Skip navigation

Hardware

2 Posts authored by: SharonS-OSM

In school, we knew we were doing OK when we passed the teacher’s test. But the stakes are higher now than in school days of yore: Testing commercial x86 designs and electronics adapted for use in rugged battlefield environments anywhere in the world is a matter of life and death (Figure 1). Same goes for the transportation industry. And, as discussed in Part 1 of this series, “Adapting commercial x86 embedded designs to harsh environments – Not as straightforward as you think,” prepping ruggedized commercial wares is not a matter of merely tacking on additional bolts or boards. So … what precisely is involved in this life-, mission-, and safety-critical testing? And must it be done on every single product that comes down the pipe, or does sample testing cover all the bases?

 

288iAA2658F640B199C7

 

[Figure 1 | Commercial-turned-rugged x86 designs and electronics can be found in rugged battlefield environments anywhere in the world. Testing is imperative to their reliability – and a matter of life and death. Pictured: An A-10 Thunderbolt II, presently engaged in combat missions over Afghanistan supporting Operation Enduring Freedom. U.S. Air Force photo by Master Sgt. Robert Wieland]


Show me the rugged … but when?

When ruggedizing a commercial x86 system, parts qualification is mandatory, and there are various schools of thought. Perhaps the most compelling, however, is the choice between: Should a company qualify the parts by test or guarantee them by design?[1]

 

Qualifying by test happens when upscreening occurs exclusively at a subsystem, board, or other high levels of integration, while the board or subsystem’s individual off-the-shelf components remain untested for harsh environments. This practice can sometimes even pass some levels of Highly Accelerated Stress Screening (HASS), but the result will always be early and unpredictable failure for the module.

 

On the other hand, guaranteed by design comprises 100\% inspection and rating of individual mechanical devices and components – before they are integrated into a board or system. This is vital when “failure is not an option.”

 

Case study: Do we have to do this every time?

Christine Van De Graaf, Kontron’s Product Manager, Embedded Modules, utters a resounding “yes” when asked if it’s really necessary to test each individual commercial-turned-rugged component. Terminology aside, Kontron, a Premier member of the Intel® Embedded Alliance, subscribes to the aforementioned “guaranteed by design” philosophy in cases when commercial-grade components must be used. Most of Kontron’s military and transportation customers request the industrial temp range of -40 °C to +85 °C, she says – a range beyond the grasp of commercial-grade embedded products (Figure 2).

 

289i85FCA4B0C6DC362E

 

[Figure 2 | Most of Kontron’s military and transportation customers request the industrial temp range of -40 °C to +85 °C, which is beyond the grasp of commercial-grade embedded products.]


But the converted commercial wares are in for anything but an easy ride. Kontron’s screening includes strenuous thermal testing while cycling the Unit Under Test’s (UUT’s) temperature, running burn-in tests, soaking the device, and conducting several other (albeit proprietary) test procedures.[2]

 

Testing is additionally performed for moisture and corrosion resistance, and Kontron uses Humiseal 1B31, inspection-friendly via its black-light visibility trait, as its products’ primary coating material. The Mil-I-46058C, IPC-CC-830, and RoHS Directive 2002/95/EC packaging specifications are additionally met.

 

Meanwhile, shock and vibe metrics are tested to the IEC 60068-2-27 and IEC 60068-2-6 standards, with custom tolerance tests run when requested.

 

And, as mentioned, there’s no sample testing here. “We do 100\% testing,” Van De Graaf asserts, “so that nothing the customer receives will be DOA. It also confirms it’s not going to fail the first time they boot up and run it in an extended temperature range. Some organizations only do a small sample testing, then certify the product and say they all can be used. But doing a small sample testing is not a valid way to say ‘This can withstand extended temperature ranges.’”

 

And, she adds, honesty is still the best policy when it comes to converting x86 products into the industrial range: Kontron only performs 100\% screen processing on commercial-to-industrial products if it's feasible; otherwise, Van De Graaf tells the customer it just won’t work. “For example, the Intel® Core™ 2 Duo SP9300 – We won’t put it into industrial temp designs or testing because we already know that one runs too hot. It would fail even in a screening format. Even some of the higher-end Core™ 2 Duos – there are some flavors we will not do a screening on because we know they will fail.” The result: back to the proverbial development “drawing board” for a new design plan.

 

But there’s always the hope that an ideal scenario arises where all industrial-temp wares can be used, without transforming anything commercial-grade into rugged. Recently, Kontron realized the dream with its first “by design” aka all industrial-temp components product: the microETXexpress-XL[3], a Computer-on-Module featuring the relatively new industrial-temp Intel® Atom™ Z520PT processor and US15WPT system controller hub, designed for use in mobile warfighter, public transportation, and outdoor POS applications. [“By design” is not to be confused with the “guaranteed by design” conversion theory described earlier].

 

If such innately and completely industrial designs were possible all the time, ruggedizing commercial components or designs would become a thing of the past, but is that feasible? What is holding processor vendors back from providing more industrial temp versions of their products?

 

Written by Sharon Schnakenburg-Hess, an assistant managing editor at OpenSystems Media®, by special arrangement with the Intel® Embedded Alliance.


References:


  1. When failure is not an option: Test and qualification help ensure reliability in mission-critical environments,” by Doug Patterson, Aitech Defense Systems, VME and Critical Systems Magazine, April 2009, www.vmecritical.com/articles/id/?3888
  2. “Commercial Off The Shelf (COTS) Configurable Systems and Modules,” Kontron, March 2010.
  3.  “COM designed to meet the environmental challenges faced by outdoor POS systems, public transportation vehicles, mobile warfighter and more”, Kontron, Jan. 2010. 

 

Face it. The ubiquitous Intel® x86 architecture has been around a (relatively) long time, and it provides myriad benefits, especially in portability and ease of legacy code use. However, though x86 thrives and drives benign commercial benchtop and desktop environments, what about the rugged designs required in the military arena? Sometimes a customer-requested military temp component is simply not manufactured, or it is extremely price-prohibitive, thus commercial wares must be used. But converting a commercial x86 design to a rugged one is not a straightforward proposition.

 

The following discussion centers around these issues:

 

  1. At what stage should designs be adapted from commercial to rugged?
  2. Thermal considerations for the Intel® Atom™ and Core™ i7
  3. Building in shock and vibe tolerance
  4. System-level allowances for commercial-to-military adaptation

Timing is everything: When to transform commercial into rugged

Curtiss-Wright Controls Embedded Computing and GE Intelligent Platforms, two military embedded industry heavyweights, both employ an “adapt first, not last” mantra when ruggedizing commercial designs. They are also both quick to point out that ruggedizing a commercial design differs from merely using ruggedized commercial components.

 

Curtiss-Wright, an Affiliate member of the Intel® Embedded Alliance, utilizes a “built rugged” concept when it comes to using commercial wares in rugged designs, meaning “designing and manufacturing circuit cards to be rugged (i.e. harsh environment capable) at the outset. We have always had this philosophy, and we have used and still use commercial temperature components,” explains Ivan Straznicky, Principal Engineer and Technical Fellow 
at Curtiss-Wright.

 

A similar outlook is adhered to at GE Intelligent Platforms, an Associate member of the Intel® Embedded Alliance. “Our philosophy is ‘designed to be rugged’: We don’t believe it is possible to create a truly rugged product ‘after the fact.’ Ruggedness needs to be designed in from the very outset,” says Frank T. Willis, Director, Military/Aerospace Product Management at GE. He further explains, “Adapting a commercial design is not the same as adapting commercial hardware.  It’s incredibly difficult to adapt a commercial design [often 0 ˚C to 60 ˚C] to rugged requirements [GE’s rugged temps are typically -40 ˚C to +71 ˚C]. …  The design approaches are generally so different that it makes more sense to start [the design] over,” he adds. 

 

Design procedure must begin with analysis of thermal and structural components, in addition to component selection and signal integrity consideration, requiring advance determination of PWB routing and layout protocol, Willis explains. “These factors are more difficult – in many cases impossible – to add later in the product manufacturing cycle. … This can only be done from the ground up, not by just adding a metal frame to a board and hoping for the best.” GE also uses commercial components as necessary within their harsh environment designs, and both companies use careful testing to ensure reliable commercial-to-rugged hardware adaptation. (See also Part 2 of this series, “Adapting commercial x86 embedded designs to harsh environments – Can it pass the test (every single time)?” for details on commercial-to-rugged testing practices.)

 

Intel® Atomand i7: Some like it hot … or maybe not as much

As Intel’s relatively newborn, commercial-temp versions of the Atom™ and Core i7™ processors gain market traction, their thermal design capacities and limits must be considered when using them in ruggedized warfighter technologies.

 

Adapting commercial processors suited for forced-air cooling in low ambient temps to a conduction-cooled design can be a “very involved” process, says Straznicky. “The conduction cooling solutions employed must have high efficiency, i.e. the thermal resistance from processor to ambient must be low.” The most critical heat removal path is the contact resistance between the card edge and the chassis rail contact, and a 5-10 ˚C thermal contact resistance reduction allows a 5-10 ˚C chassis rail increase to raise permissible ambient temperatures, he adds.

 

Meanwhile, Willis says both the Intel® Atom™ and the Core i7™ can be used effectively in conduction- or rugged air-cooled solutions, but Atom’s 10 W power dissipation makes it the easiest to manage of the two. Effective thermal conductivity for the 30 to 95 W Core i7™, though, necessitates “proprietary heatsinking methods … between the part itself and the heatsink (not forgetting a good heat path through the PCB to the heatsink also).”

 

Shaking it all up: Shock and vibe

“Shock and vibration consideration for rugged products using commercial components is much more than simply tweaking commercial designs,” asserts Straznicky. Some of the unique x86 card failure mechanisms and modes induced by shock and vibe include socket contact fretting corrosion or pad cratering underneath processor solder balls. Other types of mechanical failures can additionally occur with various processor types[1] during environmental testing (Figure 1).

 

284i4F5AD508904F988E

Figure 1 | Starting at top left, going clockwise, failure analysis shown: 1) Excessive local strains induce pad cratering underneath the BGA in vibration tests; 2) Direct module exposure to 500 hours of salt fog results in a salt bridge; 3) A vibration test produces excessive micromotion and resultant connector contact fretting corrosion; and 4) Insufficient cooling fried this processor during a high-temp test. Figure courtesy of Curtiss-Wright.

 

Learning about vibe and shock effects to the circuit card and processor is pivotal in the rugged equation, as is their mitigation. Thus, Straznicky offers these insights:

 

  • “Circuit card stiffeners are ubiquitous on conduction cards to reduce displacements, strains, and stresses imparted by shock and vibration.
  • Lighter materials such as aluminum (0.0975 lb/in3) are used instead of copper (0.321 lb/in3) to conduct heat, even though copper has a much higher thermal conductivity.
  • Shock and vibration isolators are often used on or as chassis mounts to reduce the shock/vibration levels imparted to the chassis and enclosed cards.
  • Chassis parts are structurally solid and brazed together to provide a rigid construction that will not fatigue and reduces vibration amplification.”

Additionally, pins versus mounting points is also an important consideration: Merely meeting electrical requirements via pins doesn’t work, as enhanced vibe and shock capability begins with components featuring several mounting points instead, Willis explains.

 

Systemic issues for commercial-to-rugged adaptation

When mitigating high component temperatures when allowances have been made for adapted commercial products within a rugged system, Willis suggests these steps: 1) Make sure that the PWB-to-component thermal interfaces are effectively designed, as mentioned earlier. 2) Check that the thermal resistivity is minimized by the chassis-to-PWB card edge interface.

 

“Two-phase or spray cooling can cool the highest power densities,” adds Straznicky. Single-phase liquid cooling can also effectively cool extremely high power levels, in excess of 650 W on a 6U liquid flow through card, for example. Other methods have also shown promise in rugged designs (Table 1).

 

290i5EFEE412E54505B1

 

Table 1 | Several cooling methods have proven effective in modern ruggedized designs.


Worth the trouble?

Though an ideal rugged design would contain only mil temp components, the practice of adapting commercial temp wares into rugged designs will always be required, says Straznicky, but it might cost vendors. So … is it truly worth it?

 

Written by Sharon Schnakenburg-Hess, an assistant managing editor at OpenSystems Media®, by special arrangement with the Intel® Embedded Alliance.


References:


1. “VITA 47: Environmental considerations for VITA technologies,” by Ivan Straznicky, Curtiss-Wright Controls Embedded Computing, www.vmecritical.com/articles/id/?3748

2. “Conduction Cooling Techniques for Rugged Computers,” by Serge Tissot, Kontron, www.mil-embedded.com/articles/id/?4476

 

 

Filter Blog

By date: By tag: