[Nut-upsuser] Testing UPS themselves: Best Practices

[Jim Klimov]

Adding to that reply, for collection of UPS data from NUT you may want to
use the `upslog` client which takes care of running in a loop at a
frequency you would want, and reporting the data points you may be
interested in. Then you can use that wall of text as CSV for graphing, etc.

Jim


On Tue, Sep 9, 2025 at 7:37 PM Greg Troxel via Nut-upsuser <
nut-upsuser at alioth-lists.debian.net> wrote:

> Bruce Pleat via Nut-upsuser <nut-upsuser at alioth-lists.debian.net>
> writes:
>
> > Over time and houses, I have accumulated several UPS of different brands
> > and capabilities.
> > I've been trying to figure out a way to test them in a somewhat
> > standardized way. (Test ~ see how long they last and how quickly the
> > decline on a given load so that I can determine how the batteries will
> > perform in "real life", at least relative to each other.)
> > I was wondering if I missed any obvious existing references on the topic,
> > and/or if anybody has any thoughts on the matter?
>
> Your logic seems sound.
>
> The real question is what load do you care about.
>
>
> When doing this, use an outlet strip as the input and turn it off, don't
> unplug it, so the UPS still has ground.  Use a good one with UL etc
> ratings, rated for 1500 W.  It's it's a 1500VA UPS, think harder and ask
> your electrician how to do this safely.
>
> While the discharge test is running, stay in the room.  Better yet, do
> it outside, and also the recharge.  But not if raining :-)  Seriously,
> think about the safety issues.
>
> > To date, my thought was, wait for temps to stay below 30/86 (hahaha), run
> > an Edison bulb off one outlet (bought a 300W for this purpose), and
> monitor
> > via NUT.
>
> Easy! It was 8 outside when I woke up.  But indeed, depends...
>
> > For each UPS:
> > Charge fully
> > Take readings (via NUT of course)
> > Plug in bulb, monitor readings
>
> "charge fully" is tricky; I'd say plug it in and leave it powered on for
> a week.
>
> > Run to 0%
>
> so you mean
>
>   don't configure nut to do FSD/power-off when battery becomes critical.
>   monitor only
>
>   check configuration of the UPS to be sure it's in "run as long as you
>   can" mode
>
>   just let it run until the UPS itself powers off
>
> > Go to the next UPS
> > Repeat the set
> > (If >5% difference between runs on any, run a 3rd set; if 2nd and 3rd
> > align, throw away data from 1st)
>
> I suppose you can repeat but you really need to let it sit plugged in a
> very long time to get recharged.  I really mean a week.    I'd say after
> 2 tests on each UPS, step back and if they aren't close, post here and
> let's talk about it.   This is bad for the batteries; some amount of
> validation is necessary.
>
> > Then repeat down to just 50% instead of 0%
>
> Why?  What you really care about is runtime, and '50% is just some
> measure of what the UPS thinks is 50%, not truly meaning same time
> before/after.
>
> Publish graphs of %/time-remaining/voltage during each discharge.  Go
> nuts and make a blog post of it.
>
> > Then test to see if the lightbulb is still consuming energy at the same
> > rate.
>
> Sure, but probably that's ok.   Keep logs of output voltage too.
>
> > Please chime in on a better (and less energy-wasting) process if you
> have?
>
> No, you really have to extract energy to see if you can extract energy.
> It's not really that much.
>
> > My original planning notes:
> > - UPS include 550VA to 1500VA/900W, some others; APC, CyberPower, Amazon
> > basics, other brands. Record info on all...
> > - Check all for any acid/leak due to age/storage. Do outside of house,
> just
> > in case...
>
> good!
>
> > - test which allow muting of beeping, schedule accordingly (family,
> calls,
> > sanity). Quick test...
> > - Normalize thermals for at least 48 hours before testing
> > - Test that all can be monitored in NUT; preferably all USB, all
> recognized
> > as such in NUT UPS by USB driver, to avoid any variability in drivers or
> > other. Configure/Test...
> > - Use a fixed-rate load to avoid any variability; tech equipment itself
> > might not be best, maybe a curling iron or iron or light bulb? Be aware
> of
> > heat. Brightest bulb I can find is 300W, will that drain all my UPS
> > sufficiently for this test? Is this a valid substitute for tech gear or
> is
> > the consumption different enough? Research...
>
> Get a power mere and measure your actual loads.  Look at the VA reported
> by NUT for your actual loads.  For me, it's 61 VA and 134 VA on two
> units, each rated for 660 VA.   I use a 60W bulb for load testing.
>
>
> If your 1500 VA unit feeds 300W, and your 550 VA unit feeds 60W,
> consider having a "big UPS" and a "small UPS" test.
>
> > - Run high-power extension cord outside, bulb should run outside. Make
> sure
> > cord and fixture support lengthy durations at this power. Does outside
> > temperature matter significantly? Research...
>
> Bulb and outside probably not.
>
> > - Linux or Windows NUT? Linux: N150 or Raspberry or my heavyweight
> Laptop?
> > Local, VM, LXC, Container? Likely not relevant. Ask NUT User group...
>
> > - Focus on 10%+ variances, not the tiny variances.
> > - not scientific, just ballparking, but don't want to rely on bad
> batteries.
>
> Sure, just don't get upset at 75min runtime and then 60min as a bad
> test.  Could be batteries harmed by test.
>
> As for scientific, there's nothing like figuring out what the UPS will
> see when it really gets used and doing that.  Then you have tested your
> equipment, as opposed to varying lots of things and figuring out
> patterns.
>
>
> Another thing you can do is test batteries separately, but that's not
> the same thing.
>
> See the end of
>
>   https://s1.lexort.com/topics/ham-battery-howto/
>
> for tests of new Raion 12V 9Ah batteries.
>
> _______________________________________________
> Nut-upsuser mailing list
> Nut-upsuser at alioth-lists.debian.net
> https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/nut-upsuser
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://alioth-lists.debian.net/pipermail/nut-upsuser/attachments/20250910/daca2468/attachment-0001.htm>