Server Fault is a question and answer site for system and network administrators. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

We placed a EATON Ellipse Max 1500 (900W) as the UPS for our HP Proliant ML350 G6. Upon first power failure (actually we only moved the UPS' input plug to a different socket), server immediatelly turned off, and the Health LED turned red and started blinking. UPS was in operation for about a week before that, with battery fully charged to 100%. Since our server's hot-plug supply is 460W, we are pretty sure we haven't overloaded it, the server was completely idle at that time (no web or win apps running except Windows Server core services).

Then we tried to do the same with a different, no-name older PC (Core 2 Duo, 2Gb RAM) with a generic power supply (not sure what the power is) and it continued working when we pulled the plug out. UPS load was less than 15% (measured in the provided Eaton utility).

We measured the UPS' output voltage using a smart oscilloscope and the THD of the UPS output waveform turned out to be 40%.

Did you have similar experiences? Could this be a faulty UPS? Or a faulty power supply? Or some HP sensors configured to trigger too strictly?

I wouldn't like replacing this UPS with the same brand, to get same results.


I also tried to do this while the server is turned off. While the UPS is working on battery, server will not start - as soon as I press the power button, Health LED starts blinking red.

share|improve this question
up vote 6 down vote accepted

I've seen this before and it's come down to the newer power supplies do not like the less expensive UPS systems.(it's not just HP) These new power supplies can detect fluctuation of 3ms and I've seen the UPS's response times to be from 3-5ms (on the low end). If you want to stay with Eaton try the PowerWare 9130

share|improve this answer
I am up voting this question and answer, I want to see if this is the case. I have never used an Eaton UPS and will probably stay away from them going forward. – Not Kyle stop stalking me Nov 9 '10 at 14:16
I've seen this as well with cheaper UPS's and HP servers. I read somewhere it has something to do with the output and sinewaves, I'm sure someone with an electrical engineering background could say more. – CarloBaldini Nov 9 '10 at 14:41
It is probably related to the shape of the output waveform, rather than response times (if what you you consider by response time is time needed to switch over to batteries), because server doesn't even want to turn on while the UPS is on battery supply, with Health LED blinking red. As soon as I plug the UPS input cable back in, the server can be started and Health LED goes green. Since RMS of the wave is ok, I can only conclude that the problem is with the waveform (which has a THD of 40%). – Groo Nov 9 '10 at 20:39

There could be multiple problems here. Assuming your server was not plugged into any "surge protection" (non-UPS) plugs on the UPS.

1. Your power supply could have a defect and poor hold up time. If this is the case, it cannot survive the 3-5 mS that it takes a typical UPS to switch over to batteries. Hold Up time should be around 15mS or more.

2. The power supply's Power Factor Correction (PFC) might not be compatible with the "Modified Sine Wave" output of the UPS. I'll explain how this happens.

A typical UPS output looks like a square wave, with positive and negative components, and a short zero component in between. Exactly how the UPS matches the equivalent power of a true sign wave is up to the implementation. For example, it could hold a lower voltage of +/- 90V for a longer step time, or a higher voltage of 100V for a shorter step time. The average delivers the wattage advertised, assuming all is working to spec.

Modern PFC power supplies draw power at the top(bottom) of each sine wave, so as to draw less current than old-style transformer power supplies that draw large current amounts during the low voltage portion of the sine wave (leading and trailing edges). If the voltage of UPS "modified sine wave" is not high enough, some modern power supplies just don't see the "top" of the "sine wave" and shut off. So, the UPS is delivering the power, but the Power Supply is just not seeing it.

If you don't have the equipment to discern the problem, call HP and Eaton each and ask them if the other equipment is qualified to work together.

share|improve this answer
We can exclude the first problem, because the server didn't even want to start after the UPS had already switched to battery. Regarding the second issue, the signal indeed looks exactly like that: a square wave with voltage zero for a while (between positive/negative transitions), while its true RMS corresponds to that of a utility socket. But the statement about PS drawing power at the top is not exactly true: PFC supplies merely take care to charge the DC capacitors in sync with the input voltage, in order to keep the power factor low. They shouldn't care about the peak voltage value IMO. – Groo Nov 9 '10 at 16:56
They SHOULDN'T care. BTW THD on a square wave is pretty meaningless if the meter is expecting a sine wave, it could be a very clean square wave but the meter is trying to compare it to the sine and marking all deviance as distortion. If you truly have Harmonics, interharmonics, notching, RF noise, frequency variations, etc. in the 40% range, that would be an instant RMA on the UPS, it is defective. – kmarsh Nov 10 '10 at 20:41

We´ve got the same Problem with a APC Smart UPS SC1000. It´s a Problem with the HP Powersupply it is too sensitive and the server turns off when the power changeover happens. Now we changed the UPS to a Eaton 9130 and now it is properly working.

share|improve this answer
Thanks for the info. Datasheets for all the "enterprise-level" UPS devices we've considered state they produce "pure sinewave output", while with low-range devices this is not the case. – Groo Jan 11 '11 at 9:25

I assume your power supply is an HP 460W CS HE Gold Power Supply. Either way all HP hotplug supply's are redundant, that means two actual units inside the supply so 460+460=920, and it is just barely overloading your UPS. If you tested the UPS on other PC's this definitely what the problem is. Plus your description sounds like every overloaded UPS I have ever seen. Try a 1000W UPS I'm 99% sure that will work for you.

share|improve this answer
No, we only have a single 460W power supply, no redundancy. Second socket is empty. Nevertheless, rated power is the maximum power that a supply unit can draw during a longer period of time. It does not correspond to the power actually being drawn from the network when the server is idle (Eaton utility shows around 10% load when the server is online). – Groo Nov 9 '10 at 14:25
I stand corrected, according to the HP site they do make a 460W single. I wasn't able to find it before, sorry. – Not Kyle stop stalking me Nov 9 '10 at 14:54

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.