I saw the following on the Beagle Bone wiki page…
"Improper Power Down…All Revisions
… Based on the latest information we have gathered, this appears to be an instance where the PMIC does not power down the processor in the right order due to the violent removal of the power cables, either USB or DC. So, we are recommending that the power button be used to power off the board or by issuing the Halt command. After that you can remove the power cable. If everyone follows this process, we should see a decrease in these issues, but it will take a while for the results, either good or bad, to show up. This will also help prevent the contamination of the eMMC and the SD cards by allowing the kernel to shutdown properly before power is removed."
I also saw in one post that Gerald noted this was thought to be an issue, but now has been dismissed.
Can you confirm that this is still an issue or not.
The reason that I ask is that we are having Beagle Bone Black failures in the field where they go dead and we are trying to nail down why.
If this is still an issue would unplugging a power supply from the wall be considered “violent” removal of the power?
It is still an issue. We have seen it on .003% of the boards shipped. The issue can also be related to improper design of capes that violate the power up sequence. It may be a while before that shows up as a failure.
Thanks for your fast reply.
I’ve seen it on two of our boards - out of about 15. Only one of them had a cape (the RTC cape). Any idea if that one causes the issue?
I admit that the boards are installed over in Africa where clean AC power sources are not common. We are putting a UPS on the system to clean up spikes and avoid brown outs.
Does powering down a unit via unplugging a supply from the AC (the wall) also constitute a “violent” interruption in power, or is it just unplugging the power connectors from the board?
We are looking at the idea that maybe the PMIC gets confused when there is a power dip on power up, basically a power supply that can’t handle the surge when the board powers up. The idea is the PMIC starts shutting down and then tries to power up again when the dip goes away. I have yet to capture this myself, but there is one person he says he sees this on his board.
If you have power than can brown out for a short period of time, this could be your issue based on this idea we are looking out. You might try adding a battery to the battery lead or maybe a super cap that can at least stop the shutdown from starting when it thinks the 5V is going away…
Yes, we do have brown outs and are putting UPS on the system to avoid this, so that should take care of it for now.
One other thing that we’ve been concerned about for a while is that immediate shutdown could cause a problem if it interrupts a MySQL operation or even a filesystem operation.
It would great to have a power down detection cape with a super cap that keeps the board running for 30 seconds and triggers an input/interrupt that shuts down the system using the OS command. I plan to contact some cape makers to see if there’s interest in creating such a cape. Any suggestions for who to contact?
I don’t know of anyone I could recommend, but I would think that others in the community would have something like this.
The Power Cape will perform automatic switch-over to a battery on DC power failure. A script can monitor “DC power good” over I2C and shut the system down as required (I usually wait until the battery hits 3V since power here is usually restored before then). By default, the cape will also power the BB back on after DC restoration. It might be over-kill if you’re looking for a “super cap cape” but it’ll solve brown-out problems. I created it so feel free to hit me up with any questions…