[geeks] Memo to self: try not to destroy UPSs

[James Fogg]

On Tue, 19 Jun 2001, THOU SPAKE:
> > > That's one thing I hadn't thought of lately: MOVs. Need to get some of
> > > those into the mix. 
> > 
> > They are the problem, not the solution.
> 
> Would you mind clarifying a bit?
> 
Warning, long reply ahead:

MOV's clamp at a specific voltage (usually 145Vp). This clamping is actually a
short across line (hot to neutral and hot to ground) in most circuits. MOV's
are rated at x voltage at x amps over x time. TIME is the key here (measured in
nanoseconds). If you get the spec sheet on an MOV you will usually see another
time value describing how many times it can clamp within parameters per second
or minute and not become a torch.

Utilities generate power mechanically. Voltage rises and falls over time based
on the position of the stator in relation to the field windings in the
alternator. There is nothing at the power station that the utility can do to
produce a spike repeatedly. The sudden rise/fall of a spark during switching
will be amplified inductively in transformers during normal utility load
distribution. This (and lightening) is the usual cause of voltage spikes.
Severe spikes are caught by discharge horns (seen on pole-tops) and
gas-discharge arrestors (a vacuum tube or noble-gas filled tube with an anode
and a plate at ground). An example of a simple gas-discharge arrestor is a neon
lamp (not its purpose, but same capability), sometimes used for the purpose. A
discharge horn looks like this  ___>  <___   (air-gap between > and <).

A UPS produces its output with A/B switching transistors (square wave). A
modified square wave UPS produces its output in steps (step-wave). A sine wave
UPS does the same with an added LC circuit (inductance/capacitance) to modify
the output. Only very expensive UPS's have true sine out (through the use of LC
networks and a  ferroresonant transformer). Under heavy load, or when the batts
are dying, the output of expensive UPS's can become non-sine.

We are accustomed to speaking of AC current as Vrms (voltage - root mean
square). The typical U.S. outlet provides 110Vrms. RMS refers (roughly) to the
average voltage across one cycle (remember, in one cycle the voltage
is zero for a while and begins to climb). The actual Vp (voltage - peak) is
closer to 130 Vpp (peak - to - peak).

When a UPS switches its output transistors there is a momentary transient where
the total output of the source voltage is present on the output. This is
usually on the order of 150Vp. In poorly designed UPS's this is inductivly
amplified further. This spike is extremely narrow (as seen on a oscilloscope)
with dramatic rise time. In fact, it takes an excellent 'scope to even see it
(200mhz or better). Step waves suffer this problem to, as do modified
(smoothed) step wave UPS's (sometimes called sine wave). This spike occurs 60
or 120 time a second, depending on design.

When an MOV clamps, it shunts current to ground. It then tries to recover its
thermal balance and release heat. An occasional spike is no big deal. If it has
to do this repeatedly, it becomes an fire hazard (I am surprised that UL hasn't
required warnings on home UPS's). Most UPS's have MOV's on their input, so
daisy-chaining UPS's is a bad idea. Removing the MOV's from the downstream
UPS's won't help since most UPS's use a switchmode power supply and the
rectifier might not like the abuse either. This is also the reason that you
don't want a "surge arrestor" located after your UPS.

There are also some conditions of inductance that can cause an MOV to fry
equipment without the above scenario. An out of spec MOV in a circuit with
inductance can cause the MOV to repeatedly clamp (briefly) and the resulting
magnetic field collapse downstream can wreak havoc. I had a "surge arrestor" do
this once and it ate 3 server motherboards before I suspected the MOV problem
(I'm talking schmoked/roasted/flamed).

-- 
=======================================================
	 James D. Fogg, Network Engineer
	Vicinity Corporation - Lebanon, NH

     DESK (603) 442-1751 - CELL (603) 252-1864
     PAGER (802) 742-0280 - HOME (603) 526-7729
            EMAIL jfogg at vicinity.com

If you can read this e-mail, Thank a Network Engineer!
=======================================================