bad power very VERY bad power! dont want server damage

[cheapie]

Re: bad power very VERY bad power! dont want server damage

The voltage here is *checks voltage display on fancy new UPS* 122V right now. It seems to very from 120V to 125V, with the frequency exactly 59.9 Hz.

This reminds me... one time, I built a double-conversion UPS using a battery charger, lawnmower battery, and inverter (the kind that plugs into the power outlet in your car).

[senz_90]

Re: bad power very VERY bad power! dont want server damage

I was logging through my UPS that reports the voltage through the serial port. Which reminds me I need to go get new batteries for the UPS, they're shot...

I dont know why UPS have failed in many cases because just a leakage batteries.. i have two UPS lay down fine, just has a leakage batteries. it was given by friend that thought those UPS circuit board failed already. maybe the manufactures have to built in a good quality batteries.

[eccerr0r]

Re: bad power very VERY bad power! dont want server damage

Yeah during the summer the voltage variation through the day as people turn on their air conditioners make the voltage go down significantly during the peak hours of the day and highest in the wee hours of night/morning. During the winter it's much more stable with most people using natural gas heat.

I was logging through my UPS that reports the voltage through the serial port. Which reminds me I need to go get new batteries for the UPS, they're shot...

[theOracle]

Re: bad power very VERY bad power! dont want server damage

The voltage at my house, well, at my workbench, is pretty much spot-on at 120V :P

It doesn't fluctuate either, we had the transformer powering our house get replaced after the old one arced over internally.

yes it does, put an accurate true RMS A/C multimeter on it, and set it to register min/max and you will see a 24 hour fluctuation of say, something on the order of 118-122

[ben7]

Re: bad power very VERY bad power! dont want server damage

The voltage at my house, well, at my workbench, is pretty much spot-on at 120V :P

It doesn't fluctuate either, we had the transformer powering our house get replaced after the old one arced over internally.

[eccerr0r]

Re: bad power very VERY bad power! dont want server damage

I've got buried power lines going to my house and for a while I was logging voltages. It was also going from 111 to 118V after they fixed the transformer when it blew out one year (it went down to 105 before that). I've never had any power supplies failing because of power though, most failures were due to poor design/bad capacitors/bad fans...

If anything, frequent blackouts are a bigger problem if case your data didn't quite make it to disk before the lights go out.

Pretty much an on-line UPS with voltage regulation is the only real solution for the paranoid.

For the past almost century, power lines have been regulating frequency for synchronous motors - specifically for clocks - and likely because of this, the frequency will not be changing much. I don't know if there are plans to drop this as many clocks no longer depend on AC frequency for timekeeping. Only potential issue is noise but that input cap in SMPS's will tend to low pass filter the noise away.

[senz_90]

Re: bad power very VERY bad power! dont want server damage

On my town, they are usually using an UPS and voltage regulator like this.

i have use the voltage regulator on my CPU, voltage regulator rate is 500VA because im just using one CPU. My mains is 220V but when i measure, it just about 205-215, at night the mains makes fluctuation on my analog VOM needle. I don't have any problem until now using these voltage regulator.
Make sure to using the voltage regulator that has a toroid transformer controlled by a board that move the crankshaft automatically. Some manufacture makes a cheapo suck regulator just from a simple circuit design without transformer those are very lightly compared with the heavy one has a transformer, my father have ever use those when i was a child and my regulator with TV both has burned out after the lighting strike my antenna. the one those has a transformer can protect our devices, on many cases when something bad happened, just the regulator or UPS burnt. not our devices.
Some cases, it doesn't guarantee our devices safe if home mains get hit straight by lighting (God will ), but it is very rarely happened as my experienced.

[kingsapper]

Re: bad power very VERY bad power! dont want server damage

All the more reason to run a real file system like ZFS

Enterprise power supplies in real servers are built pretty hard-core. And most modern servers have dual redundant power supplies that are hot swappable by staff in the dork fridge (datacenter, oops.. )

If you're not running "real" servers but desktop machines configured as servers it's not a bad idea to consider setting up battery busses of 12 and 5v, and then park the chargers/high current power supply on the batteries. If you're worried about charger/battery supply noise all you have to do is ensure that the impedance of the batteries is lower impedance than the chargers.. A low pass filter on the positive and negative leads configured for common and differential mode interference will do this nicely. Cost you about 20$USD in parts (not counting etching a PWB and a plastic box to put it in.. )

That way if power goes out, computers are running off of batteries anyway. Sure you'll have to make a new wiring loom or two, but big deal right? Find some old dead PSU's and tear the bundles out. That's how the network storage computer in my lab is set up, as there are occasional blackouts during the hottest parts of summer here and it's a real pain in the ass.

Figuring out how much battery you need is a simple calculation or two if you have a high-range ammeter, and remembering that conventional lead-acid batteries don't like being discharged 1.75v per cell..

Past all of that, UPS is going to be the way. I prefer offline systems personally.

Cheers,
KS

[shovenose]

Re: bad power very VERY bad power! dont want server damage

True server/enterprise grade power supplies are extremely well built and you should have no issues dealing with unclean power, at least the worst that could happen is a shutdown.
That in and of itself might be an issue for a server due to filesystem corruption... and RAID won't protect you from that, even though if it's hardware RAID with a BBU chances are it won't be quite as bad.

[joshnz]

Re: bad power very VERY bad power! dont want server damage

In 06 I lost a power supply to bad power caused by bad power mainly flickering.
I wish they had a seperate 33kv bus for the town and another for the region

[budm]

Re: bad power very VERY bad power! dont want server damage

Most good UPS have AC AVR (Automatic Voltage Regulator) by using tap switching to maintain the output voltage within +/-5VAC if the UPS is Line interactive type. If you get true on line type (more expensive, and will be sine wave output instead of step sine wave), it will maintain 120VAC output and you will not be hearing relay switching tap (in the Line interactive type UPS with tap switching) if the input voltage varies a lot.

[cheapie]

Re: bad power very VERY bad power! dont want server damage

Believe it or not, but fire is conductive!

It can also be used as a diode. That's how flame sensors in many powered gas appliances (furnace, clothes dryer, power vent water heater, etc.) work.

voltage fluctuating from 111v to 118v

SMPSes (especially with APFC) don't care (unless it gets too low, then they derate).

and i'm guessing horrid frequencies but i don't have the tools to find out.

SMPSes don't care.

[kc8adu]

Re: bad power very VERY bad power! dont want server damage

get a good ups system yesterday!
you should have heard mine earlier this morning with the derecho that went through here.
about 100 blinks then out about an hour.nasty storm!

[ben7]

Re: bad power very VERY bad power! dont want server damage

really? fwew! the only reason this is happening is because the electric infrastructure of colorado is VERY fragile because of these 4 wildfires and the pulsing is an attempt to prevent rolling blackouts by xcel.
edit: colorado is under a redflag warning to..

No, the pulsing is very likely due to short circuits/arcing from the powerlines in the wildfires. Believe it or not, but fire is conductive!

[goontron]

Re: bad power very VERY bad power! dont want server damage

An isolation transformer would do absolutely nothing in this situation. Interestingly, I have played around with several PSUs on a variac and never managed to kill any of them by pulsing and dropping the AC.

really? fwew! the only reason this is happening is because the electric infrastructure of colorado is VERY fragile because of these 4 wildfires and the pulsing is an attempt to prevent rolling blackouts by xcel.
edit: colorado is under a redflag warning to..

[tom66]

Re: bad power very VERY bad power! dont want server damage

An isolation transformer would do absolutely nothing in this situation. Interestingly, I have played around with several PSUs on a variac and never managed to kill any of them by pulsing and dropping the AC.

[theOracle]

Re: bad power very VERY bad power! dont want server damage

Isolation transformer

[mariushm]

Re: bad power very VERY bad power! dont want server damage

Buy an UPS or get an inverter and 12-24v batteries.

[goontron]

rite now it colorado we have BAD power continuous pulsing of power that can be seen from CFL's and HID lamps, voltage fluctuating from 111v to 118v, and i'm guessing horrid frequencies but i don't have the tools to find out. the point is i dont want my servers down or damaged! [/rant]
so, no UPS, how can i prevent damage to my servers and towers?