[mythtv-users] Power outages and UPSs

[f-myth-users at media.mit.edu]

    > Date: Sun, 29 Jul 2007 13:53:31 -0400
    > From: "Michael T. Dean" <mtdean at thirdcontact.com>

    > The one that described an approach in an article on the 'net that Lionel
    > said could be done better with a DC-DC PSU.  The one you obviously
    > haven't read because you're trying to convince /me/ that "my ideas"
    > (which are not my ideas, but are straight out of the article)

Don't get your knickers in a twist; I really don't care whose ideas
they are, nor have I been bothering to track their provenance.
Whether an idea is good or bad has nothing to do with whose it is.
Let's please skip the personal attacks and attack the technology instead.

				> 				  are
    > wrong.  Perhaps you should write to Dan and set him straight (or at
    > least have him remove his blasphemy from the 'net).

Not my job.  (AKA "I don't care enough.")  But see below.

    > http://www.gossamer-threads.com/lists/mythtv/users/281185#281185

Hmm.  Apparently your original message never went through my mailbox
(interesting; I'll have to see if it arrives vastly out of order, or
not at all, but certainly it it hasn't arrived yet and your Dan's URL
is nowhere in anything I've ever received), but I've read it from
gossamer and looked at Dan's article.  Comments on his strategy are
below.

    > And I'm saying this probably isn't a problem because my power is almost
    > always available.

Okay, so no bulk-charge phase.  Very different from what many people
get UPSes for.  Certainly wouldn't have worked in my area, where we
(used to) get 4h failures once/month as 4kV overhead lines fell apart,
and still occasionally get 1h failures once every six months or so,
and occasionally a pair of 30-60m failures separated by a few hours
(no doubt as they replace something).  Obviously riding out a 4h
failure is unreasonable w/o a genny, but pairs of 30m failures
separated by a few hours are exactly what a bulk-charging UPS will
save you from and a trickle-charging one will not.  Yes, for a while
I used to complain that I was living in a third-world country.
Fortunately things seem to have improved.

[This also means that you're wasting your money on that spiffy AGM
battery; if you only need to ride out a tiny outage and are willing to
have things die in a few minutes, then almost any cheap half-dead
battery will do.  In fact, several cheapies in parallel may work
better if you can keep them balanced, because that way one dead one
with high resistance won't hose you during a 10-second outage.  Come
to think of it, maybe you can use the one from your car the next time
you have to replace it?  Just 'cause it can't crank your engine
doesn't mean it doesn't have enough amperage to run your machine, and
if it dies in a few years, well, how long before the new car battery
needs replacement? :)  Though I guess in FL car batteries may last
longer than in cold climates, if nothing else 'cause they can be
"deader" and still be good enough because you're never using them at
0F and your engine oil is never sludge.]

    > >     > * The UPS doesn't restore power after an outage, so my system is down
    > >     > the 4 days while I'm traveling rather than the 10 minutes the power was out.
    > >     > * The UPS decides that the battery isn't performing as expected, so it
    > >     > cuts power--while there's still power to the mains--causing an outage
    > >     > that wouldn't have existed without.
    > >     > * The circuit breaker in the UPS triggers and the /only/ way to reset
    > >     > the UPS is manually by holding the power button for 4 seconds or something.
    > >
    > > WTF?  All of these sound like broken or misconfigured UPSes.  I've
    > > used a couple dozen midrange consumer APC units (Back-UPS Pro &
    > > ordinary Back-UPS; SmartUPS, and also much older ones like the 1200VX
    > > and others) for 15+ years and I've never seen any of these misbehaviors.
    > > I assume other vendors' UPSes are similar or we'd hear lots of screaming.
    > > (a) & (b) are just dead wrong;

    > Both of my APC Back-Ups did that.  Perhaps they were broken, but after
    > just over a year, I felt they shouldn't be (and I refuse to keep buying
    > new  garbage to replace the old).

I agree that they shouldn't have done that.  I've also never seen that
behavior, and all of my APC UPSes have been consumer-grade with a
couple of exceptions.

Is your power exceptionally spiky?  APC UPSes have surge clamps in
them, but simple MOV-based clamps can only absorb so many joules over
their lifetimes before becoming useless (they can get "used up"); I
don't actually know how the APC clamps are implemented in any given
UPS.  [I ask because we had an issue in a lab once where a 480V feed
started arcing and generally playing hell w/the 120V feeds; that
caused permanent damage to some UPSes and soft upsets to others.]

You say your router is unlikely to overload the UPS, and I agree, but
you've had 7 UPSes; were the only UPS failures on very low loads, or
is it possible that you were near overload on one or more of them?
I'm wondering if the UPS had issues trying to come off the inverter
at full load and wound up oscillating or worse---or simply went into
overload the instant it went -on- inverter and turned itself off then,
rather than failing to switch back to mains once power was restored.
Those tripped circuit breakers are a real gun on the mantlepiece---
I've -never- seen one trip in an APC except for the one that got
completely blown up by the laser printer.

In my experience, midrange APC's (dunno about others) also do a
battery test every 14 days, and make obnoxious beeping noises and
light their battery-fault light if it fails.  Even when this happens,
it won't drop the load.  However, by the time the battery is this
trashed, it also instantly drops the load if power -does- fail if
you've got more than 10-20% load on it, and I don't recall if it will
bring the load back online when power is restored, e.g., it might turn
itself completely off.  I assume you weren't ignoring bad-battery
alarms.

[Recent experience w/a midrange APC that had a battery that was
-almost- bad enough to fail selftest, but not quite, was that a
1-second failure caused it to instantly dump its (hefty) load but
it didn't turn itself off, so the load came back the instant the
power did, and putting a tiny load on it let it ride out similar
brief (less than a minute) outages.]

				>        I don't know if the problem was
    > buying consumer-grade UPS's, but I'm convinced that consumer-grade UPS's
    > aren't what I want.  I may spend several hundred dollars on some
    > SmartUps UPS's, but the smarter my previous UPS's have been the more
    > they've failed (i.e. they think they're smarter than they are).

Do you recall what models?  Were they perhaps particularly small?

			 >     If you're willing to spend a couple hundred bucks
    > > for that AGM battery you were talking about, you could get a few thousand
    > > F of supercap and that might be good enough.  Send me the VA of what you'd
    > > like to power and I'll spend a couple minutes computing whether it's even
    > > -possible- to survive more than 5s of flicker without going broke...:)

    > I'll keep that in mind.  I'm still in the research phase (as I have been
    > for about 2 years--the second of which I've been operating without any
    > UPS's).

I'll make a quick guess of 100W, which is probably way too low but
gives a multiplier.  Say you want to ride out 10s.  That's 1kJ, so at
2.5V, you'd need 320F assuming perfect efficiency.  You can't get that
(in part because even with a buck/boost converter, not all of the
discharge curve is going to be usable), so let's assume 500F.  You can
get 50F at 2.3V for about $7 qty 10 (totally random choice of DigiKey
part # P11066-ND; no doubt there are many better and/or cheaper
choices [e.g., different component -and- different vendor] for what
we're doing but searching elsewhere will take more than the 5m I'm
going to spend), so you can get 500F (which is 1322J) for $70.  If you
needed 500W for 10s, then you'll spend $350 instead, which is pricier
than your battery but will last forever; note that the qty 100 for
this part is only $4.40, which would give you 13kJ for $440, meaning
you run 100W for maybe 100s or 1kW for maybe 10s (if your inverter can
-handle- 1kW :).  [You may well be able to get very lucky on eBay if
you're patient, and Digikey is not the cheapest distributor around.]

Let's hope I didn't drop an order of magnitude in there, 'cause I was
doing this quickly; I'm sure if I did someone will let me know.

[You can easily get caps in the 5000F range and up and they may well
be cheaper than a zillion little caps; there's a very large search
space out there.  OTOH, lots of little caps can be put in series
configurations so you can use the string at (say) 25V or 100V or so,
which means you don't have to use busbars to connect things (lower
current & lower voltage drop) as long as you balance them (resistors
will work fine here).]

Now, for -this- solution, your bench supply is perfect---it's current-
limited, so it won't blow up trying to charge these up (they'll look
like a dead short for a -long- time on the way up), and as long as
it's nonbackdriveable (or you toss a blocking diode in there that's
rated for the peak forward current of your supply, and the peak
inverse voltage of your cap string) the caps won't discharge back
through the supply.  But, really, you could use any total-POS source
of voltage whatever as long as you're sure it won't go overvoltage and
won't explode against a short (and has the amperage to drive your
normal load forever :), so a lab bench supply is overkill.  (Like, a
transformer and a full-wave bridge is probably plenty---the supercaps
are your filtering stage... :)

Fusing is absolutely critical.  As is a closed nonconductive box
around them.  And the fuse goes -inside- the box.

[Man oh man, I'm not kidding here.  Supercaps will usually not be
damaged themselves by a short (check the specs), but -you- certainly
will be, and unlike batteries, they're -more- than happy to give up
all that energy in under a second if your interconnects are up to it.
Even if you don't electrocute yourself directly, dropping something
conductive onto the array can vaporize it, and being in the path of
molten metal being blown around at transsonic speeds is not my idea
of a good time.]

    > Also, I'm anxiously awaiting your, "Better than Dan's DIY UPS" article.  :)

Being snide to those trying to help you is a bad idea.  So is trying
to prove you're never wrong.  You do a lot of that.  I could have just
said, "Hey, he wants to mischarge his batteries, what's that my concern?"

Since I've designed high-cap Pb chargers commercially, my "better than
Dan's" would be "if you're an EE, here are some -real- circuits that
have actual charge regulation in them so you don't have to babysit
them" and "if you're not an EE, just go buy commercially, and you'll
pretty much get what you pay for."  But, more to the point, it'd
be "do this if your time is free---otherwise, you cannot beat what
a professional [and perhaps cheap Chinese labor] can possibly do
without spending 10-100x the cost of your own time doing it."
[Where -I- got involved were places where there wasn't already some
commercial solution because the applications were bizarre.]

So let me just give a brief sketch of issues I see w/his strategy:

(a) I worry about safety.  He doesn't have a fuse anywhere in his
design that I can see.  At the very least, it would be a good idea to
put an automotive fuse of almost any ampacity (how about 50A?) as
close to those battery terminals as possible, and then to insulate
everything exposed before the fuse.  Otherwise, one dropped wrench can
really ruin your day.  [I recall a DEC service technician who reached
inside a VAX and shorted the multi-hundred-amp +5V supply across his
metal wristwatch (remember those?); the burns were not very pretty and
neither was the downtime on the machine.  And waiting for a new guy
with a dead machine while the first guy got taken to the ER.]

(b) He assumes that people are never fallible.  The obvious failure
modes here are forgetting the topping charge (easy or hard, depending
on how organized you are about keeping a calendar) and forgetting to
turn OFF the topping charge after a while and instead cooking your
batteries (remarkably easy to forget to UNtweak a knob unless you have
an excellent memory or never misplace a reminder---or are willing to
set as many kitchen timers as it takes to get you to go deal with the
charger).  A better design would allow an input to the supply so you
can set a timer on that topping charge.

(c) Yes, as whoever-it-was complained, online conversion means you're
going to dissipate heat in the inverter constantly.  (Actually, his
complaint was the CPU PSU after the inverter, I think, but whatever.)

(d) I haven't translated the AU$ to US$ etc, but certainly if you
didn't already have a lab supply and an inverter around, this ain't
necessarily a cheap solution.  Even if you did, while AGM batteries
are truly excellent, it's not 100% clear to me they're worth the
money in -this- application---I don't know if the lifetime benefit
or the capacity benefit is worth the price.  But it's great if you
happen to have 2 or 3 of the 3 major components (supply, inverter,
battery) hanging around anyway, of course.  (Where AGM's really shine
is in their charge-acceptance, e.g., your ability to just JAM as much
charge into them as you have available from your power supply.  If
you're just trickling the thing all the time, well, eh...)

Okay, I'm way out of time on this thread.  Hope this has been helpful.
If you find that supercaps are the way you want to go and need more
advice on using them, ping me, but I'm going to tune out of the rest
of the battery/DC-DC/etc discussion.