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

Back to Subject List

Old thread has been locked -- no new posts accepted in this thread
???
02/13/03 08:59
Read: times


 
#38947 - Redundant controller system
I am working on a project that requires a redundant controller system. Using three controllers in an identical design allows the systems to control each others performance and vote if certain decisions are to be made.

I have found two different kinds of systems so far: one that does the voting in hardware at instruction level, and one that does it in software.

The first system has a big advantage in that the voting is done via a dedicated chip, which basically is a triple comparator; it monitors the data- and address busses of all three circuits and 'sees' what the processor is doing. If one processor is doing something different than the other two, it is voted out and the unit is effectively stopped. I like this method because it has a minimum overhead in software when compared to the software voting, so at this moment I am going for the hardware option.

The problem I am facing at this moment is synchronisation. Obviously, the clocks of the three controllers need to run synchronized. I have only briefly looked into this problem, but I think it wouldn't be that difficult by using PLL circuits. For the time being, I use a single oscillator with three synchronous clock outputs so that I am sure that the processors receive identical clocks.

As the modules need to be "hot replaceable", a module that is reset at any moment has to synchronize with the two units that remain running. This requires a sync routine in software, but first and for all the processor needs to be restarted at the correct moment. My initial idea was to count the clock pulses with external counters and create a "sync" pulse at each 6th clock that drives the reset after a power-up. However, I don't get that to work properly. Sometimes it works, most of the time it doesn't. Analyzing the ALE signals and the counter sync pulses I can see that the instruction cycles don't coincide, i.e. the processors don't run synchronous. FYI: clock frequency = 12 MHz, processor is 6 clock type.

So now for the big question: is there anybody out there who has faced a similar problem and found a good solution for synchronizing more than 2 controllers in a way that I indicated? Any suggestions are welcomed.

Regards,

Menno


List of 5 messages in thread
TopicAuthorDate
Redundant controller system            01/01/70 00:00      
   RE: Redundant controller system            01/01/70 00:00      
   RE: Redundant controller system            01/01/70 00:00      
      RE: Redundant controller system            01/01/70 00:00      
   RE: Fault tolerant system / Michael            01/01/70 00:00      

Back to Subject List