I have to buy surge suppressors for a lab.  What are the criteria I
should look for in a surge protector?  In particular, if I buy 1
750-Joule Belkin  suppressor per computer for 15 computers, is that as
good as a single 10,000 Joule supporessor for the whole lab?

---

Request for Question Clarification by krobert-ga on 05 Nov 2002 17:07 PST

dispensa-ga,

I have read through the comments and have some questions to ask.  Have
you considered the ramifications of having one point of failure
compared to having fifteen?

Assume you have one surge unit installed in the lab: if it happens to
fail, you now have fifteen unusable computers (unless a student
decides to bypass the system, not that that ever happens!). With
fifteen units, you now have multiple points of failure and more uptime
for your network.

Also, you need to consider what is cheaper to install. Does the 10,000
Joule unit install in the electrical system (in the wall for example)
or does it install at an outlet? Does that installation method require
some heavy duty cabling to join all the computers to that protected
circuit?

krobert-ga

---

Clarification of Question by dispensa-ga on 08 Nov 2002 08:20 PST

Regarding the single point of failure,  yes, and I'd rather have
distributed points of failure.  Also, I'd rather have something that
is cheaper to install.  As you point out, it would be much more
expensive to install a big suppressor.

I'm really leaning toward individual suppressors.  Now the question is
this:  for a relatively low-budget operation, what does a 750-joule
surge suppressor buy me?  What would a 1500-joule unit buy me that a
750-joule unit doesn't?

dispensa-ga,

Well... after reading your clarification and all the comments below,
I'm going to bite down and try to provide you with a satisfactory
answer on this one.

The simple answer to your last question is this: Imagine the
electrical system in your lab was to receive a line spike due to
nearby lightining striking a power transformer that feeds your lab.
The power of this spike is 1000 Joules at the outlets feeding your
computers. 10 of your computers are on 750-Joule suppressors and 10
are on 1500-Joule suppressors.  Each computer with the 750-Joule
suppressor will fail permanently... they weren't able to absorb the
spike.  The computers with the 1500-Joule rated suppressor will keep
on functioning normally.  If they have fuses that get blown (e.g., the
suppressors have fuses that protect the varistor), they may shut off
your computers temporarily, but they didn't fail... they WORKED! You
don't have to replace those. ALL of the suppressors did what you
needed them to do... they protected the computer equipment from
getting zapped.  Bye bye to the word document that you were working
on, but your computer can still boot up!

I would suggest this... If you don't know it personally, talk to
someone who has used or supervised the lab over a period of time.  If
they can tell you that they have been replacing a lot of suppressors
with 750-joule ratings... you need a better rated suppressor. Simple
as that.

Think about what you are buying... suppressors are -risk management-. 
If you are unsure about things, get good quality suppressors for the
expensive computers in the lab.  If things are stored locally on any
particular computer, you probably want to protect them better than the
i486-computer in the back of the room that acts as a dummy terminal to
a major server.  Get it?

If you need a solution that allows the computers to keep running while
absorbing a spike, your going to need to look at Uninterruptible Power
Supplies (UPS).

I hope this provides you with some guidance on this topic.

Be sure to request a clarification if you need it.

krobert-ga

---

Request for Answer Clarification by dispensa-ga on 08 Nov 2002 12:46 PST

Thanks for the detailed answer.  One clarification would be helpful,
though:  I realize that a 1500-joule protector is twice as effective
at absorbing energy as a 750-joule protector; my question is more
practical than that - what is the practical difference?
 - All strikes/surges/etc are >750 joules, and most are >1500, so
individual units don't offer any practical level of protection.  Buy a
10KJ unit for the whole lab.
 - All surges are >750J but <1500J, so a 1500J unit is (for practical
purposes) infinitely better than a 750J unit.
 - Most surges are <750J, so they offer equivalent protection

This requires knowledge of typical surge levels.  If you're saying
that this sort of information is impossible to find other than by
trial-and-error, that's fine, too.  Then the question becomes:  what
would you specify as an ideal surge protector?  100,000J might be
great, but it's probably (i hope!) totally overkill.  What is the
lowest joule rating that you would be comfortable would filter >95% of
surges?

Thanks for the information.

---

Clarification of Answer by krobert-ga on 08 Nov 2002 13:33 PST

dispensa-ga,

In all my searching, I don't think that there is such a thing as an
"average" surge. We are talking everything from mother nature
(lightning) to human error (oops... that shouldn't spark that
brightly!). As far as "What is the
lowest joule rating that you would be comfortable would filter >95% of
surges?" goes... that requires the "average" information that we don't
have.  I think that averages in this case depend too much on your
particular situation to have an overall average.

To expand the metaphor of risk management, maybe think in terms or
earthquakes... there are lots of little ones all the time, ones that
instruments can detect but humans can't. Occasionally there's a big
one. That big one is planned for, but depending on how much money
you've spent on protection, your house won't necessarily survive it.

Simple solution -> big bucks is better.

Belkin and Fellowes and APC must have this kind of informaion in their
own databases.  A simple glance through these products can give you an
idea of how much protection you are getting.  From what I've seen, a
surge suppressor with a rating of about 700 Joules generally has about
$5000 of insurance(?) protection on it from the manufacturer. Say this
puts you in the 95th percentile of protection.  Maybe one in 10,000
gets a claim paid out to them.  Up from there are the suppressors
rated at 1250 to 1500 Joules. These have claims of protecting you up
to about $50,000 worth of damages.  Use similar logic. You added a
zero to the end of that insurance figure, so... maybe your chances of
a loss have gone from one in 10,000 to one in 100,000.  So, double the
protection MAY give you a tenfold decrease in chance of loss.

Here's another twist.  What is protecting this lab right now?  If
nothing, then I would say that either the 750J or 1500J suppressor
would be better!

Best Regards,

krobert-ga

You probably want to go beyond a surge suppressor for
a lab environment, and maybe look at UPSs with power
conditioners, etc.

As is often with such things, the amount that you have
to spend will greatly (and unfortunately) determine the
type and amount of protection you get.

No, putting a 750 Joule suppressor on each computer is
not the same as a central 10,000 Joule suppressor; from:

http://www.suttondesigns.com/doctext/surgco.htm

Definition: 1 joule = 1 watt-second = 1 (volt*amp)-second.
This is a measurement of the energy absorbing capacity of
a given unit.


The 10,000 Joule suppressor will afford greater overall
protection that the individual 750 Joule suppressor.

More info at:

http://frentzandsons.com/content/howto/wiring_part4.htm

But there are things to be concerned with other than
lightning strikes, such as low voltage, high voltage, power
"flickers", etc.  There are things to take into consideration
other than Joule ratings, such as the design of the suppressor
and its features such as alarms, etc.

Depending on your budget, you can get into line conditioners
and UPS with AVR (Automatic Voltage Regulation) such as ones
that Belkin makes, etc.  These things can be applied to an
entire lab at one central point (regulating/conditioning, etc.
the power before it gets to the lab) or on a "per station"
basis, again, depending on budget.

Actually surge suppressors are meant for protecting electrical equipment
from transient surges(It cannot protect against continued high voltages or
against brownouts-continued low voltages).  It is more suitable for glitches
on the power line caused by transmission equipment malfunctioning or due
to lightning strike(not a direct one).  It is connected in parallel to the
appliance and acts as a short circuit when excess voltage appears across the
power terminals and dissipates the excess energy as heat(which is being
referd by the joule rating of the device).  It does not help in connecting
muliple devices in parallel(since the strike voltage is not very precise
and will endup overloading a single device instead) to increase the joule
rating.  It would be just sufficient to add one device(with very high joule
rating) at the power panel where the power enters the premise.  This will
not offcouse protect the equipments due to high voltages produced by one
of the equipment malfunctioning(which is rare, unless the connected equipment
is highly inductive such as a motor to produce high back emfs).  A combination
of the spike suppressor(one in the panel) and individual circuit breakers
and a fast acting fuses at the equipment would be a better choice for protecting
a large number of computer equipment.  It might also help to add a single
line filter before the spike suppressor to slow down the response of large
spikes and to reduce rf/emi noise from entering supply line and result in
malfuctioning of the computer devices attached.

To clarify, my hypothesis is this:  Assume a computer lab with 15
computers.  There are two different surges that can occur if I
understand the situation:  a current (i.e. amp) surge and a voltage
surge.  The two are related, of course, so maybe that previous
sentence doesn't make much sense.  Regardless, both current and
voltage divide proportionally over the number of paths.  If I had only
one computer plugged into the power feed, it would absorb the entire
current/voltage spike.  If, on the other hand, I had 15 computers
plugged in (in parallel) to the same power feed circuit, I'd have 1/15
of the voltage/current distributed to each workstation - electricity
101, right? (or wrong?)  Anyway, by that math, to get surge
suppression for the entire lab equivalent to 750 Joules at each
station, I'd need 15 * 750 = 11,250 Joules of protection.

All of the suppressors I'm looking at are <330V leak-through voltage
models, so that basically factors out.

As usual, I'm on a tight budget, and it really will play a big role in
my decision.  All things being equal, the 15 750-Joule Belkin
suppressors are cheaper than a 10KJ panel-mount unit.

One last thing - What order of magnitude of
watts/joules/volts/amps/whatever is a lightning strike?  What numbers
do people have in mind when they say "direct lightning strike" or "not
a direct strike", etc... ?

This is a more interesting question than I thought. :-)

Actually you are right on the mathematics AND wrong on the assumed parallel
distribution of the surge.  The spike suppressor is equivalent to a zener
diode(if you know what I mean), You cannot parallel two spike suppressors
to provide a combined joule dissipation.  Only the weakest(the one having
the lowest leak through-say 321V) of all is going to take the entire load.
Also surge happens in a very short time(that being the reason for a the 
nano second switching time specified for suppressors). Also surge does not
get distributed as a normal current would(do you know why the ground conductor
for a lightning arrestor need to be as straight as possible?).  Having parallel
protection would help provided the surge is well within the joule capacity
of a single surge protection unit and is not recurring. Having many surge
protection units just gives a false sense of security(ie. if one surge protector
malfunctions, atleast the others will work).  If you are really inclined to
reduce the cost, you could build one using parallel surge protectors separated
by very low inductance coils, the surge suppressor elements itself being 
shunted by low value high voltage bipolar caps.  This arrangement will provide
better protection both for single event/recurring surges and also will help
reduce the cost of replacing burnt out suppressor elements(few at a time).
Eventhough a current surge is possible, it does not occur without a voltage
surge(current is equal to the amount of water flowing in a tube, where the
pressure on the line is the voltage.  Without sufficient pressure on the
line, you cannot increase the water flow on the line without a change in
the pressure).
Now coming to the lightning strike, by direct hit I mean the lightning directly
passes through the connected equipment to the ground.  In general it will
be several thousands(if not millions) of kilo joules and you can do nothing
about it. A indirect strike is the sudden increase in the supply line due to
the ligtning completing its path somewhere else on the same line, or due to
sudden burnt out of a supply transformer leading to inductive surge.  The
light dischage tubes(that are part of television tuners) does not protect
against this type of discharge but does help in the one that occurs prior
to active lightning(corona discharge due to charged cloud passing over the
antenna, but has not accumulated enough charge to cause a strike).  If
you really want to protect your equipment from a direct strike, lightning
arrestor is the only choice(apart from totally disconnecting the equipment
from the electrical outlets).

BTW the link mentioned by denco-ga
http://frentzandsons.com/content/howto/wiring_part4.htm
says the surge suppressor uses a capacitor that acts as a sponge.
This is totally incorrect.  Surge supressors use electronic
metal oxide varistors(varistor meaning variable resistance).
As long as the surge is within joule rating of the surge protector
it will continue to function without "soaking" up any charge.
When an excessive surge strikes, it overwhelms the heat dissipation
capacity of the varistor element, so that it overheats leading to
physical damage(burnt connection within the element)

You could follow this link for more information:
http://www.extremetech.com/article2/0,3973,54336,00.asp