Open Standards-based Rugged Systems—A Third Way?
The 6th generation Intel® Core™ processor is part of an approach that suggests a way to potentially lower the overall risk of creating rugged solutions for industrial and mil-aero applications.
As device complexity has increased and development resources reduced, customers are looking to use more open standards-based hardware to create rugged embedded systems. There is a wide choice of open standards, (some would say too much choice) each of which has benefits useful for specific application needs. This article looks at the compromises that should be considered when architecting solutions consisting of a processor and some user specific I/O that are both rugged and compact. For the purposes of this article, let’s define “compact” to mean “no bigger than a shoebox.” Let’s say that “rugged” equates to being able to operate at a wide temperature range and sustain shock and vibration similar to the levels defined in ANSI/VITA 47 for conduction-cooled equipment.
Solutions Based on Off-the-Shelf Components
A low risk way to construct a rugged solution is to source all the building blocks off-the-shelf. The good news is that this is possible using 3U sized boards connected by a backplane. Historically standards like PICMG 2.0 CompactPCI augmented with the ANSI/VITA 30.1 mechanical definition of the heatframe and wedgelocks might have been used. More recently 3U VPX™ has become the de facto standard for rugged solutions, especially since the ANSI/VITA 65 OpenVPX™ initiative to improve vendor interoperability using predefined profiles. There is now a vibrant ecosystem around 3U VPX; the author’s company offers Intel processor boards that can be mated with myriad FPGA, GPU and I/O boards plus backplanes, power supplies and rugged chassis, all available off-the-shelf. Such a solution can be capable of sustaining high levels of shock and vibration at a wide operating temperature range.
This solution just described definitely has an advantage when it comes to scaling up. Additional processing and I/O boards can be simply installed providing that there is provision for expansion in the initial solution. However, solutions based on 3U VPX boards are not particularly compact because the fundamental board area is 100 x 160mm before taking into account the additional space required for a backplane (Figure 1).
There are a number of options to construct more compact systems based on smaller form factor open standards. Probably the simplest way to reduce space is to remove the backplane where scalability is not required. An off-the-shelf processor module can be directly connected to a custom carrier in a hybrid off-the-shelf/custom configuration. The premise for the PICMG® COM.0 COM Express™ standard is that the customer can design their carrier and then fit successively more modern processor boards as needed. Computer on Module standards were originally designed to be cost effective in high-volume embedded applications. The COM standards have largely delivered on this promise, with widespread deployment in applications like medical scanners and transport based digital signage that can run into tens of thousands of units per year. They offer defined pin outs for the connector(s) to the carrier card aiding compatibility but leaving little scope for functional differentiation. COM Express modules are available in a number of size formats. Generally as the processor module size decreases, the higher the level of differentiation on the carrier. Differentiation on the carrier has the advantage of allowing the equipment manufacturer to design and control the application-specific part of the solution, especially if there is safety certification involved, for example, in medical scanners.
Widely used and popular standards tend to go through regular update cycles to keep them in line with modern I/O technologies and interfaces. Another good example is the PC/104™ family, which has undergone several updates to stay relevant over its long life cycle. PC/104 boards are consistently 90×96mm, with each update of the specification bringing a faster interconnect option. PC/104 is unique (as far as the author is aware) in being stackable: instead of a single computer on module processor mating to a custom carrier; PC/104 allows a user to create a stack of modules in a cube-like structure.
Whilst COM Express and PC/104 modules are physically small, they can require the user to take more responsibility for the rugged solution. Some of the challenges that may need to be addressed include:
- Devising and engineering a heatframe to enable conduction-cooling.
- Making sure the processor and carrier solution is stiff enough to meet any required shock and vibration testing without resonating.
- Confirming that the connectors are not susceptible to contact bounce and fretting issues during vibration testing.
- Verifying that any additional cables are securely fixed to the boards. This is especially relevant for PC/104 modules as they typically include I/O connectors around the board edges.
Whilst COM Express and PC/104 are fundamentally smaller they do require more user intervention and expertise to build a rugged solution. They may struggle to meet the required qualification levels, especially if they use commercial memory modules and connectors.
A Third Way?
Is there another solution that can provide a lower risk path to ruggedization and yet provide a more compact solution compared to 3U VPX boards?
One alternative and fully ratified standard worth considering that has widespread adoption is the ANSI/VITA 42 XMC module. This is a progression from earlier ANSI/VITA specifications including ANSI/VITA 20 conduction-cooled PMC, based on the IEEE-1386 standard for a Common Mezzanine Card standard. Many PMCs were low power I/O modules designed to fit onto processors or carriers and were widely used to add functionality to VPX and VME boards. However, instead of having a processor card fitted with an XMC I/O module, it is possible to turn this around and use an XMC processor mounted onto an I/O carrier, effectively turning a processor XMC into a type of Computer on Module. The carrier card is likely to be a custom design so the end solution would be partially off-the-shelf, leading to the same advantages as for other Computer on Module formats.
XMC modules and carriers must have one primary connector: this is used for power, module control and PCI Express® for data communication. Most modules implement a secondary connector for unspecified user I/O rather than allocating exact pins to generic PC-like interfaces such as SATA, USB and DisplayPort. Use of the secondary connector can be beneficial, as it provides the flexibility to use application-specific interface signals but, conversely, does reduce the chance of processor XMCs being interchangeable between vendors.
For XMCs, a key differentiator is the definition of thermal interface connections between the module and carrier, which makes it easier for an XMC and carrier solution to be fitted with a combined heatframe suitable for use in a conduction-cooled environment. The user will have to devise a heatframe to fit their custom carrier, but at least there is a reasonable starting point based on open standard formats like VPX. The mechanical design of the combined heatframe would incorporate stiffening bars, and providing these are in line with the equivalent off-the-shelf products, there is a better chance that the solution will pass rigorous shock and vibration tests.
This XMC-based alternative leverages an off-the-shelf processor module but will not generate as compact a solution as might be possible using more commercial Computer on Module formats. However, the key advantage of starting with a processor module that is designed for rugged use is that it reduces the overall risk of creating a rugged solution. The risk instead becomes something closer to that of a completely off-the-shelf solution like 3U VPX.
Nigel Forrester is Technical Marketing Manager at Concurrent Technologies plc with responsibility for product strategy and promotion across VPX, AdvancedMC, CompactPCI and VME form factors. Forrester has previously held a number of product and vertical marketing roles in the embedded space and is an accomplished speaker and author. He has a B.Sc. (Hons.) in Computer Science and Statistics from Reading University, United Kingdom and can be contacted at firstname.lastname@example.org