Email: Password: Remember Me | Create Account (Free)

Back to Subject List

Old thread has been locked -- no new posts accepted in this thread
???
07/26/07 08:53
Read: times


 
Msg Score: +1
 +1 Good Answer/Helpful
#142333 - the value of such test...
Responding to: ???'s previous message
Doug,

Well, I believe it is fully justified to run such a test for the reasons you mentioned - especially to make your customers happy. I would go further in this direction, and stress the visual side of the even: I would take pictures and tape a video of an example testing setup, with red glowing IR-heater-based furnace in the background, the chip sitting atop a nice scientifically looking test fixture, with all cables (color coded!) neatly tied up - plus a PC-monitor, with the current temperature in huge font, rising all the time; and a window showing the test running (rapidly scrolling numbers - have you seen Matrix? - stopping at the moment of failure) - all in the view at the same time. No kidding!

On the other hand, you (and your boss) should be aware that the real value of this sort of tests is questionable, for various reasons. Just some of them:

  • as mentioned already, you cannot guess the coverage of your test unless you have intimate knowledge of the chip's internals (down to the transistor/interconnects/topology level). You hinted that this is an ASIC - in this case, you might have access to these data, but to calculate or at least estimate coverage of a particular test - or to develop one with a fair level of coverage - will still be a challenge.
  • there are also modes of failure which you won't cover by a simple "logic" test. For example, leaks might develop across the CMOS structure (increased IDD), FLASH retention might suffer, etc. These might lead to various "soft" failures under e.g. marginal voltage (not only high but also low), low temperature etc.
  • testing a new chip simply cooking it to a high temperature also tells nothing on long-term impact of elevated temperature/voltage operation. Think Arrhenius.
  • testing a few specimens tells you nothing on reliability on the long run (e.g. in a new batch etc.)
  • testing from the "top" is not the best approach in such complex systems as ICs are. There are many things which can go wrong - individual transistors, interconnects, bonding etc. - the manufacturer tested them all individually under various conditions and the overall reliability picture is put together in this "from bottom to up" manner
  • there are more factors leading to IC failure than just simply cooking under maximum voltage. Think of chemicals (mainly moisture), mechanical stress (vibrations and shocks), rapid thermal cycling; and all of these have their long-term implications, too.

So, if you want real data, I think you should go to the manufacturer and ask for process qualification data and quality assurance procedures.

I might be wrong, of course - please note that I have no, zero, nada real-world experience with the matter - I am just repeating here what I have learned in school, 15 years ago.

JW



List of 25 messages in thread
TopicAuthorDate
Processing capability algorithm            01/01/70 00:00      
   At your computer, within a half an hour?            01/01/70 00:00      
      re: Processing capability algorithm            01/01/70 00:00      
         what is "fails outright"            01/01/70 00:00      
            re: Processing capability algorithm            01/01/70 00:00      
               Idea...            01/01/70 00:00      
               you can't do that            01/01/70 00:00      
               hummm, Oxford...            01/01/70 00:00      
   Look In the Space Program            01/01/70 00:00      
      there are some rad-hard '51's            01/01/70 00:00      
   re: Processing capability algorithm            01/01/70 00:00      
      the value of such test...            01/01/70 00:00      
         a bit iof emphasis            01/01/70 00:00      
            rethinking the value...            01/01/70 00:00      
            ESD, 100C operation, and corrosive gas            01/01/70 00:00      
               not operational burn-in and testing            01/01/70 00:00      
                  Pushing chip limits            01/01/70 00:00      
                     HUH?            01/01/70 00:00      
                        Authority vs Certainty            01/01/70 00:00      
                        the misunderstanding squared            01/01/70 00:00      
                           Misunderstanding cubed and baked            01/01/70 00:00      
                              OK I am not much of a chipmaker...            01/01/70 00:00      
                              also, the 100°C was an arbitrary number...            01/01/70 00:00      
                                 Proving the negative            01/01/70 00:00      
                                    this is the point.            01/01/70 00:00      

Back to Subject List