Even a good quality PSU can fail

[philo]

About 10 years ago I built my wife a top of the line XP_64 machine which
included an Antec 500 watt PSU

After I built her a new machine a few years ago, I kept the old one in
my office just for use as a backup machine. Woke up yesterday morning to
a slight burned smell. That good 500 W psu apparently died in it's
sleep. (Unit was plugged into a UPS, so not likely to have been a surge.)

Replaced it and when I turned the machine on the video was scrambled
even before it started to load the OS...so replaced the video card and
tried again.
Dang..that PSU also took out the DVD and hard drive!

One of the worst disasters I've seen. The machine is now running Win7,
so XP_64 is a goner. (The Win7 drive was fortunately not hooked up at
the time of failure.)

Good spare machine.

 

[Paul]

About 10 years ago I built my wife a top of the line XP_64 machine which
included an Antec 500 watt PSU

After I built her a new machine a few years ago, I kept the old one in
my office just for use as a backup machine. Woke up yesterday morning to
a slight burned smell. That good 500 W psu apparently died in it's
sleep. (Unit was plugged into a UPS, so not likely to have been a surge.)

Replaced it and when I turned the machine on the video was scrambled
even before it started to load the OS...so replaced the video card and
tried again.
Dang..that PSU also took out the DVD and hard drive!

One of the worst disasters I've seen. The machine is now running Win7,
so XP_64 is a goner. (The Win7 drive was fortunately not hooked up at
the time of failure.)

Good spare machine.

Antec doesn't make their own power supplies.

They contract others to do it.

I've lost two Antecs made by Channelwell. If
you look inside the power supply (lid off, don't
touch), you may see "CWT" printed on a transformer
or two. Stands for "ChannelWell Technologies".

They do use other suppliers. Perhaps Delta was used
for the Earthwatts ?

I think Channelwell can make good product when you
pay them. If you insist on a low build price, they
put the nastiest caps in there they can.

Now, as to your symptoms, those are disturbing, because
it means the supply had no OVP (overvoltage protection).
The easiest way to do what happened to you, is have
the feedback path on the supply cry out "more coal, more coal",
causing the switching circuit to "go to the wall" in
terms of driving the primary side. All rails rise
on the output side. A well designed supply senses the
out-of-spec output condition, and removes drive
from the switching transistors. That should be a
separate control path, a "monitor" that has a
"kill switch" it can use.

It pays to check the specs, to see if OCP, OVP, and
the like, are listed in the specs. The last time
we heard of a lot of failures of that type, it
was a Bestec branded design doing it (with no
OVP at all).

*******

In the case of the ATX supply, there are two switching
transistors on the primary side of the isolation
transformer, that make an AC waveform. Power can't
pass from the primary to the secondary, unless the
waveform remains AC (alternating current). So to
implement an OVP/OCP, all you need to do is stop
the switching circuit from making AC. Even if a
primary side transistor fails ON, if the second
transistor is not alternating, then no power ends
up on the other side. It means though, that the
control path has to be opto-isolated, as
anything passing from (isolated) secondary
to primary side, has to honor the isolation.
The supply can be tested with up to 1100 volts (HiPOT),
to make sure the secondary is well-isolated
from the primary.

Next time, *explicitly* check the supply specifications
for overcurrent and overvoltage protection. Now that
you know what an Antec is really worth.

*******

Now that hard drive PCBs are inverted, you'll have
no chance to check for burned components on it.
There are two transient suppressors near the
power inlet terminals. They're there to stop
15v inductive ringback from damaging stuff
(on the 12V motor rail). When a sustained overvoltage
condition exists, the protection device on the 12V and
5V rails, should get burned. If you've given up on
the hard drive, you can flip the PCB over and take
a look.

If you could find a matching PCB, it's possible a
new PCB would make the drive work. But what are the
odds of finding the exact right PCB ? The heads
might be far enough from the event, to have survived,
so all your data is still there.

Paul

 

[philo]

On 02/12/2015 09:24 AM, Paul wrote:

In the case of the ATX supply, there are two switching
transistors on the primary side of the isolation
transformer, that make an AC waveform. Power can't
pass from the primary to the secondary, unless the
waveform remains AC (alternating current). So to
implement an OVP/OCP, all you need to do is stop
the switching circuit from making AC. Even if a
primary side transistor fails ON, if the second
transistor is not alternating, then no power ends
up on the other side. It means though, that the
control path has to be opto-isolated, as
anything passing from (isolated) secondary
to primary side, has to honor the isolation.
The supply can be tested with up to 1100 volts (HiPOT),
to make sure the secondary is well-isolated
from the primary.

Next time, *explicitly* check the supply specifications
for overcurrent and overvoltage protection. Now that
you know what an Antec is really worth.

*******

Now that hard drive PCBs are inverted, you'll have
no chance to check for burned components on it.
There are two transient suppressors near the
power inlet terminals. They're there to stop
15v inductive ringback from damaging stuff
(on the 12V motor rail). When a sustained overvoltage
condition exists, the protection device on the 12V and
5V rails, should get burned. If you've given up on
the hard drive, you can flip the PCB over and take
a look.

If you could find a matching PCB, it's possible a
new PCB would make the drive work. But what are the
odds of finding the exact right PCB ? The heads
might be far enough from the event, to have survived,
so all your data is still there.

Paul

As always, thanks for all the great info.

Up until yesterday, the only time I've seen a PSU take out a lot of
other components has been with eMachines. It has been very common to see
bad mobo along with a bad PSU.
Foolishly I thought with an Antec, I'd have a superior product.

I ended up putting a PSU in the machine from a Compaq server. It was too
large for me to use anywhere else. so at least I had a chance to get it
out of my "inventory".

All data on the drive were backed up so there is no need for me to
recover anything...but I pulled the control board off the drive just to
see if it was as simple as some blown TVS...however two of the I.C's
were visibly burned.


I also saw a video stating that replacing the board with another from
the same model drive...will not necessarily work... as calibration data
is stored on the board. The video I saw was made by a data recovery
outfit so they may have done that to encourage folks not to
try...however the video did look authentic.


Considering the machine is about ten years old, the fact that it failed
is not quite the end of the world. BTW: The Asus mobo was not damaged in
the least so I assume that it must have over-voltage protection.

 

[Paul]

On 02/12/2015 09:24 AM, Paul wrote:






As always, thanks for all the great info.

Up until yesterday, the only time I've seen a PSU take out a lot of
other components has been with eMachines. It has been very common to see
bad mobo along with a bad PSU.
Foolishly I thought with an Antec, I'd have a superior product.

I ended up putting a PSU in the machine from a Compaq server. It was too
large for me to use anywhere else. so at least I had a chance to get it
out of my "inventory".

All data on the drive were backed up so there is no need for me to
recover anything...but I pulled the control board off the drive just to
see if it was as simple as some blown TVS...however two of the I.C's
were visibly burned.


I also saw a video stating that replacing the board with another from
the same model drive...will not necessarily work... as calibration data
is stored on the board. The video I saw was made by a data recovery
outfit so they may have done that to encourage folks not to
try...however the video did look authentic.


Considering the machine is about ten years old, the fact that it failed
is not quite the end of the world. BTW: The Asus mobo was not damaged in
the least so I assume that it must have over-voltage protection.

The motherboard can survive, because it uses secondary regulation.

The 12V on the ATX12V, goes through the VCore switcher before
it gets to the CPU. The CPU might never see the insult (so
your CPU survives).

When the motherboard needs 1.5V or 1.8V, those aren't available
from the ATX supply. They're made by local regulator chips.

Modern DIMM supplies are now switchers, working the same
way that Vcore works. Something like 12V is their input,
and the switcher makes the lower voltage the DIMM needs.

The motherboard does not have as a design objective
"no damage from overvoltage". But, because of the amount
of local regulation, portions of it are resilient. And
you may find, even if the motherboard is toast, the
CPU and DIMMs were well protected, and they can be
reused.

So even an Emachine with a Bestec, can't blow everything.
You might be able to keep the CPU and DIMMs. While the
hard drive is blown for sure. That Bestec taught a lot
of techs about the importance of supplies with OCP
and OVP. So many fried computers ...

Paul

 

[philo]

The motherboard can survive, because it uses secondary regulation.

The 12V on the ATX12V, goes through the VCore switcher before
it gets to the CPU. The CPU might never see the insult (so
your CPU survives).

When the motherboard needs 1.5V or 1.8V, those aren't available
from the ATX supply. They're made by local regulator chips.

Modern DIMM supplies are now switchers, working the same
way that Vcore works. Something like 12V is their input,
and the switcher makes the lower voltage the DIMM needs.

The motherboard does not have as a design objective
"no damage from overvoltage". But, because of the amount
of local regulation, portions of it are resilient. And
you may find, even if the motherboard is toast, the
CPU and DIMMs were well protected, and they can be
reused.

So even an Emachine with a Bestec, can't blow everything.
You might be able to keep the CPU and DIMMs. While the
hard drive is blown for sure. That Bestec taught a lot
of techs about the importance of supplies with OCP
and OVP. So many fried computers ...

Paul

Yep, even the eMachines with bad mobos still had good cpu and RAM

The only CPU I ever cooked was an AMD-450 where I had the mobo jumper
set for too high of a voltage...and it still took six months to kill it/

 

[philo]

Antec doesn't make their own power supplies.

They contract others to do it.

I've lost two Antecs made by Channelwell. If
you look inside the power supply (lid off, don't
touch), you may see "CWT" printed on a transformer
or two. Stands for "ChannelWell Technologies".

Just opened it up....the transformers are Viking



BTW: Now would be a good time to clone the drive on my main machine.


Are those WD "red" drives any good?

At any rate I'm not planning on going SSD

 

[Paul]

On 02/12/2015 09:24 AM, Paul wrote:


Just opened it up....the transformers are Viking



BTW: Now would be a good time to clone the drive on my main machine.


Are those WD "red" drives any good?

At any rate I'm not planning on going SSD

WD has too many colors for my liking. Marketing fail.

*******

http://forums.anandtech.com/archive/index.php/t-2079127.html

"Antec has used FSP, Delta, Seasonic, CWT, Enhance, HEC"

But I haven't a clue what "Viking" should be telling me.
Doesn't ring any bells.

I've had an Antec HEC here, and that was OK. Still have
it around here somewhere. For a time, HEC tried to sell
under their own name. I think they were bought by
Compucase ?

Since I stopped buying Antec, I haven't had an opportunity
to experience some of the other items in the list. You
can get Seasonic under its own brand name. And Fortron
Sparkle, you may have run into them somewhere too (they're
pretty good and probably won't pop on you - that's why I bought
an SPI supply here). The Fortron stuff usually have
short cables on them (i.e. not good if using a bottom
mounting case and needing to run cables diagonally up
to the other corner).

http://en.wikipedia.org/wiki/FSP_Group

The supplies with the most operating hours on them
here now, are Enermax. I got a couple of those when
they were on sale years ago, and they've been
fault free.

Paul

 

[philo]

http://forums.anandtech.com/archive/index.php/t-2079127.html

"Antec has used FSP, Delta, Seasonic, CWT, Enhance, HEC"

But I haven't a clue what "Viking" should be telling me.
Doesn't ring any bells.

I've had an Antec HEC here, and that was OK. Still have
it around here somewhere. For a time, HEC tried to sell
under their own name. I think they were bought by
Compucase ?

Since I stopped buying Antec, I haven't had an opportunity
to experience some of the other items in the list. You
can get Seasonic under its own brand name. And Fortron
Sparkle, you may have run into them somewhere too (they're
pretty good and probably won't pop on you - that's why I bought
an SPI supply here). The Fortron stuff usually have
short cables on them (i.e. not good if using a bottom
mounting case and needing to run cables diagonally up
to the other corner).

http://en.wikipedia.org/wiki/FSP_Group

The supplies with the most operating hours on them
here now, are Enermax. I got a couple of those when
they were on sale years ago, and they've been
fault free.

Paul

I bought two new EVGA PSUs about six months ago.

Used one for a repair and I have one left

Specs specifically say over-voltage protection (among other things)

 

[danilo]

philo, these power supplies have never earned a good reputation in my area because of the reason mentioned above, they don’t make it themselves, they contact someone else to make it for them in a given unit price. The issues that you have been facing are mostly those of over voltage. Since OVP module of the supply might have failed and instead of OVP at each output, I think they might have just been using one for all outputs.

Regards,
Danilo C.
http://blog.7pcb.com