Need someone who knows more about UPS systems than me.

[mhenley]

I'm aware of the three primary types of UPS systems, and simply put, I cannot afford an Online UPS, but I will spend the extra for a Line Interactive system. My most recent purchase, which was ~5 years ago, was a Cyberpower 1000AVRLCD model, and it's been working like a champ until recently when I realized something odd.

House meter was upgraded to this new digital smart meter blah blah doesn't matter. Point is, it was a scheduled blackout, and my computer reset. I decided to test the unit by starting the computer, and then unplugging my UPS from the wall, and it did it again. Ordered a new battery, installed it, let it charge overnight, tested it the next day, same thing.

FYI, watt meter says I'm pulling ~322w while conducting these tests. Nowhere near overload. Also, if I unplug from the wall, system boots and runs fine on the battery until the battery runs out of power.

My question. Do line interactive system's 'interter/converter' unit slow down over time? Are there some brands that are known to have faster switches than others?

 

[ghost6303]

My question. Do line interactive system's 'interter/converter' unit slow down over time?

they shouldnt, and the amount of time it takes to switch from line to battery power should be small enough that the DC-side of the power supply never sees a thing. the first large set of capacitors in your power supply should be able to absorb the few miliseconds without incoming power.

if you didnt have a ociliscope to accurately measure the delay, i would try putting an inductive load, like a heater or something, on the UPS and see if you can see a momentary drop in power when you pull the plug on the UPS.

the two possible problems i would think of would be; either the UPS isnt switching to battery fast enough upon power failure, or your power supply is very weak and cant even handle the smallest drop in voltage without causing problems in the output.

edit: according to cyberpower's website, the switching time on that model *should* be 4ms. that *should* not be long enough to cause any problems, and 4ms should not be noticable to the human eye.