Low Voltage from the Central Air Crashes My Computer

[JRoughgarden]

When our A/C comes on, it pulls a lot of power. The lights dim
momentarily, but all the computers in the house weather it fine, except
one: my homemade Asus 2.67 GHz dual-proc system.

I purchased a Belkin UPS and an APC Line-R voltage regulator. I can see
the indicator light on the APC regulator show low voltage on occasion
when the AC is on. But bottom line is that the APC device cannot make
up for the problem.

Does anyone think the problem is in the power supply on this computer?
If so, is there a model or type that could withstand the voltage drops
better? I can't recall the output of my current model, but I do recall
searching for a quiet one with adequate margin. Would a very much
oversized model do any better?

Any advice is much appreciated.

 

[Paul]

When our A/C comes on, it pulls a lot of power. The lights dim
momentarily, but all the computers in the house weather it fine, except
one: my homemade Asus 2.67 GHz dual-proc system.

I purchased a Belkin UPS and an APC Line-R voltage regulator. I can see
the indicator light on the APC regulator show low voltage on occasion
when the AC is on. But bottom line is that the APC device cannot make
up for the problem.

Does anyone think the problem is in the power supply on this computer?
If so, is there a model or type that could withstand the voltage drops
better? I can't recall the output of my current model, but I do recall
searching for a quiet one with adequate margin. Would a very much
oversized model do any better?

Any advice is much appreciated.

Looking at the installation manual for one of the Line-R products,
it mentions a selector switch. The 110V setting is the lowest one:

80 - 130V (Selector Switch set to 110V)

Now, if your line voltage drops below 80V, even the Line-R is
not going to be able to help.

The power supply by itself, will probably drop out at 90V
or so. Power supplies have a "hold-up time" spec (maybe
17 milliseconds), and usually the hold-up time is sufficient
to withstand the startup transient of a motor starting
elsewhere in the house. But it sounds like your problem is
sustained voltage is not right.

You should be aware, that power companies have been known to
share too many customers with one pole transformer. If the
power company is not applying brown out conditions (reduced
line voltage) on purpose, as part of summer load shedding,
it could be that the transformer on the pole is undersized.

Also, since many devices in a household, will draw extra
current to compensate for reduced voltage, you should
realize that the situation in your house is doing
neither the power company, nor you, any good. Even small
voltage reductions have been known to damage small electric
motors (electric weed whacker brush assemblies). The
motors in your air conditioner could be drawing way
more current than they are supposed to - sure, they
probably have a thermal cutout, but still elevated
temperature is not good for them.

If you own a multimeter and know how to use it safely,
you could measure the line voltage as a function of
time. Try measuring your house voltage when your AC
is off and when it is on. Perhaps if your neighbours
have AC, listen for when their unit goes off, and
make another measurement. Or simply complain to your
utility, and let them figure it out. (Maybe they'll
explain that they have been using brownout in your
area, to fix their summer loading problems.)

If you can demonstrate to the utility, that the fault is
an undersized pole transformer, they should be only too
happy to install a beefier one.

But get it fixed! Trying to hide the problem, could
result in an expensive repair when something else breaks.

Paul

 

[kony]

When our A/C comes on, it pulls a lot of power. The lights dim
momentarily, but all the computers in the house weather it fine, except
one: my homemade Asus 2.67 GHz dual-proc system.

What power supply is it using; make, model, rated wattage
and current per rail?

I purchased a Belkin UPS and an APC Line-R voltage regulator. I can see
the indicator light on the APC regulator show low voltage on occasion
when the AC is on. But bottom line is that the APC device cannot make
up for the problem.

You need another circuit for the AC, there are more devices
than just that one system that shouldn't have such low AC.

Does anyone think the problem is in the power supply on this computer?

It's all shades of grey. Yes, if the system load was less
on that PSU, or if another PSU was capable, that would be
somewhat of a countermeasure for what is still an ongoing
problem with your home wiring.

If so, is there a model or type that could withstand the voltage drops
better? I can't recall the output of my current model, but I do recall
searching for a quiet one with adequate margin. Would a very much
oversized model do any better?

I would have thought you bought a pretty beefy PSU for a
dual 2.7GHz system... a beefier one might help, possibly
even more helpful is one with active PFC but we don't even
know just how low the voltage is getting. I was only
assuming above that all the parts are on the same circuit,
the AC in addition to multiple other systems? Or are the
other systems which survive this problem, not on the same
circuit?

 

[Davy]

For the light to dim (voltage reduction) when the A/cndx comes on is
due to the initail surge taken by the A/cndx motor... the usual cause
is either the wiring or a 'resistive' joint...

this could be a fuse with bad contacts the connection from the outside
mains etc etc all not capable of passing that initial surge current
as the motor starts up from a stall or stopped position.

The wiring and anything in between such as fuses, switches etc should
be able to handle this load at start up.... abit like using thin wire
on a starter motor in a car... you'll eitther get voltage drop or the
wire will 'blow' ~ abit like a fuse would.

This problem is usually related from the outside connection to the
fuse or distributer box in the house.


Davy

 

[meow2222]

When our A/C comes on, it pulls a lot of power. The lights dim
momentarily, but all the computers in the house weather it fine, except
one: my homemade Asus 2.67 GHz dual-proc system.

I purchased a Belkin UPS and an APC Line-R voltage regulator. I can see
the indicator light on the APC regulator show low voltage on occasion
when the AC is on. But bottom line is that the APC device cannot make
up for the problem.

Does anyone think the problem is in the power supply on this computer?
If so, is there a model or type that could withstand the voltage drops
better? I can't recall the output of my current model, but I do recall
searching for a quiet one with adequate margin. Would a very much
oversized model do any better?

Any advice is much appreciated.

As Kony says there are several partial causes, any of which might
account for it alone, or one may be the whole problem. Thankfully
though there is a simple solution in many cases. First, if your pc is
on the same electrical circuit as the ac, move it to another one. Now
if necessary open the psu, discharge the mains reservoir cap, and
connect a 2nd much larger mains reservoir cap to it. This will carry
the psu through brownout for a very limited time. By very I mean a
fraction of a second, but that will generally be enough to get past the
initial surge of the ac compressor.

If you dont want to open the psu, you could use a separate unit that
rectifies the mains plus a big capacitor, and feed the pc off the
resulting dc.

There is a final time-honoured bodge if all else fails: an
autotransformer to add 10% to the Vsupply. Not recommended, not good
practice, but if you really get stuck it does work.

There are also several other approaches, all of which are not as easy
to implement, such as running the pc off a motor generator, using a
brownout detector to momentarily drop cpu freq right down, cut 12v
drive feeds and so on - dont go there, just waffling.


NT

 

[Rod Speed]

As Kony says there are several partial causes, any of which might
account for it alone, or one may be the whole problem. Thankfully
though there is a simple solution in many cases. First, if your pc is
on the same electrical circuit as the ac, move it to another one. Now
if necessary open the psu, discharge the mains reservoir cap, and
connect a 2nd much larger mains reservoir cap to it. This will carry
the psu through brownout for a very limited time. By very I mean a
fraction of a second, but that will generally be enough to get past
the initial surge of the ac compressor.

If you dont want to open the psu, you could use a separate unit that
rectifies the mains plus a big capacitor, and feed the pc off the
resulting dc.

There is a final time-honoured bodge if all else fails: an
autotransformer to add 10% to the Vsupply. Not recommended, not good
practice, but if you really get stuck it does work.

There are also several other approaches, all of which are not as easy
to implement, such as running the pc off a motor generator, using a
brownout detector to momentarily drop cpu freq right down, cut 12v
drive feeds and so on - dont go there, just waffling.

A decent full time UPS would make more sense if a second circuit isnt feasible
and the electrical supply authority cant be monstered into fixing the sag.

 

[w_tom]

Paul, Kony, etc define why that problem may exist. IOW computer
power supply should work just fine even when an incandescent lamp on
that same circuit dims to less than 40% intensity. This power supply
resiliency is even required by industry specs. What voltage - how low
- is that electricity dropping during air conditioner startup?

For example, that wire from wall receptacle probably makes many
connections through other wall receptacles to breaker box. If wire to
each receptacle is pushed into rear, then the connection is often poor
- insufficient. Sufficient for human safety and not a problem for
incandescent bulbs. But insufficient for computer and air conditioner.
Then when air conditioner starts, line voltage drops too low due to
wires not wrapped tight around wall receptacle's side screws.

You can inspect for this quickly by only removing each cover plate.
Is that wire wrapped around side screws on receptacle? If not, then
voltage may drop excessively during air conditioner startup. Therefore
computer might shutdown AND excessive strain is put on air conditioner.

Clearly your one power supply is not as resilient. However is it in
violation of industry standard specs? Learn by lowering voltage until
computer shuts down, then measure that voltage with a meter. A variac
is used to adjust voltage. You probably don't have a variac and may
not have that 3.5 digit multimeter. Numbers from that test are
necessary to properly answer your question. Better is to start with
wall receptacle inspections.

Your symptoms imply a larger problem that may also cause pre-mature
air conditioner failure. Inspect how wires are connected to each wall
receptacle on that circuit. Maybe also verify that wire is firmly
connected to its circuit breaker and that circuit breaker is properly
seated in breaker box - or get someone who knows how to inspect this.
A first and simple inspection that may eliminate the problem and avoid
future failures.

Computer power supply should work just fine even when AC voltage
drops to 90 volts, as Paul noted. Some inferior power supplies (that
sell only on price and watts) don't do this when marketing to people
who then fix a defective power supply with a UPS.

Computer should work just fine even when air conditioner starts and
without a UPS. If not, find reason for that failure before damage
occurs even to air conditioner. Start with easiest first step; wall
receptacle inspection.

 

[w_tom]

Paul, Kony, etc define why that problem may exist. IOW computer
power supply should work just fine even when an incandescent lamp on
that same circuit dims to less than 40% intensity.

Then as I posted this, I realized this is a central air conditioner.
IOW it should not share power with anything else; should be on its own
separate circuit. Therefore the loose connection may be inside breaker
box. Those main wires to mains circuit breaker must be tight. If some
incandescent bulbs dim, and others get bright, then problem may be a
loose neutral.

Also this may be a symptom of a loose or failing wire from utility or
its transformer. After confirming no obvious problems on your side (ie
inside breaker box), then utility may put a monitor on your line to
identify their equipment is OK. IOW that utility testing might find a
failure before failure happens; such as an exploding transformer or a
failed service wire.

Central air starting should not cause lights to dim significantly
which is also why a central air has its own dedicated circuit in
breaker box. That low voltage on startup would be symptoms of another
(possibly more serious) future failure. You don't want to learn
about that failure during a winter blizzard or during a weekend night.

 

[Rod Speed]

Paul, Kony, etc define why that problem may exist. IOW computer
power supply should work just fine even when an incandescent lamp
on that same circuit dims to less than 40% intensity. This power
supply resiliency is even required by industry specs.

Post that bit in the ATX specs, you cant, it doesnt exist.

What voltage - how low - is that electricity
dropping during air conditioner startup?

Unlikely he has got what is necessary to measure that properly.

For example, that wire from wall receptacle probably makes many
connections through other wall receptacles to breaker box. If wire to
each receptacle is pushed into rear, then the connection is often poor
- insufficient. Sufficient for human safety and not a problem for
incandescent bulbs. But insufficient for computer and air conditioner.

Mindless silly stuff.

Then when air conditioner starts, line voltage drops too low due
to wires not wrapped tight around wall receptacle's side screws.

Have fun explaining why it affects the lights.

You can inspect for this quickly by only removing each cover plate.
Is that wire wrapped around side screws on receptacle? If not,
then voltage may drop excessively during air conditioner startup.

Not a clue, as always.

Therefore computer might shutdown AND
excessive strain is put on air conditioner.

Not a clue, as always.

Clearly your one power supply is not as resilient.
However is it in violation of industry standard specs?

Which 'industry standard specs' are these ?

Learn by lowering voltage until computer shuts down,
then measure that voltage with a meter. A variac is
used to adjust voltage. You probably don't have a variac

Yeah, tad of a problem there. Or even know what one is either.

and may not have that 3.5 digit multimeter.

Dont need one of those for that particular test.

Numbers from that test are necessary
to properly answer your question.
Nope.

Better is to start with wall receptacle inspections.

Good way to get him electochuted.

Your symptoms imply a larger problem that may also cause
pre-mature air conditioner failure. Inspect how wires are
connected to each wall receptacle on that circuit. Maybe
also verify that wire is firmly connected to its circuit breaker
and that circuit breaker is properly seated in breaker box

Good way to get him electochuted.

- or get someone who knows how to inspect this.
A first and simple inspection that may eliminate
the problem and avoid future failures.

Unlikely that the problem is that easily fixed.

Computer power supply should work just fine even when AC voltage
drops to 90 volts, as Paul noted. Some inferior power supplies (that
sell only on price and watts) don't do this when marketing to people
who then fix a defective power supply with a UPS.

Computer should work just fine even when air conditioner
starts and without a UPS. If not, find reason for that failure
before damage occurs even to air conditioner. Start with
easiest first step; wall receptacle inspection.

Just ignore this fool's shit.

 

[meow2222]

snip

A decent full time UPS would make more sense if a second circuit isnt feasible
and the electrical supply authority cant be monstered into fixing the sag.

Why do you say that? A capacitor is far cheaper and simpler.


NT

 

[meow2222]

Mindless silly stuff.

not again


NT

 

[Rod Speed]

Why do you say that?

Because its true ?

A capacitor is far cheaper and simpler.

And it isnt trivial to find room for in most modern supplys.

 

[Bob Day]

When our A/C comes on, it pulls a lot of power. The lights dim
momentarily, but all the computers in the house weather it fine, except
one: my homemade Asus 2.67 GHz dual-proc system.

I purchased a Belkin UPS and an APC Line-R voltage regulator. I can see
the indicator light on the APC regulator show low voltage on occasion
when the AC is on. But bottom line is that the APC device cannot make
up for the problem.

Consider an "on-line" "double conversion" UPS, such as the
Tripp Lite SU1000XLa at
http://www.tripplite.com/products/product.cfm?productID=3180

-- Bob Day
http://bobday.vze.com

 

[Paul]

As Kony says there are several partial causes, any of which might
account for it alone, or one may be the whole problem. Thankfully
though there is a simple solution in many cases. First, if your pc is
on the same electrical circuit as the ac, move it to another one. Now
if necessary open the psu, discharge the mains reservoir cap, and
connect a 2nd much larger mains reservoir cap to it. This will carry
the psu through brownout for a very limited time. By very I mean a
fraction of a second, but that will generally be enough to get past the
initial surge of the ac compressor.

If you dont want to open the psu, you could use a separate unit that
rectifies the mains plus a big capacitor, and feed the pc off the
resulting dc.

There is a final time-honoured bodge if all else fails: an
autotransformer to add 10% to the Vsupply. Not recommended, not good
practice, but if you really get stuck it does work.

There are also several other approaches, all of which are not as easy
to implement, such as running the pc off a motor generator, using a
brownout detector to momentarily drop cpu freq right down, cut 12v
drive feeds and so on - dont go there, just waffling.


NT

None of these approaches is something suitable for DIY (do it
yourself) solutions.

First of all, the ATX PSU has rectifiers, an inrush limiter of
some sort, and the main caps. Adding capacitance arbitrarily to
the main capacitor, in an attempt to increase the holdup time,
means the rectifiers are going to get hammered harder when the
PSU is switched on at the back. Maybe this would cause a premature
failure of the inrush limiting solution, or blow the rectifiers,
with fiery results.

I happen to own a 30 year old central air conditioning system.
I find that in fact, my sump pump motor causes a larger flicker
of the lights, than the air conditioner. The air conditioner
uses an interrupter (relay) to switch the power on to the two
motors inside the air conditioner. Then there is a third motor
in the furnace for air circulation. I believe at least two
of those motors are impedance protected, and that leaves the
compressor as the "ugly load".

I'm not sure what kind of motor is in my sump pump. It doesn't
have an external capacitor for starting, and almost reminds
me of an induction motor. But, in any case, the sump pump motor
makes the lights flicker more than the AC does.

And with these loads, I have never had even the slightest hint
of a problem with computers. And I've had a pile of computers
in the house.

That is why I say, the electrical system should be checked
for problems. Sure, it could be a distribution problem within
the house (too small a panel, amateur wiring job etc). Or
it could be a pole transformer problem.

What I'm trying to explain, is a home owner should not be
using a Line-R to fix a problem which is normally handled
adequately by modern computer design. Something is wrong
with the power, and finding the cause is more worthwhile
than placing "bandaids" on every appliance that is having
problems.

The autotransformers I've seen, are unregulated devices.
That means, if you dialed the knob on the autotransformer
to get exactly the right voltage to the load, the instant
the input line voltage changes, the autotransformer will
pass the bonus on to the load. Since the autotransformer
has no regulation, it is not a solution for a sag problem,
if the input voltage changes with time. A device like the
Line-R is closer to being correct for the job, as at least
it attempts to take a wide range input and create a stable
output voltage. But if the Line-R is also dropping the
ball, then the input power is simply too "dirty" to be fixed.

Getting the electrical panel or the pole transformer fixed,
will cost a whole lot less in the long run.

Paul

 

[Rod Speed]

(e-mail address removed) wrote

None of these approaches is something
suitable for DIY (do it yourself) solutions.

True. And there is a significant risk of someone
killing themselves with some of them too.

First of all, the ATX PSU has rectifiers, an inrush limiter of
some sort, and the main caps. Adding capacitance arbitrarily
to the main capacitor, in an attempt to increase the holdup
time, means the rectifiers are going to get hammered harder
when the PSU is switched on at the back.

Thats only theoretically correct. In practice the rectifiers have
so much surge margin in hand that its never a practical problem,
essentially because the rectifiers will always be able to handle
much more surge current than they will experience and the size
of the cap will have been determined by economics with the
PSU that cant handle a mains sag adequately.

That supply will in fact normally have the same rectifiers used
as the ones which can handle the mains sag properly, which
have bigger caps that are the reason they handle the sag fine.

Maybe this would cause a premature

failure of the inrush limiting solution,

Nope, for the reason above.

or blow the rectifiers,

Nope, for the reason above.

with fiery results.

Nope, if the rectifier dies there is no fiery result, it just dies
silently and at worst takes out the input fuse in the process.

I happen to own a 30 year old central air conditioning
system. I find that in fact, my sump pump motor causes
a larger flicker of the lights, than the air conditioner.

Because the A/C is normally on its own circuit because
of its current draw, and the sump pump isnt usually.

The air conditioner uses an interrupter (relay) to switch
the power on to the two motors inside the air conditioner.
Then there is a third motor in the furnace for air circulation.
I believe at least two of those motors are impedance protected,
and that leaves the compressor as the "ugly load".

Sure, but that isnt relevant to the problem that he is seeing.

I'm not sure what kind of motor is in my sump pump. It doesn't
have an external capacitor for starting, and almost reminds
me of an induction motor. But, in any case, the sump pump
motor makes the lights flicker more than the AC does.

Usually because it doesnt have its own circuit.

And with these loads, I have never had even
the slightest hint of a problem with computers.
And I've had a pile of computers in the house.

Yeah, its possible that the only one that does have a problem
with it is a poorly design supply that cant handle the sag.

That is why I say, the electrical system should be checked
for problems. Sure, it could be a distribution problem within
the house (too small a panel, amateur wiring job etc). Or
it could be a pole transformer problem.

What I'm trying to explain, is a home owner should
not be using a Line-R to fix a problem which is normally
handled adequately by modern computer design.

You dont know that is true with the only power
supply that cant handle the sag properly.

There's some pretty shitty designs out there at the low end.

Something is wrong with the power,

Not necessarily, it could just be a poorly designed power supply.

and finding the cause is more worthwhile than placing
"bandaids" on every appliance that is having problems.

There is only one appliance with a problem.

The autotransformers I've seen, are unregulated devices.

Yes, but one will certainly ensure that the minimum
voltage seen during the sag will be increase above what
it would otherwise be. That may be enough with a poorly
designed power supply to see it handle the sag adequately.

The main problem with this approach is that its lousy
value, an autotranformer would normally cost more than
a decent UPS and the UPS will also allow the system to
be shut down gracefully when the mains fails completely
and to ride thru short term loss of mains.

That means, if you dialed the knob on the autotransformer
to get exactly the right voltage to the load, the instant
the input line voltage changes, the autotransformer will
pass the bonus on to the load. Since the autotransformer
has no regulation, it is not a solution for a sag problem,
if the input voltage changes with time.

It can be for the reason Iisted.

A device like the Line-R is closer to being correct for the job,
as at least it attempts to take a wide range input and create
a stable output voltage. But if the Line-R is also dropping
the ball, then the input power is simply too "dirty" to be fixed.

Or it cant handle the sag actually seen.

Getting the electrical panel or the pole transformer
fixed, will cost a whole lot less in the long run.

You dont know that either are actually faulty.

 

[kony]

And it isnt trivial to find room for in most modern supplys.

I believe he was talking about a separate *box* with this in
it. While that would counter excessive short-duration
fluctuations it wouldn't help if the persistent low voltage
condition occurs the whole time the AC is running.

It's also beyond the scope of what most people can safely
implement. While it might be possible to swap in larger
capacitors in place of the originals in the PSU, this would
also provide a slightly longer reserve against short term
voltage depression.

We really need to know more specifics about the problem, he
needs provide more data and possibly take some AC
measurements if it can be done safely. For all we know he
might have a dodgy AC outlet in the wall or a severely
frayed AC power cord though the odds are against it.

 

[Rod Speed]

I believe he was talking about a separate *box* with this in it.

I doubt it, and that would be a terminally stupid approach to suggest if he is.

While that would counter excessive short-duration
fluctuations it wouldn't help if the persistent low voltage
condition occurs the whole time the AC is running.

That last is very unlikely, because he appears to be saying that the PC
runs fun with the AC running, it just reboots when the AC is turned on.

It's also beyond the scope of what most people can safely implement.

Yep, either approach with an extra cap is dangerous to suggest
to someone who you dont know what their capabilitys are.

While it might be possible to swap in larger capacitors
in place of the originals in the PSU, this would also provide
a slightly longer reserve against short term voltage depression.

Yes, and since the other PCs handle the AC coming on fine,
that may be all that is needed with the PC that has a problem.

Corse another power supply is the safest way to get that.

We really need to know more specifics about the problem,
Nope.

he needs provide more data and possibly take
some AC measurements if it can be done safely.

It isnt that easy or cheap to work out what the
mains is sagging down to when the AC comes on.

For all we know he might have a dodgy AC outlet in the wall

Its possible, but unlikely, and trivial to check by trying a different outlet.

or a severely frayed AC power cord though the odds are against it.

Very unlikely and again completely trivial to try that possibility.

Its MUCH more likely that the power supply just cant handle the
mains sag as well as the power supplys in the other PCs can,
and maybe the pole pig is being shared too much etc too.

 

[meow2222]

None of these approaches is something suitable for DIY (do it
yourself) solutions.

Well, perhaps true for some people, but lots of people have done bigger
electrical jobs than that without any problem. I would not consider it
outside of diy by any means for a lot of people. If you read the old
readers sigest household repairs book for example, a standard household
manual for many decades here, they recommend various appliance repairs
that go well beyond fitting new caps.

First of all, the ATX PSU has rectifiers, an inrush limiter of
some sort, and the main caps. Adding capacitance arbitrarily to
the main capacitor, in an attempt to increase the holdup time,
means the rectifiers are going to get hammered harder when the
PSU is switched on at the back.
yes

Maybe this would cause a premature
failure of the inrush limiting solution, or blow the rectifiers,

It is possible, but the chance of this is small. Silicon power diodes
are normally fine in this respect. The OP can either replace the psu or
fit a cap and live with a 1-ish % chance of needing to replace the psu,
so its an obvious win.

with fiery results.

Of the various power diodes I've seen go I cant remember any that went
fierily.

That is why I say, the electrical system should be checked
for problems. Sure, it could be a distribution problem within
the house (too small a panel, amateur wiring job etc). Or
it could be a pole transformer problem.

What I'm trying to explain, is a home owner should not be
using a Line-R to fix a problem which is normally handled
adequately by modern computer design. Something is wrong
with the power, and finding the cause is more worthwhile
than placing "bandaids" on every appliance that is having
problems.

In an ideal world, sure. In the real world the OP has a choice of
various options, and an overhaul of their electrical installation,
upgrading the supply transformer and so on is a hell of a lot more
time, work and money than fitting one capacitor.

I suspect its more likely to be an underperforming psu than the mains
wiring, since its the only appliance that has problems. And since it
runs fine except during the ac start surge, we know a decent sized cap
will fix it.

Since the autotransformer
has no regulation, it is not a solution for a sag problem,

Its not a great solution, but it does work. I'm well aware of the
tolerance / regulation problems they introduce, and the reasons why in
most cases they are successful regardless. I'm not recommending this
approach, a new psu would be better all round.


NT

 

[JRoughgarden]

Thank you all for your advice. I've decided to call an electrician
since I'm not at all comfortable with home wiring. Some of the other
computers that don't fail are on the same circuit, and others are not.
And, it is a central air system on its own circuit. However, it was put
in by some free-lancers, and it may not be wired perfectly. Hence my
call to an electrician. The problem is reproducible, so he ought to be
able to troubleshoot it.