[Nut-upsuser] UPS Battery internal resistance matters.

Julian H. Stacey jhs at berklix.org
Tue Dec 11 02:29:25 UTC 2007

Hi NUT list
For any who don't know:
   Voltage is an insufficient indicator of { health, time to support
   load, of if batteries will even support the load at all } . Internal
   resistance is important to calculate, but imperils load while
   testing to calculate (***).

I mention it 'cos ~2 years ago when I was using NUT, I didn't notice
much if anything about internal resistance, but its's important.
& people I suspect often think about just battery voltage/ charge
as affecting performance / load time.

I have an APC, & another brand, & another same APC as Charlie, &
have an electrical engineering background & replaced batteries in
various UPS, & took measurements. Below is based on APC 2200 &
2000, but similar principles apply generaly.

Maybe some UPS _might_ measure both open circuit voltages & voltages
under load at test times, & current, to calculate state of batteries,
I dont know, but I suspect UPmany/most UPS may go for a simpler
voltage measurement, perhaps mapped through a hysteresis curve,
then fed to NUT as some percentage figure.

   ( Lead Acid batteries charge & discharge on a hysteresis curve,
   flattish in the middle, but curving more both ends. I looked
   at some manufacturers data sheets for 17AH 12V dry batts as in
   this model, & they do the same I recall. )

Problem is you can't test internal resistance except under some load,
& if the batteries have aged & can't take the load for the duration
one is testing, the computers will drop out.

On my APC (a Smart 2000) only the voltage of all 4 batteries in
series is available to the UPS, so APC can only calculate serial
resistance of all 4 batteries, (if they might calculate it ?)  ...
Whereas only some batteries may be going half bad, but all 4 may
still add to about the right voltage.

   ( On the APC Smart 2200 there's 2 connections between battery
     level & UPS level, so maybe APC can measure 2 x 24 V (I'll
     unscrew case again later to see wiring ) or maybe it too is
     just 1 serial measurement of 48V. )

Taking just 2 values from Charlie:
	battery.voltage: 52.95
Seems to be alright.
As Arjen wrote, can't calculate time on no load :-)
But even if the APC 2200 was loaded, although one could guess
rough nominal times from capacity, as UPS doubtless float charges
we get no figures to calculate internal resistance until we push
the test button, (or better: pull the power cord for long enough
to get a good sample). ... & maybe then we dont just test the
internal resistance, but find it's too high, & the load is dropped.

Ideally UPS owners need to test periodically, & calculate internal
resistance too, & warn before it gets too bad.

Open Question: Whether various UPSs could accept commands from NUT
(if NUT gave them) to go offline periodically, long enough to sample
the voltages, And current, to calculate the resistance (maybe needing
to take into account power factor if the UPS only reports KW out ? -

There is a danger in automatic periodic sampling:
If done regularly, & If the human monitors the results, OK,
but if internal resistance is getting worse, Suggestion ...

(***)  It would be sensible for NUT to automatically reconfig the
UPS to no longer auto sample.  (My guess is NUT perhaps doesnt do
that yet, but I don't know, I'm just returning to NUT after maybe
2 years).

Extracts of notes to owner of `my' APC Smart 2200 follow as background,
for those who might want more on internal resistance:
           On front:       Smart UPS 2200 http://www.apcc.com      (not apc.com)
           On Back
                   Model SU2200/Net
   errata.pdf@ -> ASTE-6Z8LAX_R0_EN.pdf
   main.pdf@ -> ASTE-6YWRZE_R0_EN.pdf
   operations_manual.pdf@ -> ASTE-6Z8LC7_R0_EN.pdf
   quick_ref_guide.pdf@ -> ASTE-6Z8L9H_R0_EN.pdf
   safety.pdf@ -> ASTE-6Z8LFX_R0_EN.pdf
   serial_extra_firmware.pdf@ -> ASTE-6Z8LC3_R0_EN.pdf
   xl_addendum.pdf@ -> ASTE-6Z8LB2_R0_EN.pdf
   Simple voltages I measured a few days ago on no load looked OK, but
   Now having found time to measure internal resistances, they don't
   look good.      When I had this same problem with my APC smart UPS
   I replaced some batteries & UPS been OK since.
   Battery Number Zero below for reference is mine (a Conrad ),
   bought a year ago, used once to power 12V flourescent lights (&
   known by me never to have been abused by heavy load).
   Table below is:
   Col 1:  Battery Number
   Col 2:  Voltage open circuit.
   Col 3:  Current in Amperes, on load (x 5 )
           X 5 as I used a shunt, as my meter only goes to 3A.
           Not particulalrly accurate measurements as it doesnt matter.
           Load = 4 x 12V car headlamp bulbs in parallel.
   Col 4:  Voltage (DVM accuracy) while on load measured right by battery,
           (not measured on long resistive lead to lamps).
   Col 5:  Voltage open circuit after, after perhaps 15 seconds recovery.
           I did not ensure identical recovery times while each rising.
   Col 6:  Internal resistance calculated from V/I = R (V = Col 2 - Col 4)
   0       12.98   1.5x5   11.9    12.8    0.14
   1       12.65   1.38x5  9.79    12.5    0.41
   2       12.75   1.3x5   9.0     12.5    0.58
   3       12.56   1.45x5  10.6    12.3    0.27
   4       12.64   1.45x5  10.43   12.45   0.3
   When I had poor batteries like this, I bought new ones, & had no
   more problems.  Expensive, though Conrad stock them at 52 Euros per bat.
   (12V 17Ah) BUT Conrad own batteries don't have physical regular studs,
   alignememnt of some may be problematic for the bridge fuses.  More
   expensive batteries dont suffer this random alignement.  
     > Do you have any idea, why the UPS apparently gave some power
     > outs or shocks on the computers?  all computers fed by the UPS
     > rebooted about one or two times per month.
   APC Quote:
      Automatic: The UPS performs a self-test automatically when
      tuned on, and every two weeks thereafter (by default). During
      the self-test, the UPS briefly operates the connected equipment
      on battery.

      > and our batteries were to weak even to last that "briefly"
      > interuption  ... maybe that switched off the computers shortly
      > every 2 weeks.
      > Do you think , that's a reasonable explanation?
   Yes. I do.
   I remember turning power off on wall fuse last year: demo for a
   friend, & all PCs stopped, nasty suprise for me.  I was more surprised
   later as open circuit voltage seemed OK, yet internal resistance
   was too high.  But can only measure internal resistance on load.
   I must try NUT, it might be it can measure V & A to calculate R,
   in which case we do periodic test, sample with NUT, & calculate R
   to get prior warning for the future.

Julian Stacey. Munich Computer Consultant, BSD Unix C Linux. http://berklix.com
	Ihr Rauch = mein allergischer Kopfschmerz. Dump cigs 4 snuff.

More information about the Nut-upsuser mailing list