Checked a job that the voltage on one circuit will run about 117v, then will drop down to 112-113v for a couple seconds, the current jumps up from about 12a to 17-18amps at the same time not long enough to trip the (battery back-up, (kicking on I believe), 119-121v on other circuits at panel. They have a Server with a battery back-up and it causes the the bat back-up to kick on.
I'm thinking of just adding a 20a separate circuit. But it still seems weird that the voltage would drop that much. I checked connections everywhere for possible bad connection, causing resistence, not that.
I can see overloading a circuit and the amps jump up, but the voltage dropping that much, (unless there was a bad conn or something), plus the breaker isn't tripping. possible bad breaker too?
Anyway, the separate circuit should solve the problem. Any thoughts?

Do I understand this correctly? Measured at the panel, a circuit with a 12 A load, measures 117 V, which then drops down to 113V when the load jumps up to 18 A. Other circuits on the SAME PHASE read about 120 V.

If all of the above is true, I would suspect a loose connection between the breaker and the buss bar, a defective breaker, or loose connection at the breaker.

Other possabilities include measuring Voltage or Current with a non true RMS meter. I included this because you stated that the load is a UPS feeding a server.

Larry C

What I sounds like based on what you saying is possibly some sort of motor load possibly a fan or AC that is on the same circuit as the UPS. When it starts, the voltage drops long enough that the low voltage drop out is hit so the UPS kicks in. Is the UPS big enough? is there other loads besides the server on the circuit? When was the UPS last serviced?

It sounds like you are measuring the voltage right at the equipment supplied and not at the panel correct?
I am going to take a stab at it and say there is another load on the circuit, but that the load(s) are located some distance from the panel, and the wire isn't sized to take into account the distance..
Just my thoughts..

A.D

Depending on the model of UPS, you may be able to glean a lot of information from it. For starters, higher end UPS will augment line current with the battery- if it's getting 12A of 112V and it's set for 115V, it will use the battery to bring it up to 115V. Some cheaper UPS (the one on my home entertainment center, grrr...) will also sense an undervoltage, but switch entirely to battery. Considering this is a server drawing 12A, I'd imagine they have a high quality UPS. You're probably not lucky enough to have one of the REALLY good ones with interactive LCD panels with menus and data logging, though, huh?

Most common commercial-quality UPS have monitoring systems including an integrated voltmeter and clamp ammeter- APC Smart UPS will have a green bar graph right on the front showing system load, and red lights showing the status (undervoltage, overvoltage, overload, etc). Thus, you can pretty easily tell if the server is spiking to 18A (they can and do draw more current under load) or if it's coming from elsewhere. If you talk to the ITs, they might even have it hooked up to log the data and can give you detailed readouts of everything that UPS has seen.

Is this an online UPS? True sine wave output on the UPS?

Most on line UPSs take line power, convert it to a DC buss, and then into an AC component. This may be a true sine wave or a nasty stepped square wave...yuk! Look at it with a scope. I'd guess you are trying to measure the AC component of a stepped square wave.

For what it's worth. I use off the shelf office type UPS in an industrial settings. These UPSs are fed from control transformers. I've noticed they have problems returning back to line once on battery. I'm thinking the spike to the autotransformer causes a buck/boost event which the UPS sees as a sag and then returns it to battery; alternating back and fourth. There's a sensitivity switch on most mid range APC models that allow them not be so fussy.

Can you bypass the UPS just for testing purposes?


RSLater,
RSmike