Slashdot Mirror
PC Not Booting Until a Different Phase is Used?
2by4 asks: "I run at IT Dept for a small firm, our network room houses production &
development servers. Some machines are plugged straight into a strip with
no UPS. Here is the Mystery Problem: When the power glitches, the strip
machines go down, and some of these machine WILL NOT come up again until I switch
them to a new outlet. Once this happens, I can put them back on the original outlet
and they will work. Unplugging & replugging on same outlet is not enough. I have seen this on at least 5 machines so far, with independent confirmation. We can narrow the 'fix' to plugging into an outlet of a different phase (there are 3 separate 120v phases powering the room). The symptoms vary from no powerup, to frozen at the BIOS (depends on motherboard make), etc, but consistently, switching to a new phase fixes them. I tried the 'unplug-wait-&-replug' cycle, to no avail. Using a new outlet w/ a different phase is the only solution. Any theories? I assume the new phase is causing something to 'reset', but what? I can provide more details, but I am wondering if anyone has seen this before? I am completely and absolutely stumped. Our power is healthy, lightly loaded, evenly distributed and the power strips are new. I know I should have at least a simple UPS, but this mystery is causing me to lose sleep."
Obviously, I mean that you should do this with the plug UNPLUGGED.
Try resetting the circuit breaker on your "strips".
--
Twoflower
We had a customers Gateway machine that often did that exact same thing. Machine would refuse to boot or crash at the BIOS with invalid memory errors. Swapping the outlet to a plug across the room would cause the machine to boot just fine and stay running for months on end. Even moving the plug back to the original outlet would be fine for a while. The kicker is, it wasn't just the computer. Plugging in his palm would cause the palm to reset while sync'ing and glitch during regular use.
Our Fluke meter showed nothing special on the line and an APC UPS showed no spikes nor higher than normal voltage levels.
To this day we call it the haunted outlet and tend to just keep things away from it.
Why do overlook and oversee mean opposite things?
"Mystery Problem: When the power glitches"....
then you say:
"Our power is healthy, lightly loaded"...
Not contributing to the solution of your problem, but my office doesn't get "regular" giltches like yours seems to, even though our power is "healthy" too.
Sounds like you need to call your power company.
Modesty is one of life's greatest attributes
Gnomes?
(Flunking, in case you couldn't tell)
Test the ground line on the power strip.
You mention when your power 'glitches'...brownout or blackout or spike?
We are a light industrial building in a heavy industrial park, and I swear the power goes glitchy 2-3 times per year.
We'll get brownout and blackouts, and when the power comes back it SEEMS like it's on, but only 2 of the 3 phases of the A/C actually comes up, meaning (depending on how it's wired at the *circuit box*) some circuits are dead, some are full, and some are semi-brownout (our flourescent ballasts LOVE that half-state.....not).
That third phase sometimes doesn't come back up for hours.
I have no idea if this is of any help, that electrical stuff is arcana to me, I'm just reporting what we've discovered.
-Styopa
I think starting the article with "We have since gone out and bought some fairly inexpensive UPS's to eliminate this problem but nonetheless the phase detection has piqued my curiosity
....... Thus ends my attempt at wit or whatever
Unplug the machine, hold the power button for about 10 seconds, plug it back in. You really don't want your machine bouncing on/off during power problems. Either that or go get a cheap UPS.
Did you modify the phase variance?
We just got a 30 amp circuit installed for a 3kv ups. The UPS once powered up had a "check building wiring" light on the back that came on and stayed on. The cheapy circuit tester indicated that the wiring was fine along with an electrician verifying that all the wiring from the panel to the outlet was correct. One more symptom of this area is that light bulbs blow out much more than normal, although the PCs have not had anything unusual happening.
We had the original electrician who installed the line back out to test. His voltmeter was showing about 20-30V between the neutral and ground. According to code (IANALE), these lines are supposed to be connected at the panel. Apparently without this connection, the two sides coming off the transformer can float in the voltage which may have been responsible for the light bulbs blowing. Once the neutral and ground were connected, the wiring fault light went out on the ups and everthing has since been fine.
SHORT RESPONSE: Have a licensed electrician check out your circuits.
As someone already said, when you have weird electrical problems, suspect the ground. Remember that ALL devices and connections on a computer system need to be connected to the same ground. The exception to this is Ethernet network connections, which are very well isolated.
Printers must be connected to the same ground, for example. Check the integrity of the ground; their should be low resistance, as measured with an ohmmeter after you have turned off the power, of course.
Also suspect that there is some weird voltage riding on the power. Is your power clean? The only way to check this is to look at it with an oscilloscope. Oscilloscopes make an instantaneous on-screen graph of the voltage.
All computers should be connected to battery backup power supplies, too of course.
Boy, talk about the long arm of coincidence. This just happened to me *last night* with my Replay TV. It wasn't showing any life whatsoever, but I checked other devices in the same outlet and they were getting power. So I figured the Replay TV was cooked. I pulled it out, and thinking I might try and fix it (i.e., kick the damn thing a few times) I plugged it into a different outlet and it worked just fine. I took it back to the original outlet, and it works just fine. This hardly answers your question, but another data point never hurts.
"My girlfriend's got sodium laureth sulfate hair."
I've run into this type of issue several times. The issue is'nt so much moving the plug, it's that the MB maintains power for up to a couple of minutes after the power is lost. If the Power glitched, the MB may be in a non-useable state. The way I solve the problem is this:
1) Unplug the Power supply.
2) Hold down the power button (on the front) for 10 seconds.
3) plug in the power supply
4) Turn on the computer.
I solve this issue with most of my systems by connectiong them to a UPS. Some crash on their own so often that they're not worth a UPS.
This sounds like it could be a ground problem. I'd check (if you have the capability) the hot-to-ground and the hot-to-neutral with an oscilliscope on the effected outlet. Barring that you can check it with a multimeter, you may find that the neutral or the ground is inductively coupled to a hot phase.
Some other things to check:
The continuity between the outlet and the electrical box (all three wires).
That your grounding rod is correct for the type of soil in you area.
A different power supply.
Also, like a previous poster noted, try shorting the input to the power supply (when it is unplugged), that may give you a temporary fix.
- you may well have a grounding problem either on the strip or the outlet that the strip is plugged into.
- when a computer is in the "won't work on this strip" state: unplug it and ground each of the 3 terminals momentarily. You can do this by simply touching all three prongs to the metal face plate of a switch or plug. (this is similar to the suggestion that someone gave about shorting out the prongs with a screwdriver but better). Then try the computer. If this works then you may have a floating ground on the strip or somehow voltage is getting induced onto the ground.
- you can buy a simple tester to test the wiring on a duplex plug. Home Depot usually caries them for less then 10 bucks. The plug into a 3 prong outlet and have 3 lights on them (2 orange and a red if I recall correctly). Depending on which lights light up tells you about the wiring and suggests what the problem is. I have even found high resistant neutrals and other such strange wiring problems with it.
Of course once you have tried this (or not) and have a better idea where/what the problem is, it is time to call a qualified electrician.Merlin.
There should be some kind of inline tool that measures the 'quality' of the power coming from the line and flowing into devices. It could have like 5 levels or so and you could check the outlet for problems during real-time use. It could have a bunch of different functions, you know, like checking for electrical problems at the same time.
Unclean power is the problem that causes more crashes than people would like to admit. I've had my parents on the other side of the house start a vaccum cleaner and I've bluescreened at the same time..quite a few times..before. Obviously not a coincidence.
Job? I don't have time to get a job! Who will sit around and bitch about being broke and unemployed then?
Usually 3 phase outlets are completly different looking plugs. If that is the case. It could be that the a phase is on one side, and b phase is on the other side of a panel. I would go along with what other people have said. Power will get you without out you knowing. If you know what you are doing. Make sure the all the phases are good and tight. Especially the neutral. If you are using 3 pahse, you might read up on it a bit. That would make sense. (If you havent already). IF you have an electrician come in, then you can blame them. Besty 100$ OR so you will spend. If you think that you can do this on a couple of machines. Try different ones in different strips. Use a differnt extension cord to the wall. See if there is still your problem. It oculd be a loose grounhd or neutral in the strips themselves. Make sure that you follow the adivce in balancing the load. Not sure your production numbers. If you have 6 pcs, make sure that there is 2 in each phase. You get the idea. The other piece of useful stuff that you might use is a digital ammeter. This will tell you how much power that you are drawing on each leg. You can make sure that you are balanced. I'm sure that there are some that are digital and soime that even have a little pager-like gadget that goes with you. Then its easy. I work in the entertainment industry and I use phase all the time. The big movie lights are upwards or 12kw of light. Its very critical to balance the load. Especially for generators. The easiest way is to get some big worklights that you can plug in and leave on to balance it out. Im sure that there are others.
I have mod points and I am not afraid to use them.
1) As other have pointed out, bad grounding on one or more phases.
2) Bad neutral on one or more phases.
3) Voltage drop on one or more phases.
For the first three get an electrician out, don't mess around with three-phase power.
4) Electrical noise on one or more phases, this may not be caused within your computer room.
5) Equipment connected on different phases affecting each other, for example a printer on phase A and a server on phase B.
6) Borderline power supply in a machine affected by one or more of the above conditions.
7) Faulty power strip(s).
8) Incorrectly wired plug(s) or socket(s).
Ed Almos
The more corrupt the state, the more numerous the laws. - Tacitus, 56-120 A.D.
Your parents must have very very unusual power. Three phase power is hardly ever found in houses. In the US, we don't even have two phase power - we use a single split phase.
Fascism trolls keeping me up every night. When I starts a preachin', he HITS ME WITH HIS REICH!
(Or some other kind of scarlet swimmy thing...)
Turning off and on at the same mains outlet will generally be "off, count 10 elephants/mississippis, on". Moving outlets usually involves physically moving the damn thing around, or unplugging a wire from one socket, taking the wire across the room and plugging it in somewhere else. Consider the time it takes to do this - could the power-off time be the significant factor, and the phase thing is a coincidence?
Or on a similar theme, how about disconnecting the mains cable (and waiting some time) so that the mobos are fully powered down? That happens naturally when you disconnect the cable and plug it in somewhere else. Maybe try repeating the same action, but on the original phase.
Not to doubt the fault-finding you've already done, but just adding a bit of devil's advocacy to suggest possible alternative situations with the same symptom.
Grab.
If you have 220 in the house (typically for an electric stove or a clothes dryer), you have two phase coming in from the street.
Not really. House 240V power is created by connecting one phase of a three phase feed to a transformer with a center tapped output. The center tap is grounded and voila, you get split phase 240V power. It consists of a ground, plus one leg at +120V and the other at -120V, referenced to ground of course. Take the potential across the two of them and you get 240V. While these legs are separated in phase by Pi radians, this is not what is referred to as two phase power.
"Nature doesn't care how smart you are. You can still be wrong." - Richard Feynman
Have an electrician come in and do some wiring checks and fixes. You have a combination of two problems causing you to see this.
1: You have poor to non-existent grounds on one or more of the phases. This can be tested for by the electrician.
2: The HOT and Neutrals are swapped around and generally this isn't an issue except that *modern* power supplies are getting cheaper and cheaper and this usually means cutting out *some* parts like full-wave bridges for half-wave diode sets, and similar tricks. This then makes the circuit more effected by incoming AC and it's phase against ground or what little there is of it. (hence the reason for #1) There more than likely is a potential difference between the neutral and ground that is excessive and this is causing a cap-start circuit to ignite the switching and hence, a dead PS.
Cheers;
All content in this message is copyright (c) 2008. All rights reserved. RIAA is prohibited here.
I've experienced similar problems in the past, due to current leakage. Especially with different phases, but also commonly with poorly grounded devices or devices which had peripherals or cables running close to sources of high levels of common mode noise.
Most commonly, when plugged into a peripheral or another device which is separately powered or grounded.
What was happening was small amounts of current leaking in through peripheral components was affecting the power supply... Usually to stop them starting up. ie, hit the power button, no start, no lights. The leakage was in the micro-amps region, but was enough to leak back through the motherboards into the power supply, and cause a false fault condition reading on start up, and the PSU would shut down before it got started.
The solution? In one instance, I decoupled the power rails. In another, better grounding. Another? Changed phase. The best solution was usually to find a better power supply that wasn't affected, but was not always possible.
Often in these circumstances you can feel the current leakage, as it's often at high voltage but very low amps. Sometimes it feels like a slight tickle when you touch rivets on the case.
Additionally, I've also encountered similar problems due to engineering faults, where a high impedance section of the circuit was acting like a radio antenna and was getting enough "reception" of a local signal (any strong electromagnetic radiation source) and causing a fault condition on power up that was not present during normal operations (when the applied signal was significantly stronger than the picked up signal).
Solutions there include EMF shielding and redesigning the circuits.
Problems like this are difficult to diagnose, as they are not always obvious, and there is very little you can do to test or troubleshoot directly. Often it involves experience and a little lucky guesswork.
GrpA
Enjoy science fiction? "Turing Evolved" - AI, Mecha, Androids and rail-gun battles. What more could you want?
If you opt to test the outlet with an O-Scope, first find out if the ground pin on the cord is tied to the ground on the o-scope inputs and chassis! Most analog, corded models are and you run the risk of putting 120-volts on the chassis or shorting the outlet through the o-scope. More than a few amateur electronics techs have missed this feature and blew up an o-scope or shocked the hell out of themselves. Electronic techs often use isolation transformers to protect the equipment and themselves for this reason.
There are two ways this can happen:
Note that this can be induced in otherwise properly bonded circuits by the use of daisy-chained power strips/bars. It is the act of plugging in a (typically high-power) item on the end of the line (so the draw through the whole line is high) and connecting it via a signal cable (Ethernet, serial, USB, etc.) to something plugged closer to the junction box (electrically) that then ends up routing some of the ground current through the low-voltage signal line.
In general, the cure is to return to the bonded-circuit of yesterday designated by the orange plugs where they've been installed. These consist of a single plug per circuit (never more than one!) where ground is bonded to neutral at the connection box so there is no possibility that there may be a voltage drop due to current between individual recepticals.
We seem to have gotten away from the specification and use of these (more expensive to install) power recepticals. I for one continue to specify them for most commercial installations and have yet to see any of the kinds of things mentioned here when they have been properly installed and used.
The worst case I saw of this was in an office that was long and skinny in a brand new building (they were the top floor) with retail below. The office went in first, and the retail later proved to include a dry-cleaner that used quite a bit of power off two phases of the 3.
A Unix computer in the center of the long office fed dumb terminals the length of both directions. The reception area was the farthest out - about 100' by wire - RS232 shield.
The terminal at reception kept doing wierd things: hanging, mystery characters, and in fact died - 3 times! Lights on but nobody home!
It turned out the serial interface was dieing - due to about 5 volts between ground and neutral which was pulling current through the cable's cladding and buring out the chip. The electrician poh-poh'd it saying "5 volts on a 120volt line was nothing to worry about" but in fact it was 5 volts on a 12 volt circuit and carried current in what should have been a voltage (high impedance) system. No wonder the interfaces were flaky and buring out.
Replacing the power plugs with "home-run" single bonded ones fixed the problem.
Been there, done that, paid for the T-shirt
and didn't get it