UPSs and surge protection.

[nothing to write home about]

hello everybody.

i was hoping i could get some advice from you lovely people.

i work as tech support for an NGO in guatemala. on one of our sites,
we've been having a lot of problems with PSUs dying. this is mostly
from before i was here, and i haven't been able to do any kind of post
mortem on them.

i've been reading some old threads on these two groups about UPSs and
surge protection, but i wanted to ask if anybody here knows whether the
general advice that was given for national grids in europe and america
has any bearing here. all electrical appliances are run off a single
generator: about a dozen computers, lights, an air compressor, a small
air con unit, a few big refridgerators and some probably
inconsequential medical equipment like small centrifuges and so on.
there are an awful lot of thunderstorms and buildings here are hit from
time to time.

a couple of the computers here are attached to cheap (unearthed) surge
protectors, and none of these have had problems, although it could
easily be coincidence. also, at another of the organisation's sites
there is a similar setup, but all computers are connected to UPSs and
we haven't had any problems with the PSUs.

the weather's very hot and humid, which also causes a lot of general
hardware failures, so it's difficult to pin down what causes what.
although i don't think that the PSU problem is entirely climatic,
because the other site is hotter and more humid, and no problems with
PSUs there.

so what kind of electrical problems are most likely to be causing these
failures and what's the easiest/cheapest way of protecting the
computers we have?

any ideas anybody can throw around would be much appreciated.

ntwha

 

[kony]

hello everybody.

i was hoping i could get some advice from you lovely people.

i work as tech support for an NGO in guatemala. on one of our sites,
we've been having a lot of problems with PSUs dying. this is mostly
from before i was here, and i haven't been able to do any kind of post
mortem on them.

Are they decent name-brand supplies or low-end generics?

i've been reading some old threads on these two groups about UPSs and
surge protection, but i wanted to ask if anybody here knows whether the
general advice that was given for national grids in europe and america
has any bearing here. all electrical appliances are run off a single
generator: about a dozen computers, lights, an air compressor, a small
air con unit, a few big refridgerators and some probably
inconsequential medical equipment like small centrifuges and so on.
there are an awful lot of thunderstorms and buildings here are hit from
time to time.

A primary consideration may be whether the generator has
suitable output, and/or if working properly. An UPS alone,
is by definition not a surge protector, only a power source.
A proper "line conditioner" can be necessary when there are
persistent power problems as would be the case with constant
use of same power (generator). An UPS alone is more of a
power/no-power type device which should not be expected to
continually fix this type of problem. There are higher-end
UPS that may have some continual power conditioning, but
perhaps not as good (and wasted money for the "UPS" part of
the device) given the generator.

a couple of the computers here are attached to cheap (unearthed) surge
protectors, and none of these have had problems, although it could
easily be coincidence. also, at another of the organisation's sites
there is a similar setup, but all computers are connected to UPSs and
we haven't had any problems with the PSUs.

A typical low-cost surge protector has little more than an
MOV, perhaps a slight be of filtration but nothing much. If
yours are more elaborate, then you could try similar
strategy elsewhere, but it would be good to know what the
power state is that you're trying to filter/condition, to
know what strategy to employ.

the weather's very hot and humid, which also causes a lot of general
hardware failures, so it's difficult to pin down what causes what.
although i don't think that the PSU problem is entirely climatic,
because the other site is hotter and more humid, and no problems with
PSUs there.

Are they the same power supplies though? I mean,
make/model, and the systems loading them?

so what kind of electrical problems are most likely to be causing these
failures and what's the easiest/cheapest way of protecting the
computers we have?

any ideas anybody can throw around would be much appreciated.

We'd need more detail on exact failure modes. Some junk
power supply completely omit the incoming AC filter circuits
altogether, it is a very common cost-savings "cheat" on the
bottom-of-the-barrel, junk PSU. If you are able to open and
inspect a few and find the fault, that might be helpful.

 

[Bob]

Are they decent name-brand supplies or low-end generics?

I have a Tripp Lite 750 from Directron:

http://www.directron.com/internet750u.html

An UPS alone,
is by definition not a surge protector, only a power source.

The above-referenced UPS is a surge protector as well as a power
source. It has 4 recepticals for battery power, 4 for utility power
and a telephone line - all surge protected.

I have seen my unit trip into action because of slight variations in
utility power caused by storms. The fault was so slight that the other
computers did not even respond to it, but they did experience it.


--

Map of the Vast Right Wing Conspiracy
http://home.houston.rr.com/rkba/vrwc.html

"Whatever crushes individuality is despotism."
--John Stuart Mill, "On Liberty"

 

[=?ISO-8859-1?Q?Vidar_L=F8kken?=]

hello everybody.

i was hoping i could get some advice from you lovely people.

i work as tech support for an NGO in guatemala. on one of our sites,
we've been having a lot of problems with PSUs dying. this is mostly
from before i was here, and i haven't been able to do any kind of post
mortem on them.

i've been reading some old threads on these two groups about UPSs and
surge protection, but i wanted to ask if anybody here knows whether the
general advice that was given for national grids in europe and america
has any bearing here. all electrical appliances are run off a single
generator: about a dozen computers, lights, an air compressor, a small
air con unit, a few big refridgerators and some probably
inconsequential medical equipment like small centrifuges and so on.
there are an awful lot of thunderstorms and buildings here are hit from
time to time.

a couple of the computers here are attached to cheap (unearthed) surge
protectors, and none of these have had problems, although it could
easily be coincidence. also, at another of the organisation's sites
there is a similar setup, but all computers are connected to UPSs and
we haven't had any problems with the PSUs.

the weather's very hot and humid, which also causes a lot of general
hardware failures, so it's difficult to pin down what causes what.
although i don't think that the PSU problem is entirely climatic,
because the other site is hotter and more humid, and no problems with
PSUs there.

so what kind of electrical problems are most likely to be causing these
failures and what's the easiest/cheapest way of protecting the
computers we have?

any ideas anybody can throw around would be much appreciated.

ntwha

To really get rid of those problems, You need a on-line UPS (Read:
expencive). They run from battery all the time, and thus feeds a
constant and extremly clean voltage.

 

[kony]

I have a Tripp Lite 750 from Directron:

http://www.directron.com/internet750u.html

I've sold those. Don't have any here at the moment but do
have a 500 and 350VA versions. They are the same but for a
larger transformer (and smaller battery). They aren't
continual conditioners, pass AC straight-through with
minimal filtering, nothing more than found on the typical
under-monitor surge suppressor, ie- not sufficient for much.

We may soon hear from W_Tom, who has much info on this
topic... sometimes a little TOO much info but more is better
than too little. ;-)

The above-referenced UPS is a surge protector as well as a power
source. It has 4 recepticals for battery power, 4 for utility power
and a telephone line - all surge protected.

It's not enough. I wouldn't trust it for surge protection
on a known "good" power line, let alone one already suspect.

A proper conditioner is always passing AC thorough the
transformer and it's inverter. Those you linked do not- an
important distinction.

I have seen my unit trip into action because of slight variations in
utility power caused by storms. The fault was so slight that the other
computers did not even respond to it, but they did experience it.

The problem is, that is only a temporary situation, soon
enough your line must, necessarily, work properly again as
they don't have the capacity to run for more than few
mintues (depending on system/load). With the generator OP
has, there could potentially be a constant need for
conditioning. I'm not familiar with state-of-the-art
conditioning but at one point the typical strategy was to
use a mutli-tapped transformer, with most of the taps as
shorter end windings such that the appropriate tap "level"
was switched in to result in correct voltage range. This is
of course in addition to further inductor, capacitor, and
more elaborate subcircuits depending on the quality of the
unit.

 

[Leo Fellmann]

I've sold those. Don't have any here at the moment but do
have a 500 and 350VA versions. They are the same but for a
larger transformer (and smaller battery). They aren't
continual conditioners, pass AC straight-through with
minimal filtering, nothing more than found on the typical
under-monitor surge suppressor, ie- not sufficient for much.

We may soon hear from W_Tom,

If so /please/ set follow-ups to alt.comp.hardware only, because we've
had about enough of him here.

 

[CBFalconer]

i was hoping i could get some advice from you lovely people.

i work as tech support for an NGO in guatemala. on one of our sites,
we've been having a lot of problems with PSUs dying. this is mostly
from before i was here, and i haven't been able to do any kind of post
mortem on them.

i've been reading some old threads on these two groups about UPSs and
surge protection, but i wanted to ask if anybody here knows whether the
general advice that was given for national grids in europe and america
has any bearing here. all electrical appliances are run off a single
generator: about a dozen computers, lights, an air compressor, a small
air con unit, a few big refridgerators and some probably
inconsequential medical equipment like small centrifuges and so on.
there are an awful lot of thunderstorms and buildings here are hit from
time to time.

a couple of the computers here are attached to cheap (unearthed) surge
protectors, and none of these have had problems, although it could
easily be coincidence. also, at another of the organisation's sites
there is a similar setup, but all computers are connected to UPSs and
we haven't had any problems with the PSUs.

the weather's very hot and humid, which also causes a lot of general
hardware failures, so it's difficult to pin down what causes what.
although i don't think that the PSU problem is entirely climatic,
because the other site is hotter and more humid, and no problems with
PSUs there.

so what kind of electrical problems are most likely to be causing these
failures and what's the easiest/cheapest way of protecting the
computers we have?

any ideas anybody can throw around would be much appreciated.

The combination of generators and equipment (motors etc) can mean
monstrous transients. You want good grounding and transient
filters. This means filters, not just MOVs to do some clamping.
Suitable combinations of chokes and capacitors will form PI
filters. You can control common mode stuff with isolation
transformers.

Those healthy transients are very good at punching little holes
through semiconductors, and letting all the good stuff leak out.

After you have the basics handled, you can then add in surge
suppressors and UPSs.

 

[Bob]

I've sold those. Don't have any here at the moment but do
have a 500 and 350VA versions. They are the same but for a
larger transformer (and smaller battery). They aren't
continual conditioners, pass AC straight-through with
minimal filtering, nothing more than found on the typical
under-monitor surge suppressor, ie- not sufficient for much.

All I can go on is the Tripp Lite specifications, which clearly state
that the unit I have does implement surge protection on all its
connections.

The problem is, that is only a temporary situation, soon
enough your line must, necessarily, work properly again as
they don't have the capacity to run for more than few
mintues (depending on system/load).

I got over 15 minutes on my 750 VA unit and the battery was still
charged. I probably could have gone 30 minutes or even longer.


--

Map of the Vast Right Wing Conspiracy
http://home.houston.rr.com/rkba/vrwc.html

"Whatever crushes individuality is despotism."
--John Stuart Mill, "On Liberty"

 

[kony]

All I can go on is the Tripp Lite specifications, which clearly state
that the unit I have does implement surge protection on all its
connections.

True, unfortunately many (even $3 outlet strips) may claim
surge protection with there being little more than an MOV
and fuse.

I got over 15 minutes on my 750 VA unit and the battery was still
charged. I probably could have gone 30 minutes or even longer.

It is a fair thing to do for a system when AC power scenario
is just good/no-good for short amount of time but if the
generator output is constantly (or quite frequently) bad
that may not be sufficient. If this generator is damaging
PCs it may even be better to tackle this at the
generator-end, a "whole site" filtering rather than at each
computer.

 

[w_tom]

Surge protected from which type of transients. They
conveniently forget to mention which type of transients it
does and does not protection from.

 

[w_tom]

Insufficient information means we don't know why those power
supplies are failing. Therefore we don't even know which kind
of transient might be causing damage. For example, a
longitudinal mode transient would pass right through (or
bypass) the always on-line UPS. That UPS would not even see
the transient that moved on through to damage computers.

Much to be learned from a dead body. Which components in
power supply were damaged? Meanwhile, if the generators are
creating a constant spike that is destroying power supplies,
then power strip protectors would be degraded quickly. This
is also why UPS manufacturers also quietly recommend no power
strip protectors on a UPS output. The constant spikes from a
'defective' generator would turn a protector into nothing more
than a power strip - without any external indication even from
the "OK" light.

Much is gained by having everything in the facility
connected to a common earthing ground. This may or may not be
helpful. But again, without knowing which kind of transient
is suspected, then only speculated solutions can be
suggested.

Using the recommendation from CBFalconer, learn if series
mode filters do eliminate generator (or other) transients.
Obtain a series mode protector from companies such as
Zerosurge, Brickwall, or Surgex. These are heavy filters. If
they do or don't stop damage, then important facts have been
learned. Less filters may still leave one guessing.

Another critical fact that Kony asked. "Are they decent
name-brand supplies or low-end generics?" Not all power
supplies contain sufficient internal protection that is
required by industry standards.

 

[nothing to write home about]

Are they decent name-brand supplies or low-end generics?

they're a mixture. there was a brand new dell that lasted about a
week.

A typical low-cost surge protector has little more than an
MOV, perhaps a slight be of filtration but nothing much. If
yours are more elaborate, then you could try similar
strategy elsewhere, but it would be good to know what the
power state is that you're trying to filter/condition, to
know what strategy to employ.

it would be great to know what the power supply is really like, but
there's little prospect of it. i don't have any tools, much money and
i'm not here for much longer. after a *long* time my predecessor
managed to get a little bit of money to spend on the problem so i
basically have to take a guess at what the problem is and find a cheap
way of making it not quite so bad. right now i'm thinking that buying
a
stack of spare, mid-range PSUs seems like my best guess at a way of
improving the situation.

Are they the same power supplies though? I mean,
make/model, and the systems loading them?

the power supplies at the other site are a different make and model,
but
i'd say almost all of them are cheaper and nastier than here. any
given
piece of hardware at the other site is working about 50% of the time.
but the PSUs seem to manage okay.

We'd need more detail on exact failure modes. Some junk
power supply completely omit the incoming AC filter circuits
altogether, it is a very common cost-savings "cheat" on the
bottom-of-the-barrel, junk PSU. If you are able to open and
inspect a few and find the fault, that might be helpful.

okay, i'll try and get my hands on some dead PSUs and let you know if i
find
anything.

also, i don't know if it's relevant, but i almost always get a mild
electric
shock from touching a metal part of a computer case. does that suggest
no
or very poor earth ground on the whole site, or is that just a normal
consequence of not earthing appliances?

thanks for your help.

ntwha

 

[kony]

also, i don't know if it's relevant, but i almost always get a mild
electric
shock from touching a metal part of a computer case. does that suggest
no
or very poor earth ground on the whole site, or is that just a normal
consequence of not earthing appliances?

Usually the opposite, that it's a high-static area and that
the things you're touching ARE closer to ground potential.

A multimeter (at minimum) would be useful to test grounding,
voltage of the generator, output of the power supplies, etc.

 

[twenty-eight days for thirteen grand]

Insufficient information means we don't know why those power
supplies are failing. Therefore we don't even know which kind
of transient might be causing damage. For example, a
longitudinal mode transient would pass right through (or
bypass) the always on-line UPS. That UPS would not even see
the transient that moved on through to damage computers.

yeah, this is a problem: it's quite possible that i'm not going to be
able to test the power supply properly. i realise with the little
information i've given you you can't give me any certain answers, but
i'd be happy if you could suggest some ways of improving the supply
based on what you think could be the problem. i should think your
guesses as to what is wrong would be much better than mine. although
if you can help me to figure out exactly what's wrong that's much
better, of course.

Much to be learned from a dead body. Which components in power supply were damaged?

okay, i've gotten hold of one of the failed PSUs. i also have a
multimeter and a widget for testing the output of power supplies.
please forgive my ignorance of what i'm doing here and if you can give
me any guidance as to what to look for, then that would be appreciated.

according to the power supply tester widget, the +5VSB line and ground
(labelled PG -- that's ground, right?) on the inside of the supply are
working, but there's nothing from anything else. after the PSU had
been plugged in for thirty seconds or so it started smoking and
sparking. the fuse is still intact. there's a patch of the main PCB
which looks like it's suffered a little heat damage, but not much, and
it's not near anything particularly identifiable (just a few resistors
and a capacitor). the fan didn't spin up. there's a little of what
could be corrosion on one of the resistors in the head-damaged area.
there was the usual amount of organic crap in there. i could post a
photo somewhere if that might be useful.

is any of that any help at all? it's quite possible that this one is
just a victim of heat, humidity, age and cheapness.

Meanwhile, if the generators are
creating a constant spike that is destroying power supplies,
then power strip protectors would be degraded quickly. This
is also why UPS manufacturers also quietly recommend no power
strip protectors on a UPS output. The constant spikes from a
'defective' generator would turn a protector into nothing more
than a power strip - without any external indication even from
the "OK" light.

is there an easy way that i could test what state the surge protectors
here are in?

Much is gained by having everything in the facility
connected to a common earthing ground. This may or may not be
helpful. But again, without knowing which kind of transient
is suspected, then only speculated solutions can be
suggested.

sure, i understand that.

thanks for you help,

ntwha

 

[w_tom]

Provided was some very useful information. First to explain
what we now know. Your +5VSB voltage (purple wire) apparently
is good. This voltage is constantly applied to motherboard
even when computer is off but power cord connects to power
source. (+5VSB is why power cord must be disconnected before
doing anything inside a computer.) +5VSB powers the power
supply controller located on motherboard. The power supply
controller tells rest of power supply to power up when !Power
On (green wire) goes from above 2.4 volts to below 0.8 volts.
The three major voltages should then appear on red, yellow,
and orange wires for up to 2 seconds. IOW a meter reading on
each (in same order, it is 3.3, 5, and 12 volts) immediately
when power switch is pressed would provide more useful
information.

If any one voltage does not rise in that less than 2
seconds, then the power supply detects its failure, shuts down
all voltages, and never reports Power OK to the motherboard.
Power Good is the gray wire that should rise above 2.4 volts
if all three voltages are good.

Two reasons for any one voltage not going stable in less
than 2 seconds: 1) internal power supply failure, or 2) a
short circuit or circuit failure on some other part inside
computer that draws too much current from the supply.

Now for getting further information. First, where are the
sparks coming from? Internal power supply board generally has
AC power on one side and DC output voltages on other. In
between are cube-like blocks that are small transformers, and
heatsinks that cool power transistors. Those heatsinks would
electrically be on the AC power side of transformers. Where
in relation to these devices are sparks or smoke coming from?

A failing power supply typically does not spark or outgas.
They fail internally, or they contain circuits that short out
or limit failure - without sparking, smoke, or flame. For
example, could the power supply be sucking in some type of
conductive material such as iron ore dust or steel wool
debris? A likely reason for sparking might be external
contamination getting into the power supply.

Identifying the region of sparking or smoke would be
helpful. Quite normal for some high power components to
darken the PC board. Sometimes a funny color glue is spread
about so that components don't move before being soldered.
However, if a powdery substance found at the base of
electrolytic capacitors (blue or black cylinders printed with
a number of uF and a voltage number), then that might be
another useful fact. Most electronic failures occur without
any external indication. But when a part is clearly damaged,
then useful evidence equivalent to a bullet hole is obtained.

Surge protectors are generally not testable. But then
little reason exists to test a properly sized protector.
Those plug-in protectors are typically undersized. If
internal protector parts vaporize (fail catastrophically),
then protector was grossly undersized when purchased, provided
no effective protection, and was operating well outside of the
MOV manufacturer's specs. For example, if the "OK" lamp says
it is defective, then the protector typically was not
providing protection and typically was defective when
purchased. Properly sized protectors degrade; do not fail
catastrophically. Their degradation must be measured by
applying excessive voltage and measuring resulting current.
Not typically possible in the field; nor should it be
necessary.

BTW, this figure demonstrates how that "OK" indicator lamp
really works. All MOVs (the active protector components)
inside were removed and still the "OK" lamp says the protector
is OK. Picture demonstrated the "OK" lamp only reports one
type of catastrophic failure; reports failure but does not
report a protector as good; does not report normal
degradation:
http://www.zerosurge.com/HTML/movs.html

Furthermore, if plug-in protectors are being used to protect
from constant spikes of a generator, well, those protectors
are too grossly undersized; would degrade to uselessness in
days or months. Extreme spikes from generators must be
solved at the generator either by using better equipment OR by
installing superior filters - such as series mode protectors
cited previously. The "Cadillac" of generators are Hondas.
Not only do they use fuel sparingly. Quality of electrical
output is superior. However computer power supplies should be
more than robust for most generator outputs as long as the
generator properly regulates voltage. 120 volt output should
never exceed 130 volts on the multimeter.

 

[kony]

okay, i've gotten hold of one of the failed PSUs. i also have a
multimeter and a widget for testing the output of power supplies.
please forgive my ignorance of what i'm doing here and if you can give
me any guidance as to what to look for, then that would be appreciated.

according to the power supply tester widget, the +5VSB line and ground
(labelled PG -- that's ground, right?) on the inside of the supply are
working, but there's nothing from anything else.

This "widget" is a supposed power supply tester that is
known to turn on the PSU by itself? Probably so, just
clarifying it.

after the PSU had
been plugged in for thirty seconds or so it started smoking and
sparking. the fuse is still intact. there's a patch of the main PCB
which looks like it's suffered a little heat damage, but not much, and
it's not near anything particularly identifiable (just a few resistors
and a capacitor). the fan didn't spin up. there's a little of what
could be corrosion on one of the resistors in the head-damaged area.
there was the usual amount of organic crap in there. i could post a
photo somewhere if that might be useful.

A good picture (posted elsewhere then linked here) may
indeed help, after as much of that "organic crap" as
possible has been blown out. However, if it is more than
just dust, for example shards of paper or plastic (remains
of an exploded capacitor perhaps) do mention that as well.

is any of that any help at all? it's quite possible that this one is
just a victim of heat, humidity, age and cheapness.

Sure it's possible. What brand and system was it powering?
A good picture will also tell us a fair bit about it. That
is, a high-resolution top-down shot with the cover off.

 

[nothing to write home about]

thanks everybody for the time you've given up to help me, it really is
appreciated.

okay, i have a picture of the failed supply posted here:
http://home.silencescape.net/~slack/

i plugged it back in just now, and the smoke was coming from an area
just to the right of the bottom of the fan, as you look at it in the
picture i posted. somewhere between a capacitor which is almost
entirely obscured by dirty black and white cables and a what i take to
be a small yellow transformer marked MITAC [numbers] I-MAC 9946. being
a bit of a pussy and not having safety glasses i didn't get too close
to it whilst it was smoking but i guess i can further risk my eyesight
if that's not specific enough.

i've also found a multimeter and had it plugged into the mains for a
couple of days. i know this won't tell you anything about transients,
but the power is pretty steady around 110V +- 2V except when the air
compressor (for a dental clinic) is turned on or off. when it's turned
on, the power usually dips into the mid-eighties, worst i've seen was
83V. when it comes back on, it usually peaks at a round 120V. these
transients last for a few minutes, gradually returning to 110 as the
generator recovers.

another potentially interesting thing is that the dell that originally
blew a PSU within a week has been failing again, this time it just
switches itself off when the power dips below 90V or so. i tried
adding a UPS that may or may not be working (the battery is fully
charged at least) and it didn't make any difference -- still turns
itself off when the power dips. i don't know whether this implies a
problem with the UPS as well as an oversensitive PSU, or whether a
working UPS would fail to clean the supply enough for a PSU that's so
tetchy. i've just got my hands on some more UPSs temporarily, which i
have slightly more confidence in, and i've hooked one of those up to
the dell. it hasn't failed yet, but it's only been running an hour or
so. my monitor also turns itself off for a few seconds if the power
dips below 90V.

the only brandname-like-marking i could find on the generator is
"tradewinds". there does appear to be an attempt at an earth grounding
a few metres from the generator with a copper pipe inserted into the
ground. there is absolutely no chance that we could find money to get
a new generator.

This "widget" is a supposed power supply tester that is known to turn on the
PSU by itself? Probably so, just clarifying it.

yeah, that's right.

is any of this of any use?

dan

 

[w_tom]

Can't really tell what that damaged part is doing although I
suspect it is a diode. IOW the part identifier may be Dx. Is
it D7?

A guess: diodes are part of the 'always on' +5VSB power
supply. Always powers motherboard if power cord is
connected. Those diodes may even connect to that green
electrolytic capacitor - a cylinder just right of the diodes.

One possible scenario. If the generator drops to 83 volts -
has a very slow response time - then it is also likely that
the generator voltage rockets above 130 volts for a second
when a major load is removed. Industry standards say a power
supply should withstand 145 volts for up to 0.5 seconds. If
145+ overvoltage is longer, then that overvoltage may have
damaged (for example) the green electrolytic. Now the
electrolytic capacitor is shorted therefore shorting out and
vaporizing D7(?).

Some more expensive UPSes will lower or cut off that
overvoltage quickly. Cheaper ones may not. One suspect
reason for failure: generator is too slow to respond to a
major load loss, goes into overvoltage as its internal coils
try to discharge (load dump) into a load too small (the
computer), and then create a short overvoltage spike that
damages power supply.

Either the load must be removed slowly (and I don't know
from here how you can do that), or computers must be
completely unplugged (disconnected at power cord) before a
major load is removed.

If your meter has a Hi/Lo function, it would lock onto and
display only the highest voltage. An overvoltage number would
confirm that load (air compressor) removal is indeed creating
excessive and destructive voltage spikes. Spikes destructive
to the computer but way too low to even be seen by a power
strip (or UPS) surge protection circuit, and too short to be
observed by a meter without that Hi/Lo latching display
option.

Voltage dropping below 85 would cause computer shutdowns.
IOW if the computer's power supply Power OK (gray) wire drops
below 2.4 volts, then power supply detected insufficient line
voltage. CPU then shuts down in response to that power supply
(gray wire) signal. Computers (and monitor) are expected to
shut down (without damage) if line voltage drops to a voltage
somewhere below 90 volts. If the UPS still lets voltage drop
below 90, then the UPS is not working properly. Either UPS is
defective or the generator output is too distorted (un-sine
like) for that UPS.

Earth ground alone is not sufficient. Each appliances
safety ground (the third prong) must be connected together AND
connect to that earth ground. The earth ground would provide
human safety. All safety grounds wired together would also
solve other transistors problems.

Notice those parts labeled KP1010. Those would be
optocouplers. They provide galvanic isolation between high
voltage mains and low voltage DC circuits. Notice also the
slice in PC board beneath each optocoupler. This space and
slice are part of the protection system that keeps 120 VAC out
of computer. IOW those blown diodes are on the 120 volt
side. The transformer and optocouplers make galvanically
isolated connections between DC output voltage and AC mains
input. Damage occurred on AC mains side of power supply.

thanks everybody for the time you've given up to help me, it really is
appreciated.

okay, i have a picture of the failed supply posted here:
http://home.silencescape.net/~slack/

i plugged it back in just now, and the smoke was coming from an area
just to the right of the bottom of the fan, as you look at it in the
picture i posted. somewhere between a capacitor which is almost
entirely obscured by dirty black and white cables and a what i take to
be a small yellow transformer marked MITAC [numbers] I-MAC 9946. being
a bit of a pussy and not having safety glasses i didn't get too close
to it whilst it was smoking but i guess i can further risk my eyesight
if that's not specific enough.

i've also found a multimeter and had it plugged into the mains for a
couple of days. i know this won't tell you anything about transients,
but the power is pretty steady around 110V +- 2V except when the air
compressor (for a dental clinic) is turned on or off. when it's turned
on, the power usually dips into the mid-eighties, worst i've seen was
83V. when it comes back on, it usually peaks at a round 120V. these
transients last for a few minutes, gradually returning to 110 as the
generator recovers.

another potentially interesting thing is that the dell that originally
blew a PSU within a week has been failing again, this time it just
switches itself off when the power dips below 90V or so. i tried
adding a UPS that may or may not be working (the battery is fully
charged at least) and it didn't make any difference -- still turns
itself off when the power dips. i don't know whether this implies a
problem with the UPS as well as an oversensitive PSU, or whether a
working UPS would fail to clean the supply enough for a PSU that's so
tetchy. i've just got my hands on some more UPSs temporarily, which i
have slightly more confidence in, and i've hooked one of those up to
the dell. it hasn't failed yet, but it's only been running an hour or
so. my monitor also turns itself off for a few seconds if the power
dips below 90V.

the only brandname-like-marking i could find on the generator is
"tradewinds". there does appear to be an attempt at an earth grounding
a few metres from the generator with a copper pipe inserted into the
ground. there is absolutely no chance that we could find money to get
a new generator.

 

[kony]

i've also found a multimeter and had it plugged into the mains for a
couple of days. i know this won't tell you anything about transients,
but the power is pretty steady around 110V +- 2V except when the air
compressor (for a dental clinic) is turned on or off. when it's turned
on, the power usually dips into the mid-eighties, worst i've seen was
83V. when it comes back on, it usually peaks at a round 120V. these
transients last for a few minutes, gradually returning to 110 as the
generator recovers.

Both you and a dental clinic are both running on generator
power that regularly drops to 90V & below?

You need a whole-site conditioner or a better generator.
Simply buying another power supply is not a solution, nor is
an UPS or two.

I'm not certain but it may help to buy further power
supplies with Active PFC feature.

 

[Guest]

okay, i have a picture of the failed supply posted here:
http://home.silencescape.net/~slack/

i plugged it back in just now, and the smoke was coming from an area
just to the right of the bottom of the fan, as you look at it in the
picture i posted. somewhere between a capacitor which is almost
entirely obscured by dirty black and white cables and a what i take to
be a small yellow transformer marked MITAC [numbers] I-MAC 9946. being
a bit of a pussy and not having safety glasses i didn't get too close
to it whilst it was smoking but i guess i can further risk my eyesight
if that's not specific enough.

That supply has an elaborate surge filter, consisting of the 2 coils at
the top of the picture and the blue rectangular capacitor between them,
plus the 2 blue disk capacitors at the lower left, on the AC
receptacle. But it may not work well without a grounded electrical
system.

Some switching power supplies are prone to breaking down with high
input voltage, and it seems the +5V standby part breaks down most,
probably because it's left on all the time. I'd replace the burnt
parts, the power transistor (next to the smaller transformer), and the
green electrolytic capacitor, using a low ESR type with a slightly
higher voltage rating than the existing one, and see what happens.