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

Back to Subject List

Old thread has been locked -- no new posts accepted in this thread
Per Westermark
02/22/12 08:02
Read: 445 times
Sweden


 
#186139 - I just can not agree with that reasoning
Responding to: Kai Klaas's previous message
Sorry, but I don't agree at all.

1) You basically claim that a watchdog doesn't do anything. That is bull. It can't help in 100% of cases. But it most definitely can help. A good watchdog have a very simple state change diagram. So it is way harder to make it lock up.

2) You basically claim that you trust 100% your fuses/arrestors, ... They are no better than a watchdog. They can only improve the situation. Not give 100% security. So availability of input protection is no argument for removal of watchdog.

I haven't heard any claim that it was impossible to fail a curve with the Audi Quattro. Only that it did stay on the road, until it couldn't anymore. So people didn't have any feedback how near the limit they where. And no margin for recovery when it finally passed the limit. But the car did deliver, if the requirement was to stay on the road as long as possible. And that is what we want our hardware to do - keep working as long as possible. Until the hardware gets so abused that we just can't keep ticking anymore.

"But it gives you only the illusion of safety."

That requires that you can prove that a watchdog can never help. Because if one in 100 failures gets recovered by a watchdog reset, they your sentence is directly wrong.

But to use the word "illusion" correctly, I can claim that it is an illusion to believe that you can design a sellable product where input protection can be 100.000% guaranteed to protect the unit from a hang. As Erik notes, you can't power up a device and expect 100% reliability.

"No, the only way to handle these problems is proper filtering, shielding and grounding."
Totally male bovine manure, to use an expression loved by another forum visitor. There are no "only way". Engineering is about looking at "all ways". You are ignoring a large part of all research done about the design of reliable devices with your "only way". You are putting all your energy on a single link of a multi-linked chain.

No one in this forum have made a claim that a watchdog can replace good protection circuitry. But you are making the claim that good protection circuitry can replace a watchdog. You can't even sell your products in some product areas with such a design rule.

List of 44 messages in thread
TopicAuthorDate
Pointless watchdog?      Christoph Franck      02/18/12 12:27      
   Sounds Like From What You Said...      Michael Karas      02/18/12 15:00      
      Didn't want to point fingers.      Christoph Franck      02/19/12 02:33      
         A?      JecksonS Ben      02/19/12 04:50      
            Don't forget AVR      Andy Neil      02/19/12 05:44      
               One more argument ...      Christoph Franck      02/20/12 06:14      
                  Independently testable      Per Westermark      02/20/12 11:00      
                     Slow but critical.      Christoph Franck      02/20/12 14:13      
                        Yes      Per Westermark      02/20/12 17:59      
                           ESD      Kai Klaas      02/21/12 05:38      
                               You can't      Per Westermark      02/21/12 06:04      
                                 You can...      Kai Klaas      02/21/12 06:27      
                                    But emi-filter, overvoltage protection etc may not be enough      Per Westermark      02/21/12 06:46      
                                       ESD ...      Kai Klaas      02/21/12 07:51      
                                          It's not just ESD, it's "the works"      Christoph Franck      02/21/12 08:47      
                                             I do not trust the watchdog...      Kai Klaas      02/21/12 09:44      
                                                The simpler, the more trustworthy.      Christoph Franck      02/21/12 11:08      
                                                   Less complex does not mean less probable to fail!      Kai Klaas      02/22/12 06:48      
                                                   WDT and debug breadcrumbs      Oliver Sedlacek      02/23/12 06:14      
                                                Lightning strikes      Kalpak Dabir      02/22/12 04:33      
                                                   A watchdog should never be needed - but should be there      Per Westermark      02/22/12 05:55      
                                                   A watchdog isn't a fuse...      Kai Klaas      02/22/12 07:19      
                                                      correction      Erik Malund      02/22/12 07:35      
                                                      I just can not agree with that reasoning      Per Westermark      02/22/12 08:02      
                                                         Oh please, don't...      Kai Klaas      02/22/12 17:04      
                                                            so?      Per Westermark      02/22/12 17:26      
                                                Why you need a watchdog      Oliver Sedlacek      02/23/12 04:07      
                                                   And even if you had perfect software and hardware ...      Christoph Franck      02/23/12 05:03      
                                                      Reset versus power cycle      Kai Klaas      02/23/12 05:34      
                                                         Mars Pathfinder      Christoph Franck      02/23/12 06:27      
                                                         many modern chips      Erik Malund      02/23/12 07:05      
                                                            The problem with bidirectional reset pins.      Christoph Franck      02/23/12 07:26      
                                                               The uC in question      Erik Malund      02/23/12 08:07      
                                                            Latchup      Per Westermark      02/23/12 07:29      
                                                               ONLY clocked circuits?      Erik Malund      02/23/12 08:15      
                                                                  That kitchen exhaust fan ...      Christoph Franck      02/23/12 08:42      
                                                            link...      Kai Klaas      02/24/12 05:01      
                                                               Sales talk      Per Westermark      02/24/12 05:55      
                                                   certified applications      Erik Malund      02/23/12 07:00      
                  Using external watchdog      Oliver Sedlacek      02/21/12 03:11      
                     It's a feature, actually.      Christoph Franck      02/21/12 04:16      
               Wet      JecksonS Ben      02/24/12 01:04      
   Featuritis...      Kalpak Dabir      02/20/12 19:14      
   Looking at it from another perspective      Jez Smith      02/23/12 22:38      

Back to Subject List