Ruggedization advancements enhance memory module design

Managing memory module design is a whole new ball game in rugged environments.

3Embedded OEMs are looking to the latest memory technologies to solve their specific design needs and market demands. But which memory modules provide the most optimal solution for excessive shock and vibration or increased thermal dissipation? And what new testing and validation techniques are being used to reduce overall design risks and increase reliability? Designers must evaluate these factors and other key considerations when specifying memory devices for embedded systems in rugged environments.

Embedded systems designers historically have had limited memory products for applications that need to operate in rugged high-shock and -vibration conditions. That is because advancements in memory technology and its associated standard Dual In-Line Memory Module (DIMM) and Small Outlined DIMM (SODIMM) form factors have been largely driven by PC, telecom, and server market requirements. Memory modules designed for these market applications typically do not meet crucial embedded application specifications that must allow for space-constrained layouts while also providing high reliability and performance with long-term operation in rugged or harsh environments. In the embedded market, memory products must support long product life cycles and be cost-effective as well.

Some memory module suppliers focus on the needs of the embedded market and continue to develop memory technology advancements. Memory suppliers have come together through various standards groups to make these advancements in commercially available memory modules, giving embedded systems designers access to rugged devices in a wide range of capacities. This standardization also brings the added benefit of consistent availability from multiple suppliers, which helps OEMs accelerate time to market while reducing overall system cost and project risk.

Making strides in rugged memory technology

Memory technology innovations have made a variety of ruggedized options available to embedded system OEMs, including lower-profile module designs, Error Correction Code (ECC), thermal dissipation, extended-temperature operation, and the addition of thermal sensors to monitor module temperature.

Embedded system OEMs look to Double Data Rate type three (DDR3) SODIMM memory modules as the mainstay for rugged embedded system design. Adding to the ruggedness of DDR3 SODIMM is new low-power, low-dissipation DDR3L memory modules, which solve a key embedded system design challenge. JEDEC stipulates that systems running memory beyond +85 °C must double the DDR3 self-refresh rate. DDR3L memory modules resolve the double refresh rate requirement by selecting the lowest total electrical current, incorporating thermal-relief copper pour methodology PCB design, reducing chip count, and utilizing 1.35 V DDR3 Dynamic Random-Access Memory (DRAM). Compared to current DDR3 designs, DDR3L memory can save up to +10 °C per module and remove the double refresh rate requirement. Supplier-based testing has shown that depending on the components used, DDR3L modules contribute to significantly reducing power, thereby helping increase performance (see Table 1).

Table 1: Virtium internal test data shows that, depending on the components employed, OEMs can realize up to 50 percent power reduction with an 8 GB ECC memory module.
(Click graphic to zoom by 1.9x)

Blade VLP is a lower-profile (17.78 mm) alternative to the JEDEC standard VLP with a height of 18.75 mm (see Figure 1). Reducing the height of a DDR3 VLP memory module to a lower-profile 17.78 mm solves the space-constrained limitations found in many telecom and networking applications, where it is difficult to accommodate the memory required for both an industry-standard DIMM or Mini DIMM socket plus a standard VLP. This approach allows designers to reduce the total power in systems that use multiple memory modules and those that must run above +85 °C, which is a typical design challenge in a wide range of -based telecom and Ethernet blade switch networking applications.

Figure 1: Virtium’s Blade VLP allows more airflow in the system and utilizes low-power DRAM to reduce thermal dissipation up to +10 °C on the DRAM surface.
(Click graphic to zoom by 1.9x)

Designers of telecom and networking blade systems typically face strict limitations on system height. In addition, these systems require spacing on top of the memory module to enable airflow for effective thermal management. Incorporating a reduced height DDR3L VLP memory module helps improve airflow and provides a low profile, allowing OEMs to offer higher-reliability products that reduce total cost of ownership. Specific DDR3L VLP modules also offer single refresh rates, which are now essential to maximize performance in high-temperature systems.

Multiple methodologies contribute to ruggedization

To help OEMs meet extreme requirements for vibration, temperature, or other harsh environmental conditions, memory suppliers offer manufacturing advancements such as side retainer clips to ruggedize DDR3 SODIMM modules (see Figure 2). These universally adaptable clips can be easily implemented on a variety of applications. In the recent past, designers typically were limited to weaker, commercial-grade retainer clips to keep memory modules in place. Under certain conditions, these retainers can pop open and result in system-level failures. Other alternatives that involve mounting holes require significant modification to the mainboard, frequently resulting in a nonstandard COTS-based design that does not adequately address the problem.

Figure 2: Virtium offers two types of locking devices that are designed for modules with and without heat sinks and require no holes or physical modifications to the motherboard for installation.
(Click graphic to zoom by 1.9x)

In addition, OEMs can take advantage of the underfill option that provides increased shock resistance of components populated on standard FR-4 PCBs. Conformal coating is another option that can be qualified under MIL-I-46058C to provide enhanced protection from environmental degradation.

Along with mechanical enhancements, numerous electrical upgrades are available to OEMs, including extended-temperature screening and burn-in and the addition of thermal sensors to monitor module temperature. Designers can typically select from three temperature options in memory modules:

  • Industrial temperature: -40 °C to +95 °C
  • Extended temperature: -25 °C to +95 °C
  • Standard temperature: 0 °C to +95 °C

Testing is essential to assure that modules adhere to temperature specifications. Thus it is important that a standard set of temperature test parameters is defined and that memory suppliers collaborate with OEMs to adjust testing methods per specific equipment and testing time needs. Embedded systems often perform mission-critical operations, so after testing definitions are established and validation is completed, it is advisable that memory modules be 100 percent tested according to the defined plan.

The optimal testing method to ensure extended-temperature operation is accomplished via production testing using customer motherboards or on approved motherboards with identical chipsets and setup. These tests can also be performed using specifically developed ovens that match the size of most form factors, enabling temperature testing at full system performance.

System testing is critical to catch defects such as ECC errors that cannot be discovered using standard test systems. It is important to note that depending on the application or system specifications, system-level tests might also be necessary using Temptronic ThermoStream units for temperature requirements as low as -45 °C, or tested using an under-temperature oven for up to 12 hours at +85 °C ambient.

An analysis of the failure modes of DRAM in memory modules has determined that DRAM components with suboptimal reliability tend to fail during the first three months of use. As newer DRAMs advance to smaller process geometries, there can be a greater risk for chips that contain weak bits (a microscopic defect in an individual cell). This is not enough to cause a DRAM failure outright, but could exhibit a single-bit error within weeks after initial field operation begins.

Using Test During Burn-In (TDBI) helps eliminate any potential early failures and improve the overall reliability of memory products. Although most DRAM chips undergo a static burn-in at the chip level, TDBI offers a more comprehensive testing approach that implements a 24-hour burn-in test at the module level while dynamically running and checking test patterns as the module is performing under stress conditions. Studies conducted by various memory manufacturers show that using TDBI chambers can reduce early failures by up to 90 percent.

New standards and form factors

Multiple industry organizations such as JEDEC and the Small Form Factor Special Interest Group (SFF-SIG) are actively involved in standardizing memory devices for today’s embedded systems. Standardization brings the added benefit of consistent availability from multiple suppliers, which helps OEMs accelerate time to market while reducing overall system cost and project risk.

ECC has become a mainstay in embedded systems. However, the JEDEC membership initially did not recognize the need to accommodate ECC when it was developing the DDR2 specification on the SODIMM form factor because most laptop chipsets did not support ECC at the time. Seeing the need for ECC that could be implemented on faster DDR2 memory modules in embedded systems, Virtium sponsored the ECC SODIMM specification within JEDEC, which has been extended now to DDR3 and DDR4 modules.

The XR-DIMM specification from SFF-SIG is another example of a memory device defined to meet embedded system needs to reliably operate in excessive shock and vibration conditions. Designers of these systems needed a small form factor, extremely rugged DDR3 module. This standard relieves designers from the former limitations of commercial-grade products that required soldering, straps, glue, or tie-downs to secure the module.

A collaboration among Virtium, Swissbit, and LiPPERT Embedded Computers working through the SFF-SIG resulted in a module with a pin definition that closely resembles that of a DDR3 standard DIMM. The pin definition leverages a high-performance 240-pin SMT connector system that uses standoffs with screw attachments to firmly hold the XR-DIMM memory module to the motherboard. Additionally, the pin definition includes a SATA interface to enable the development of dual-function modules that contain both DDR3 and NAND flash for a combined memory and solid-state drive storage implementation from a single socket. Future standards for combination SATA and DDR3 modules are planned.

Satisfying rugged memory requirements

While demands for ruggedized embedded devices continue to rise, memory module suppliers continue to make technology advancements and associated manufacturing enhancements that meet the needs of OEMs. DDR3 SODIMM, DDR3L, and lower-profile DDR3L are all examples of new technologies that help satisfy rugged memory requirements. These advancements solve many design challenges, including low power, enhanced thermal dissipation, and extended-temperature tolerance while delivering the performance needed for today’s complex embedded systems.

Standards for XR-DIMM and ECC SODIMM have also contributed to advancing a ready supply of rugged memory products. Furthermore, designers have access to underfill side retainer clips and conformal coating manufacturing options along with advanced testing methodologies to help ensure robust designs.

The challenges to maintain the highest reliability and availability in rugged will continue, but memory module advancements are set to keep pace with these requirements, helping enable OEMs’ continued competitiveness and future innovation.

Phan Hoang is VP of R&D at Virtium Technology.

Virtium Technology

Follow: Twitter Facebook Google+ LinkedIn