Where is energy stored in my PC?

I have an HP Pavilion Elite desktop computer, model HPE-490t . I like it because it didn’t cost too much, boots itself from an SSD, came with 16 GB of RAM, and has 6 CPU cores for editing video and camera RAW images. It has one behavioral quirk that I cannot explain, however. The recent power interruptions here in the Northeast got the machine into a state where it could not be restarted. It would power up for a second or two, shut down, and then power up again, never being able to get to the point of showing anything on the monitor. I unplugged it for about 10 seconds and plugged it back in. Same behavior. I unplugged it and walked away for an hour, then plugged it back in and it worked perfectly.

I think something similar happened after installing a second hard disk drive into this machine.

So the question is why does the computer behave differently depending on how long it has been unplugged? Where is energy stored that affects the machine’s ability to boot? Capacitors in the power supply? Battery on the motherboard (there is one for the clock, but that wouldn’t be exhausted by being unplugged for an hour, I don’t think)?

21 Comments

  1. Brian

    November 2, 2011 @ 11:21 pm

    1

    Completely rampant speculation here, but it could be that your RAM needs some time to decay to a randomized state

  2. philg

    November 2, 2011 @ 11:29 pm

    2

    Brian: Why does RAM need to be randomized? Following a cold boot, the computer doesn’t try to read anything from RAM that it has not recently written into RAM, does it?

  3. Dave Kammeyer

    November 2, 2011 @ 11:41 pm

    3

    Probably thermal. You probably have a cracked trace on some board or chip that is sensitve to thermal expansion.

  4. presidentpicker

    November 2, 2011 @ 11:50 pm

    4

    It sounds like the “Power Good” signal from the power supply was not asserted. When this happens the motherboard does not power up and in turn sends the power off signal back to the power supply. a micro controller inside the PSU probably needed a full reset and that may not happen until the voltage drops to less than 1v. My guess is that it takes 10-15 minutes for the smoothing caps to drain down to the microcontroller “reset” voltage

  5. Jitesh

    November 2, 2011 @ 11:52 pm

    5

    I think in my Dell when I pull the power plug, a green LED on the motherboard takes 10-15 seconds to turn off. I’d assume it’s all stored in capacitors.

    There’s a pretty big gap between 10 seconds and 1 hour. Maybe 30 seconds was all you needed. I’d wonder why you’d have to unplug it to begin with.

  6. Jeffrey Friedl

    November 2, 2011 @ 11:55 pm

    6

    Dave’s thermal idea is probably the most on track, but capacitors could well be taking time to discharge. Both suggestions are just triggers for some separate problem (e.g. a crack somewhere, or a connection made when something expands) so once it boots I’d be looking to doublecheck your backups.

    (BTW, “raw” in reference to image files is not an acronym… it is the English word meaning “not cooked”, so should be written as “RAW” only when yelling.)

  7. Shane

    November 2, 2011 @ 11:56 pm

    7

    There are definitely capacitors in various places, including the power supply. I can’t speak to where they all are or why there isn’t some mechanism to drain them upon power failure. But whenever I cut power to my PC (usually prior to cracking it open to hack on the hardware), I hit the Power button afterward. It revs up for a fraction of a second before stopping, at which point I assume everything is drained. Interestingly, there’s an LED on the motherboard that stays lit until I do this; I wonder if it was placed there for this purpose.

  8. Roger Bigod

    November 2, 2011 @ 11:58 pm

    8

    I blame things like that on the cache. (I’m not a hardware person, so I have a very abstract, even ethereal concept of the cache.) If it’s power fails in the middle of a calculation, some temporary memory will contain crazy intermediate results. The power up program has a test that says “if contents of temporary memory are crazy, there must be a hardware problem, so exit”.

    There may be a way to check this. If there’s an alternate way to start up that explicitly clears all temporary memory, you should get an immediate startup without the wait of an hour.

    On my Macs there’s a magical incantation of holding down “p” and “r” plus a couple of special keys on startup. This clears “parameter RAM”, whatever that is. I do it every couple of weeks when I check disk permissions. It may not be accomplishing anything, but it makes me feel virtuous.

  9. philg

    November 3, 2011 @ 12:02 am

    9

    presidentpicker: I like your http://en.wikipedia.org/wiki/Power_Good_Signal explanation the best. It does seem as though the computer is waking up just enough to immediately shut itself down.

  10. Richard Miles

    November 3, 2011 @ 2:32 am

    10

    Have the identical issue with an HP Pavilion dv9500. Following guidelines in this HP support page http://h10025.www1.hp.com/ewfrf/wc/document?docname=c02798426&cc=us&dlc=en&lc=en&product=3255934&tmp_track_link=ot_search
    I found that disconnecting the power, removing the battery, and then holding down the power on switch for more than 15 seconds to remove any residual current, then reconnecting the power cord but leaving out the battery, did at least get it past the immediate shutting down – the led lights stayed on. But that’s as far as I got. Screen remained black. Same black screen after connecting to an external monitor. Prior to failing the computer had been getting hot, if allowed to automatically go into sleep mode, it would get extremely hot. So, at this point it looks as if it got fried. I’ll try pulling the RAM and H.D. and report if there was any success after doing that.

  11. Tom

    November 3, 2011 @ 9:15 am

    11

    Have you considered an UPS to at least let you shut the computer down in an orderly way? It will also let you ride over momentary interruptions.

  12. boozedog

    November 3, 2011 @ 11:38 am

    12

    Similar to Richard’s comment, I’ve found that switching off the power supply (or unplugging the machine) then pressing the power switch will discharge whatever juice is remaining. Depending on your model u may have to press the power button a couple times to get it fully discharged.

    As for the battery removal step that Richard mentions, I don’t think you should have to worry about that on your desktop computer. Just try discharging the rest of the juice in your power supply before reconnecting the AC power.

  13. Gordon Richardson

    November 3, 2011 @ 1:09 pm

    13

    I have a vaguely related and sometimes amusing problem, when I shut down my (very old) PC, sometimes the ATX power supply decides to wake up 30 seconds after I have assumed it has all gone quiet for the night!? A bit of a ‘ghost in the machine’ (mine is almost certainly related to power_good and shutdown/restart options).

  14. Dave D

    November 3, 2011 @ 2:08 pm

    14

    I believe the residual energy is stored in capacitors in the power supply, specifically on the +5VSB output. SB stands for stand by. It is a +5V low current output from the supply that is always on, whenever the computer is plugged into the wall. It supplies power to the small circuit that monitors the power button and turns on the main supply when it is pushed. In some computers, +5VSB powers that motherboard LED that is always on. +5VSB stays up for several seconds after power is removed because the current draw of that little button monitor circuit is quite low. One minute probably would have been sufficient.

  15. Mike Warren

    November 3, 2011 @ 2:34 pm

    15

    Your power supply is near failure. There is a current surge at startup. The power supply is less efficient when it is warm. Powering down lets it cool off, then the voltage drop is not as bad.

  16. philg

    November 3, 2011 @ 2:41 pm

    16

    Mike: Near failure? The machine runs for weeks at a time continuously. I use it every few days to edit and compress HD video and I think all six CPU cores run for up to a few minutes at a time.

    Tom: A UPS for the computer? I don’t want to litter the house with crud that is seldom useful. I’ve never in my life had a UPS on a desktop computer and never really had a problem.

  17. Dan Packard

    November 3, 2011 @ 3:29 pm

    17

    It sounds to me like the cmos battery is getting low. That’s the flat pancake shaped battery on the motherboard. How old is the pc?

    Otherwise, the computer may have taken a hit during the power fluctuations you experienced (another good reason to have it plugged into a ups – also serves as a big heavy surge protector). The voltage regulation circuitry may have been adversely affected, hence the flakey behavior. Get a nice solid (and quiet) new power supply, or get a Mac (hee-hee).

    On another note, I just heard about and purchased your “Guide to Web Publishing” book. I love it! Great writing and philosophy. Thanks!

  18. philg

    November 3, 2011 @ 4:10 pm

    18

    Dan: The PC is a year old. It exhibited the same behavior when it was about one week old (we were installing a second hard drive). You bought one of my books? They’re all online for free! I guess the hardcopy is more fun…

  19. Joshua Levinson

    November 3, 2011 @ 4:12 pm

    19

    Phil,

    I wouldn’t dismiss the notion of a failing power supply out of hand. QC for power supplies is notoriously poor, it may have been poorly made, or have had faulty components, from the get go.

    Don’t forget that the power draw at start-up is often very high. Spinning up a hard drive draws a surprisingly high amount of current, for the first couple of seconds. If you have more than one HD, just multiply. That’s why larger servers have the ability to power up the hard drives sequentially, rather than all at once.

    It might be worth the $100 for a new power supply. A failing power supply can easily fry a motherboard and disk controllers.

  20. philg

    November 3, 2011 @ 4:16 pm

    20

    Joshua: But it works perfectly nearly all the time! And I don’t think startup can be very costly for this machine because it need not spin up the second disk. It pulls the OS from the SSD and loads it into RAM. That should consume less power than running all six cores to do video compression, shouldn’t it?

  21. Joshua Levinson

    November 3, 2011 @ 11:19 pm

    21

    Phil,

    The overall power consumption is probably not as high, I think you’re right there. However, the momentary spike of current may be enough to overwhelm the system. First off, the disk is spun up, even though it isn’t needed. When the system is powered on, all peripherals are powered up; it’s the OS, once it’s loaded and running, that tells the second disk to spin down to conserve power. Servers usually have a more intelligent boot-up system that powers on each hard disk individually, to avoid this current spike from many disks spinning up at once, but very few, if any, desktops have such a system.

    That doesn’t detract from the original theory, though. It’s not about the overall power consumption, as much as it is the rapid change in current draw, from spinning up the hard drive, cooling fans, and all the other components in the machine. Sure, the total power consumption ends up being low, but for that fraction of a second, the current draw from the power supply ends up being pretty high. The power supply has a few inductors in it, and the DC motor in the hard drive will probably have a pretty substantial back EMF as it spins up (it’s just a big inductor, after all). There is certainly the possibility of a voltage spike during that current draw, especially if a power capacitor that is supposed to be helping supply that instantaneous current has failed, and the current is instead all being pulled from the power supply transformer directly.

Log in