available to target all major applications, the benefits are tremendous.
Standardisation leads to highest design security as designers can rely on the future availability of modules with the same interfaces. They can also develop second source strategies and are not dependent on a single vendor. This benefits not only the design security but also provides commercial advantages due to competitive pricing. There is also greater room for module vendors to offer more services in an attempt to separate themselves from competitors by improved support of customers’ demands.
Standardisation further delivers the capabilities to offer a broad ecosystem of commercially available accessories, ranging from heat spreaders and carrier boards to cable sets and housings. This makes it easy to purchase components from third parties so that NRE costs are reduced to a minimum. Finally, a large community of designers working with the form factor ensures continuous standard improvements.
Suitable where no other form factor fits
Having said all this, computer-on-modules are really only suitable if no other embedded form factor fits. Engineers consequently need to check the specifications and market trends of other embedded form factors before choosing a module approach. As the forecast from IHS indicates, checking the availability of standalone boards which directly fit the application is most important.
The relevant form factors in this growth cluster are the Mini-ITX and Pico-ITX boards, as well as the new eNUC standard, as they offer small form factors perfectly suited for space-constrained embedded system designs. In the market segment of passive backplane-based systems, only VME/VPX shows good growth perspectives — due to intensified spending in the military market — while CompactPCI and xTCA technologies are declining.
Unsuitable for ultrahigh-volume productions
Engineers also need to check whether a full custom design might fit better in the end. This is always the case in ultrahigh-volume productions, where every single component is a cost factor to be considered for economising. While the connector of a module may cost only $1, when you’ve got 10 pieces this adds up to $10 and the mounting of the module is a cost factor too. So when it comes to very high-volume productions, the breakeven point between a COM/carrier concept as opposed to a full custom design needs to be determined.
Calculating this breakeven point is complex, as R&D costs and investments in future upgrades also need to be taken into account. Module vendors can help OEMs with these calculations, as in most cases they also offer embedded design and manufacturing services for full custom boards where they can often re-use the layouts of the carrier board designs manufactured for the evaluation of the boards.
Spoilt for choice
After having evaluated these options and finding that a computer-on-module approach fits best, engineers need to evaluate the right computer-on-module standard as a final step. Today’s state-of-the-art technologies include the specifications from two worldwide standardisation bodies: the PCI Industrial Computer Manufacturers Group (PICMG) hosting the COM Express standard and the Standardization Group for Embedded Technologies e.V. (SGET), which is responsible for Qseven and SMARC.
COM Express
The COM Express standard defines a family of different module sizes and pinout types covering a broad range of designs from low-power small form factor devices up to powerful embedded servers. COM Express sizes include:
- Mini (84 x 55 mm)
- Basic (95 x 125 mm)
- Compact (95 x 95 mm)
- Extended (110 x 155 mm)
COM Express Type 7
PICMG’s COM Express Type 7 specification is tailored for modular server designs, which are being deployed at the edge of the IoT and Industry 4.0 applications as cloud and fog servers, or in cloudlets at the edge of the carriers’ base stations for high-bandwidth mobile communications. What is most interesting from a feature set point of view is the support of up to 4x 10 GbE bandwidth and up to 32x high-speed PCIe for high-performance storage, and dedicated interfaces supported by 440 signal pins to the carrier board. Target processors that can be found on the basic sized 95 x 125 mm modules include the Intel Xeon D processors and the upcoming successors from both x86 server processor vendors, Intel and AMD. Larger modules are possible as well, as COM Express already specifies the Extended format measuring 110 x 155 mm.
COM Express Type 6
The established PICMG COM Express Type 6 specification is state of the art for the high-end sector of embedded computer systems with implemented processors ranging from Intel Core, Pentium and Celeron to the AMD Embedded R-Series. These modules measure 95 x 125 mm (Basic) or 95 x 95 mm (Compact), provide 440 pins to the carrier board and offer a comprehensive set of state-of-the-art computer interfaces with everything needed to build powerful PLCs, HMIs, shop floor systems or SCADA workstations in control rooms. Further application areas are high-end digital signage systems, high-end gaming machines and complex kiosk systems.
COM Express Type 10
PICMG’s small form factor COM Express Type 10 rounds off the set of COM Express specifications. It comes with the credit card sized Mini form factor. These modules measure only 55 x 84 mm, offer 220 pins and are dedicated for low-power x86 SoC processors such as Intel Atom and Celeron as well as AMD G-Series processors. Thanks to the unified connector technology and design guides used within the entire PICMG COM Express ecosystem, developers can re-use as many features as possible. Designers have one standard they can leverage to scale their designs on the basis of COM Express, from Mini Type 10 modules with Intel Atom processors up to Intel Xeon D processors for the server segment.
Qseven and SMARC
Engineers that are targeting not only x86 but also ARM-based designs are best served with Qseven or SMARC 2.0 modules as they incorporate both processor architectures. The difference between Qseven and SMARC 2.0 can quite easily be explained. On the connector side, Qseven offers 230 pins and SMARC 2.0 offers 314 pins. SMARC is more orientated towards feature rich multimedia applications, whereas Qseven offers more I/Os as required by the deeply embedded and industrial arena.
All the other benefits are comparable. Both standards enable slimmer designs compared to COM Express because of their flat edge connectors. Both have reliable connector vendors: the Qseven connector is currently supported by three and the SMARC 2.0 connector by two vendors. So for all those who have criticised Qseven in the past for only having one connector vendor, it needs to be underlined that this vendor bottleneck has now not only disappeared but changed to a slight advantage compared to SMARC 2.0.
The difference in the number of interfaces between Qseven and SMARC 2.0 is also kind of a price indicator: Qseven is designed for less complex designs and SMARC for the high-end of small form factor applications that demand credit card-sized modules. In general, any decision therefore depends on what the task of an embedded system will be.
Conclusion
The benefits of computer-on-modules are so substantial that a majority of embedded system designs are already using these application-ready building blocks. As the number of IoT and Industry 4.0 applications multiplies, many new designs are forecast to be also based on computer-on-modules and the new class of server-on-modules for edge computing. Identifying the best form factor is the next major step within the design evaluation process where module vendors can help. As long as they offer all the relevant form factors, they can provide better consultancy as well as better options to migrate from one form factor to the other.
When choosing the right vendor, it is key to have a look at the BSPs, firmware and communication middleware, as they are getting more and more important in a connected world. This does not mean that the vendor should complement its offerings with an entire cloud for the system because it will never meet the needs of a customer entirely. It is more important to have a closer look at what is offered on the board and module level itself. For example, is the board management controller proprietary? Then take care as it could prove to be a dead end. Better to choose open, non-proprietary APIs because openness and standards are the fundament for most efficient and simplified re-use of existing engineering efforts. Check that integration support is offered for ARM and x86, because it is better to get one engineer who supports both architectures for a unified product family instead of two different engineers with two different product lines. This also requires unified APIs.
Finally, check the provided documentation. It is better to have more pages of content instead of only the bare minimum. And think also about relying on local manufacturing capacities wherever you or your customer reside. This will not only allow you or your customer to buy local but can also help with potential government trade restrictions. Fabless board-level vendors such as congatec, with subsidiaries all over the world, can offer you all these advantages.
Unlocking next-gen chip efficiency
By studying how heat moves through ultra-thin metal layers, researchers have provided a...
Ancient, 3D paper art helps shape modern wireless tech
Researchers have used ancient 3D paper art, known as kirigami, to create tuneable radio antennas...
Hidden semiconductor activity spotted by researchers
Researchers have discovered that the material that a semiconductor chip device is built on,...
