system excercising/burn-in tests?

Jim Kuzdrall gnhlug at intrel.com
Wed Aug 17 16:18:00 EDT 2005


Greetings Paul,

    Before you worry about comprehensive software tests, you should be 
aware of the most common failure modes for modern electronics.

    The semiconductor processes are so consistent and the testing so 
sophisticated that a "leaky" dynamic ram is not worth testing for.

    The most common reason for failure is detachment of the wire bond at 
the chip, attachment of the chip to the PC board, and cracks in the 
printed circuit wiring.  All of these are aggravated by thermal 
expansion and contraction.

    The military has found thermal cycling as the best way to accelerate 
these failures if they are going to happen.  Thermal cycling is just as 
effective with the power off, greatly reducing the cost.  It is also 
relatively fast.  You can complete it in a day or less.

    Basically, you take the equipment to it lowest rated storage (not 
operating) temperature.  Let is "soak" there until the "cold" has 
penetrated to the center of the components with the lowest thermal 
conductivity.  The calculation is quite simple and usually is about 20 
minutes for an open board.  A hour is a good safe time for most 
enclosed equipment.

   After the low temperature soak, bring the equipment up to the highest 
storage temperature - again, un-powered.  Use the same soak time.

   Repeat this procedure 5 to 10 times.

   When finished, the bad computers will no longer work or won't get 
through boot-up without an error.

   As a final check, run a thorough memory test at the highest operating 
temperature (not storage temperature).  Few offices run computers at 
-40C (they warm up above that quickly if so started), but it is common 
for them to run hot.  SuSE includes a memory test on their CD, but 
Linux must have something someplace.

   You might want to thermal-cycle the CDROM drives and disks separately 
from the motherboard if the storage temperature range of the mechanical 
stuff is very narrow.

   Anyway, that is how I have done it for the military.  You should have 
no trouble finding an idle temperature chamber in the area.  Some of 
them are big enough to hold a satellite, so you can do all the 
computers at once.

Jim Kuzdrall

 

On Wednesday 17 August 2005 11:46 am, Paul Lussier wrote:
> Hi all,
>
> Does anyone here have any experience with creating system burn-in
> suites?  Our vendor does a very rudimentary burn it (if it doesn't
> burst into flames on it's own while sitting in the 100-degree+
> warehouse on a shelf, it must be okay), and is puts the onus on it's
> customers to provide them with a burn-in test to be run (all for
> small, per-system fee, I'm sure :)
>
> So, now I'm looking to come up with a burn-in suite.  My initial
> thoughts were starting with a knoppix-like CD and hacking that to run
> various utilities, however, I'm not even sure what utilities would be
> good to run.
>
> We basically want to stress-test all components and busses.
> Especially the drives.  We're not doing performance testing at this
> stage, however...
>
> Any ideas?
>
> Thanks,



More information about the gnhlug-discuss mailing list