Do you know your Micro-ATX from your COM Micro? Or your Qseven from your XTX? The small form factor arena is one of those areas where it is very easy to get lost in the jargon, and, unless you are an enthusiast (and, if you are, then please forgive me), it can be very hard to get excited.
There are at least 150 different sizes and shapes for small form factor boards, according to expert (and enthusiast) Hermann Strass. And there seem to be almost as many organisations working in the area. As well as organisations like the Small Form-Factor Committee, the SFF Club, Small Form Factor Special Interest Group, and so on, each form factor has its own consortium devoted to promoting its interests and developing the format.
What is a small form factor? About the only definition that seems to work is “Not Large.” PC motherboards have shrunk over the years, with the ATX that was inside early computers shrinking down through Micro-ATX, Mini-ITX, and Nano-ITX to Pico-ITX. ATX is 305 mm by 244 mm (actually 12 in by 9.6 in) while pico-ITX is only 100 mm by 72 mm (approx 4 in by 3 in). But PCs are only a tiny part of the board universe. 10 billion embedded microprocessors are going to be produced this year, and each one of these is going to be mounted into a board.
For many applications, putting computer functionality (processor, memory and some peripherals) on a separate small board with an array of communication options is a way of providing design flexibility. The processor on the board can be changed, and the functions of the peripherals can also evolve. This approach, Computer-on-Module or COM, has developed over the last few years, and it is possible to draw up a massive matrix of board sizes, processor types and architectures, and interfaces, both physical interfaces with different numbers of pins, and the protocol interfaces.
Even with this huge choice, people still design their own COM, but the question that has to be asked is “Why?” There may be good reasons – the existing products don’t precisely match the requirements of the system, for example. But unless the volumes are high, or the requirements are so precise that no compromise is possible, it must surely be sensible to devote people and time to getting the most out of an existing commercial design.
One of the first formats to use the name COM was COM Express in about 2003. This standard was initially thrashed out by a consortium of companies (like most of the other formats) including board companies Kontron, Radisys, and PFU Systems along with Intel (although the standard is processor agnostic). In 2004 PICMG took over co-ordination and published the standard in 2005.
The specification was for a small rugged board that would use low voltage differential signalling (LVDS) while still providing legacy support for established bus and communications protocols. The original standard includes a board-to-board connector with two rows of 220 pins. One provides pins for PCI Express, LPC Bus, SATA, LVDS LCD Channel, VGA and TV-out, LAN, system and power management, and power and ground interfaces, while the second row provides SDVO and legacy IDE and PCI signals with additional PCI Express, LAN, and power and ground signals. So together it will support 32 PCI Express lanes (80 Gbps aggregate), x16 PCI Express graphics, 2 LVDS Channels, 2 Serial DVO Channels, 4 SATA-150 links (600 MBps aggregate), 3 10/100/1000 Ethernet ports (10G provisions in the future), and, deep breath, 8 USB 2.0 ports.
The board dimensions are 125 mm x 95 mm for Basic and 155 mm x 110 mm for an Extended Form Factor. (Extended has more memory, can take bigger processors and chipsets, and has appropriately uprated power.) Don’t worry – we are not going into that level of detail again for all the other formats – but if you call your friendly local board supplier, they are happy to talk for hours on the topic.
The developers saw COM Express as going into applications that need high bandwidth, high I/O, and small size, such as industrial automation, test and measurement, military/aerospace, medical, and similar markets.
But aren’t there single board computers (SBC – like the motherboards we saw earlier)? Why use a COM? Single board computers have the complete set of functions, including keyboard and mouse interfacing and graphics interface, and control that you would expect from a desk-top machine. They may even include expansion slots for ISA or PCI, etc. and can come heavily ruggedized for extreme environments. They are available off the shelf and ready to run. All of this is useful and may be ideal for small and medium volume applications. COMs provide a lot more design flexibility but still need work to design-in and require a carrier board. They are probably more suited to high volume solutions. (Although when you talk to some companies they claim that their COM is an SBC, muddying the waters often for their own marketing ends.)
Apart from COM Express, there is a raft of other COM boards, including, but not limited to, COM Express Type2, ETX Rev. 2.x and Rev. 3.x, XTX, ESMexpress, Qseven, nanoETXexpress, ESMini, and CoreExpress. Each of these normally starts with a company identifying a market gap, designing something to fill the gap, and getting together with other companies to create a broader standard. The arguments are the same here as with other standards. If you are the only supplier, you get 100 per cent of the market – whatever its size. But potential customers don’t like to be locked into a single source. By spreading the standard, you gain broad acceptance: while your percentage is smaller, the pie you are slicing may be significantly bigger. (And, as the initial creator of the standard, you should be able to retain a competitive edge.)
An example of how this process works is the Qseven module. Congatec, a spinoff from the German board giant, Kontron, saw the need for what they called an “ultra-mobile COM module” with very low power consumption. It is only 70 mm by 70 mm (thus Quadratic (square) seven — for 7 cm) or, in imperial units, 2.75 in by 2.75 in, and uses the MXM edge connector, designed for use for PCI Express in notebook computers. It was also designed to take advantage of the Intel Atom processor. Once Congatec had begun design work, they involved MSC of Germany and Seco of Italy.
Today, there are 18 participating members from Europe, Asia, and the US, and 9 supporting members listed on Qseven-standard.org. Ten boards are listed on the website and others pop up on Google (although some of these may be from resellers re-badging). You can also download design guides, pin-outs, and schematics to roll your own board.
The other COM standards are evolving in the same way. One driver for the evolution of multiple COMs is the development of different communication standards and protocols. PICMG was originally founded to look after PCI technology, and now, with over 250 members, it works on a whole range of standards for telecoms and industrial areas, including, for our purposes, CompactPCI, CompactPCI Express, RapidIO, and MicroTCA. And, for each of these standards, manufacturers have built or modified COMs.
Within each format, individual vendors will vary processors and speeds within the types. They will add different types of memory in different quantities and provide support for different peripherals and different protocols.
Do we need so many standards and variants for COM? The logical, engineering answer is that we don’t, probably. Just as we probably don’t need so many different kinds of shampoo or flavours and types of potato snack. But while the array is available, it should be possible to dig among the spec sheets to find a COM that matches the requirements of a particular project.
Historical note. The earliest version of the COM that I can find is the TRAM (TRAnsputer Module), introduced by INMOS in 1987 – 22 years ago.