Pushing the envelope: VPX SBCs fly on the cusp of military bandwidth and power/performance needs

VPX has a bright future in the defense arena, considered by many as next-generation VME. Mil/aero designers now have the flexibility to evaluate both when looking at their design requirements, assessing the legacy environment and considering rugged application demands and the types of data being processed now and moving forward. Key considerations include power-efficient performance and bandwidth.

Editor’s note: The following article argues that VPX and its switched serial fabric interconnect flexibility is the natural next step beyond VME64 and VME320. But where’s the mention of OpenVPX? It’s only due to our long editorial lead-times that it appears that author David Pursley has omitted OpenVPX, the new VITA 65 specification that defines interoperable “profiles” to VPX’s flexible backplane. Nearly six months have passed since this article first arrived in our queue, and OpenVPX had very little public information available at that time: One had to be an OpenVPX Industry Working Group member to get the real skinny. As you read this article, keep in mind that today’s now-public OpenVPX is a subset of VPX as David describes it.

Today’s modernized battlefield could best be defined by the increasing role of rugged embedded systems delivering high-performance computing in mission-critical defense applications. Rooted in incredibly unique environmental conditions, integrated battlefield management can include intensive signal processing, secure multimedia and battlefield communications, avionics and sensor processing, shipborne servers, and a multitude of other demanding computing tasks and environments. This integrated battlefield management essentially creates an embedded design environment sharply focused on power-efficient computing.

At the same time, designers must address increasing bandwidth demands in a maximum uptime system. Performance and features must be designed-in without affecting energy variables such as thermal output or shock and vibration levels to the extent of system failure. Rugged SBCs – particularly the VPX (VITA 46) platform, an advancement of its VME predecessor – are rising to the challenge, combining higher levels of power-efficient performance and bandwidth amidst the growing requirements for the lower power dissipation and higher performance so critical to military warfare applications.

VPX: A closer look

VPX, as compared to VME, fulfills higher mechanical requirements in terns of robustness, resistance to mechanical stress, corrosion, and insensitivity to bending or incorrect board insertion. In terms of system casing, 6U VPX is compatible with VME, both using the 6U Eurocard form factor, albeit with a different backplane connector. Because VPX acts as an extension of legacy VME standards, VPX is recognized as the logical successor to VME for applications requiring high communications bandwidth and EMC and signal integrity within a COTS infrastructure.

VPX additionally satisfies the military’s need for modular serial switched fabrics, such as Serial RapidIO, PCI Express, or 10 GbE, featuring higher bandwidth and increased immunity to electromagnetic interference in extreme physical environments. Thus, the VITA 46.x standard – a family of specifications for extremely robust serial switched fabrics (see Table 1) – was born.

21
Table 1: Serial data signal types defined in VPX subspecifications
(Click graphic to zoom by 1.8x)

As VME’s progenies, VPX SBCs combine the fundamental strengths of the VMEbus architecture (for example, robustness) with new high-bandwidth connector capabilities for high-speed differential signaling over the backplane. VPX also provides for advanced cooling techniques, such as Liquid Flow-Through (LFT), allowing higher-power boards to be used in the system. While currently a more costly option than VME, VPX’s combination of features in a smaller 3U form factor gives VPX the advantage for a wider range of real-time, data-intensive, network-centric applications requiring reliable performance in harsh environments.

VME’s adaptability, simple maintenance, and rugged performance have made it the standard of choice in a wide range of mil/aero applications on ships and aircraft, such as missile launch control, acoustic data processing, or sonar systems. In contrast, applications such as radar or threat detection, the increased use of imaging and graphics in situational awareness, digital signal processing, and real-time communications including voice, data, and video are constantly delivering high levels of raw data that have to be transported, processed, and distributed with zero errors. As a result, complex military data networks are seeing a drastic increase in bandwidth and power-efficient processing capacity for VPX SBCs.

Power-efficient performance

Computing performance today frequently requires serial data connections or serial differential signal transmission, demanding higher transaction and aggregate bandwidth, lower link latency, options for different bus systems, increased scalability, and less real estate for signal routing. VPX’s support for serial switched fabrics allows for higher performance over VME’s bus-centric architecture, but also creates issues in power consumption and heat dissipation.

Improving I/O bandwidth necessitates that the processing elements within the system (CPUs, DSPs, FPGAs, and so on) scale in performance accordingly. Otherwise, the bandwidth capacity will go unused. Traditionally, this means that the power of processing elements will increase. While huge amounts of power can be delivered in VPX (up to 768 W per blade), more often power efficiency is the design goal. Improved power efficiency can enable high performance in harsh environments without compromising reliability.

With more performance per watt, today’s low-power processing technologies such as the PowerPC e500 and the Intel Atom architectures are greatly improving power-to-performance ratios, while at the same time directly supporting the communication protocols used by switched-serial architectures like VPX. The Freescale MPC8544E, for example, delivers high performance (2240 Dhrystone MIPS @ 1.0 GHz) combined with low power consumption (<4 W) and minimal cooling requirements. The 32-bit processor is based on the e500 core and includes an on-chip DDR2 memory controller, a security engine, two GbE controllers, and PCI Express interfaces. When this single-chip technology is implemented in a VPX processor board, such as the Kontron VX3230, the result is a high-performance board consuming less than 10 W, including CPU, memory, and I/O.

21
Figure 1: Kontron’s low-power VX3230 3U VPX processor board consumes as little as 10 W between -40 °C and + 85 °C.

Bandwidth speeds data transfers

The VPX platform has emerged as ideal for data-intensive applications, especially where high bandwidth is an issue, for example sensor data acquisition onboard a UAV. These high-performance boards can ease transition from VME, when the design goal is to reduce data transmission time, increase intelligence-gathering capabilities, and improve situational aware- ness. For example, VME has long been the major standard in military avionics. However, with new requirements and the new serial interconnects being used to transfer greater amounts of data faster, VME’s bandwidth represents a limiting factor for today’s designs.

The main communication mechanism for VME systems, VMEbus, can theoretically attain a maximum bandwidth of 320 MBps via Two-Edge Source Synchronous Transfer (2eSST). VITA 31 augments this by allowing GbE over the P0 connector. And VITA 41 (VXS) augments bandwidth further with a new MultiGig RT2 P0 connector that provides each board with 16 high-speed differential pairs for communication via the backplane. In contrast, VPX uses multiple MultiGig RT2 connectors to maximize bandwidth. At the same time, the higher-density connector can allow the system to shrink from a 6U to 3U implementation. 3U VPX implementations provide each board with 64 differential pairs and 12 single-ended I/O connections, or 32 differential pairs and 84 single-ended connections. 6U VPX allows up to 192 differential pairs or 404 single-ended connections per board, or a combination of both.

Each differential pair in VPX can handle signal speeds up to 6.25 Gbps. Combined with a high number of differential pairs available on both the 3U and 6U connectors, VPX can offer orders of magnitude increase in bandwidth over VMEbus. At these data rates, the CPUs and sensors become the limiting factors in terms of throughput. In other words, VPX’s high-capacity infrastructure currently exceeds the capacity of the elements using it. While a VPX-based board will also require a new backplane, it is ideal for high-bandwidth, I/O-intensive military applications. These include high-speed radar signaling systems, real-time video processing systems such as UAVs, or other video surveillance applications – or even new missile control systems that are capitalizing on 1080p video for the most accurate imaging and targeting. The platform also enables designers to incorporate mixed backplanes (backplanes with both legacy VME and VPX slots), allowing integration of the most current connector, bus, and serial fabric technologies (for example, 10 GbE, Serial RapidIO, PCI Express, and SATA II) with an existing VME product.

Mechanically valid concepts for military design

VPX is essentially an extension of the legacy VMEbus architecture, and both are well-suited to meet the needs of current and future applications. Whether to choose VME or VPX is not quite as simple as “older” design versus “newer” design, however.

For new programs 3U VPX is an attractive alternative to VME, providing higher bandwidth connectivity in a smaller form factor. Overall, VPX offers numerous advantages to the designer building rugged military embedded systems. VPX is tried and tested, and its reliability as a standard can be considered optimal with little or no learning curve for the system designer familiar with VME. Add in the possibility of gaining VPX’s power-efficient performance and bandwidth requirements, in addition to several other factors, and designers will be able to extend the architecture potential of COTS systems and help preserve many existing VME implementations. CS

David Pursley is a Field Applications Engineer at Kontron. He holds a Bachelor of Science in Computer Science and Engineering from Bucknell University and a Master’s degree in Electrical and Computer Engineering from Carnegie Mellon University.

Kontron 888-294-4558 www.kontron.com