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

Back to Subject List

Old thread has been locked -- no new posts accepted in this thread
???
11/02/06 14:05
Read: times


 		 #127261 - it's not just the matter of "kitchenware"...
Responding to: ???'s previous message
... programmers. E(E)PROMS do suffer from "memory loss" on the long run.
Just one story for all:
http://www.8052.com/forum/read.phtml?id=78564

There are also issues with increased radiation, chemical/humidity/temperature impact, mechanical shocks etc.etc.

Then, there are issues with detecting deliberately modified code, but I'd bet you don't want to hear too much about this one... :-)

Erik Malund said:
I have worked and am working with 'inherited' products. Some have checksum verification. I have seen NO catches by a checksum routine in my life (except verification immediately after programming). I have seen chips that 'lost their mind' but none that had enough left to tell.


Aaaaaah, to paraphrase Erik Malund, "testing does not prove absence of other modes of failure"... :-)

A CRC check upon startup costs little. But, I repeat myself, the appropriate-ness of particular algorithm etc. is strongly application dependent.

Jan Waclawek



PS. After preaching for so long, I must admit, I have similar experience and take a similar view than you and I usually don't do startup CRCs nor other checksums unless it has a particular reason, e.g. in one application I have to determine which of more sockets contains EPROM with valid code+data; recently I was forced to do it because of legislation requirements...


List of 26 messages in thread
TopicAuthorDate
Global Checksum            01/01/70 00:00      
   Checksum of what?            01/01/70 00:00      
      Rom Contents            01/01/70 00:00      
         why checksum???            01/01/70 00:00      
            The OP intended to use CRC-32...            01/01/70 00:00      
            ... but it's a good point otherwise...            01/01/70 00:00      
               timestamp vs crc            01/01/70 00:00      
                  timestamp, serial number, neutral 16 bit checksum            01/01/70 00:00      
                  who said 'automatic' and yes I have            01/01/70 00:00      
                     You didn't, Jan did            01/01/70 00:00      
   A bit of a logical fallasy            01/01/70 00:00      
      not at all... depends on application...            01/01/70 00:00      
      Statistics            01/01/70 00:00      
         ferrite beads            01/01/70 00:00      
            \"legitimate\" uses of checksum            01/01/70 00:00      
               I have seen NO catches by a checksum            01/01/70 00:00      
                  it's not just the matter of "kitchenware"...            01/01/70 00:00      
                     One thing that should not be handled by checksummi            01/01/70 00:00      
                        blah blah            01/01/70 00:00      
                        That's not how I remember it ...            01/01/70 00:00      
                           that was me            01/01/70 00:00      
                  I\'ve got a box that contains a few ...            01/01/70 00:00      
            links to CRC            01/01/70 00:00      
               I still say that crc's and checksums only work            01/01/70 00:00      
                  Most            01/01/70 00:00      
      I'd like to learn more            01/01/70 00:00      

Back to Subject List