[Nut-upsdev] Any UPS hardware hackers ?

[Russ Romano]

> And now for something completely different... :-) And yes, 
> it's off-topic to this particular forum as it's not about NUT 
> development. <ducking>

Let me preface my statements with the fact that I have no familiarity
with the inner workings of APC UPSes.  However, I am a circuit designer
who does chip-level analog design for a living.

> Is there anyone around here who knows his way inside the 
> hardware of an APC Smart-UPS? The APC support division will 
> not reveal _any_ details or service information, and the only 
> descriptive documentation I've found on the Web is in Russian.
> 
> The problem: I have an APC Smart-UPS 500 which has slowly 
> increased its battery float charge voltage over the years 
> until now the 24V battery is being float charged with 28.5V. 
> This destroys the battery in short order. 
> 
> Presumably some component has drifted off its nominal value 
> over the years. Finding out which one is difficult, though. 
> The inverter's power MOSFET stage doubles as a charging 
> voltage regulator. The MOSFETs are pulse width modulated by a 
> driver signal coming from the CPU and at least two chips with 
> custom type numbers. Debugging the processor system and PWM 
> hardware is difficult because of the custom chips on which I 
> have no info at all.

I know that APC designs custom ASIC (application specific integrated
circuits) for their UPSes.  This means that you won't find ANY public
documentation on the workings of those devices at all, and considering
what they pay to design and manufacture them, you certainly won't get
them to share any details with you.

You have already determined that the UPS uses PWM on a power mosfet to
charge the batteries.  There are only a couple of variables here.  

1) Power supply voltage feeding the power mosfet in charging mode.  This
will obviously be higher than any voltage you want to place across the
battery, but if it is too high, it could be a source of trouble.
(Perhaps measure the DC voltage on that side of the mosfet and compare
it with a measurement of another APC user?).

2) You can also measure the digital pulses on the mosfet to be sure you
see a reasonable pulse stream (do you have a scope?).  

3) There has to be some circuit that feeds back the battery voltage into
the ASICs/CPU.  I'd imagine there is some kind of filtering circuit, and
perhaps a voltage divider (to get the 24V battery signals back in the
realm of what we like to see on-chip).

Of these... I think a dead capacitor or resistor in the filter/divider
feedback circuitry would be the most likely culprit.  Effectively
causing the system to "sense" a lower battery voltage and continue
charging.  Unfortunately, this circuitry could be integrated on-chip...
In which case there is nothing you can do to fix it.  Chips have a
limited service life, and stressed components in them do vary as they
age.

[to help bring this back on-topic]
Does the UPS report accurate battery voltages when monitored in
software?  I can't imagine they would use 2 different circuits to
monitor battery voltage, so if it reports accurately in software, #3 may
be a stretch.

> Any response would be greatly appreciated!

In the end... UPS hardware has gotten pretty inexpensive and probably
isn't worth the effort to debug and repair.  But it's fun to talk about
anyway...

-Russ