Small Form Factor Update
New Small Form Factor Interconnect Spec Supports Intel’s Atom and VIA’s Nano
Given the specification diversity for COM modules, there is a need for a connector system that will combine the self-sufficiency of a single board computer with the flexibility of a COM module and ease the integration burden for developers.
COLIN MCCRACKEN AND DAVID FITZGERALD, SMALL FORM FACTOR SIG
Page 1 of 1
As ETX and COM Express form factors prepare to penetrate the broader embedded systems market, a new wave of tiny Atom-oriented COMs (computer-on-modules) tries to take the beach. However, an apparent obsession with achieving the smallest module sizes leaves system OEMs once again facing most of the same carrier board design problems as their predecessors. Rather than allowing that burden to be passed to system designers unchecked, the Small Form Factor Special Interest Group (SFF-SIG) introduces a simpler approach
To date, there are numerous specifications for COM-type modules that define various expansion buses, connectors and module form factor/sizes serving many diverse markets. This targeted market approach and specification diversity is also the “Achilles’ heel” of broad acceptance for virtually all specifications published so far–certainly those that do not bring in large scale design wins (10-100K units annually).
COMs offer interchangeability of modules–on paper, at least. Commoditization goes hand in hand with standardized carrier board interfaces. Improvements in chip-level integration along with smaller transistor geometries mean that two-chip and three-chip x86 solutions provide much of the circuitry needed for many embedded PC applications. Facing competition and eroding prices and margins, COM architects keep stripping away extra features, leaving only what’s “free” in the chipsets. The downside is that system manufacturers are left holding the bag as they try to integrate their carrier board circuits with the “black box” module’s interfaces and BIOS.
Getting these minimalist modules to actually boot and interface to the rest of the system I/O has become quite a chore. From power management to power supply impedance to interface circuits, true interoperability has been kicked to the curb in favor of bargain basement pricing. In addition, connector technology has advanced substantially, yielding a more robust and higher-bandwidth solution and thus a better value for the price compared to four to eight-year-old architectures.
In addition, there is no existing specification that (1) adequately addresses modern ultra-low-power chipset requirements like those for the Intel Atom and the Via Nano / VX800 families, (2) looks beyond to PCI Express Gen2, USB3 and SATA600, or (3) specifies a small, cost-effective and technologically capable connector system to enable next-generation modules.
Focus and inclusion of these key elements in a specification that defines connector type(s) and pin placement, not connector placement or board size, will begin to bring convergence and interoperability to a COM implementation in the small form factor (SFF) embedded computer arena.
With the limitations of existing approaches well documented, several module vendors collaborated to break with tradition and deliver self-sufficient SBC-like modules that add value and solve system-level problems. Of course, the new modules will be larger and cost more since more circuitry is present. The result of rethinking COMs is a realization that the total system cost and development time and budget should be the relevant metrics, not merely the module cost. The greater the portion of the system design that is solved by the COM, the more valuable the COM would be to the system OEM.
Saved by a COMIT
The SFF-SIG’s new solution bears the name COMIT, which stands for “Computer-On-Module Interconnect Technology.” Much like Stackable Unified Module Interconnect Technology (SUMIT) has focused on the OEMs’ design challenges independent of SBC form factors, COMIT starts with system-level requirements and then determines how best to partition the architecture. It offers a blend of signals that are most often needed for embedded system designs, going beyond what is simply “free” in the chipset.
COMIT achieves an optimal balance of new and legacy interfaces to achieve the “common denominator” of I/O that is relevant across the broader embedded market. Just as separating the connector and form factor specifications into separate definitions was done in SUMIT, the key differentiating feature of COMIT from other COM specifications is the purposeful exclusion of a myriad of application-specific expansion buses and I/O that are often found in other COM architectures.
This additional I/O makes them larger, more expensive and less portable across applications. Parallel PCI, for example, has been supplanted by PCI Express because of low speed and high pin costs; therefore it is not included in COMIT. Beyond a full serial port and another TX/RX-only implementation, which are still useful in many embedded environments, much of the I/O normally included in a system is assumed to be implemented on the carrier board of a system and is not included in the COMIT module interconnect.
COMIT instead supports legacy and future generation buses available from the latest generation chipsets, leaving the baggage of on-module I/O and previous COM generations behind with direct support for legacy devices through the connector. Once a basic interface bus list was generated and a pin budget developed, an evaluation was done comparing connectors used on other similar specifications to the proposed connector here. Cost per pin, pin density, bandwidth through the connector and total area consumed were all considered. They include the following:
• 3 PCI Express x1 lanes with clocks and card detect
• 1 PCI Express x4 with clock and card detect
• 6 USB ports with 3 overcurrent pins
• VGA + PnP
• 2 LVDS panel support (24-bit)
• SDVO (Serial Digital Video Out)
• ENET (GigE or 10/100)
• 2 SATA channels with spin up control
• HD AC97 audio
• LPC Bus
• 8-bit SDIO
• 2 Serial UART ports
• Power, ground and various control signals
COMIT is an electromechanical connectorization specification that uses 240-pin high-density (0.050-inch pitch) SEARAY SEAM/SEAF connectors from Samtec and is second sourced by Molex (Figure 1). It is designed as an open pin field array configuration organized in 6 rows of 40 pins to allow maximum routing and design flexibility. The SEARAY connector was developed for the large and growing high-density and high-performance mezzanine connector market to accommodate low-profile applications. The chosen connector system for COMIT is capable of a differential signaling rate of over 10 GHz bandwidth (at -3 dB insertion loss) in order to support current and future high-speed signaling for interfaces like PCI Express Gen2, Gen3 and USB3.
The contacts in the SEARAY system are robust and allow for “zippering” when mating and unmating the connectors. This contact design lowers the insertion and extraction forces, which is an important consideration with 240 pins. Each gold-plated pin is rated for 2.7 amps and the connector will operate over the temperature range of -55° to +125°C. The rugged connector system is good for over 2000 mating cycles and is RoHS compliant. The mating height of the connector pair is 8.5 mm. The actual size of the 240-pin connector is only 57.9 mm x 9.67 mm.
8.5 mm was assumed as a working standard for the height due to the need for the smallest modules to place taller power supply components on the COM PCB between the COM and carrier board, as well as some minimum space for components on the carrier board in the same vertical areas. Other heights are possible. This height also conveniently would allow a COM with short or no heat sink requirements to fit underneath a SUMIT 15.24 mm module stack.
Since COMIT is a connector specification, it can be added to a variety of SBC, COM, or custom board form factors and is flexible and compact enough to meet a very broad range of application requirements. Even custom board designs can benefit from the architecture, chipset mapping and engineering analysis that have been completed already.
One of the key breakthroughs with COMIT is the ability to power modules directly from a system power supply, rather than forcing power supplies to be generated on the carrier board and passed through numerous expensive connector pins. This eliminates much of the guesswork of power supply sequencing and stiffness associated with “black box” COMs, as well as the inefficiency of power-up current surges passing through tiny carrier board connector contacts.
The Advanced Configuration and Power Interface (ACPI) is an open industry specification supported by many of the popular operating systems that standardizes levels of system and processor power-down. It was originally designed for laptop / notebook PC motherboards, which are single boards. However, as the system architecture becomes more complicated it tends to sacrifice module interchangeability when power management is split across two boards (i.e., the COM and the carrier). The philosophy in creating COMIT is that the module is completely responsible for generating all of the power planes on-board and for supplying the power management handshaking signals to the carrier boards. Once again, tightly locking down ACPI requirements saves the carrier board designer from trial-and-error board spins to get second-source modules to work.
Most COM form factors are proudly touted as “legacy-free,” ushering in the latest desktop interfaces like PCI Express, Serial ATA (SATA), USB, Gigabit Ethernet and so on. Again, this serves to minimize module costs, but ignores the popular usage of legacy peripherals such as UART serial ports. The Low Pin Count (LPC) Bus can replace much of what the ISA Bus has provided for so many years, but the dependence on the module BIOS to initialize carrier board super I/O devices proliferates custom BIOSs. COMIT’s solution to this is simple, yet beyond what the other new COM form factors provide–integrated serial ports on the module.
First Implementation of COMIT
The first board being developed as a technology demonstrator is called sffCOM (Figure 2). It is a 62 mm x 75 mm module with the COMIT interface on one end of the board. An sffCOM board includes processor/chipset, RAM, power supplies and clock source for the module only, and connectorization for “standard” interfaces available from the chipset. The pin definition has been assigned based on test routing both Intel’s Atom and VIA’s Nano platforms and is a “best practice” approach for ensuring that signal integrity (SI) requirements will be adequately addressed for board implementations. Samtec assisted in the connector selection and SI decisions that led to the connector and placement choices made.
Signaling for additional features like video overlay input, etc. is not included in the main connector and should be included in a small connector from the COM to the carrier board if needed. This reduces cost for the vast majority of these systems that will not use this feature. Conversely, LVDS signaling is included in the main connector because it is foreseen that many applications will continue to use this signaling, and the aggregate cost of having a second connector pair and cabling for this is too high to encourage broad adoption. In the future, DisplayPort can be used directly off the COM using the standard connector, since it is cabled directly to displays.
The primary COMIT connector includes all signals except slow parallel buses like PATA and parallel PCI. The cost-per-pin is too high and the signaling bandwidth required is too low for these interfaces to be included with this connector density. These optional features are connected through secondary connectors. A low-cost solution is available from Samtec for PATA or other interfaces that consists of a 2 mm SMT pin header and high-reliability SMT board connector, which will meet both cost and rugged application requirements. PATA is not required, and that connector simply goes away once IDE storage devices are replaced by SATA, USB and/or rugged SDIO storage devices such as the new MiniBlade standard from SFF-SIG.
After the development of the sffCOM module, a carrier board is required to host the I/O or other special interfaces needed for each unique customer’s application. This can be done with a custom design for a specific application or by using an industry standard SBC form factor. For example, sffCOM can fit on a 90 x 96 mm industry standard module (Figure 3). This allows state-of-the art processor technology to be mated with existing SUMIT expansion cards and enclosures. A COMIT connector can be placed on an EPIC size board as well (Figure 4).
Currently the COMIT and sffCOM specifications are being completed within the Modules working group of the Small Form Factor Special Interest Group (SFF-SIG). This group has charted a course to develop, adopt and promote circuit board specifications and related technologies that will help electronics equipment manufacturers and integrators reduce the overall size of their next-generation systems. The group’s philosophy is to embrace the latest technologies, as well as maintain legacy compatibility and enable transition solutions to next-generation interfaces.
The SFF-SIG has formed three working groups to address different product categories. The SBC Working Group is discussing new small form factor single board computers. The Modules Working Group is developing a specification for a new small computer-on-module (COM) form factor. The Stackables Working Group is examining approaches to embracing new high-speed serial technologies into legacy systems in a smooth manner that preserves investments in I/O, cabling and enclosure designs. This activity is not limited to circuit boards. Related small form factor technologies and components such as MiniBlade rugged latchable peripherals have incubator groups now as well. There is a broad community and ecosystem of technology leaders served by the SFF-SIG.
Small Form Factor Special Interest Group.