Help!! Power Quality Issues Multiple Office UPS's Going Off Randomly!

Status
Not open for further replies.
141114-1725 EST

5 S electric:

I have never worked with your Amprobe instrument, but looking at the manual you need to have it setup for Voltage Anomalies. I don't believe you have done that. I would try 120 V nominal and 5%. If that produces too many small variations, then change 5% to 10%.

Anomalies has a time resolution of 10 mS, and records the time and date of an event that goes outside the % limit. At a setting of 10% and 120 V a voltage dip below 108 V or rise above 132 V would produce a recording and indicate the associated phase.

The data you have shown us is from the UPS and not from the Amprobe. What does the Amprobe data look like?

.
 
Last edited:
141115-1045 EST

5 S electric:

You have spent a lot of money and time trying to solve your problem. Had you used the correct instrumentation and correctly I believe the problem area could have been found at much less cost.

Your approach is sort of like in the old days of trying to troubleshoot a 5 tube AC/DC radio with no voltmeter by using your eyes and deduction to look for failed components and trial and error substitutions. However, I have heard a story that a final inspector at Zenith Radio would do much of his troubleshooting with a screwdriver. It served as a rough voltmeter, and as a signal injector.

We don't have any specific information on your UPSs, and what are their characteristics. For example, what are the upper and lower trigger points to go into backup, and what is the response time that triggers a change of state?

According to the Fluke 87 manual, https://physics.ucsd.edu/neurophysics/Manuals/Fluke/87______umeng0800 ,
min-max has a normal response time of 100 milliseconds (0.1 second), and in the Peak min-max a response time of 1 millisecond. In peak mode it is not clear from the manual exactly what happens on an AC waveform relative to the minimum measurement, or even what is measured.

Your Amprobe instrument has a 10 millisecond RMS measurement period. This was obviously designed around 1/2 cycle of 50 Hz. This is the shortest integrating time. Disturbances shorter than 10 milliseconds will still show up in the measurement, but will not produce as large a change as would a change over the full 10 milliseconds.

I don't know the characteristics of your UPSs, but I doubt they would switch with just the full loss of voltage for 1/2 cycle (8.3 milliseconds). This kind of variation would be very obvious with your Amprobe instrument. There is mention of 2 to 3 millisrconds response for some UPSs in http://www.tri-sysdesigns.com/Articles/UPSprimer.html . What does this statement actually mean? With a simple averaging circuit on a full wave rectified sine wave a sub-half cycle response is unlikely. Your UPSs are apparently 3 phase so shorter response times might be possible. Do you need response times of shorter than 100 milliseconds? This depends on how much energy storage is in the computer.

The correct instrumentation used correctly should greatly expedite your finding the cause.

.
 
Last edited:
141117-1415 EST

Eh40:

The TED 1000 had a 1 second integration time, the TEDs since then are 2 seconds or worse even though TED states 1 second.

From the manual the Amprobe that 5 S electric has will resolve 10 milliseconds if properly set up.

.
 
gar said:
141114-1725 EST

5 S electric:

I have never worked with your Amprobe instrument, but looking at the manual you need to have it setup for Voltage Anomalies. I don't believe you have done that. I would try 120 V nominal and 5%. If that produces too many small variations, then change 5% to 10%.

Anomalies has a time resolution of 10 mS, and records the time and date of an event that goes outside the % limit. At a setting of 10% and 120 V a voltage dip below 108 V or rise above 132 V would produce a recording and indicate the associated phase.

The data you have shown us is from the UPS and not from the Amprobe. What does the Amprobe data look like?

.
Click to expand...

I will attach some Data that was recorded with the Amprobe. You will have to go to Amprobe and download there software though.
http://www.amprobe.com/Amprobe/usen/Service/Software/default.htm Any Download Suit 3.0.0.05 will work.
It won't let me attach it to this forum reply. Send me an email and i will forward the info that i collected to you. sbunning@admin.nmt.edu
 
141125-1125 EST

5 S electric:

Use info@beta-a2.com .

I mocked up a circuit to produce a momentary loss of AC input. This consists of an OAC5 SSR and a 555 monstable pulse generator. At present I have a 100 k adjustable resistor in series with 1.91 k fixed, with a 1 ufd timing capacitor. Time constant range is 0.002 to 0.102 seconds. Approximate pulse duration is about 1.1 times the time constant.

A pulse as short as 2 milliseconds to the SSR input should produce an output change from the SSR of about 1/2 cycle of a 60 Hz signal. I haven't looked at waveforms of the switched AC power yet. Nor have I checked the actual duration of the pulse to the SSR.

The quick test results.

On a CyberPro CPS425SL UPS the shortest interruption was sensed. Probably 1/2 to 1 cycle. To be determined if the actual input drop to the SSR is this short.

On a mid to late 1990s IBM desktop the 0.1 second, about 6 cycles, drop in voltage to its input did not cause a reset of the computer. Note, the UPS is not part of this test.

.
 
141126-0652 EST

A post I made was lost.

Checked actual pulse duration range. Minimum was 2.04 milleseconds, and maximum was 102.6 milliseconds.

That I only see full half cycle, 8.3 milliseconds, drop outs is confusing. If the pulse drive was directly to the Triac gate in the SSR, then drop out near 2 mS should appear. This means the OAC5 gate drive circuit makes a big change in the effective gate drive signal.

I don't plan to do more tests at this time. There is at least a 100/8.3 = 12 to 1 ratio between the response time of the UPS and drop out of the computer. The UPS could be less sensitive to drop out time and not cause the computer to reset.

This brings up the question of whether the three phase UPSs have an adjustable sensitivity control, and then if all UPSs on the different floors have the same response time. This could relate to whether the AC line noise is associated with just the one floor or not.

.
 
gar, many UPS are unnecessarily sensitive, especially APC branded ones. It's not just the voltage, but the micro-controller is designed to transfer on excess rate of change even if the depth of sag is not that deep. The software parameter for steady state transfer points and sag/swell are different settings.

5 S electric said:
1- 1 big computer lab and some offices
2- Manley offices
3- 2 Med sized computer labs and offices

sorry i would take a while to count all of them.
Click to expand...

"UPS goes off"
"setting to low quality setting"

I smell 120v standby plug-in UPS that you put near the computers rather than a hard-wired units.

Computers are not what they were in 1970s and 1980s. They don't like power surges but very flexible on voltage as long as long as it doesn't brown out long enough to reboot. Have someone put a desk lamp on their desk and hook it up to the same power source that is causing the UPS to trip. Have them report the flicker.

The out of range transfer should be set for 90-95v on low and 135v on high. Power supply is meant for continuous use from 100v -10% to 120v + 10%.

I bet its the fuser in printers and photocopiers. In response to energy conservation standards as well as user expectations for warm-up time, they use a very powerful heater and control it using thyristors, so they cause distinctive dips that trip out the UPS. It's not the depth of the dip. It's the speed.

Put the UPS back to original settings and use some 10 or 12 gauge extension cords to move the photocopiers to outlets fed from circuits that do not share power with computers like maintenance outlets. If this resolves the issue, printers/photocopiers have to find a new dedicated circuit.

gar said:
141126-0652 EST
This brings up the question of whether the three phase UPSs have an adjustable sensitivity control, and then if all UPSs on the different floors have the same response time. This could relate to whether the AC line noise is associated with just the one floor or not.
Click to expand...

Larger UPSs are usually double conversion or delta conversion online type. Not standby offline type. APC has a lot of white papers about the difference if you search the web. They don't "transfer". The input and output are linked by PLL so things like line sync clocks don't drift.
 
Last edited:
In some cases of generator power supply, the sensitivity window of a UPS need to be widened because of poor quality of output voltage from the generator. But a utility transformer usually does not have such an issue. So the harmonic loads in the facility, most probably, are the issue. So widening the sensitivity window of UPS might solve the problem.
 
Sahib said:
In some cases of generator power supply, the sensitivity window of a UPS need to be widened because of poor quality of output voltage from the generator. But a utility transformer usually does not have such an issue. So the harmonic loads in the facility, most probably, are the issue. So widening the sensitivity window of UPS might solve the problem.
Click to expand...

Read, then post.
 
Status
Not open for further replies.
Top