2 Replies Latest reply on Feb 22, 2017 2:27 AM by SpiderKenny

    Protecting Edison from power-off


      I'm working on a product design with Edison.

      The product could be switched off by the user at any time, without a chance to "tell" Edison that the power will go off. Or the mains power might just fail at anytime.


      When we did a power-off test during a prototype testing stage yesterday, Edison wouldn't boot again afterwards. It would get to "press any key to stop autoboot" but then would crash and re-start, so I can only assume that the OS partition got corrupted during the power-off. I restored the OS using "flashall.sh" and copied the applications back on and it all worked again, so no permanent damage.


      So, what are my options for protecting Edison for being corrupted during a power-off event?

      I know I could use the "poweroff" command from the shell - but we won't always have the opportunity, occasionally the power may just go off without warning.

      I could add a couple of large capacitors to give us a few seconds of runtime, but what's the best way to detect that the power is going off, and what's the best action to take?


      I'm thinking some kind of ADC with a system kernel driver that just does a processor "halt" - would that work? I'b be worried that it gets triggered by transients though.


      Any suggestions? The design goal here is commercial grade prevention of corruption to the OS during a power off event.

      I don't care about preventing damage to the application data, it's never gets persisted anyway.


      Thanks everyone!